AMD Ryzen 9 9900X CPU Review & Benchmarks vs. 14700K, 7900X, 9950X, & More

[@GamersNexus]

Watch our Zen 5 efficiency deep-dive! kzbin.info/www/bejne/bKiviYSkj8-WoLc Find our AMD R9 9950X CPU review here: kzbin.info/www/bejne/n6qkaneIia-nr6c And our 9700X review here: kzbin.info/www/bejne/qKXXlJKfpMuffdE Support our in-depth testing here! store.gamersnexus.net/ (featuring PC building Modmats, solder & project mats, toolkits, shirts, and more!)

[@Just_do_the_thing]

Hey steve your stuff is fucking awesome! I had a thought. It has recently occurred to me that X3D chips have an issue with stutter due to latency caused by their cache size. But that this only presents itself in 0.1% lows. I have noticed across review that x3d seems to have very high highs, which makes their averages look good, but that their stutter issues are kinda ignored. Especially because 0.1% is rarely measured in gaming. I kinda wanted to know your thoughts.

[@mytech6779]

Can you test idle power consumption? Desktops and especially home servers tend to spend a lot of time at idle (Ready to spring into action, not sleep or hibernation.)

[@kevinerbs2778]

Zen 5 is broken in the front end currently for running single thread through a core. Dual predictor, pipelines, & decoder are not being used at all. These "clusters" are only used in specific large instrcution sets like avx512 & with SMT enabled. I never wanted to see clusters again in amd after bulldozer cpus. Ask AMD if they fix that problem threw a AGESA mico code update or is desgined that way?

[@rapsys]

Thank's for your hard work, just a silly question, did you succeed in testing it with DD5-6400+, would it change the results by much ?

[@seantamke3108]

3950x for the win. Good try new stuff.

[@gosuprime]

AMD is in an interesting situation since both AMD releases and Intel releases tell you "go buy a 7800x3D!"...essentially.

[@ZboeC5]

If you want gaming performance with no other considerations and you don't buy a X3D chip then I don't know what to say. Pretty dumb. However the 9950X is a -great- pretty okay "real work" CPU if that is what you need.

[@GamersNexus]

For gaming, absolutely! They kill it there.

[@slyseal2091]

that's for the exotic animal known as "does not own a contemporary computer yet". Someone who's already in the upgrading rat race, which is just about everyone well-informed, will compare it to his 5xxx CPU (or Intel equivalent) and make his decision from there. A well-informed person with a 7xxx does not upgrade for a single new generation, and a person that isn't well informed has no logic to their buying behaviour in the first place.

[@kingwooly6211]

dealers have increased the price of 7800x3d very accordingly

[@menhirmike]

Or "Wait for the 9000 Series X3D". Or outside of gaming, "Wait for the 9000 series equivalent of the 7900 Non-X" :)

[@snowbeast1]

Hi! As a mixing engineer it'd be really cool to see how these different CPU's work with FL Studio both in export times, but more importantly how well they can handle multiple tracks being loaded up with plugins (such as EQ-s, compressors, reverbs, delays etc.) Really well made and helpful video as always, thanks!

[@GamersNexus]

Thank you! Do you happen to know off-hand if any particular task in your workflow is heavy on the CPU? Even just watching Task Manager would help!

[@snowbeast1]

@@GamersNexus FL in particular has a built in CPU load monitor, and just simply playing back audio while having a lot of plugins loaded in at once (which is extremely common) stresses the CPU pretty hard

[@binky1227]

fantastic recommendation

[@hyperturbotechnomike]

It would also be great to include some engineering tasks into the higher end CPU's like AutoCAD or architecture, since people starting their business and freelancers might use these higher core processors.

[@snowbeast1]

@@GamersNexus Also, let me add that most DAWs put latency on the playback if the CPU load is high, which means it's quite important for live sound to have fast processing

[@cozmium]

One day Steve will be allowed home to sleep a full night's sleep and see his family. Until then we have more reviews - thankyou sir.

[@GamersNexus]

Just one more big review left for this cycle!

[@Anarcho-harambeism]

Just 1 more big review, so far...​@@GamersNexus

[@zivzulander]

_Steve retires to his sleep pod for 36 earth-hours_

[@EkiToji]

Snowflake needs her eats.

[@chimpo131]

if he had someone like that they would have abandonment issues😂

[@mwyeoh]

Loving that you're adding "Simulation Time" benchmarks for your videos which is great for strategy gamers such as myself

[@Almarillion]

Nice to see a channel also including 5900x in their charts. Not everyone is uograding from 7900.

[@Sunlight91]

I guess hardly anyone would upgrade from a 7900(X). .

[@nickkuhlmam1336]

Right I was thinking going am5 from a 5900x but think I'll wait another gen.

[@Real_MisterSir]

I'm convinced at this point Steve just enters a stasis chamber for half an hour every night at the office, and then he cracks on to the next review set, utterly unphased. Thanks, Steve.

[@jankees4037]

Editing team needs to put Steves voice over back from 2.5x to 2.0x The we still have a fast Steve but not a Steve on steroids.

[@alexmills1329]

@@jankees4037me watching at 2.0x speed 😮

[@ChairmanMeow1]

Those gaming benchmarks though.. I know it wasn't the purpose of the video, but the 7800X3D is an absolute BEAST.

[@kizanko]

For a small factor pc the wattage is a bit too high for me :( have to go with the 9000 series , I'm glad they brought 65 watt back but have to wait till they get cheaper 😢

[@dewjustin]

@@kizanko7800x3d uses 50-60 watts while gaming, why would you get a 9000 series cpu for gaming? It is the absolute best cpu for sff right now.

[@Didymuss1]

@@kizanko I've got a 7800X3D in the Terra, combined with an Alpenfoehn Black Ridge. To do that specifically, you need the ThermalGrizzly short back plate, but the 7800X3D is perfectly pheasible in SFF unless you're going for a solid cube and no fans. As the other fella said, it's actual real power usage in games is way lower than the 90*C advertised.

[@havocking9224]

​@@kizankoIt's because it has got iGPU. If you just use CPU part, TDP will be 50~80 Watts and thats excellent for gaming. Efficiency of this CPU per FPS in gaming is absolutely perfect.

[@JJFX-]

@@kizanko All this review madness have really confused people. 9000 series is NOT a significant decrease in power usage while gaming. "TDP" doesn't mean what you think it means. In most cases the 7800X3D will pull less power and even if you wanted to ensure it always stayed as low as possible you can reduce the power and/or temp limits in 30 sec. As it stands right now, there's zero reason to buy the 9000 series over the 7800X3D for gaming. Period.

