Why Reviewers Benchmark CPUs @ 1080p: Misconceptions Explained

[@puokki6225]

Anyone saying 1080p is "dead" needs a serious reality check. According to Steam hardware stats, ~65% of their userbase still uses it. 1440p is at 11% and 4K at a tiny 2.7%. For reference, 1366x768 is at 5.9%. The high end market, which realistically almost all of the DIY PC market currently is, is very niche even when it comes to gaming in general.

[@vsaucepog7829]

Just people looking for confirmation bias lol. 1080p was and still will be king for a long time.

[@SpartanDusk]

I blame the console advertisements and TVs, consoles advertise running games at 4K, while true sometimes, is not always true. That said, my greedy mind would want a PC that would outperform a console so a PC with at least a 5600X/12100 and a 3060Ti, with either a 1080p or 1440p monitor.

[@dante19890]

1080p is dead for enthusiasts unless ur heavy into competetive esports with a 360hz monitor. ITs enthusiast and not the average consumer who are interested in these benchmarks

[@vsaucepog7829]

@@dante19890 yeah, but that's a very small minority of the community.

[@lucianocortes8636]

I bet those same idiots use a 23" monitor to play at 1440p or higher, rofl.

[@chrishexx3360]

This video should be classed as a mandatory watch before buying new CPUs. Absolute gold standard content.

[@mircomputers]

and GPUs

[@MoChuang343]

19:48 I also understand why viewers want to see what kind of performance upgrade a new CPU will provide for their specific configuration but you can quite easily work that out with just a bit of digging. All you have to do is work out how much performance your graphics card will deliver in a chosen game when not limited by CPU performance this can be achieved by looking at GPU reviews simply check the performance at your desired resolution then cross-reference that with the CPU data from the same game. Can you please clip this and post it like in all of your video descriptions?

[@seppomuppit]

mandatory before getting your youtube benchmarking commentor's licence

[@TechRIP]

@@MoChuang343 So you want people to do the work for the reviewers? HUB gets paid to do reviews. Hold them accountable for doing them correctly instead of being lazy.

[@MoChuang343]

@@TechRIP I would argue its HUB's job to provide the resources and data in order for consumers to make an informed decision and that is exactly what they do. Reviewers cannot possibly make a video on every single possible hardware configuration. Even with just 3 generations of CPU and GPU releases, that would be an insane combinatorial number of possibilities. But by making videos that shows the upper limit of a CPU, and the upper limit of a GPU (at given settings), then you can easily determine if a system with that combo is CPU or GPU limited in that game. And with that info, you can pretty quickly determine whether a CPU or GPU upgrade will help.

[@antonyhermens9985]

Wow Steve... you have a enormous amount of patience explaining these tests. Thank you for the effort and don't get burnout man.thanks again

[@seylaw]

This saves him a tremendous amount of explanations in the comments though. :D

[@justinvanhorne9958]

@@seylaw It really doesn't. the new human standard of internet dweller that is being targeted here simply is incapable of listening. As I stated elsewhere, you could produce 100 of these videos consecutively and one benchmark gets released and the comments would be flooded. This is what happens when you attempt to assign explanation and logic to situations that ask for none. People are stupid, gamers that walk into a shop and drop 6k USD on a prebuilt are by definition ignorant. It's not their fault, and its not their entire existence, its just a way of making you understand what you are up against here. With the ever increasing gap of knowledgeable techies to ones that were simply brought up in it, believe me I can look at my kids and see it. Intelligence, logic, complex thoughts can take many forms, when things are taken for granted, they just want them "to work". Now if you'll excuse me, that's enough PC time for today, time to go back to the real events of impending financial collapse of "first world" economies and all the other good things that go with capitalistic societies.

[@seylaw]

@@justinvanhorne9958 Sure, but he can always point to this video. You can't be safe of ignorant people, right. But I support it that he tries to explain this to educate people that are willing to learn and to listen.

[@justinvanhorne9958]

@@seylaw that is the unfortunate thing, those educated enough to know better, probably aren't spouting off in the comments about something passionately that is the equivalent of putting balls in holes.

[@bestrum1188]

Videos like this are so important and I just want to thank you for all the content you produce. This has been my favorite hardware channel for many years and it is so great to see you getting close to the 1mil milestone.

[@AM-qc3vk]

The Call of Duty Modern Warfare 2 benchmark has a cool feature that shows you how many fps the cpu delivers vs the gpu. Pretty cool to see that data directly in a benchmark and to see where the bottleneck is in your system.

[@invalid8774]

Well its usually not that hard to spot the bottleneck. If your gpu is at over 95% load, its a gpu bottleneck. If not, its most likely a cpu bottleneck.

[@pendulum1997]

@@invalid8774 What?! My 3070ti is a bottleneck with your logic then?

[@ungoyone]

@@pendulum1997 Why wouldn't it be? It's not a high-end card.

[@dnalekaw4699]

@@pendulum1997 well yeah of course it is

[@MoltenPie]

@@pendulum1997 depends on the resolution and settings. Your 3070ti can bottleneck even an i5 2500k if you push for 8k ultra. And vice versa if you set 720p with 25% resolution scale on low settings even an i9 13900ks would probably become a bottleneck with your 3070ti.

[@SashimiSteak]

In a nutshell we are trying to compare "maximum performance potential" of CPUs, so we need to do our best to create a CPU bottleneck scenario. Edit: to elaborate: The test is an artificial scenario designed to compare, not to show "real world performance". Because CPU/GPU utilisations varies from game to game too much, "real world performance" will swing up and down so much and will never show any meaningful patterns. The CPU you'd want to buy depends wholly on what games you play, as such it's impossible for reviewers to tailor made their review based on your needs. "maximum performance potential" is a far more universally meaningful metric. You can actually extrapolate real world scenarios by taking the "maximum performance potential" shown here for games you play, find a review for your GPU somewhere and find out how many frames it can push for those games, then match the closest performing CPU / GPU. This is one way to optimise your purchases. But you still need to do some calculations, can't expect the data to be spoon fed to you.

[@jason6296]

^^^^^^ ??????? lol

[@chadhowell1328]

@roenieI’m sorry but what?

[@GewelReal]

@Roenie It will absolutely matter. Games not only get more GPU demanding but also CPU demanding. Otherwise we'd be still using Pentium III's

[@araqweyr]

​@Roenie It's not like games only use GPU, or rather CPU isn't needed only to shove data into GPU. There are also some game related computations on the CPU and it also has to be done for each frame. We might not be able to see the difference right now but if (or rather when) developers start better utilizing CPU then gap between CPUs will start to widen

@Roenie Here in PAL region we used to play at 24fps in 576i, so we've not just went a long way up in graphics quality and resolution, but framerate as well (I use to target at 60fps in my PC, but my screen is capable of going up to 144). Different starting framerate and resolution in NTSC, but same history.

[@Cinetyk]

The thing to understand is that one thing is a CPU review/benchmark, another thing is building a "reasonable" configuration (CPU+GPU) and see what kind of performance you get. They're two different kind of "studies", looking into related but different things. Like Steve said, you can combine both CPUs and GPUs reviews to inform on what kind of performance you can get with your desired pairing. I hope you continue to do the "scaling" videos where you see how a CPU scales with increasing GPUs and vice-versa - this truly makes it able to get a clear picture about performance.

[@markhackett2302]

It is far worse than that. WHAT GPU is there? "Oh, it has to be an NVidia one, because they have like 10x the market share of AMD!!!", and since the speed of a 6600XT and a 3060 is about the same, there won't be an AMD card on the roster. And any balanced system will spend money on a GPU rather than a CPU, and bottleneck to "the CPU is largely irrelevant", because even in "cpu tests", the GPU bottlenecks first. Indeed the GPU is rated not to the CPU but the monitor, so the CPU+GPU combination is 100% irrelevant, it is the GPU+Monitor that should be combined. So it isn't "who buys a low end 120$ CPU to use with the 1600$ 4090", it is "who buys a 4090 but runs it on a 1080p monitor?"

[@Spido68_the_spectator]

@@markhackett2302 they used the 6900 XT for 1080p testing for a while

[@aratosm]

No one likes you.

[@MarcABrown-tt1fp]

@Mark Hackett You simply don't understand that in order to find out how powerful any given Cpu is in single core or multi core performance, you can't be GPU bound. The resolution is irrelevant in Cpu testing.

[@GeneralFuct]

​@@MarcABrown-tt1fp the data he presented literally shits on everything you just said

[@4zur3]

The concept is to never limit one tested part by its counterpart. Always test a gpu with most performant cpu. Vice versa for cpus, testing at low resolution so you are not limited by gpu. This concept holds for decades. Every gamer should deeply understand by taking time to learn this.

[@r1pperduck]

Takes no time at all, it's a very simple concept and you explained it perfectly.

[@meurer13daniel]

Agree. If you are entering the pc gaming market, you should at least know how hardware works. It doesn't take time to learn and you will buy the best product for you.

[@sammiller6631]

performant? shouldn't that be "performing"?

[@YurandX]

@@sammiller6631 artist can be performing, cpu is just most performant. idk if its correct but it makes sense to me

[@meurer13daniel]

@@sammiller6631 Performing = verb Performant = adjective Performant in this context is the correct word

[@PK-lk5gs]

Great piece of testing! "The games you play aren't the same games everyone plays" - I'm shocked to realize how this very basic point is not obvious to everyone. Thank you very much, Steve!

[@flameshana9]

It is a problem when _everyone_ assumes you need 90+ fps to game.

[@masudparvejsani]

@@flameshana9 It is a problem when everyone assumes you need to max out everything you see in graphics settings.

