E-Cores For Gamers, Are They Useless Or Not?

[@ConfusedStu]

You've listened to feedback, reflected on what happened, explained your position & thinking and revisited with a different outlook. This is how journalism and science is supposed to work, and very good on you. A lot of more mainstream journalists would do well to follow your philosophy more often.

[@rozgalang831]

which is why I will stick with my current cpu.

[@PainterVierax]

agreed with OP but just a remark for Steve: I don't think the Manichean thumb ratio matters that much because I personally didn't vote on the last video since I had mixed feelings about it and I felt that content didn't really justify a non-nuanced thumb down even if I voted up on most of the critical comments.

[@joewallace9030]

@@rozgalang831 this is why I'll stick with windows 10 Also privacy issues make me more hesitant until we know more

[@MrWarface1]

Lmao my god you are delusional. He got caught shilling for amd and is now trying to salvage the minute bit of credibility he has left.

[@PainterVierax]

@@MrWarface1 you must be very new to even think he left his entire credibility on one single video that feels incomplete and partially prone to misinterpretation. Sound like someone that judgemental should just stfu.

[@Dj0rel]

14:56 - You could test that. Gaming while streaming, 6p+8e vs 8p configuration. I think there's a number of people who would like to know how that turns out.

[@EvanidusEvan]

exactly

[@nvignesh]

I have been asking from this from very first reviews, its even in Intels ad for 12th gen. none of these reviewers are doing sht

[@ryanmalin]

I think that's a great idea. Use the e cores for streaming and game on the p cores

[@alexspeed8888]

Don't worry, the 13600k will answer your question, you just have to wait a while

[@carlomorischi3435]

Yes please

[@effeb2889]

When you disable the e-cores ring bus frequency shoots up, which is probably the reason for at least part of the performance increase when running no e-cores. It should show up in a full p-core enabled config too (8p/0e would do better than 8p/8e).

[@tablettablete186]

Yeah I was expecting something like that. Same amount of P-cores and change the count of E-cores to see if they help.

[@toddsimone7182]

That's an important point but can you lower it back to where it was for an apples to apples comparison? I wonder if Steve knows about that.

[@foch3]

@@toddsimone7182 You can but it hardly makes any difference.

[@goa141no6]

If that true how Intel would solve the problem? Do they have their own cache system like AMD?

[@jaturnley]

@@goa141no6 You assume there *is* a problem here. There isn't - this is how the CPU is supposed to work. The ring bus speed is a trade-off for having e-cores, and if you aren't using them (i.e. you only use the PC for games), you can just turn them off and get a tiny boost. Likely Intel will eventually sell a version of the chips with the e-cores disabled once there is sufficient stock of chips that fail to meet QC in that part of the die much like they do the failed GPU dies now, and those will have the higher speed ring bus by default. That's what Steve was referring to with the "cut" Core i5's - they will be CPUs that had major failures in the CPU core section of the die (and likely GPU as well), but they think they can salvage enough of to still make money on.

[@jonathanellis6097]

It would be interesting to use process lasso to force the likes of discord to run on the E cores, while benchmarking a game and then compare to letting the CPU manage its self. It would also be interesting to test streaming in that way. A for science test of course as most users are not going to bother assigning specific tasks to specific cores.

[@jemborg]

Interesting, and worth trying, but I got the impression that utilising the E-cores _at all_ caused memory bottlenecks that impeded fps. I'm talking about desktop gaming here.

[@jonathanellis6097]

@@jemborg I wasn't aware of that, I guess if the memory controller is having to wait for the slower E cores that makes sense? Or is it that they are taking bandwidth away from the P cores? Either way I would be interested in seeing some benchmark data.

[@niks660097]

@@jonathanellis6097 never make conclusion about cpu memory controllers that easily, they are most weird and hard to predict part(search layout bias), memory controllers is usually aware of program request is being made and is designed around usage patterns, and not a problem for tasks like gaming(arm still hasn't caught up with amd/intel in terms of memory controller latency i.e 50-100ns compared to 150-300ns in snapdragon or apple chips),

[@jemborg]

@@jonathanellis6097 I believe it was in the body of this video, many times. Do you remember Steve discussing the cheaper i5s coming with the E-cores fused off, also suggesting 2 extra P-cores would perform better than 8 E-cores? He also said he wasn't sure why incorporating the E-cores slowed fps suggesting bottlenecks may be the cause. It's my belief E-cores are of use only in mobile solutions really. But maybe things will change with the next generation or two.

[@PainterVierax]

@@jemborg not really only for a mobile solution. As shown in the launch reviews, e-cores are very good in productivity tasks like compiling/rendering jobs or containerization/virtualization. It's just that games/emulators are very dependent on ipc and overall latency and they can't be further multithreaded as easily.