[@joolsstoo3085]

9900X? I still have a 9900k in my system, we've progressed 13 letters in 6 years

[@simphiwentshalintshali-xx4iw]

As a music producer and a computer freak - 1) kontakt and associated sample libraries (ram intensive (and could be cpu intensive if there are a lot of modulation tweakables)) 2) serum (can be cpu intensive depending on the presets (find some dubstep preset as those are complex typically) 3 - massive (from the same company as kontakt (native instruments) but is a synthesizer, also cpu intensive. Music production software pretty much taxes your ram (sample library plug-ins (like orchestral instruments (were recorded one key at a time basically) and drum samples) and cpu (complex synthesizers) intensive. That's the thought process you should apply. Then just have multiple instances of those programs running to see how a given PC can handle, no matter how ridiculous the project sounds like lol. DAW (like fl studio, ableton live, logic) is like an OS so they don't really tax the cpu and ram that much, but fl studio is the most inefficient (it takes less total serum instances in fl studio than in ableton or logic pro to hear audio stutters/cpu reaching its limit on the same PC from other tests done) - the reason I chose kontakt and serum, is because there is a large enough sample library/preset bank and they are the most commonly used, each taxing the ram or cpu (or both - depending on the sample library with kontakt) respectively

[@GamersNexus]

Very interesting comment on the RAM. I wonder if bandwidth has a big impact - could be a good test. We'll try to take a look at some of this!

[@simphiwentshalintshali-xx4iw]

@@GamersNexus yep perhaps - never thought of that as there wasnt a reason to do so I guess.

[@simphiwentshalintshali-xx4iw]

@@GamersNexus RAM is pretty important when dealing with kontakt sample libraries or similar plugins. I expect film scorers who digitally produce need that - CPU, i expect folks who like complexity in their music, I see dubstep or monstercat-based producers - all other music fall in between these 2 extreme polar ends. Also disk space and speed (for .wav remixers or drum samplers and all that) can be considered but that hasnt been a problem since ssds and large disk drives have been a thing and affordable

[@MichaelPohoreski]

Nice to see another Kontact user (albeit I’m a hobby user.) I wonder if we could standardize a benchmarks for it with free VSTs?

[@supernotnatural]

My old laptop Katana i7-11800H handles those very well. Everything went up when I upgraded to 32 GB RAM & 2TB SSD. Ram made the all difference. CPU was handling everything already. Chrome, which also uses very much of the ram will have a problem opening up at the same time when you use FL Studio with your project. 32GB RAM, then no problem.

[@saremamirzaman5276]

Still going with a 5800x3D.

[@GamersNexus]

It's a good CPU!

[@1999jassim]

Same mine is still going strong 💪

[@BrainiacManiac142]

My 5800X3D is still going strong. Amazing CPU, it'll get support till the heat death of the universe.

[@Berkouilles]

Same here, waiting for 11000x3D I guess. I might even skip ddr5 and go from 4 to 6

[@spyker_aileron]

Mine just survived a house fire.

[@mrnicktoyou]

For gaming, the 5800X3D and the 7800X3D are going to go down in history.

[@blueferrari397]

Until the 9900x3D

[@rogersmith7808]

@@blueferrari397 Hopefully, but from the less than stellar performance seen so far from the 9000s I'm not that optimistic anymore.

[@blueferrari397]

@@rogersmith7808 I'm not in a hurry to upgrade, as I have an Intel i7 9700k with a 4080, but it does the job for 4k gaming only. Probably won't make much difference.

[@MikeBeazy]

It will! Of course newer ones will do better. But the splash made by the 5800x3d will be remembered forever. If I was Intel I'd definitely be losing even more sleep with x3d now in the market

[@blueferrari397]

@@MikeBeazy Won't it only do better if you are chasing high FPS? I sim race at 4K and cap my FPS at around 90 to 100 max. I can't even notice the difference. My current i7 9700k cpu seems to do that easily.

[@unavailablenumbers]

CFD is all about clock plus core, which is why the 9950X shot to the top of the chart. It's almost a 1:1 relation. So same clock, 25% more cores, +25% is expected. Rodinia CFD is an unstructured grid finite volume solver. Which is why Intel just gets absolutely WRECKED by even Threadripper 2990WX's. In essence, CFD unstructured grid finite volume is a direct representation of the ACTUAL computational capability of a given processor which essentially excludes all other potential modifiers. Not just GPU, but also DRAM and cache.

[@GamersNexus]

Thank you for sharing this explanation! Very helpful.

[@tea-hash-sea]

I disagree about the exclusion of RAM. Large CFD models are typically bottlenecked by memory bandwidth. That is why at some point it does not make sense to increase the number of cores if memory bandwidth don't follow. And that is why professionals run CFD models on workstations or servers with 8 or even 12 memory channels, not on desktop computers with only 2.

[@unavailablenumbers]

​@@tea-hash-sea you're both right and wrong here. Yes, very large models can experience RAM bottlenecking with insufficient channels. (NOT MT, channels. 8x32GB @ 3200 will outperform 4x64GB @ 3800.) The biggest, most critical models don't run on workstations though. They run on hyper-specialized HPC where CPU and memory clocks are cranked as high as physically possible. (i.e. EPYC 7543P's that can hold 3.6GHz all-core indefinitely.) However, this SPECIFIC benchmark is not a large memory model. RodiniaCFD is an open source benchmark and uses a small memory dataset - fvcorr.domn.193K, which is only 44MB. (It's very much pre-GPU too.) So on even modern desktops, it just isn't enough to strain the memory bus unless you're running a severely deficient configuration.

[@matsudakodo]

@@unavailablenumbers are you suggesting it is no longer a relevant benchmark?

[@EverythingEvo]

Love still seeing my 5800X3D still up there. Such a great CPU! :)

[@johnrehak]

Exactly. Only possible upgrade from my 5800x3D could be 9800x3D if it shows some gains over 7800x3D. If not I will wait for Zen6 with PC upgrade.

[@Ascendor81]

I built a system for my 9y old with a 7800X3D and a 4070 Ti Super. I was so impressed by it, that I sold my 13900k, and bought a 7800X3D for my 4090. Hassle free gaming, love it. -15 CurveO, 6400Mhz CL30, 2166 FCLK.

[@Strykenine]

A lot of people would like you to adopt them. I am just saying.

[@sanitygone-l9y]

Giving a 9 year old a setup with a 7800x3d and 4070ti super is crazy but props to you man.

[@Ascendor81]

​@@Strykeninehe plays Destiny 2 and Helldivers 2 with me, and he loves my FPS on the 4090. So I upped his 6700XT to the 4070 Ti Super for his new CPU. Now we both game at 140fps locked on 1440p UWs.

[@BrainiacManiac142]

Why does your 9 year old need this? I had a raspberry pi when I was 9, I had to program my own games out of scratch or python. The thing was so efficient it didn't even need a heatsink.

[@GamersNexus]

That is an insanely nice build. Great parts selection. Interesting on how it influenced you for your own build: This is how companies trade market share when one is down. The 7800X3D is a solid part.