[@TheDude50447]

As an RTS player I know this better than most. Some people find it hard to believe that I generally favor cpu over gpu performance for gaming.

[@Machinens]

​@@TheDude50447 same thing with MMO's, I need more CPU horsepower for VFX and loading into the areas and players as well

[@darthdadious6142]

The absolute most useful info was his recommendation to cross reference cpu vs gpu charts to match up your cpu/gpu combo. I was really wanting to know how to find the maximum GPU my CPU will benefit from. Now I know how to do that. I just wanted a simple chart with all cpu/gpus combos to make it easy. Is that too much to ask? lol, kidding. That would be a nightmare for anyone to compile every time a new cpu/gpu came out... But a quick look, my 3800x wouldn't need more than a 3070ti or a 6800 (16 game average at 137 fps at 1080p). If I upgraded my cpu from my 3800x to a 5800x, I could utilize a 6950 or a 3090 ti (to get a 180 fps 16 game average at 1080p). While I'd have to jump to a 7900xt or 4070 ti to get the 200 fps 16 game average on a 5800x3d. Sadly, I couldn't find any recent benchmarks to see how my 5700xt handled today's games. But based off of the original review of the 3070ti, I'd see a 45% improvement at 1440p or 55% at 4k. But no 1080 numbers, which is what I game at...

[@TheAllhailben7]

Honestly, you don't want to be CPU bottlenecked in a game. If you're GPU bottlenecked, the FPS might go from like 90 to 80 sometimes. If you're CPU bottlenecked, you could be bouncing back and forth between 75 and 100 while your gpu struggles to be fully used constantly which is not a good experience. If you're running anything like a 3080/6800xt or better, at least a 1440p monitor is recommended because you get all of the performance you're paying for.

[@The_Noticer.]

I can't believe we're still having this conversation. I'm so glad i'm no longer on computer related webforums, because i'd have pulled out my hair trying to explain people this. I've done it so many times, made the same tired arguments.. but people see a result they dont like and they immediately revert back to this grug mentality. Same with the "bottleneck" argument. Brutal, just absolutely brutal trying to tirelessly explain this. Bless your patience, I would've long lost that patience by now.

[@umamifan]

Computer and tech forums are the worst places to be on the internet. Too many armchair experts who get upset the moment a disagreement happens...

[@YeaButCanUDoABackflp]

For real. It's not even that complicated.

[@Bareego]

People are not stupid, they're just not as rich as you to make that argument. Until you get this there's not point arguing.

[@mcbellyman3265]

@@Bareego They are stupid. It doesn't require money to understand basic logic. In fact, the less money you have, the more important it is to understand this stuff, as it requires you to be more selective in buying the appropriate equipment. Rich people can just buy the best CPU and best GPU and ignore all this.

[@The_Noticer.]

@@umamifan Just like with most things now, people are hardwired to see people having different opinions as an enemy that is beyond reproach. There are SOME acedemic arguments against using 1080p data. Such as it not being very good to measure ABSOLUTE performance of that specific CPU in that specific title, as increasing the resolution also has a detrimental affect on CPU load. It has to process more geometry data, especially now with ray-tracing. However, when used as a RELATIVE test against other CPU's using the same load, its still perfectly valid. As they all are tested under the same conditions. But when something as simple as what steve is explaining is lost on these people, the finer points will be lost on them even moreso... As I said, im glad I no longer dwell on LTT-like forums.

[@MA-90s]

This guy doesn't pull his punches and delivers outstanding content. We need this. Thank you.

[@RNG-999]

Doesn't*?

[@MA-90s]

@@RNG-999 Does not*?

[@zenstrata]

he's wrong though. because eventually 1080p will be a useless test benchmark, because nobody will use it. Otherwise they should be using 144p or even 480p for 'testing'. If he was entirely right, then he should go to the extreme, and be 'more right'. but instead he's only 'partially' right, and that won't last, because eventually 1080p will be consigned to the dustbin of history. Just like 144p and 480p is today.

[@MA-90s]

@@zenstrata "eventually" dont cut it. This is the here and now. Right now, he is 100% correct.

[@Kuriketto]

@@zenstrata No, he's not wrong, and you've completely missed the point of the video. Testing at 1080p with top-end GPUs gives the audience an idea of how much a given CPU might scale with GPU upgrades years down the line, because it's by far the resolution that most gamers use today and will be using even several years from now, likely even after multiple GPU upgrades. Games will continue to become more demanding, and as an example, somebody building on a budget may need to keep upgrading their GPU just to maintain an acceptable level of performance at 1080p, while their CPU remains untouched. The idea that 1080p might one day fall out of mainstream usage is irrelevant.

[@Beefy-Tech]

I tested a 4090 at 1440p competitive settings in warzone 2 with a 5800x3d, to test cpu bottlenecks in warzone 2, and me, a guy with 169 subs got comments saying 4090 is not meant for 1440p, but for 4k only...as if 360hz 1440p does not exist, or is it assumed that most professional play at 4k res vs 1080/1440p high refresh xD? The point of a cpu test IS to bottleneck it to see which ones faster or how far in FPS it can take the GPU.

[@elirantuil5003]

Probably children. 2080ti was also a "4k only" card, let's see them gaming on 4k with it now.

[@mrbobgamingmemes9558]

@@elirantuil5003 expecially on forsebroken (forsespoken)

[@Threewlz]

360hz is copium, isn't it? Hell even 240hz is kinda overkill. How many games are young getting 240fps plus (apart from valorant and csgo) without spending thounsands of dollars into a pc? 144hz is enough and should be the norm a few years from now on I believe

[@TheGreenReaper]

@@Threewlz They tested this with fighter pilots. 144Hz is not the limit of human perception. 240 Hz is closer. 360 Hz is likely beyond a noticeable return.

[@BlahBleeBlahBlah]

@@Threewlz ​ projecting much? Sure it isn’t a concern for you (or me TBH). However I understand there are people who want and can afford the hardware to give them super high frame rates @ 1080p.

[@qlum]

While I was fully aware that there is a difference in cpu performance, this does really show it's time to upgrade my aging 8700k@5ghz now that I upgraded my gpu to a RX 7900 XT. Especially since a lot of the games I play may be more cpu bound than the games regularly benchmarked. However, I do tend to spread out my upgrades a bit as not to spend a load of money at the same time. So probably something that's gonna wait till middle of the year. The gpu upgrade was definitely more needed (GTX 1080) not only for performance but also because Nvidia's linux drivers are hot garbage.

[@MrHeHim]

Generally CPU bottlenecks can be felt/seen in game as hard stutters and GPU bottleneck of course would be lower FPS but still relatively smooth. I personally would match it with a 7700x, if you're lucky enough to live next to a Micro Center they are giving away 32GB DDR5 6000 if you buy one. BUT the motherboards are over 250 so it comes out to around $600 still.. BUT (more butts) they have a $600 bundle with a 7900x, G.Skill 32GB DDR5 6000, and a ASUS B650E-F ROG STRIX GAMING WIFI 😅 I'm personally happy with my 6800 XT and 3700x, so although i almost made the purchase twice i backed out last second because then I'm going to want to get a 4080 or 7900 XTX to match it 🥲

[@qlum]

The paradox games I play benefit greatly from x3d I as for microcenter, I live in the Netherlands so there is nothing of the sort here. In checking I did notice the 7700x dropping quite a bit in price to € 300 in the last few days and that does look interesting. Still with the first x3d parts around the corner, I can at least wait for reviews / benchmarks of those. With a decent am5 motherboard and a decent kit of memory that would be around € 720 including vat or comparing it to usd roughly $635 excluding vat. Not too terrible either with no bundles.

[@michaelrobinson9643]

In a world where people say "I feel" as an expression of thought, I'm not surprised many do not understand scientific or engineering rigour. The concept of isolating variables to objectively examine each is too horrible to think about for many I expect.

[@JarrodsTech]

Great explanation, I will definitely be linking to this video in a number of my own comments 👍👍

[@pf100andahalf]

Hi Jarrod!

[@martineyles]

Only half helpful though. In the desktop space, it's good to test the CPUs with a high end GPU. However, reviewers have acknowledged that there is increased CPU workload when using ray tracing at 4k, so it's possible a low end CPU might not provide the framerate we want at 4k, and the CPU benchmarks at 4k are still useful.

[@tyre1337]

always better to link to reviewer that can be trusted and isn't manipulated by nvidia

[@dandylion1987]

Uh oh. You do realize all viewers here have 150IQ or more and know the intricacies of PC technology inside out, right ?

[@Eternalduoae]

Dude, I have an IQ of above 500 at 360p!

[@godblessbharat708]

@@Eternalduoae sounds like frame rate

[@lukeshackleton4775]

Sounds like a joke that went over a head

[@stephenbull2026]

@@lukeshackleton4775 Sounds like a joke which, if you don’t get, means your IQ is the bottleneck. 😂

[@petrnovak7235]

Sadly, the comments shown at the beginning of this video says different story 😞

[@andrewcross5918]

I just wish there were more non FPS metrics tested like tic rates in Stellaris or turn times in Civ 6 and other 4X titles. I also wish there was a wider array of game types tested like late game ARPGs like Path of Exile or MMOs. The kind of load these titles put onto systems is totally different to many AAA style games so seeing how these perform would add a lot of breadth.

[@HexerPsy]

MMOs are particularly tricky to test due to the uncontrolable variable of the other people in the game. No busy city is the same, and testing in an empty field is kinda pointless. As for the rest... idk - if enough people want Civ 6 in the comments, then it ll happen? Guess the reviewer mostly tests the games that the audience wants to see / plays?