[@gamergod9182]

I think a great test would be to see if the E-cores actually help with background process. So, one system with a fresh Windows install versus one with the usual stuff like Discord, Anti-Virus and game launchers like Steam, GOG Galaxy etc., and then a worst case with a hefty amount of bloatware.

[@squelchedotter]

I would expect that it is like with previous tests in that those things are not enough to fill up even a single CPU core

[@soppaism]

It may be difficult to test reliably though, as those background process load spikes are probably rather sporadic in nature and out of users control.

[@c.f.m.8133]

How do you guys imagine to do this, you have no control over chosing what cores games use. All you can is disable e cores completely, so they wont hurt your performance. The software you are describing doesn't need and wont max a single p core.... Marketing...

[@wile123456]

Don't forget corsair icue with 100 periphials with custom lighting

[@Valla451]

You should also test with manually forcing all non essential background tasks to the E cores. And with games disabled and enabled on e cores at the same time. And compare. Things like discord, a chrome window. Steam. Spotify. One test with heavy background and one with light, what you should have.

[@giorx5]

Another factor to consider trying to explain why disabling E-cores gains performance is the power limit for the package that increases a bit to be used for the P-cores only.

[@alpha007org]

I think when E cores are in use they lower ring bus speed and the ringbus is connected to all cores, P and E. So using only P cores gives you higher ringbus speed. That is also big factor here. But I don't know which has bigger effect, Power or Ringspeed.

[@RobBCactive]

Another reason can be reduced cacheline contention, running extra threads may slow down original ones by modifying data. Most people talk as if programs were embaressingly parallelisable, when the reality is usually only parts like rendering are.

[@argasasrgetthe9032]

12700k with e cores ring 3.6ghz, e cores disabled 4.6ghz. and they draw around 10-15W

[@andersjjensen]

He ran a fixed 4.2GHz all-core "overclock" in all scenarios, so power budget is disabled entirely.

[@saricubra2867]

Enabling AVX512 can boost IPC

[@EvernooBE]

One advantage of the E-cores is that your power consumption when idle is somewhat lower than if you only had P-cores. While Alder Lake is very power hungry at peak load, for most workloads you will spend more time idling and use less power. Very interesting for laptops on a battery. Less so for desktops.

[@Lishtenbird]

Since HU is into "for science" videos currently, it would be pretty interesting to have an FPS-normalized power consumption benchmark of various combinations of E/P-cores.

[@jhanninnen]

Would be great to see power consumption chart, between E and P cores, I really like the hybrid design, even on Desktop.... I think it also makes the CPU cheaper

[@williaamlarsson]

Of course it makes sense on desktop, because I bet the alternative would much more expensive or with fewer cores.

[@MasticinaAkicta]

So a Little/Big approach the ARM chipsets use. The Little cores use less power and allow under light loads for the battery the last longer. The Big cores are the real workers that if something needs to get done, it gets done quick. Which makes LITTLE sense on a desktop CPU. But for a laptop? It makes a lot of sense!

[@z-junlai3804]

Alderlake-P/M laptop and ultrabook CPUs run the gamut of 2 to 6 P cores, with 8E cores. So as much as these desktop configs are "for science", it's actually not a far stretch.

[@Ravic]

Please compare the 6 pcores with ecores off with the same setup but with ring/cache set to the higher you can get likely 49/50 since you can get a higher ring with the ecores disabled. I'm curious if it would be an additional benefit for gaming.

[@GewelReal]

OMG It's Ravic!

[@concinnus]

The comparison would have to between each's max ring, not default with e-cores and max ring without. If you overclock just one side it's not a fair or realistic comparison.

[@Ravic]

@@concinnus Not sure what you mean. There is the default ring they are running with this benchmark having 6 pcores and the ecores off, all that I'm asking is if there are differences between that when compared to running 6 pcores and the ecores off but with the ring at the maximum they can get. Since the idea of this video or comparison was already about gaming performance or gaining fps by turning the ecores off I'm curious if there is additional performance to be gained by also increasing just the ring while the ecores are turned off since it has been shown to be possible with Alder lake when the ecores are disabled.

[@concinnus]

@@Ravic If you overclock the ring with e-cores disabled, but leave ring at stock when e-cores are enabled, you are not making a fair or realistic comparison. Overclock ring to its maximum for each scenario. Even with the e-cores enabled, you can overclock the ring more than the cores. I doubt it's that much of a bottleneck.