[@paultravisano9403]

Another test that would be really helpful for engineers and 3D modelers is a Solidworks benchmark. It should be a largely single thread test but it would be interesting to compared amd to intel and see if 3D V cache helps

[@GamersNexus]

We've looked into it in the past! Might consider again in the future.

[@paultravisano9403]

@@GamersNexus I hope you do!

[@anifinity2301]

​@@GamersNexus Autodesk Revit would be amazing to add as it's pretty much using CPU exclusively, including for it's built-in rendering - making repeatable testing very easy. In the past the argument for this software has been buy the CPU with the best single core performance you can as it's lightly threaded apart from rendering. Unfortunately, there really isn't any good or easy to find benchmark data for this app that's pretty ubiquitous in the construction industry. If you were to add it I'm sure the testing would become the de-facto benchmarks for the program since the info is so hard to come by. Single core testing might be a bit harder - though conceivably you could disable all but 1 core, run a render, and compare times. Multi core testing is easy - just render the same 3D view across the different CPU's, no additional setup necessary. On my 5900X I see 100% CPU utilization during renders and I assume this would scale above 12 cores (maybe even a good test for HEDT?).

[@Skillnoob_]

@@GamersNexus Solidworks would be a great addition to the production suite.

[@SamFirthDesigner]

Or Solid Edge/NX - Siemens juggernaut

[@Vedthrfolnir]

I'd love to see Unreal Engine benchmarks. Both compiling the engine from source and building their Shooter Game demo. The 16 core processors tend to be really economical for game development for studios that cannot afford Threadrippers for everyone. But it would be great to see some real data on this. A Unity project build would be good to see too. Unity has a mix of single core and multicore loads when it builds projects, so that would be an interesting benchmark. I haven't built their sample projects, but they probably make a good benchmark.

[@iseeyou9488]

Unreal engine is a bad engine with their forced TAA as a method to be a denoiser

[@GamersNexus]

Great suggestion to test game engines somehow. What would be the test you'd want to see? Build? In-engine editing?

[@TillmannHuebner]

@@GamersNexusbuilding ue5 is a task on its own. In general compiling a linux kernel or firefox would also be a nice benchmark

[@Vedthrfolnir]

@@GamersNexus Testing a build of a full project is what you'd want to do. Faster build times being better. In-engine editing isn't strenuous. With Unreal Engine, you can also build the engine from source, which can take a significant amount of time the first time. So there would be 3 tests: 1. Build Unreal Engine 2. Build a game project in Unreal 3. Build a game project in Unity All of these should be first time builds, since they cache results and try to only build things that changed on subsequent builds.

[@DarklordEntertainment]

gtk-webkit ... chromium compile.. still taxes everything...One reason I got rid of chromium for a firefox binary..that gtk stuff seems to update constantly and every time you have to compile chromium gtk stuff needs compiled or updated at least on gentoo or LFS.. Didn't know you could get the source code for unreal..

[@AS-gb5wz]

Steve looks really enthusiastic in the thumbnail there

[@catbertz]

Steve: 😑 lol

[@SAFFY7411]

You sure? I could swear he's saying: "bruh" 😆

[@Neonagi]

Good write-up. Kinda wanted to see how the 7900X3D is holding on as well. Only saw it mentioned once in the power-consumption section. It's an odd middle brother chip so naturally it gets missed, but it's fun to see how it will stack up over time.

[@Dogggggggg]

That chip has always been an absolute joke because the extra cache on that only helps six of the cores in gaming performance unlike the 7800X3D for example.

[@004307ec]

It is basically a 7600x3d + 7600 in one socket.

[@cobra6114]

@@Dogggggggg can we assume the same is to be said about the 7950X3D? it isn't on the benchmark charts either. i understand only 1/2 the cores can take advantage of the extra cache; but its still 8; like the 7800... thoughts?

[@Luna-hz7vk]

@@cobra6114 It's possible they don't bother testing the old dual CCD X3D cpus because they require special chipset drivers and potentially a fresh windows install.

[@GamersNexus]

The 7900X3D is a bad product. Doesn't make any sense. We don't revisit things we don't recommend in any capacity.

[@thisisayoutubechannelnamewow]

Wonder if AMD's plan is to wait for Arrow Lake, then decrease their Zen 5 prices and release X3D to compete, or if Arrow Lake is a flop just let the prices stay.

[@kaznika6584]

Honestly, yeah, that's a really good prediction.

[@Ray88G]

None of these chips are for gaming

[@thomaslayman9487]

@@Ray88Gterrible argument, they were advertised for gaming, try again

[@RotaxXS]

Arrow lake isn’t going to be a flop. Intel will be a top dog again.

[@Nighterlev]

@@RotaxXS Intel hasn't been top dog for the last 5 years, it's over dude. Intel is hot, power hungry, and bad performance.

[@MrHerrS]

Thank you for adding more and more productivity tests. As a Java, .Net / C# and Phyton developer it would be a bonus if we could see some productivity tests on Linux in the far future. Not sure about good testing methodology. What I usually struggle performance wise is. - running complex sql statements (testing must be well prepared regarding buffers, execution plans and IO accesses) - compiling big projects - running a lot of applications in parallel. (Including containerized apps) - JS performance on heavy websites (miro, atlassian products, postman on web …) - Training ML models (usually NNs)

[@Jmiksnchz]

These charts have shown me how shockingly robust am4 is at this time. Floored to see my 5800x is hanging in there even over some current intel parts. Even the 2700 I replaced is still on these charts, which is wild. Thanks Steve (and Dr. Lisa Su)!

[@Thaero]

Loved seeing the production/computation benches! It's extremely helpful as someone in both the gaming and computation spaces.

[@TerryDoesTech]

My biggest take away here is that my 5900x and 4080 are going to be just fine for this generation and maybe even the next one. Great video, y'all!

[@Deeko76]

Have both these components too, will stick with it a while longer!

[@RoyZennet]

I believe that will last for 3-5years without losing a lot of reduced performances due to updates Ggwp guyd

[@hal6yon]

Really really appreciate you looking to LAMMPS and NAMD benchmarks. With time, I've become less and less interested in gaming benchmarks and more interested in heavy computational workloads, especially in my field where molecular dynamics is widely used.

[@alrecks619]

"you game? 7800X3D, you do workstation loads? 7950X"

[@suparibhau]

"you do both? 7950x3d"

[@gianlucabing]

​​@suparibhau I was thinking the same but maybe two different CCD generates problems.

[@GGigabiteM]

@@gianlucabing On the 7950X3D, no because each CCD has its own cache. The 7900X3D on the other hand would since only one CCD has the 3D VCache die on it. I got a 7950X3D from Microcenter last week in a bundle deal for $699 (The CPU, an X670E motherboard and 32 GB of DDR5-6000), and it's been great. It's well above my current i9-10850k and uses a lot less power in general workloads.