[@lexiconprime7211]

I, for one, would love MMO testing in raids and dungeons (WoW, FF14, BDO, Gw2, etc...). But I think the reason they don't is because there's no way to eliminate variables that could potentially skew the data.

[@rk9340]

@@lexiconprime7211 It's a bit pointless. You just lower the settings where it can run 60/144 fps lows consistently. Things like random texture loading times are influenced by your Harddisk more than anything else and the rest of your system can actually freeze to a halt while waiting for that data.

[@andrewcross5918]

@@rk9340 in plenty of MMOs there are no GPU settings that will do that in large raids or PvP because the CPU is the bottleneck. Same for Stellaris et al. Low tic rates means you need to have the patience of a saint or stick to smaller maps with fewer other civs for it to remain playable to end game and beyond. Just checking steam and Civ 6 is 22nd in the top player list, Heats of Iron 4 (Paradox grand strategy) is 34th, Cities is 37th, Path of Exile is 42nd. All ahead of CP2077 in 54th. Then you have EU4 in 69th, Stellaris in 74th and CK3 in 88th. I don't see Hitman 3 or Rainbow 6 siege or tomb raider or horizon zero dawn on the top 100 yet they get tested over far more played games. Edit. Adding up how many people are playing those 4 grand strategy titles (I believe they use the same base engine) and you are in and around the player count of Call of Duty at 80k ish. So on the whole very very popular.

[@lexiconprime7211]

@@HexerPsy That's kind of why it would have to be primarily a cpu test in controlled environments like raids or dungeons, where the number of players and NPCs is always the same. But even then, you would need a consistent crew of dozens of people all repeating the same actions ad nauseum, which is why I think no one would do it. It'd be VERY useful for someone like me though, since I play a lot of MMOs.

[@reaperofwaywardsouls]

Great video. I think that including 4k benchmarks is a great way to remind some folks that they don't necessarily need a CPU upgrade (people that game at 4k). No need for the full suite of benchmarking games though, maybe 2 or 3 should get the point across. You may be surprised about the amount of people who look at 1080p graphs and transpose it to 4k in their minds, and then are disappointed when they don't see a comparable performance uplift when they upgrade (and game at 4k).

[@charlestwoo]

I just want cpu gaming benchmarks to have 3 games shown in 4k so I can see it clearly, that is all. I know 1080p is where u lean on the cpu the most but I still want to see 4k damn it.

[@Strix-gp9xg]

That is exactly why they should always include a 4K benchmark results when testing CPUs.

[@5i13n7]

I find the 'mixed data' chart very interesting. I would greatly appreciate if you guys could do a video showcasing the expected gains from historical cpu architectural upgrades at various resolutions on some of the more popular gaming titles. eg. 4090 with each 2700x, 3700x, 5700x, 7700x, 8700k, 9700k, 10700k, 12700k, 13700k. Even just a couple games to compare would be great to see actual expected gains for a cpu generation upgrade at a user's target resolution. I, and likely many others, tend to stretch my pc build as long as it's legs will carry it, and a lot of the charts for new hardware include only the last generation or two to find the generational uplift. Enthusiasts may be upgrading with each generation, but I'd wager more people are considering upgrades from several generations back, and could use this information to choose a most suitable upgrade for their use case.

[@domm6812]

Good vid, but I suspect a lot of the people complaining about 1080 testing may not understand the definition of what a bottleneck is in this instance, how the CPU becomes one, and the fact that CPU performance is masked/hidden at higher resolutions. In short ...go back to the very basics so there can be no confusion.

[@pf100andahalf]

Exactamundo. He needs to make a 60 second short too. It probably wouldn't hurt to make several follow-up videos to this one showing the concept in different ways for those in the back... of the short bus.

[@UnStop4ble]

my problem with this video is that it couldve been atleast 50% shorter and still wouldve gotten the point across, maybe even better cause of peoples low attention span

[@moldyshishkabob]

@@UnStop4ble Frankly, I think the people with such a low attention span that they don't understand this may not want to put in the effort to understand it in the first place. Because why are they watching a channel that has an average run time on videos that's well beyond 15 min? There's a reason LTT has millions more subscribers than HUB or GN, as S-tier as their testing is.

[@JACS420]

@@UnStop4ble yeah, tbh I’ve built enough pc’s I’ve sold graphics cards once overclocked performed like the next tier none ti/super variant. Mean while others gpu/CPU’s can’t handle 150+mhz oc’s. One customer I sent along the way with a 3060ti boosting up to 2050mhz. I warmed him, but he popped in my discord not to long ago claiming it’s still kicking.

[@huggysocks]

Nope I have literally zero respect for people commenting on how people do their work when they haven't bothered to learn the most basic things about the topic. It's like backseat driving from a toddler only dumber cause they are older and should know better.

[@chrzonszcz323]

For future i would suggest including strategy games like Stellaris, rimworld or Factorio. These are very CPU intensive games where you will be able to see the difference between different CPUs much easier

[@rENEGADE666JEDI]

But in factorio processors with 3d cache are in a different league. Although, for example, a star citizen would open people's eyes to how much this technology gives ;)

[@TleafarafaelT]

im tired of watching AAA games in CPU conparision when we know the indies games that have a insane hungry ass wanting a chonky cpu

[@Daeyae]

My r5 1600 and my friends r5 5600x are night and day in stellaris, his is orders of magnitude faster.

[@thentil]

What metric exactly do you want them to show for these games? The productivity charts should give you a good idea of how they'll scale in these games.

[@saricubra2867]

@@TleafarafaelT Indie games don't have good optimization unless they do.

[@tkllluigi]

When someone can explain something simply means that they full understand the topic. One of the best hardware educated videos, well done!

[@jaymacpherson8167]

Thanks Steve for making it clear how the hardware compare using a baseline standard. That apples to apples comparison empowers the individual PC builder when mixing and matching their hardware.

[@grizzleebair]

Great video! I did know this, but bless you for doing this video for your newer followers. That is a heck of a lot of repeated testing. For you new followers, stick around, these guys know their stuff and do quite extensive testing. They also have a good working relationship with some other of the top Tech YTers, and aren't afraid to do collabs or refer us to someone they trust if it helps us out or helps them reinforce a point they have made. On another note, I know Australia is big, and this may not be feasible, but it would be nice to see a collab with several of the Aussie Tech YTers. Maybe with just the ones in your corner of Aussie Land. Just a thought.

[@russellmm]

this was a great video! nice job. I think this also shows why doing productivy testing is so important as it also isolates the CPUS from the GPU.

[@zakelwe]

Exactly All the cpu testing in games this site does is rather wasted because if it is not gpu limited who cares if it is 200 or 150fps apart from a very small minority. What's more important is how long it takes to do something useful such as video conversion etc. If you are playing games then the vastly more important thing is to spend money on the gpu. The 5800x3d was better at gaming but worse at everything else but cost more. It was just not balanced.

[@albundy06]

@@zakelwe Everything you post is a waste. Don't watch it. Don't comment.

[@tortuga7160]

1080p is still alive and healthy. 65% on steam survey, and LG is coming out with a 500hz 1080p monitor. I'm pretty sure LG did the calculus before deciding to make that panel.

[@OutrunCid]

The Steam survey is far from representative. Its 1080p numbers are more likely to be pushed upwards by ancient PC's and laptops, than by ultra high framerate monitors. Unfortunately it does not measure the refresh rate used. I think the numbers shown by Steve are more representative for those looking into high-end hardware than those presented by Steam though.

[@alsamuef]

Moronic product.

[@Hyperus]

@@alsamuef How so?

[@Kuriketto]

@@OutrunCid How is the Steam Survey far from representative when it literally just pulls system specs? It might be nice to know what refresh rate is used by most gamers, but the hardware survey largely focuses on computer hardware and baseline software, not in-game settings or monitor makes and models. LG also is likely very aware that extreme refresh rate monitors are a niche category and haven't churned out millions of them, but one also needs to recognize that with today's technology, producing a 500Hz monitor is only achievable at a max resolution of 1080p anyway, and while using a TN display at that. In that context, the Steam hardware survey resolution results are slightly irrelevant, but the survey as a whole can at least inform what percentage of the userbase has hardware capable of driving a 500Hz display to its full potential, though realistically only a handful of games can even achieve 500+ FPS given modern hardware. And as far as benchmarks on this channel go, it has people with ancient hardware, someone like a pro player who wants the absolute best and everybody in between, covered. As Steve pointed out, just do a little cross-referencing.

[@CelestialGils]

@@OutrunCid Not really. You have the misconception of steam doing the survey automatically on every computer that has steam installed. That's not how it works. The steam hardware survey is random and you have to accept it in order for your system to be part of it. You can trigger it of course, but most people don't do that. You also can see the data more in detail, based on what hardware or resolution makes up that month, not just the change where all the surveys are taken into account.

[@siewkimng1085]

Great explanation, while I certainly understand the need to avoid GPU bottlenecks while testing CPU performance, it’s great to see it articulated in such a clear manner.

[@earvin4602]

I agree with everything said here about CPU benchmarks and I'm still adamant about the fact that a CPU "scaling" benchmark is incomplete and misleading without being paired with corresponding GPU scaling benchmarks. Let's take Steve's example about wanting to retire the 8700k: While the 4090 benchmark suggests a 98% uplift potential in Watchdogs when switching to a 13900k, only 6% will be achieved without retiring the paired 3060 alongside. And for that information the presumed meaningless GPU bound benchmark is actually very valuable.

[@thisguy317]

Awesome video guys. We need more videos like this I'd love to see more GPU reviews include old CPUs mixed in to the results, just to get a feel for what the FPS-cost is associated with not upgrading my CPU in tandem.