[@Ravic]

@@concinnus Where did I say to leave the ring at stock with the ecores enabled? I'm just talking about comparing against pcores with default ring against pcores at maximum ring. This benchmark was about showing the performance or differences in performance in gaming with the ecores disabled so I'd like to know if increasing the ring which can go much higher when the ecores are disabled has an affect on the performance against the existing 6 pcores with ecores disabled example in this benchmark. It's a much greater difference in ring ratio then if you tried to go from stock to as high as you can go with the ecores enabled so it would be easier to tell if there were any differences or if they were negligible to the point where doing anything to the ring made no difference. For example with ecores enabled you'd likely be limited to 41-43 ring maximum under normal conditions, that vs stock ring isn't as big of a difference compared to ecores disabled and a ring of 49-50 at maximum vs stock ring. That larger gap would make it easier for them to see differences in the results. It's also to save time and make it a simple task since no benchmarking channels seem to want to test this yet with regards to gaming.

[@MEMETIZER]

Them E-Cores are pure marketing BS. I'd rather have a few more performance cores in their place.

[@whismerhillgaming]

and IMHO this is exactly why Hardware Unboxed is such a great reviewer/hardware youtube channel : not afraid to review their own previous content... listen to critics & add upon it, even if it wasn't really a mistake to begin with.... hats off.

[@oOMeowthOoPlushie]

11:56 - So the takeaway is, if you are into Alder lake purely for gaming, just disable the e-cores, this leads to the contradiction of buying into something just to disable them. And if you are both gamer and heavy threaded user, you will have a dilemma of wanting maximum performance for both world. Using process lasso is IMO very annoying, it is certainly a thing for advance users only. I don't want to see a new path being developed by an enthusiast community dedicated to find the number of threads for each game in order to maximize the performance, it is just too much work to follow, I just want to see Windows 11 to do this job itself. And you don't know how many years it will take for this issue to be ironed out, so this whole P and E cores thing is certainly early adopter technology. My question now here is, are these "Efficient cores" efficient for power consumption or just efficient for Intel's expense on big cores manufacturing? If it is efficient for power consumption, I don't see how this is suitable for desktop sectors, but power consumption is very important for laptop. I can understand why the i5 12400 doesn't have E cores, since you are going to disable them anyways.

[@lookitsrain9552]

The performance numbers were not too surprising, but what i would be most interested in, is the power consumption of the cpu in these specific configurations. Obviously, a 2 core 4 thread P core config with no e cores(not tested in this video) would likely perform terribly, but how much wattage does 8 e cores add to that config for that performance boost? TLDR performance per watt testing for different configs would be very interesting.

[@jonathanellis6097]

Intel are only really bothering with the E cores because of power consumption. If they stuck all P cores on the current line up (especially the i9 & i7) the power and cooling requirements would be insane. Intel's latest designs just serve to show how efficient Ryzen is. It will be interesting to see if the next AMD refresh will re claim the overall performance crown, both companies having competitive products is a win for everyone.

[@vncube1]

I believe it was primarily to increase multi-threading performance. Much much better performance per watt over Waste-of-Sand Lake and competitive against Vermeer of course, but that advantage seems secondary so far.

[@Heatranoveryou]

If you lock alderlake to 125w you dont lose much performance but gain a ton of power efficiency. Its just your typical small dick energy from the big 3. Cant lose by a small margin so were gonna cram as much power into it so we win by a small margin.

[@Time_Traveling_Lesbian]

Alder lake is more efficient than zen 3 so Im not sure what youre talking about. The e cores seem fine for their purpose. The 12600k manages to outperform the 5800x while using less power so the design works well

[@RobBCactive]

@@Time_Traveling_Lesbian That's not born out by the i9 8+8 vs 5950x 16 core performance and power consumption results.

[@RobBCactive]

It's true that the justification for E-core is power & die area efficiency. Similarly if it were possible to build uni-processors that were 16 x Zen3 performance for equal die area & power, that would actually win in benchmarks, because serialised sections would complete much faster and seperate threads would not contend for resources. P-cores are present to reduce latency at cost of extra power & die area; precisely because uni-processors hit a performance wall. In balanced mode, E-cores are used for background tasks NOT foreground applications, which creates 2 deconflicted processor pools. The problem for using say 16 E-cores matching 4 P-cores in gaming is programs don't scale performance linearly with processors. If just 10% of code is serialised (perhaps by memory access), Amdahl's law shows it actually takes vastly more small cores to reach near theoretical peak performance. So in most problems the weak small cores perform way below linear as more overhead is lost to resource contention. So the whole question of chip design is balancing factors in tension for current and coming programs.

[@martigrey5872]

Great to see you do an update and reconsider after receiving feedback. That's rare these days, no matter what the criticism. Great job.

[@youzernejm]

I'd like to see those e cores in a cheap, basic laptop with some new and improved celeron lake quad core. That could be a really, extremely nice chip for cheap, basic computing. I'm afraid it will never happen though, as they probably won't be able to price it where they want it.