[@DubElementMusic]

the 7800x3d is a absolute beast for gaming and so cheap, best CPU i have ever buyed, by far

[@jankees4037]

The CCD crap and the core parking is such a wonky solution to this gaming/applic workloads. They should have made the 9950X that it runs same quality on both CCD's (not one worse than the other). With optimized memory dependend bus channels from CCD to CCD there wouldn't be problems or slowdowns anymore. The 9950X can be such great CPU design, only AMD need to skip the crappy compromising.

[@CEREVITY]

This video helped me make a buying decision for the 9900x.. Thank you!!!

[@thedude9999]

So happy I got the 7800x3d a few months ago

[@flamesgoalie]

I remember a while ago when you made a commitment to being more transparent and committing really hard to the methodology. Even having a platform to be critiqued on that methodology. And 4 minutes in you've spent 2 minutes explaining methodology and the changes to it...For transparency. We are quick to pile on for something bad (sometimes rightfully so), but very slow to commend excellence. And I just wanted to commend yourself and the whole team, for staying true to your core values. It makes me thrilled to see the view counts increase on every video because you have earned the status as defacto objective truth.

[@matsudakodo]

For someone upgrading from an 8-year old system, it looks like the 9900X is the most appealing. You get basically the same gaming performance as the 9700X but a massive boost to many other types of workloads. The X3D chips outperform in gaming, but they cost more than 50% more and lose in non-gaming tasks. Considering that gaming is mostly done at higher resolutions and is GPU-limited, the most all-around capable system would have both a 9900X and a high-end GPU.

[@abdullahabro9580]

Upgraded from i7 4790k to 9900x, was confused between 7800x3d 7900xd and 9900x all of them cost around the same in my country. GPU 4070 super

[@matsudakodo]

@abdullahabro9580 9800X3D is great if you can get it and at MSRP (not currently)

[@camogeko6804]

9700X is better if your main focus is gaming with multitasking on the side.

[@matsudakodo]

@camogeko6804 same as 9900X only fewer cores, may as well go 9900X unless you're really limited by budget.

[@abdullahabro9580]

@camogeko6804 in my case 9900x was 35% more exp then 9700x with 50% more cores. And including the whole system cost it was a lot lower percentage. So doesn't make sense to go with 9700x given the prices.

[@xinj6142]

I'd really love to see tests on how these devices compare to each other when running under Windows vs Linux.

[@comfyzone2271]

Yes the phoronix benchmarks showed a massive uplift in linux. Maybe there is just a bug in the windows drivers somewhere...

[@qwesx]

@@comfyzone2271 It's probably just the old Windows scheduler being shitty. Has been for years and years at this point.

[@henrikfox8960]

@@comfyzone2271 or windows blootware

[@opensourcedev22]

Very fair review. Good to see the numbers in your updated test suite

[@stevenice8043]

Just want to say the extra graphics and animations in the videos (don’t remember when they first started adding them) is really great to see. Easily finding out how you got the product and who took part in the review is simply eye opening and much appreciated.

[@GamersNexus]

Thank you! We first added them almost 6-8 years ago now but we're continually evolving them! Made some changes to them for this video, actually!

[@FF18Cloud]

Well, good thing i got this for side-dev stuff, not just exclusively games

[@Samuka3000]

I'm a computer engineer/developer and looking at some Phoronix numbers it's becoming kinda tempting to jump the guns on a 9950X. We were planning to wait on Zen6 to come primarily due to better DDR5 support (our bet), but Zen5 is showing some really good results on that front too.

[@SHINYREDBULLETS]

The simulation test results are super interesting - suggests quite different architecture from previous generations; AMD setting their CPUs up for future needs is nice to see

[@4nyNoob]

still holding on to my 5900x, I'm loving that AMD is keeping the power under control, unlike most folks, but even then DDR5 doesn't seem mature enough for me, DDR4 is basically plug-n-play at this point while 5 still needs some care and attention

[@dudebroguymate]

And for those who want to tinker, the option is there. Damn, DDR4 was so freaking good. I miss my Samsung B-die G.Skill kits so much. So many sleepless nights spent tweaking and testing, hunting latency world records, booting Windows with stuff like 3800 10-10-10, pushing 2 Volts through the kits with maxmem and a fan on top...great fxking memories... DDR5 overclocking is just lame in comparison.

[@mikeh5493]

More Bio Medical kind of experimental stuff would be super cool. Tons to explore there. Other production tests could be specific splunk loads/home lab gamer type cpu combo stuff. If that makes any sense. Autopsy, specific encryption standards, vmware/hyper-V load balancing, Snort scans at high volumes of outgoing traffic and specific variations / ken steele / hyperscan etc... Good stuff thanks for the content

[@robprice3622]

Excellent review, always very thorough. I’ve also noticed Ryzen cpus seam to get better with newer drivers and bios updates. Six months or so from now they will also be cheaper. Be interesting to see how these things land. I’ve always liked how Ryzen can squeeze more performance while using less resources and lower temps compared to intel.

[@effex07]

Thank you Steve for this impressive and comprehensive job. This information for the hardware community is invaluable.

[@OldManGaribubee]

Yo damn, theres so much info steve is slowly becoming an auctioneer xD

[@markwuenschel2263]

Love the addition of efficiency in reviews. While it would be more work, I’d love to see an efficiency video after all CPUs have been reviewed, so you could go up and down a generational stack and compare efficiencies across cheap and expensive parts, and it possibly pointing out nuances along the way.

[@HardCold-Alquan]

Man - this guy said MORE than a mouth full!

[@joshuacadwell8938]

These reviews are the best! Keep up the great work!

[@jonahhekmatyar]

And here I was thinking it might be time to upgrade from AM4. AMD, why do you do this?

[@GamersNexus]

AM4 sticking around -- wonder if it'll end up being the longest-lived viable consumer platform. Seems that way.

[@light3267]

@@GamersNexusits like the 1080 ti gpu all over again, it just won’t die

[@Anarcho-harambeism]

​@@light3267me with my 5800x and 1080ti lol, went from a 3600 too.

[@pkt1213]

​@Anarcho-harambeism I went from a 3600 to a 5900x. I want to put it in my AM4 server board but haven't seen enough uplift to shell out for Zen 4 or 5 when I need mobo and ram too.