[@KevinM0890]

u can do it on your own. create a cpu limit on your system and u know if your cpu needs to be upgraded or not to reach the FPS numbers u need :)

[@codyl1992]

@@KevinM0890 not really because that doesn’t take into account everything. Clock speed and core count isn’t everything for this stuff.

[@aaronwhelan1]

So true!! moving from a i7 2600K to a R7 7700x with an older 980ti felt the same as when I moved from a Geforce 680 to a 980ti with the older 2600k

[@OutrunCid]

Now that 2600k is an oldie of course. Will be making the switch to the same R7 soon whilst still on an R9 Fury. Not expecting wonders to happen, but I know that in those cases where I was CPU limited, improvements will be experienced. I just need to listen to my GPU right now to know which games these are :D

[@alexm7777]

A i7 2600k and a 980 ti would make for a nice budget rig.

[@Daeyae]

@alex m sadly its just cheaper to get an old ryzen or new i3 most of the time

[@MalHerweynen]

@@Daeyae depends whats on the used market like rn I cant find any reasonable priced ryzen parts that arent like 90% of the price of buying it new

[@Daeyae]

@Malakai Hawaiian true, in my area a 2600k would be like £15 and a 1155 mobo about £30, ddr3 is super cheap too. A 1600 is about £30-£55 with a cooler A320 boards are about £25+, ddr4 is cheap but not as cheap and ryzen loves fast ram so youd want decent stuff. I guess it depends also how much £15 is to your budget, if your trying to spend under 300, £15 is 5% which is quite a bit but is it worth losing out on 2 cores and a small speedup?

[@brutlern]

You need 2 categories, the productivity people, who are extremely interested in the CPU performance, and the gamers, who DO NOT CARE about CPU performance. The gamers only want to know which is the cheapest CPU which will not bottleneck a mighty 4090 with settings cranked up to 11. If all CPUs perform the same (mostly) at 4K, then that is the information that, we, the gamers, need to know. We need to know if the old crappy CPU is still good enough for whatever extremely expensive GPU we may get, or if it's time to retire the old dawg and upgrade the CPU as well. The 4K gaming chart for CPUs might be the most boring chart ever (basically a rectangle), but it's the most crucial information a gamer needs when thinking about that new GPU upgrade. Case and point: in this video the 4K Watch Dogs Legion test specifically provides the info we need. For a 4090, you need a faster CPU, that i3 is not good, it doesn't matter that it's still 100 fps, you are leaving 50 fps on the table. But if you're only upgrading only to a 3080 than the i3 is still good enough, and you can save a buck today, and upgrade later.

[@mlqvist]

Thanks for videos like these Steve, they are the one opportunity I have to feel smart. How do people not get this stuff?

[@VPoje]

Can you show good old Intel CPUs performance compared to now? (From like i7-3770K to 9900K?)

[@Hardwareunboxed]

At some point we will revisit older hardware but we do that during quiet periods.

[@VPoje]

@@Hardwareunboxed Thx!

[@korneldenes]

Köszönjük!

[@halistinejenkins5289]

16:56 is why i started watching your channel. not really complaining but this kind of scaling has been lost it seems over the past two years. i love the cpu/gpu scaling content, because it shows you the cut off point for certain platforms at 1440p high frame rate gaming (144-165) with different gpus. i guess it's hard to get it all in, so everyone can't get what they want. dunno, maybe it's just me, but the inner nerd gets excited when i see how far the 8700k can go with today's gpus before you start leaving noticeable performance on the table. regardless, cheers! been watching since the beginning and will continue to do so.🙂

[@darkywarky]

Thank you for also testing the 8700k. I was wondering if I should upgrade it playing at 1440p with a 3070.

[@InternetListener]

No. And you don't need a 3070 because new AMD cards with same or better performance (and higher efficiency, that will save you hundreds or thousands of moneys in electric bill) are cheaper even than the used ones. Just check prices of 6700 series (6800 or 6900 series, or even 7900 with a proper discount could make you play at 2K or 4K). While you can get more than 90 fps consistently you can always make use of the supersacaling or downscaling option in new GPUs to make sure your GPU works at 100% even if you were in a CPu bottleneck scenario in a particular game.

[@GewelReal]

@@InternetListener AMD cards do not have RTX features tho 🤷

[@meurer13daniel]

I had a 9700f non K and had to upgrade to a 13600k.I dont know how your 8700k is performing know, but I was having a lot of bottleneck scenarios in modern games with a 3070. Spider Man, Cyberpunk and Plague Tale Requiem are clear exemples. For 2023 games I supose it would be even worse such Hogwarts Legacy

[@darkywarky]

@@InternetListener I already have the 3070 for a long time, but I was wondering if I was leaving performance on the table by using a 8700K in combo with a 3070.

[@khalidzouzal8417]

I got the 8700 non k paired with an RTX 3080ti, I basically play any game really at Ultra 2160p above 60 fps. Like they already mentioned in the video, the higher your resolution is, the less your cpu matters. I dont play competitively, so im okay with 60+ fps at 4K. And it works flawlessly for me.

[@kodysmith8897]

I was stuck with a 1080p monitor for some time so I appreciated the information. Now I still think of it as academically interesting, shows historical scaling, and allows me to help inform my non-pc literate friends and family.

[@matrix3509]

Seems like most people just refuse to understand what the purpose of benchmarking even is. SPOILERS: Benchmarking is about founding out the true performance characteristics of whatever part you're testing, thus it makes sense to eliminate other variables. Benchmarking is the computer equivalent of taking a car to a race track to find its lap time. Then some dipshit comes into the comments and talks about how nobody cares about lap times because the real world has speed limits.

[@kmputertechsupplies2374]

@ 11:30 you said "is that useful data, in my opinion NO" While you are correct in the fact that the 13900K is much better than the 3% uplift over the 13100, Consider the following. One question this graph does indeed answer and you are ignoring is: What CPU do i require to get the reasonable most out of a 3060? If you ignore the above question and assume showing CPU performance relative to the best GPU is best, most people will assume that if you pair an i3 with the 3060, there would be a major loss of performance as little as 30% as shown when tested with better GPUs. In conclusion, people who purchase i3s cannot afford a the highest end GPU, and showing them their CPU's performance vs better CPU's performance scaled against a better GPU will influence their buying choice. They will think an i3 is not enough when in their use case it completely is. And WORST OF ALL, they will save more to buy the i5/i7 and get the same exact performance as if they would have gotten the i3.

[@InternetListener]

you can build a full system with the price of only one component used in a YT GPU or CPU benchmark. They should add to the test hardware that can scale in a sensible way from that of the Gaming Consoles (if they are focused on gaming). But a PC it useful for many other thing and as you say people doen't get informed that with a Celeron G6900 and a RX6400 they can game, do 3d art, RT, render, photo editing, coding, video encoding (thank to accelerated GPU softaware) and so much with a low electric consumption thank to the Internet content creators forgetting to test the basic components and user cases. Who cares about Cinebench if you can run Blender in real time with a rx 6400 or 3050 ti faster than with a Mac or whatever... or encode a mp4 video in minutes using the GPU (RX or ARC), no need to go to rtx 4090 or a 1000$ cpu.

[@ChaosPootato]

It boggles the mind that this is still a thing. The CPU testing methodology is very well proven at this point, every single (serious) outlet uses low resolution because that's what makes sense. These people don't want a CPU benchmark, they want a benchmark of their specific targeted complete system and targeted resolution. I'll attribute to ignorance what could very much be attributed to selfishness

[@CriticalPotato]

Excellent name, Potato brethren.

[@kjgasdfduhigsdauf]

Didn’t realize people questioned this practice. Once again shows how little common sense gamers have

[@skinscalp222]

lol I'm currently arguing with a genius on reddit who thinks when you're CPU bottlenecked at 1080p just increase your resolution to 1440p and you will get more performance because your CPU usage goes down and the GPU usage goes up. Bro, your GPU usage goes up because it now has to process more pixels not because it's processing more fps! and your CPU usage goes down because it doesn't have to process as much frames at higher res than at lower res! It's mind boggling how many people can't understand something as simple as this 😂

[@ComputerGeeks-R-Us]

Love the work you folks are doing! I've been in IT for decades, including benchmarking and standardizing client hardware platforms for a Fortune 250 for over 20 years. The understanding of the interdependencies and limitations between components isn't exactly straight-forward. I've also run across some really weird inconsistencies that the OEMs couldn't even explain. It's nice to see Hardware Unboxed doing the hard work and sharing their knowledge. Keep up the good work!