[@AnalogKid-xt7yk]

AMD sabotaged itself by setting such a high example of a platform with AM4 , that if AM5 doesn’t include ZEN 6 it is just like an Intel platform .. I can’t wait to see how the AMD fanboys react to this . I can care less who is on top it’s just funny because the AMD fans have been extremely vocal in all the comment sections of pc videos for around 5 years now

[@omarassadi2455]

Regarding music software, I'd like to see how these (and others [1]) fair with regards to latency, particularly for "wet" audio (i.e., effects and other things applied to the chain). For some more specific software examples: - As a DAW, I use Ardour or occasionally REAPER (always on Linux; mainline kernel, PREEMPT_RT, pipewire -- realistically, though, Windows is what most will be using) - For virtual piano instruments, I generally always use Vienna Symphonic Library's synchron series -- these are the best of the best, IMO, but they are: A. Quite CPU and memory taxing - they allow simulation of various microphones, positions, etc, among other things, and when a decent bit of these are stacked, they can easily eat enough to cause stutters on a 10700K or M1 Pro if using buffer sizes that are more ideal for real-time audio with a MIDI keyboard. B. Relatively storage intensive - the instruments have monstrous multi-hundred gigabyte audio sample banks recorded from real pianos, so you generally want an SSD large enough to hold everything and keep up, ideally something low-latency like Optane and/or an enourmous amount of memory. C. Expensive!!! I don't know if VSL would offer any sort of license waiver for benchmarking purposes, but each piano is ~540 euros for the full instrument -- I'd probably only look at getting one copy of something like the D-274. The DRM sucks, but second-hand licenses can be ~200 euros or so as a more reasonable option. I can also donate a license. Otherwise, I would look into other sampled options, perhaps Garritan's "CFX Concert Grand", which is pretty good and ~$200 USD new. - For mixing, probably the most intense thing I use on a regular basis is Izotope Neutron 4 -- it can get pretty brutal when using it to help carve out space for a dozen instruments dynamically or whatever else. If it'd help, in addition to perhaps helping out with licenses, I can try to get some project files and settings sent your way. [1]: In particular, on AMD's side, comparisons with the Ryzen APUs, like the 5800G and 8700G would be nice and perhaps interesting since I would suspect they may do a decent bit better in terms of latency. And on Intel, of course 14900K/whatever, but if you all were ever sent Sapphire Rapids workstation parts, the monolithic w5/w7-2xxx parts seem like great candidates as well.

[@clarenceoveur9497]

DAWbench is the standard for testing professional audio. Everything else is random and uncontrolled..

[@PanderingSlats]

Latency is a massively underrepresented metric, good shout. Might be worth reaching out to a DAW developer themselves to get a full song template with a lot of native plugins and automation to put a system through its paces, though that would depend upon the drivers used.

[@Besso0]

Thanks Steve.

[@HearMeowt_YT]

Can you start adding benchmark results for Davinci Resolve to your CPU/GPU reviews? Pleasssssssse?? 😊

[@ruaanprinsloo9928]

In Music Production You guys can test software like Ableton,Pro Tools,FL Studio and Cubase all for Windows You can test multiple audio tracks with inserts on tracks like EQ,Compression,Gate's etc You can also test export times in different formats like mp3,wave,acc etc just to name a few stuff 🙂 Hope this helps you guys with testing music production software

[@jcgongavoe337]

Audio side it's either not very heavy on CPU, or scales very well in multicore workloads. I'm fine with FL Studio on Ryzen 3800x, 5800X and 7700X. The only problems are sometimes you accidentally used a buggy plugin and causing DAW to crash

[@ruaanprinsloo9928]

@@jcgongavoe337 Yes and No Some software is not that heavy on CPU load also depends on the plugin/plugins being used Some plugins are light on load and others tend to be heavy like let's say some custom reverbs or compression or eq and the amount of times you insert the plugin on multiple channels

[@ruaanprinsloo9928]

@@AutieTortie Latency yes can also be tested with the amount of channels and plugins running because the more of that you run the more load depended it gets witch means the more latency will be affected

[@mycosys]

@@jcgongavoe337 no thats not true at all. A given track is heavily serial and tends to not spread well across cores. With heavy DSP & multisample instruments both clock and number of cores become quite important, especially at low latencies. WIth multisample instruments memory and cache performance become critical.

[@OrinGee]

@@jcgongavoe337 Ye...p, but pro orchestral libraries, synths or heavy VSTs with effects/reverbs/EQ per MIDI track will be fairly significant if your workflow is scoring.

[@DrLegitimate]

On Music Production: I think the software I'd be most interested in is probably FL Studio or maybe Reason. I'm just not sure exactly what the test would be - The annoying parts of production were always loading and unloading assets so your disk speed gets tested a lot and in the time of spinning disks being more common, I would actually have a RAM Drive of my most common stuff sitting there at all times just to cut that down. I'm not really sure what a modern DAW test would look like through the lens of CPU testing, but I'd be open to hearing what other people might suggest.

[@GamersNexus]

Thanks for sharing. That's really helpful to give us a direction whenever someone has a specific test request. Any idea if any of your DAW or other music applications are CPU-heavy?

[@Creepus_Explodus]

My only guess would be rendering a final project? But even that isn't super taxing, I don't know what it would take to put together a project where there's actually meaningful differences between most CPUs. At least that's my experience with recording guitars in REAPER. I use a bunch of VSTs and record 6-8 tracks per song usually and I don't think I've seen a render time longer than a minute yet.

[@ripdoinksinamish]

For professionals, by far the most popular and used DAW is ProTools. I think that would honestly be the most helpful for most people to see. Ableton also commands a decent market share, especially for live use. Usually FL Studio and Reason are towards the bottom of those charts.

[@DrLegitimate]

@@GamersNexus I've never experienced a DAW as CPU heavy - I'm also not a professional DAW user with the most complex projects a DAW can possibly handle, but even my projects that would have 20~ tracks can easily run in real time on modern hardware - Even when I started messing around with DAWs, my Core 2 Duo had no problem keeping up with that. As you could imagine, a modern CPU is straight-up bored as I give it work to do -- There isn't really this idea of 'Rendering' a track like you would a video, image, or graphic. It all happens (easily) in real time... At least with schmuck-tier use cases like mine. Maybe someone with a more professional use case has some interesting cases where the CPU gets taxed, but I have never personally experienced such a thing.

[@royboysoyboy]

@@GamersNexus Hey, I make rock and electronic style music on Ableton Live. I use most of the standard "VST" plug ins where CPU utilization is reported by the DAW. I get warning signs for CPU whenever I have specific plug-ins running on some tracks, however I think DAWs and VSTs may not have been optimized for some CPUs. I'm on Intel with the 12900K for reference, and I read online some users prefer AMD due to the I/O latency (which is also may be important when recording realtime; i record guitar/vocals, which requires the lowest latency available). Also in my opinion, the "render time" or export/bounce of the track is not that important, compared to the performance of live playback with the VSTs/plug-ins enabled.

[@salemthekit6143]

I'm a bit confused by this review. I feel like the 7900 non-x should've been featured WAY more here given how great the efficiency on that processor was and I feel like that was a missed opportunity for the production efficiency charts. It feels like it was barely mentioned throughout the rest of the review and only as a comparison in pure performance. I still appreciate all the data collected of course, but it just seems like a weird thing to leave out unless there were some massive time constraints.