[@jimatperfromix2759]

I agree with @gregorymchenry1464 and many others that this is both a well-needed and extremely well done video. Most people with IT careers like Gregory and myself understand the general principle espoused in this video very well. But the average Joe or Jane buying gaming hardware typically doesn't have exposure to this general principle of performance analysis. Just to reinforce that Hardware Unboxed is right on target in this video, let me more formally state the general principle. (a) In any computerized system there are typically multiple system components, each of which *might* contribute to slowing down the response time of the entire system to a low enough level such as to disappoint the user(s) of the system. (b) Given any specific configuration of components into such a computer system, it is almost always the case that just one of the several system components contributes the most to slowing down system response time. (c) Within the discipline of computer performance analysis and system tuning, point (b) is generally very useful and is called the "bottleneck principle." (d) The rare exception to having a single bottleneck is actually literally the goal of perfect system configuration, namely to have all potentially bottlenecking components hit bottleneck state at the same system load - that is, you're not overbuying on any given component and all components sort-of run out of gas at the same time. (e) It's fairly hard to perfectly tune a system (as per (d)), yet we can use the bottleneck principle to try to at least ensure that one system component is not "horribly bottlenecked" relative to the other system components. (f) So although configuring computer systems is somewhat of an art (for which you can perhaps hire Gregory to help you out), you can't mess up too horribly if you at least keep the bottleneck principle well in mind. (g) In a computer system designed for gaming, as long as you also take care of important extra issues such as making sure you have enough system memory and making sure your primary disk (plus as many other disks as possible) are SSD disks, then your main decision has to be focused on the two major potential bottlenecking points, namely the speed of the CPU and the speed (and to a lesser extent video memory) of the GPU. (h) In looking at reviews of potential CPU/GPU purchases to configure in your gaming system, look at a wide range of reviews, but take with a grain of salt any reviews that do not properly respect the bottleneck principle as apropos. (i) For CPU and GPU (respectively) reviews in the context of gaming PCs, that means in practice that CPUs should always be tested with "best available" GPUs, and GPUs should always be tested with "best-available" CPUs. (j) Other types of sub-reviews within a broader review article are sometimes useful to provide additional scope, but if you smell that the reviewer doesn't understand the usefulness of the bottleneck principle in doing hardware reviews, run away. (k) A recent example of a review that totally botched the bottleneck principle (and from a normally highly respected reviewer) was when LTT did a gaming oriented review of new 7000X series AMD CPUs, but they didn't have the patience to wait 2 more weeks to get their hands on a new Nvidia 4090 card to do those CPU reviews with, such that the 3090 they used wasn't fast enough for a proper review of those very-fast CPUs, and they came to erroneous conclusions about the lack of speedup given by the new AMD CPUs. Bear in mind what this Hardware Unboxed video teaches, and you're much better equipped to understand the implications of all CPU and GPU reviews. I would, however, toss out one somewhat more obscure point that requires a bit more nuance in understanding better. By nuance I mean that it's a bit more complicated and might be worthwhile of additional discussion (whereas the bottleneck principle is a slam dunk). Specifically, what I've got in mind here is the additional recommendation to do primary testing at 1080p - and I would like to modify that slightly. Substitute for 1080p the most-common low-end resulution "of your era" or perhaps do double testing on both the most-common low-end resolution of your era, plus the most-common middle-end resolution of your era. This is a sliding scale across macro time, since in the next era the low-end resolution will be at 1440p, and way down in the future we might see such great CPUs and GPUs that 4K gaming would then become the "minimal" resolution for testing. I mean, if nearly everyone is actually gaming at 2K or 4K, there is no point in differentiating 1080p performance of 1000 frames per second from 1080p at 500 frames per second. That having been said, for this "current era," it looks like testing at 1080p will be fine for at least a while longer. I want to quote in its entirety the comment in this thread by @puokki6225, who says "Anyone saying 1080p is 'dead' needs a serious reality check. According to Steam hardware stats, ~65% of their userbase still uses it. 1440p is at 11% and 4K at a tiny 2.7%. For reference, 1366x768 is at 5.9%. The high end market, which realistically almost all of the DIY PC market currently is, is very niche even when it comes to gaming in general." So in our era, we DIYers can safely ignore the testing needs of the 5.9% still running at 1366x768 (since presumably they want to upgrade to at least 1080p), cover 65% of current userbase by focusing on 1080p testing, and for those 13.7% of DIYers aiming at either 1440p or 4K gaming, they should be able to apply some reduction factor to guess approximately what 1440p and 4K performance might look like.

[@mauree1618]

I guess it’s easy to forget how much of what we know is built on prior knowledge. We need more videos like these for the beginners.

[@MrNside]

I can still see the argument from the other side though. While it's nice to know the limits of a CPU, a realistic perspective is still important for someone who never plans on going near those limits, or upgrading their GPU to the highest-end. Using your examples, the 8700K blows the 2600 out of the water with a high end GPU, but costs almost twice as much. But with a more realistic GPU, the difference was only a few FPS. While knowing what a CPU is capable of is important from a benchmarking perspective, it doesn't really tell them much when they go to build their system, particularly a system on a budget. If someone were building with the GT 1070 or the RTX 3060, and wanted a good CPU recommendation to pair with it, they would have a hard time seeing that the 2600 or the i3 13100 were actually decent options. If they only saw the high-end GPU data, they would think both of those CPUs were trash.

[@jouniosmala9921]

Some graphics quality settings affect the CPU limit significantly in some games. I learned that during the year when I was using 2070 super with i7 920. (My GPU upgrade wasn't for gaming but for a project that needed it, 2060 super was ideal choice but 2070 super had a quieter partner model.) It would be interesting to see if that's actually case for your normal set of games tested. Pick fastest GPU, slowest CPU and check if quality settings affect performance.

[@cellanjones28]

Graphic quality settings would impact the CPU by how much the CPU needs to work ( I think)

[@calisto2735]

No shit, Sherlock? Quick! We need to call NASA!

[@AerynGaming]

They certainly do. Resolution is a good knob to tweak because it almost never has any kind of meaningful effect on the CPU, but usually massively changes the GPU throughput.

[@jouniosmala9921]

@@cellanjones28 Yes, some settings do. Other's I was able to keep at maximum settings since those only affected the GPU load. But what was interesting is that some graphics settings just turned games that were playable with settings used with my old GPU into completely unplayable and I had to tune it down to find playable settings but all my games at the time were older games. This is something that would be interesting topic worth further exploration when they are not busy working hardware that's about to release or just released.

[@cellanjones28]

@@jouniosmala9921 just depends on the resolution. i run 1440P so it's all gpu based.

[@philipreininger2549]

Really great video, I 100% get where people are coming from with the "unrealistic pairing argument" but for testing it's just a different story than for actually purchasing parts and recommending certain combos. Keep up the great work :)

[@CanIHasThisName]

I would like to say that I really appreciate that you also include midrange GPUs in your CPU benchmarks even though it's a lot of extra work for you. It helps the viewer to quickly get an idea where the difference still matters and where it doesn't. Theoretically you could use the presented graph and compare it to another graph that focuses on testing the GPU, but then you also need those benchmarks to be done under the same conditions in the same games and sometimes you'll need more than one outlet for that which really complicates things. And then you have what's seen at 7:53 - based on the 3080/4090 data you'd assume that the 2600X would get you the same result as 8700K with an RTX 3060, except it doesn't.

[@clickbait-g4o]

Well... choosing a handful of modern games within your preferred genre, comparing the performance of various CPUs and GPUs, and then selecting components with fitting numbers is one of those ideas that seem so obvious once you've heard them, you wonder why you didn't think of it yourself.

[@moliii8809]

Thanks for this really helpful explanation - came here from watching your recent 9800x3d benchmarks!

[@antonschaf4088]

Nice, we havent seen the i7-8700K in a while. Impressive performance of the i3-13100 compared to the old i7. You really should compare the i7-8700K and i9-9900K and maybe the i9-10900K to 13th gen in the near future! Edit: At the end of CPU benchmark videos you could mention a realistic pairing to that tested CPUs, so the "realistic pairing" dudes have their needed answer.

[@jonsummers3453]

8700k has really held up good overtime. 6 core is a sweet spot and that cpu could be overclocked massively helping keep it relevant for a long time.

[@ZWortek]

Really appreciate you guys making videos like this because since COVID began there are a lot of new people in the PC gaming space that are almost entirely clueless and borderline delusional with their logic regarding topics such as this. For example, quite a few people that reply to your tweets that I've seen both Tim and Steve face palm over LOL.

[@jdlpeke1]

A perfect explanation, between the difference in CPU performance and CPU+GPU performance, because although they go together they are worlds apart. You can see perfectly how things change when there is no GPU bottleneck.

[@ZeroUm_]

I'd say we're having an XY problem, what the ones asking for a matching CPU and GPU benchmark really want to know is what is the best match of two hardware so they can get the best performance for their budget. Most reviewers do proper CPU and GPU benchmarks, but they usually don't give direct answers which GPU matches well with a given CPU.

[@michaelrichter2528]

19:30 I would be interested in an explanation why you don't test CPUs at even lower resolutions like 720p or even 480p. In theory this should completly eliminate any remaining GPU bottlenecks without changing the tested workload for the CPU. Are there any unexpected downsides because these resolutions aren't commonly used or is 1080p simply low enough for current high end GPUs?

[@assetaden6662]

That's a good metric, but doesn't show any real-world scenario. People playing below 720p are just the competitive shooter players, who are a minority.

[@michaelrichter2528]

@@assetaden6662 But to me that was the entire point of the video. Of course nobody plays games at 720p or 480p on modern hardware just like nobody buys an i3 for his 4090. CPU tests are synthetic scenarios that only show the maximum possible performance (fps) of a CPU without any GPU constraints.

[@SiberdineCorp]

@@michaelrichter2528 There are some legitimate reasons, many games adjust pop-in and Level of Detail along with resolution (logic being that you won't see the difference at lower res anyway). Lower LOD = less draw calls = less CPU load 1080P is a good middle ground, not too low to reduce LOD by much, not too high to take pressure off the GPU

[@michaelrichter2528]

Thanks, that makes sense

[@ferasamro9735]

To be Honest as Enthusiasts we think all of your viewers understand what you do but these comments are a good think because this means that non experienced people are watching your videos, this is very good news for the hardware community and for you as whole Thanks always for your Hard work.

[@myfakeaccount4523]

Enthusiast = spends too much money.

[@Bula-Boy]

Thanks for adding 1080p in your benchmarks. Most fortnite pros are still using 1080p and likewise for other competitive titles.

[@fiendhappy6964]

yea they SUPPORT FOR COMPETITIVE GAME PLAY 1080P.

[@conkcreet]

@@terraincognitagaming facts

[@Bula-Boy]

@Terra Incognita Gaming brah think you're still lagging 2 years behind. In November, 2022 we had fncs invitational LAN and this year they have FNCS global championship with prize pool of $10 million.