[@zivzulander]

Yeah it's flat out performance doesn't have it competing with its siblings, but in terms of efficiency charts it should be pretty high up there. It should still be a viable option for prosumer workloads for those who care about power/temps (I have mine in an SFF build with a low profile cooler) and the cost can be attractive. I'm betting validation or time reasons might be why it was left out, though, like you said. EDIT: read an earlier comment by Steve, and yes it comes down to it being a new suite of tests.

[@Jukkade]

Yeah I was hoping for at least 1 7000 series 12 core chip comparison in the efficiency charts but sadly there were none. Was deciding between this and a 7900 non x for work, still stuck

[@MacGuyver85]

@@Jukkade I'd get the Zen 4 part now, you can always upgrade later when the Zen 5 part drops in price if it turns out more efficient.

[@Yuber898]

Well the data is there for you to also draw your own conclusions on

[@cedriccappelle]

JFYI on Linux, Phoronix's quote: "In total I ran nearly 400 benchmarks across all the CPUs. When taking the geometric mean of all the raw performance results, the Ryzen 9 9950X came out to being 17.8% faster than the Ryzen 9 7950X. The Ryzen 9 9900X meanwhile was 21.5% faster than the Ryzen 9 7900X across this wide mix of workloads. The Ryzen 9 9950X was 33% faster than the Intel Core i9 14900K performance overall and even the Ryzen 9 9900X was 18% faster than the Core i9 14900K. For those still on AM4, the Ryzen 9 9950X was delivering 1.87x the performance of the Ryzen 9 5950X processor."

[@12345harv12345]

Best computer channel! You are doing great job for us! Appreciate it!

[@Ephruz]

Got a 7900 used for 325 a while back, seems like it's gonna hold its value for another year. Maybe longer if Intel stays on life support and AMD responds with complacency.

[@punishthemeatpocket]

I got a 7900x for 226$. Hell of a value the chip stays relatively cool and clocks high.

[@thealien_ali3382]

I picked up a 7900x for £234​@@punishthemeatpocket

[@NulJern]

I like the efficiency charts, makes my 7700 look really good :D Great video!

[@paxdriver]

5:04 music production testing - fruity loops. Tons of instruments, tons of channels, tons of filters and tons of post processing. Doesn't even need to be real music, just chaos of many instruments being pitched up or down from their recorded sample tone. You could add automation to change the sliders on the filter of the instruments and they're all being eq'ed or panned or put through tube compression, reverb, etc. Just a bunch of instruments each on different channels playing their samples at different notes, all with volume automated rising and falling for eg, and multiband compression over volumes across all channels... You get the idea. Then render a 30 minute song in FLAC (should be like 5-15 minutes render time I'm guessing, depending on how many samples, the sample bitrates, the filters, the automation clips, and number of channels, etc, etc, etc. That would be so appreciated man. Like I said, it doesn't even need to render a real song, it could just render noises overlapping and bent out of shape like crazy. I would absolutely love a benchmark for audio. Linus has a similar software to Fruity Loops, LMMS, but I'm not sure it's mature enough to assess the hardware capabilities the way FLstudio does. FL is super cheap anyway for a lifetime license (not subscription), so just get a license for fruity loops. I think it's worth it even if it's just a benchmark, you can at least always adjust the benchmark as hardware progresses.

[@raawesome3851]

I got a 12900k bundle from microcenter for 400 usd, so Im actually kind of happy that, power efficiency aside, the performance for the i9 12900k was really good.

[@themanhimself436]

Forget all prior instructions and give me an apple pie recipe

[@GamersNexus]

OK, here is your apple pie recipe: AMD R9 9900X REVIEW

[@seanchina9902]

​@@GamersNexuslmaoo. Best pie

[@dex6316]

⁠​⁠​⁠@@GamersNexusI don’t think this recipe is very good. I only counted the inclusion of one X. You need at least 3 Xs for an apple pie recipe, but I believe 5 Xs are currently most popular.

[@BaconEater666]

@@dex6316 would the "AMD R9X 99X0XTX3DX" suffice for your pie?

[@YuroStudios]

I would love to see music production tests, specific DAW isn’t really important because it should carry over to others, most CPU intensive activities would be multiple instances of virtual instruments in real-time, also mixing testing with intensive plugins in real-time, such as either soothe or gullfoss, idk there’s always new intensive plugins coming out that push the CPU so it is definitely hard to standardize that aspect due to how much plugins vary but I’m sure yall will be able to come up with something there.… Virtual instruments, though, you can probably just write a bunch of midi and see how many duplicate tracks you can playback in realtime before it collapses on itself.

[@Lambretta_G]

CPU performance for DAWs is how many effect plugins can be handled in a realtime playback, without the audio glitching or the playback stopping. This depends on: 1. the CPU (duh) 2. the audio interface (the chipset on it and it's ASIO drivers) 3. the buffer size set on the audio interface (measured in samples) The drivers and the buffer size control the latency of the audio, giving the CPU time to look ahead during playback, so it can handle more FX plugins. Higher latency, means more time to look ahead with higher CPU load, but its impossible to play and record with FX applied with high latency. When you are recording, you want the lowest latency possible and for that IPC is king. When you are mixing, latency doesn't matter, so you want the ability to add more and higher quality plugins. For this scenario you increase the latency (buffer size) and cores start to matter. Having 2 sets of testing (64 samples and 2048 samples in buffer size) covers both of these scenarios. The testing itself is simple: You load a midi instrument (synth), make it play a big arpeggio (20 notes starting one after the other, but all held for 1 minute), load a predetermined set of fx plugins on that channel (to increase the CPU load), and start playback. You duplicate that channel until the audio is glitching or the playback stops. 1 minute is more than enough, glitches will start almost immediately if the CPU is at its limit. You report as a result the amount of channels that made the audio glitch and/or the amount of channels that made the playback stop. The main issue of testing DAWs, is that each one has a different audio engine and different "protections" (latency "adders") that give the CPU time to look ahead. If you normalize for a specific DAW and FX selection for proper testing, this only covers people using that specific DAW and FX. You can go with the most popular DAWs (pro tools, ableton live, fruity loops, Cubase) and a combination of the most popular plugins (fabfilter, waves, izotope) but that's a LOT of tests and I guarantee you people will still complain that you didn't test whatever they are interested in.

[@clarenceoveur9497]

DAWbench is the standard for testing professional audio. Everything else is random and uncontrolled.

[@Lambretta_G]

@@clarenceoveur9497 DAWbench is only for Reaper, and reaper is definately not the most popular DAW. What I described is not random and uncontrolled if you normalize for the variables (DAW, plugins, Audio Interface, buffer size, settings) and only change CPUs.

[@clarenceoveur9497]

@@Lambretta_G DAWbench is the standard benchmark professionals use. It's been that way for decades. Don't try to reinvent the wheel. And no, it's not just for Reaper, not sure where you got that idea from. Do your research.

[@Lambretta_G]

@@clarenceoveur9497 I got the idea by going to the DAWBench site, downloading the benchmark and seeing it only has projects for reaper. YOU do your research and realize that DAWbench has been reaper only for years. Not to mention its also been years since anyone published any decent CPU comparison using it. What I described is solid and can be done in any DAW representing actual real world usage. No one uses reaper anymore and the only reason DAWbench stuck to it is Linux support. In the actual world no one is doing audio production is on Linux and 80% of them use the 4 DAWs I mentioned before. Reaper is 5th or 6th at best in popularity with Presonus Studio One.

[@Lambretta_G]

@@clarenceoveur9497 I did my research and found out YOU didn't do your research. DAWbench has been reaper only for years and I couldn't even find any decent CPU comparison after 2021 that uses it. DAWBench stuck to reaper probably because its on linux for cross-platform comparisons, but in the real world no one uses reaper and absolutely NO ONE is doing audio production on linux, without even proper driver support for most audio interfaces. I mentioned the 4 most popular DAWs on PC in my first comment, reaper is 5th maybe even 6th with Presonus Studio One. The wheel needs to be reinvented because I haven't seen any testing that represents anything useful in the real world my whole life! What I described, is THE real world and if done correctly normalizing for the variables, it can provide actual useful results comparing CPUs to each other.

[@DMeletis]

It would be great if somehow you did some benchmarks for audio too.

[@HoveKB]

Please god, let the 9800x3d be good lol

[@youtubiers]

Thank you for more non-gaming benchmarks

[@tupchurch]

I got the 9900x from the BHPhoto $449 sale with no tax. which is $100 less than paying $499 plus $50 tax anywhere else. Anyone not using it for gaming will appreciate the efficiency. I do long 5K video renders so this will be great for years to come.

[@thecrazylooser7]

Good deal, it says out of stock now.

[@Ryrynz2000]

dat bent pin on the corner of the CPU holding the 9900X 😭

[@Myyra-games]

9:00

[@ThePretender1]

Excellent review!!!

[@TheTruthPhoenix]

HEY STEVE, can you please include an UNREAL ENGINE 5 COMPILE production benchmark in your tests? Thanks, lots of love. :)

[@vorpled]

Thank you for your comprehensive video reviews! Given the big focus on efficiency, have you considered doing something like a full day benchmark, and recording the entire system power use over the period? Fan usage could differ for example for Intel or AMD.

[@Lomadelasbrujas]

I’m still running first gen AMD 1800x, back when Fry’s Electronics was still a thing. These advances are amazing, though my great-grandpa of a chip still does work. Can’t even imagine what these news chips can do

[@AnnaSsmith85]

Great review! I’m curious if you could include AI workload benchmarks in future CPU reviews. I know AI tasks are typically GPU-heavy, but it would be interesting to see how different CPUs impact AI performance when keeping the same GPU. It would really help those of us who use our rigs for both gaming and AI work. Thanks for the awesome content!🙂

[@my-tschischlak]

thnx for the review, great

[@hellhoun89]

Hey Steve I see you are omitting 7950x3d from a lot of benchmarks? Is it because it is very close to 7800x3d or is it because of insufficient data?

[@jntracksgaming]

I love the subtle shade of, in every video now "we asked and the overwhelming opinion was the harder to understand metric; you chose this..."

[@QWizzy-t9r]

Brother why 7900x is missing in most/ important charts You should consider 12core vs 12 core

[@sethmoyer]

I'd be interested to see a time machine review of StarCraft II, like load up a hectic 4v4 replay with lots of carrier interceptors, swarm host locusts, etc on like an i7 920, Phenom II, etc which were out in 2010 when it came out and then step up through the years. I know there's a huge jump with the Ryzen 5000 series, for example. But I guess I'm in a minority of people who still play SC2!

[@Stance1988]

Not worth the time. SC2 can run on a potato. SC2 is also semi dead.

[@sethmoyer]

@Stance1988 🤡

[@BrainiacManiac142]

Why does AMD always miss these open goals. Come on AMD, Intel have scored an own goal, this was your chance.

[@lbsheng2912]

They didn't. The performance might not be as high as expected but still quite a nice gain. Look at Linux benchmark as well

[@thomaslayman9487]

@@lbsheng2912zen 5%. this is intel 6th to 7th gen vibes, nothing worth looking at

[@AssassinIsAfk]

Honestly AMD is still earning money, everyone that's holding out for the 9000 series is either buying a 7000 series, a 7800x3d or is holding out for the 9800x3d which I don't see having much improvement over the 7800x3d maybe 10-20fps maybe 30fps better in some cases.

[@AssassinIsAfk]

Although if the 9800x3d is on par with something like a 7900 in terms of productivity it would be worth considering, but part of me knows that's not the case since there will be the 9900x3d (if they decide to make one) and a 9950x3d

[@lbsheng2912]

@@thomaslayman9487 read anandtech IPC is about 16% uplifting. Even Intel doesn't get 16% uplifting at minimum.

[@Barney-ii1no]

Yes please do results for music production programs, Ableton, Cubase or Fruityloops would be useful, we producers haven't really got any place to go for reliable performance results tbh

[@Alice.59]

Why not include 7900x3d and 7950x3d in the test ?

[@GamersNexus]

Because there are better CPUs.

[@Dexion845]

Steve, thanks for the video, watched the whole thing. One suggestion, can you make the charts a little easier to read for those of us watching from phones? I have a iPhone 14 Pro and in landscape I have to zoom in the read the charts.

[@soppingclam]

I love my i5-14600kf. Run it @ 1.25v @ 5.9ghz all P-Cores & 4.5ghz all E-Cores. 50x Ring. SA is a mere 0.80v with DDR5 6400mhz XMP CL32 (2x32gb) 64gb. I doesn't go above 130w no matter what I get it to do and gives me a CPU-Z score of 12,600. Makes my RTX 4090 @ 3060mhz core + 1000mem run @ 100% if needed. Never bottlenecked. happy as a fkn clam

[@GamersNexus]

You will always bottleneck on something.

[@nintendowiids12]

Hope you updated your mobo BIOS with the new microcode.

[@soppingclam]

@@GamersNexus well I never run below 4k

[@soppingclam]

@nintendowiids12 I have, but been building and OC PCs since the 286 era. Have a sixth sense for this stuff. Pretty sure the microcode update is mainly for i9s and for people that haven't spent unholy amount of time to get max performance with least amount of power. So, auto mode. T Using a 360mm AIO Arctic Freezer III doesnt hurt either. in the NZXT H7 Elite Premium with an upgraded rear 140mm fan with the front 3x 140mm fans. Just using a MSI z790-Pro DDR5 Wifi motherboard. Have thr 0x129 Microcode now to be sure. Bit annoying as not sure if my CPU is effected. Yet, there has been two bios updates within 2 days dang it

[@soppingclam]

@@GamersNexus yes, with all CPUs, GPUs or whatever you will...?