[@fiendhappy6964]

@@terraincognitagaming no bruh , pubg , dota and warzone

[@MarkHyde]

Controversial ;) - love you taking the trouble to do this. More power to HU!!!

[@naswinger]

glorious hungary!

[@hillardbishop8755]

@Hardware Unboxed My rule of thumb if I have to make a compromise is go for higher tier on CPU because it is easier to swap out graphics card later. It doesn't matter how powerful your graphics card is if it is being limited by you CPU. Ideally, I want to match CPU and GPU for the build, but sometimes due to budget constraints, one must compromise while minimizing the tradeoffs and keeping in mind future upgrade paths. I agree 1080p testing shows which CPU performs better, GPU testing shows which graphics card performs better. With those results in mind, one is better equipped to pair them appropriately in a build.

[@sniper0073088]

This data is extremely useful. It makes it very easy to choose the best gpu for a given cpu. This kind of benchmark would be very nice to have in every cpu review. Making the most of a budget cpu while not overspending much easier

[@spladam3845]

This was a great idea for a video, and a good resource for new viewers, well done Steve and team.

[@RongGuy]

First of all I would like to say I agree with your testing methodology. That being said, I think it would be cool to see "upgrade guide" type videos with new GPU generations: e.g. take 5-6 popular CPU's that are a couple of generations old, and benchmark them with 2-3 GPUs people are most likely going to buy (like xx60 (Ti), xx70 (Ti) and xx80).

[@PainterVierax]

I remember such videos in the past, testing different GPUs on a single popular midrange CPU to show the bottleneck. But those videos are done when there is less launches… and when Steve doesn't have to do such a basic explanation for angry newbs.

[@allxtend4005]

belive me the i5 was way more popular then the i7 because of the price of the intel products was that high and motherboards as well from intel. and at this time when the ryzen 2600x was selling well the most people had a intel I7 3770k / 4790k and so on and not a 8700 ( not the most populated cpu for gamers at all )

[@Jansn]

I can feel these comments you guys get, I'm getting the same stuff, those people will always be there, don't worry, just keep doing what guys are doing.

[@SKHYJINX]

That 720p argument, vs 1080p... @HardwareUnboxed I was curious if there are more drawcalls at same settings but higher resolutions? I'm wondering if asset quality (number of edges to make up an object) increases also with resolution and is that indicative of more drawcalls needed/more CPU throughput (for that task) a higher resolution scene. 1440p and 4K in mind as well,....just using 720p vs 1080p as it was the topic.

[@MichaelPohoreski]

1. Resolution shouldn’t change the number of draw calls. _Quality_ does that (Low, Medium, High, Ultra) 2. With D3D12 and Vulkan the number of draw calls is largely irrelevant (extremely low overhead) so programs can do 100,000 draw calls on a modern GPU before they start to see overhead. One of 3DMark benchmarks has a feature test, *API Overhead,* that shows how the number of draw calls scales.

[@AirborneHedgehog]

Great video, it was amazing to see how wide the gap in performance is when you get the GPU out of the way. One minor thought: at 11:24, you question how valuable the data from the 13900K vs 13100 with a 3060 is. I would argue that it is valuable when paired with the charts above it: If I'm gaming at 1080p, and all I can afford is a 3060 for the foreseeable future, then I know I'm throwing my money away if I pay for anything more than a 13100. I won't see any benefit unless and until I upgrade the GPU. But to your point: if I wanted to future-proof my build against an eventual upgrade? Yeah, that chart doesn't help much. (I'm in a bit of a weird case in that my preferred genre - flight simulation - is EXTREMELY limited by single-threaded CPU performance, so I did notice a bump when I updated from a Ryzen 3600 to a 5800X when paired with my RTX 2060. Prior to FS2020? I don't think I would have bothered much with the CPU.) Edit: Looks like you address that in the final thoughts. Well done. I'm convinced. 👍

[@qubes8728]

Funny this popped up. Was commenting in another video about my gains in WZ with the 5800x3d. One guy asked how a cpu can increase frames and didn’t even know what L3cache was! Another guy with a 3090ti/ 5800x3d asked if he should enable xmp. He also wondered if he should reduce vram usage in WZ because that’s what he saw on KZbin. I told him he’s got 24gb of vram so it won’t be an issue. I think some guys get confused when amd calls L3cache, 3D v cache. They think it something completely different or didn’t even know all cpus have L3cache.

[@absolutelysobeast]

just curious? what were your gain in WZ with the 5800x3d? I upgraded to mine a few months back before i was playing warzone so i dont have an idea of what i could have expected before. When i run the MW2 benchmark i get a 1 percent CPU bottleneck result. I am running it with a 4080 at 1440p max settings. But i seem to have alot higher 1 percent lows than my friends i play with, but they are all on 30 series cards ranging from 3060 up to 3090 so its hard to really get a reference point on whats CPU and whats GPU you know?

[@GewelReal]

doesn't matter what AMD calls it. average Joe has no idea about the PCs in depth. You could explain to him that CPU is like a gearbox in a car. Better CPU = longer gears aka higher top speed (if engine aka GPU can drive such speeds)

[@qubes8728]

@@absolutelysobeast i was averaging 120fps with a 5600x in demanding areas now im averaging at least 150fps. So around a 30fps gain with the 5800x3d with my 3070ti. 1% cpu bottleneck is good. Jufes from Framechasers noticed higher 1% lows in wz with 5800x3d as well. Check out his 5800x3d video. Seems to ba a wz thing with the 5800x3d but if they have x3d's then it could be the 4080 combo. Dont stress on lows unless it doesnt look smooth. your lows are probably higher than most guys highest fps anyways. Enable the gforce fps overlay in game and see how the lows go in game.

[@lawliot]

I buy CPUs based on the number of Xs in its name, no benchmarks needed. But seriously, I see at least one comment on every CPU video (not just on HUB) about 1080p and it felt pointless to bother replying to them.

[@Eidolon2003]

XFX RX 7900 XTX

[@assetaden6662]

@@Eidolon2003 thats a gpu tho. The most x's CPU can have is one.

[@DarkReturns1]

Can you please do an 8700k revisit? That CPU was game changing at the time and hugely popular. After 5 years is a pretty normal time to upgrade a CPU, I know alot of 8700k owners looking to upgrade this year

[@BlackHoleForge]

Thank you for taking the time to show people. Some people think you're just humoring stupid people. Stupid implies they know right from wrong, but these people are ignorant, meaning they don't know. You're not humoring anybody, you're educating people.

[@austinfowler2707]

Testing a cpu with the fastest GPU is about removing as many variables as possible. You can only go slower. So say with this cpu, the most fps you can achieve with the fastest gpu is 100 fps. Well if the next step down gpu is 70% as fast as the top, then you can reasonably expect that gpu to do 70 fps. Testing is all about removing the most variables. That gives you the best case scenario.

[@GamingLovesJohn]

I am one of those Steam users that plays at 1080p 165Hz with a 3900x and an 3080. I haven’t seen the need to upgrade my monitor, as most games I play are on high refresh rates anyway. I find that switching to 1440p, you still need a higher tier card. At 1080p, it’s easier for me to hit far higher than 144FPS+.

[@Outmind01]

Serious question because I have yet to experience refresh rates so high - can you really tell the difference between 144 and 165Hz?

[@assetaden6662]

@@Outmind01 Going farther than 144 has the effect of diminishing return. It will give you an advantage, up until the point of you being a bottleneck. No reason to buy 360hz monitor, when your reaction time is 400ms.

[@KontrolStyle]

@@assetaden6662 this is a great point, if you're not shroud or a pro streamer/gamer no reason to get 360hz. @outmind01 no you can't tell a difference.

[@VirtualAirTour]

I don't need you to explain why you test at 1080p - I understand it. What I need to know is what gains can I expect at the resolutions I play at. Some of us are actively shopping now. If I upgrade my 7600x/4090 system to a 7800x3d will I see any performance gain in MSFS at 4K or in VR? I have no way to determine that if the benchmarks only include low resolution. If the game is GPU limited, I need to know that so I don't waste the money on a CPU upgrade.

[@passerby6168]

Good point.

[@itsJPhere]

Sometimes you just have to make an educated guess based on what has been tested. The more data sources you can find, the more accurate your guess will be. With that said, 4K is the great equalizer of CPUs.

[@MrHamof]

11:34 Well, it is. In that it tells us that there's no point buying anything significantly faster than a 13100 if you're going to be pairing it with a 3060. (At least in this game at very high settings) But that's what the GPU scaling videos are for.

[@speedracer9132]

I want to see more channels benchmarking resource intensive games like MSFS2020 instead of low resource games that get hundreds of frames per second

[@marka5968]

Great video and insightful look into CPU testing. That is why I tell people to take CPU testing differences with lot of thought. You won't notice the difference between a high end CPU and a lower end one until years into the future. At that point, the upgrade would be much much cheaper to do than overspend today to keep for 6 years.

[@OverseerPC]

This sort of testing is really important especially for competitive-style gamers, wherein in most instances the CPU IS THE BOTTLENECK.

[@zoltanlaky4907]

Which is quite frustrating for gamers like me, who exclusively play GPU intensive single player titles and are happy with 60 to 90 fps. I skim CS:GO, Fortnite, Overwatch, Apex, PUBG and all other similar software in every benchmark. I want to know which CPU is capable of working well with my GPU without being needlessly overpowered, so CPU benchmarks are generally not useful for me. That is not to say I don't understand why they are set up the way they are, it's just not useful for me.

[@shagadelic3000]

@@zoltanlaky4907 Then go watch "gpu" benchmarking instead of "cpu" benchmarking, 5head.

[@OverseerPC]

@@zoltanlaky4907 In your case mate, just get a decent CPU (modern 6-8 cores) and then dump most of your budget on the GPU. Personally, I like playing at 60-100 fps high fidelity on single player games because it gives a more 'realistic' feel for immersion. Getting more frames is good but sometimes I feel like it takes out the immersion. For me, high FPS really matters when gaming multiplayer or competitively wherein you don't want to enjoy the graphics but focus more on winning with less visual noise and faster response.

[@assetaden6662]

@@zoltanlaky4907 overpowered CPU will serve you long, than having a "good enough for 90fps". It's future proof, just as the i7-8700k shown in the video. You will hit the bottleneck by year 5, but by then you'll have enough money to just purchase a better CPU, or be happy with it's performance.

[@alinuofal2581]

It really blows my mind how absurd trying to test CPU with a 4090 with silly FPS count, while ignoring the option to benchmark with 4x games that are CPU-intensive like Stellaris, HOI4, ANNO that brings the most powerful CPUs to their knees at late game. All youtube reviewers seem to copy each other at this point.

[@skinscalp222]

Use that brain, bro. Using top GPU to ensure the only bottleneck happening is on the CPU is basically the same scenario as testing it on games that are CPU intensive. In both cases the CPUs are the ones working hard. Like how to get this point through all those thick skulls? 😂

[@lukeearthcrawler896]

I wish you also did 4X games performance review, like the built in benchmark in Civilization 6. They are perfect for CPU benchmarking, as these games are CPU heavy and less GPU heavy.

[@saricubra2867]

CPU heavy? Laughs in PS3 and Switch emulators.

[@lukeearthcrawler896]

@@saricubra2867 Yes, CPU heavy. Run the GS AI benchmark in Civilization 6 and see how many seconds you wait between turns.

[@odytrice]

This is also why you shouldn’t just run out and upgrade CPUs if you play latest single player games at high resolutions because at 4K most GPUs are bottlenecked and the difference between a 5900X and 7900X won’t be visible in most games

[@opinali]

Avoiding bottlenecks from components that aren't the focus of a study is obviously correct. The problem is that "real games" are highly opaque and imprecise as benchmarks. Games are not designed to be good benchmarks; source code is typically proprietary so even experts with a lot of time can only make educated guesses about their behavior. And behavior varies a ton. See your 50-game bench for 5800X3D vs. 7600X. In CoD:MW2, the game is GPU-limited since all scores are identical for the two CPUs, a tie at 4K and still a tie at 1080p including 1%-lows; but scores scale well with resolution. This game engine does the same CPU work per-frame no matter the output resolution. Same for CSGO, Dying Light 2, Battlegrounds, Battlefield 2042, probably many others as well (I only checked the games with charts highlighted in the Techspot article) where the distance between the two CPUs is not zero, but the same for all resolutions. "Well many other games are highly resolution-sensitive for CPU load" - but if you average with others that aren't, your NN-games benchmark scores will badly underestimate the difference between competing CPUs. And if you cherrypick the games that are good CPU benchmarks at low res, that's a big problem in any benchmark methodology: selection bias, which can overestimate the difference because of a few games that are excessively resolution-sensitive in ways that aren't uniquely caused by CPU speed. There is a solution to that. CPUs (and also GPUs) should be tested mostly by synthetic benchmarks. But no major game reviewer does that, despite evidence that those correlate extremely well with real-world perf of games/apps that are known to not have undesired bottlenecks or performance anomalies. I didn't want to plug my own, amateur content, but look me up on Medium, I show that the 3DMark scores are spectacularly good for GPUs, matching the best real-world game benchmarks. Some titles depend more on integer code, others FP, memory, number of cores, cache latency, etc. Synthetic benchs can handle that with a good mix of tests that include all important scenarios. There's the problem of representing the mix of operations found in games, but you have the same problems with real-game benchs as well, e.g. some games are highly sensitive to a big cache while others don't care at all (see every Zen 3D review). So I don't know any reason to not use synthetic benchs. Not saying that CPU/GPU reviews should drop real-game tests; those are useful for multiple reasons. But reviewers could add synthetic benchs to their methodology, perhaps using a weighted average with game tests. Much better than relying too much on unrealistic CPU/GPU/resolution configs, which is problematic no matter how much you justify it. For one thing, there's no objective way to decide how much weight to give to 1080p/Low-settings scores in the final conclusions/scoring. It boils down to each reviewer's empirical choices.

[@InternetListener]

As a matter of fact they are Game benchmarks that serve the purpose of being CPU or GPU showrooms. The only thing tested is how much room for improvement the game has to fit an actual combination of hardware. Isn't it amazing when "reviewers" or game "benchmarkers" show power and % usage and you clearly see that 94% of your CPU (MB+ram+CPU can't go north the 1000$ mark) and your GPU sometimes goes at 100% but you can see the 60 or 70% at 1080p (and what a mistery when they're not using RT or AI scaling, and so you have 40% of unused die with tenths of milion of unused transistors for wich you are supossed to pay 2000$ or more).

[@sceptikalmedia]

I had the chance to test the 5600 non x vs 8700k . I sold the 8700k just for the 1% lows. the AMD part ran smoother every time with 1%low more than 20% higher. I dont miss quicksync.

[@riba2233]

Awesome video Steve... I am shocked at how many people don't get that. And some of us want to know how cpus perform at eSports titles, and if they can run 360 or 500hz monitors

[@BattleToads]

One of my old "buddies" should watch this video. He would swear that if your GPU was running at 99%, the solution was to upgrade the CPU to "open up" the GPU. He would insist that if you had, for instance an RX 6800, that you would need a MINIMUM of a Ryzen 5800X or else you would be severely bottlenecking the card. He based this data off of a promotional poster that Lisa Su was standing in front of during a press conference, insisting that the CPUs and GPUs were best "matched" if their numbers lined up. (Ryzen 5600 with Radeon 6600, Ryzen 5900X with 6900XT, etc.) I would explain to him that in the vast majority of games, a 5600X is going to perform nearly identically to a 5800X, regardless of how much GPU you throw at it. He simply would not listen. I would show him video after video illustrating this point and his response was that "all these big KZbin guys get paid by [insert company here] and go into their BIOS and tweak settings to get the results they want." I'd pull up a video of Gamer's Nexus who streams the entire process, BIOS included. Still "doesn't trust him." OK, buddy. Keep wasting your money on a GPU bottleneck.

[@SaltyMaud]

Oh my god thank you for mentioning performance targets. That's exactly how I view it. If I can get the game to run at _at least_ 120 FPS, I'll make the game run at _at least_ 120 FPS, no matter if that means high, medium, or below minimum quality. You can always lower the graphical settings, but you can't make the CPU run games faster than it can. I wish outlets would take this even further, the standard seems to be to test every resolution at ultra everything - I wish 1080p low was a standard test as well just to see what kind of performance target is achievable in the best case scenario. For competitive titles I'd like as high FPS as my monitor can display. 240, in my current setup. Not just average 240, I'd like lows to stay above 240 too for perfectly consistent gameplay. GPU limited data in CPU testing is completely useless unless that happens to be your exact use case and build.

[@SaltyMaud]

On a somewhat related note; 20:48 Funny anecdote, I had my GTX1080 through 3 CPUs until a year ago before upgrading my GPU because, like I alluded to earlier, you can always lower your graphical settings to make game go faster, but if you want more frames from your CPU, you need a more powerful CPU.

[@Raglarak]

I feel that all this 1080p tests still make sense even with the lowest settings, for all the competitive & eSports players that do gaming for a living, some competitive games like COD Warzone with the lowest settings at 1080p, you still can't give you solid +240FPS 0.1% lows, because you're still GPU bound with a 4090 paired with a 13900k, while in other competitive games like Fortnite you still can't get solid +240FPS 0.1% lows but this time because you are CPU bound with a 13900K paired with a 4090. So yeah, thank you Hardware Unboxed for also including 1080p Benchmarks.

[@blasterrr]

Great Video for educating some viewers. For multiplayer titles in non gpu bound graphicssettings the Performance differences between CPUs can be even higher. Unfortunately i Don t know of any reviewers who test the most demanding scenes of a game. As a result it looks like games run super Fluid these days and the gpu is way more important than the cpu. While in reality a game in a demanding multiplayer Scenario has a 1/4 or 1/3 of the framerates that u see in the Review. Especially in the 1% framerates. So actually for multiplayer the cpu is way more important if u go for high framerates and not for visuals. It would be great if reviewers could test more the demanding multiplayer Scenarios of games. I know that this is very difficult and not very reproducable. Bir u could do a more subjective play Analysis for multiplayer Scenarios. Great cpu bond games are warhammer vermintide 2, planetside 2 and Star citizen. Of course only in the right gameplay Scenarios. But those are actually the scenarios that do matter and not the ones where there is anyway not much load on the cpu. It would be super great if u could do a Video about this topic. To further fight against the misconception that CPUs Don t really matter for most games and to Show us multiplayer focused people how much we would really Benefit from a cpu Upgrade.

[@YeaButCanUDoABackflp]

True. Hard to messure but very important. Battlefield is the worst example as far as I know.

[@rjy87]

This also applies to cpu demanding sim racing games which are often neglected by reviewers. Especially at wide resolutions, the cpu becomes the bottleneck most often.