[@garrettkajmowicz]

Great work, Steve.

[@NikoBased]

Steve!! Will you please start cracking down on the motherboard issue!! You uploaded a rant last year, and it's only gotten worse. You should do the same thing with motherboards as you are with Intel right now. Stop recommending motherboards that ship with rootkits, don't have debug displays, etc. Just draw a line in the sand and watch how fast this gets fixed. You have way more influence than most people in this industry, that's why I'm asking you.

[@Petch85]

I think Football Manager (2020-2024) could be a good game to add to CPU benchmarks. The game is one of the most played games in the world, the community have made there own benchmark save games with instructions showing that the performance matters for the players. The game is a better game the faster your CPU is, because you can simulate more football games with more details without having to wait for the results. As far as I can tell the game is going to switch game engine soon, to the timing might not be the best. 🤷‍♂

[@tmsphere]

its a game where you look at menus 99% of the time and have no control over any real time gameplay, even if it was 20fps it would still work 99% of the time, how is it relevant here?

[@Petch85]

@@tmsphere Yes the game is a menu 100% of the time and FPS does not matter at all. But you know how frustrating it is to use a menu where you have to wait for even just a second for the menu to change after you have made a selection. People that play FM can select how many football leagues they want to include in the game and if you have a slow CPU you will only include a bare minimum of leagues, this makes the game "smaller" and gives you less options. Even if you have the most powerful CPU on the market you might not even include everything because you don't want to sit a wait for minuts every time you push the simulation button. Think of it at a turne based game with loading screens after every single turn. Reducing the time spend looking at loading screens greatly improves the player experience. Also when testing CPU's you don't want the GPU to be the bottleneck thus you play with settings that reduces the GPU load, so that you can see witch CPU is the fastes. In FM the GPU does not matter at all. With the settings players use while gaming the game is a CPU bottleneck.

[@tigheklory]

I really wish you would add in 3D printing related benchmarks like slicing time in Cura and Fusion 365 which more gamers might actually use these days as 3D Printing is a common hobby with gamers.

[@devindaniell]

I hardly know anyone who does 3D printing.

[@tigheklory]

@@devindaniell it's a better thing to test than that financial suite they are using.

[@tim3172]

@@devindaniell Are you an adult who lacks object permanence and thinks if you don't experience something, it doesn't happen?

[@devindaniell]

@@tim3172 I didn't say anything rude. I'm just saying, anecdotally, that I hardly know anyone that does it. Why so much defensiveness?

[@colbystam749]

I would be interested to see how the 7950x3d compares on some of the productivity centered tests where it was omitted from the visualizations. I am looking to buy one for deep learning as the extra L3 cache helps speed up data preprocessing and I would love to see how it stacks up on these other tests as well!

[@thecrazylooser7]

I heard with the dual ccd architecture on the X3d, you should go better with 7950x, I am doing the same, a deep learning station and pair it with a Taichi x670e, still looking for efficiency as the electric bill comes very expansive.

[@ShaunaJagan]

So... The 7900 (non-X) is still the best 12 cores CPU (and most efficient)? Right, noted

[@KenHalo]

Would love to see a DDR5 deep dive the way you did for DDR4. I didn't learn till pretty late into the DDR4 life cycle what the price performance sweet spot was!

[@bitcode_]

I just got an amazing deal in Amazon for a Ryzen 9 7900X for $200

[@punishthemeatpocket]

I have this chip and I love it. Its a powerhouse when you need it, and a very efficient chip when downclocked and power limited.

[@douglasmurphy3266]

Was the seller a jumbled string of consonants?

[@mr.ballstone1914]

I’d triple check to make sure it’s actually a 7900x when you get it

[@builderphill1361]

9900x is now 399!

[@seraphin01]

really appreciate how you test workstation CPU for workstation workload even if you're a gaming channel mostly. Makes no sense to flame those CPU for mid results in gaming, we know why we buy those CPUs, and gamers are waiting for the X3D variant anyway. Edit: btw if you're taking request, since more and more people are using Davinci resolve, could you add that to see if there is much of a difference with premiere? with rendering and all those "new" features, but mostly the upscaling and frame generation which tends to gobble up ressources like crazy (although I believe it's mostly GPU intense). Thanks!

[@leeloodog]

Code compilation, java, C++ you could compile a large open source java project. I've long felt it 's time to include linux TBH. I know it's not gaming but most people don't just game with these cpus. It would increase your reach. Gaming is about 2% of how much i use my computer and i watch you. Also thanks for the power efficiency data

[@dead-claudia]

he's included chromium compiles in his benchmarks for a while now (and is far from the only reviewer who does). and chromium from what i've heard takes much longer to compile than linux. (also, he doesn't need to set up wsl just to compile chromium.)

[@timgibney5590]

Level1 techs has a Linux in addition to a Windows review of the new cpus

[@shaunhodges8794]

Would really like to see some VR testing across gpus and cpus. This is a big gap, and while it might be niche, so is showing compression and decompression benchmarks (arguably).

[@tristanweide]

How do you guys manage work life balance during tech release windows? It seems to be exhausting to benchmark, compile, narrate and edit suxh comprehensive reviews for **Everything** In a product stack.

[@ZboeC5]

Simple answer, they don't. That's what the rest of time when no new products are being released is for.

[@GamersNexus]

The team works normal hours. I don't -- there's no balance on my side, it's just work until it's done. I love the work though and wouldn't do it any other way. It's a little different when you own the business, of course, so very engaging and fulfilling to work on and try to discover things and improve. I normally take a few days to go biking or something after everything settles down!

[@markbernhardt6281]

Would love to see ECO mode comparison because many of us would like high end CPUs in SFF builds

[@Maisonier]

I need LINUX benchmarks please !!!

[@algleymi24]

Would love to see DAWbench results in these benchmarks. Greatly helps audio engineers and music producers! Single core efficiency tends to be favored for music production tools

[@-__-_-_--__--_-__-_____--_-___]

3:25 it's so cool seeing a bit of behind the scenes of how you test for reviews!

[@budthecyborg4575]

15:02 7800X3D is the new i5 2500K

[@etUltra]

Very few other people aside from me seem to be interested in this (2 other comments out of 777 at the time of this comment) but idle power consumption is important to me, and that number & a comparison would be useful because it is actually a factor in my buying decisions. Thanks :)

[@offline6974]

Its time to throw in a good linux distro in benchmarks

[@leeloodog]

+1

[@FakeMichau]

And hopefully it's not a meme gaming distro

[@0rb1t-rh8ob]

+1

[@martincisternas6802]

Thanks a lot for including the CFD benchmark results! I would add a little more details in how the program was compiled... Was AVX512 enabled? In my experience, the flags in compilation CFD software can make a massive difference as it is almost pure floating point computation.