[@manoftherainshorts9075]

Always some new blood coming in to ask the same questions. Also, I honestly don't think 1080p tests are entirely useless. So many people still play at 1080p and are interested in performance at this res, including me. I personally don't *upgrade* monitors per se, I treat them the same way I treat a pair of shoes: if it's working, I don't throw it away or change it to better one. I upgraded to 1080p once my old 1280x1024 monitor died, and I will be use my new 1080p144 monitor until it fails. Also considering the fact that many games struggle to run at 144 frames even on 1080p monitors, 1080p testing is still relevant for a couple of years in my opinion. I'd rather run game at 144Hz locked at 1080p than downgrade to 90FPS at 1440p.

[@loowick4074]

MOST people still play at 1080p. It's still the more relevant resolution.

[@DavidMiller-dt8mx]

Fascinating. While I did understand why you tested with vastly overpowered gpus, I'm a more casual gamer, and this video showed me that I'm good with a lot less power - I also have multiple windows open while playing most games, and the range of cpus I'm looking at should be just fine. Excellent video.

[@flameshana9]

Having windows open while gaming is irrelevant, just in case you're wondering. They've done a video on that specific concern because so many people think you need 12+ threads as a minimum.

[@DavidMiller-dt8mx]

@@flameshana9 I do it all the time, on four. Trust me, if it's taking significant memory and/or thread cycles, it matters.

[@cpt.mccartman]

I already got why the 1080p testing is relevant for CPU benchmarking but wouldn't it be an even better showing if the GPU strain would be even more relieved by turning down the quality settings? I always wondered why you (and many other Techtubers) turn them to the highest preset or setting instead of benchmarking in 1080p at the lowest settings. Does that just not impact the result as much as I think it would or would that make a difference?

[@cpt.mccartman]

@unsubtract ok, that makes totally sense to me :) Didn't thought about that aspect, thanks :)👍🏻

[@Kuriketto]

As someone who's still rocking an i7-8700K, just upgraded from a Vega 64 to a 7900XTX and has plans to upgrade my CPU platform to hit a performance target of at least 144 FPS in most if not all modern shooters at 1440p, this video has been fairly vindicating.

[@nhkbeast3884]

At 1080p my cpu is being used like 100 percent and gpu like 30 percent but when i switch to 4k both are at 100 percent. Most people don't realise that you guys are actually looking for cpu bottle necks and that's why y'all test cpu's at 1080p.

[@dante19890]

the cpu bottlenecks will show up in higher resolutions too but not as pronounced

[@assetaden6662]

@@dante19890 yeah, when gpu is bottlenecked harder, cpu bottlenecks can only be seen in 1% and 0.1%

[@deltazero7012]

I completely understand why you guys test the way you do, I think these people are looking for something a little different from what you are doing and they don't realize it yet. They are more than likely trying to see how realistic combinations of cpu and gpu perform as a whole. Like say a 13600k and a 4080. Figuring out what cpu you need to not bottleneck your gpu is a confusing process with most benchmarks. A new set of benchmarks might be in order, or maybe just an episode that answers the bottleneck question for the relevant gpus on the market. Just some solid answers for people new in PC building like "Going above CPU X gives no extra performance with GPU X."

[@wilfredroettger5906]

I agree that, in order to test CPUs, you must remove GPU and other bootlenecks from the chain. But in end, it is kind of misleading. If you don´t have a top GPU, you most likely will get only a small improvement from a CPU upgrade (1440p / 4k - reallz don´t core about 1080p), if any. Best to go for the best GPU you can afford and a middle tier CPU.

[@WrexBF]

It's not misleading. After seeing the performance difference between an i3 13100 and an i9 13900k when paired with a 4090, go and look up benchmark videos of a slower GPU that you have like a 3060 for example. If the fps is below what that 13100 was getting with a 4090, that means that the i3 isn't holding back that 3060. There is no need to ask questions like ''can you test with a 3060, can you test with a 1660, can you test with a 3070, etc.''

[@wilfredroettger5906]

@@WrexBF Yes. But look the chart at 17:00. A 13900K is wastly superior than a 8700k with a 4090. But if you have a 3060, it really doesn´t matter. Better to upgrade the GPU, not the CPU.

[@whosnick6506]

I would like to see raytracing focused CPU benchmarks, kind of like you did with Spiderman, but with more games that are known for becoming CPU-limited with RT turned on. Many non-RT games are mosty GPU limited at higher resolutions, but games like Spiderman, Callisto Protocol, Cyberpunk, Witcher, Dying Light 2 etc. are known to be CPU-heavy with RT on. Would be interesting if some CPUs are "better" at RT than others or if it just scales with rasterize performance.

[@Oceanborn712]

Excellent showcase. Thank you for this. I can't count the amount of arguments against people going "Lol who's gonna play like this? That's not realistic AT ALL!!!!" I've had so far. This video should be linked under each of your future testing videos.

[@okesik]

What people also do not realize is that there are plenty of purely CPU-bound games, mostly simulators or physics-heavy games (Space Enginerers, Stormworks, practically speaking most building games have a CPU limit due to drawcalls) where eventually you reach a point where more GPU != better regardless of screen res. Some even recently released games are still CPU bound on a single-thread (OGL/DX11 titles on render thread, sims on physics thread, etc.) also.

[@tortugatech]

Heavy raytracing titles like spiderman, crysis remastered, sims like Microsoft flight sim and milsims like arma, squad and hell let loose are heavily cpu dependant, there's been a proliferation of heavy cpu games in the last 5 years!

[@mrbobgamingmemes9558]

True i play cities skylines with several mods and the gpu feels like almost idle, teardown is even worse with modded weapons, about stock/vanilla games (no modding) i managed to drop to 20 fps flying on london city and maybe tokyo again due to cpu bottleneck on xplane 12

[@saricubra2867]

I bought an i7-12700K with DDR5 for music production stuff and console emulators, it also has AVX512. Perfomance for those emulators is on par with the best Zen 4 Ryzen 9s at least from Techpowerup's testing and results. Ryzen 7 5800X3D got destroyed, 3D cache doesn't matter. Raw CPU compute do. Literally the CPU bottleneck for PS3 games is so big that using the Intel UHD graphics inside the CPU is fine.

[@assetaden6662]

Also people tend to forget that you can be CPU bottlenecked, even if RivaTuner shows 30% usage. It means that 6 of your 18 cores are being used, while other 12 are just chilling.

[@tortugatech]

@@assetaden6662 exactly, you need to look at gpu utilization to see cpu bottlenecks

[@cmja09]

yeah we have a PC shop and 95-98% of our customers play on 1080p, 2k and 4k benchmarks cater for a very small percentage of PC gamers. This is also why RX 580 8gb is still our best selling GPU because we only sell it for $100 and it is the best value for price to performance ratio. And I'm here wondering why GPU makers aren't making a sub $100-150 GPU with RX 580 performance that draws less power.

[@cmja09]

@El Cactuar 1440p is 2k

[@dante19890]

ye but its exactly that small percentage who are interested in buying high end parts. The average 1080p60 gamer doesnt care about benchmarks or the latest cpu and gpus as long as their shit still work.

[@damasterpiece08]

@El Cactuar nobody uses that resolution, get real

[@cmja09]

sure. well lots of technicalities, generally, 1080p is full HD, 1440p is 2k, 2160p is 4k. Because that's how most monitor manufacturers market their products.

[@ecokecok1376]

@El Cactuar You can even check that you are wrong from this video's quality settings

[@orangecapy9494]

This looks like a good occasion to point out a major misconception, related to power consumption and efficiency, that I see in many CPU reviews from many channels. What often happens is that CPU or system power draw are measured during very high CPU usage, such as in synthetic benchmarks, and the resulting values are used as a measure of power efficiency, to then rank CPUs. The problem with this is, obviously, that CPUs aren't usually used at a very high usage all the time, so that doesn't account for how efficient the CPU is during less demanding tasks such as browsing and gaming, which I imagine describe a very common scenarios for a lot of people. Some CPUs are very efficient at low and medium usage, and much less efficient at high usage, and this tends to make the results of that way of testing power efficiency incomplete or even unreliable.

[@krazyfrog]

How much power a CPU consumes under full load is set by the manufacturer. If a manufacturer decides to set the CPU to burn as much power as they want under full load for a particular part then that's on them. The review should reflect this rather than make excuses for it.

[@CyberneticArgumentCreator]

"but which CPU draws more power during a power virus?" is such a useless stat to benchmark

[@orangecapy9494]

@@krazyfrog You're missing the point. Which is that efficiency during intense usage is not very relevant for a lot of people, or at least not by itself

[@krazyfrog]

@@orangecapy9494 Full load testing is as valid as any other. The power targets are set by manufacturers. If they want their parts to be perceived as efficient they can set more reasonable targets. Instead, they set ridiculous targets just to chase the highest Cinebench result. If we are doing all the benchmarking under full load then it's only fair the power testing is also done under full load. Besides, HUB always includes gaming results as well so it's not like that testing is completely absent.

[@HexerPsy]

There are few consumers who really care about efficiency though. To really compare efficiency you would have to measure the power used over a given workload. Very few reviewers perform such a test. The function of measuring max power draw is to determine the cooling required for such work loads. And you are right in that most consumers on every day workloads can get away with lesser coolers much of the time.

[@rocketsurgeon1349]

Good explanation. And i'm sure there are plently of videos talking about reasonable pairings of CPUs with GPUs, now or if you're upgrading.

[@andersjjensen]

You better save the script for this one Steve... Because you're going to have to make it again in 2-3 years! :P I've been in the hardware space since the mid 90s and this always comes up. People want to see the thing they already want do good and be told that it's the right choice. They also most certainly don't want to be told to think for themselves by picking the lowest bound in the CPU and GPU test and conclude that that's the limit of that combo. Between those two, with a dollop of brand preference, there's apparently room for some real funky mental gymnastics.