[@Wamboland]

I really want to see tests with more programs running. I have bluestacks (running black desert mobile), chrome and amazon music running when I play a main game. So I am interested how the p/e cores handle such a scenario.

[@mycelia_ow]

Something like that for now is ONLY intelligently handled in windows 11. I've tested this myself with a 12700k, windows 10 gets confused by the cores, the Intel thread director only helps a little. In windows 11, the OS communicates with the thread director on the CPU and there's much better performance and responsiveness when you have a lot of things open.

[@saricubra2867]

@@mycelia_ow I can't imagine how insanely snappy Alder Lake would feel for music producers while doing other stuff at the same time Windows 10's scheduler is pretty bad, even with quad core CPUs.

[@vh9network]

Steve I wish you would take another look at the last two flagship Ryzen Threadripper processors and their performance improvement under Windows 11 scheduler.

[@czbrat]

they should release a cache heavy and no E core SKU for gamers. don't care how hot it gets under stress tests as it would be for gaming.

[@saricubra2867]

That would be another boring CPU like the i9-11900K.

[@ahnilatedahnilated7703]

The ONLY reason Intel put the E cores on the current chips is to meet a power consumption and heat output limit. They can't compete otherwise as their stuff is to power hungry and throws off way to much heat.

[@flameshana9]

That's the impression I'm getting as well. They're not a great idea, more like a half-thought out attempt to play catch up to AMD. And in 8 months or so we'll see how they'll compete with Intel. They might make a chip that's just as a strong as 8 p-cores and doesn't run as hot.

[@lordof7seas]

this is the graph I wanted to see, but also include power consumption and multi-threaded workloads, so we can finally have a clear picture whether adding more E cores or P cores will give the best Perf/Watt in future CPUs.

[@saricubra2867]

Big Little is of couse the future, M1 Pro and Max as examples

[@Staarfury]

Great followup video. I was honestly expecting the 2p+8e configuration to outperform the 4p configuration in more demanding games. After all a full e-core does outperform a hyperthread in other tasks.

[@andersjjensen]

Yes, but the core-to-core communication is dog slow between individual e-cores and even slower between p cores and e cores. This doesn't matter in Cinebench where each software thread doesn't talk to the others, but it matters a lot in a game because ultimately each thread is going to need to communicate it's states back for the final "freezing" of the current frame.

[@serioserkanalname499]

Love how you're going all the way with these tests .

[@Nikhilsj33]

I want to see how’s the power consumption while playing games with E cores disabled… Also would like to see if the E cores kick-in while streaming…

[@flameshana9]

Especially since like at 14:50 Intel claims the advantage will be huge while streaming.

[@striker241186]

Hey Steve. I really appreciate you having another look at this. Thanks for those tests. Gives me a better idea now of how the two core types perform with games

[@conza1989]

Wow yeah didn't even occur to me that it wouldn't be for science, I mean we can't buy E-core only CPUs, or even that anyone would be concerned about the E-cores with their CPUs... They still will be buying P-cores. If anything, the 37ns vs 57ns, to me as a lamen, seems like an achievable gap to close over generations, so if they could do that, wouldn't that be something?

[@boshumok6399]

I suspect the e-cores are really helping when there are only 2 p-cores, which you could confirm with a 2p-core/0 e-core test. I think it's just when hitting those diminishing returns after 4 p-cores that the e-cores don't provide any extra benefit.

[@ItsDJD]

Great video again! I watched the video of der8auer in which he also talked about the e cores. And one has to see the relation between the size of the single cores and the performance. If you take the size and power consumption in consideration, then its relatively easy to say, that the e cores are amazingly powerful. The e cores only have 1/5 to 1/4 of the size of a p core but at 1/3 of the performance

[@RobBCactive]

The problem is that a 4:1 ratio often won't be enough because of the way serialisation creates bottlenecks in parallel programming. So rather than linear scaling deploying more slower cores can be slower. That ⅓ performance may in practice kill performance, it's actually worryingly low. There's a rule of core design that 4x transistors deliver 2x peak performance, but that has to be balanced against Amdahl's law showing rapidly diminishing returns from parallelism (as if only ½ of code is parallel the 4xfaster using 4 equal cores delivers just ⅜ reduction in execution time (not even 2x speed up). A ¼ area core at ½ performance case: The ½ code executing ½ as fast on 4 E-cores would mean increasing execution time by 25% (2x½ = 1 + ¼x2x½= ¼). If you check Amdahl's law, you'll start to understand why games struggle to utilise many cores.

[@saricubra2867]

It's basically one half of P-core perfomance.

[@EraYaN]

4 E-cores still have more raw throughput that 1 P-core, so if games actually cared about heterogeneous chips like that they could absolutely leverage that for less latency sensitive threads within the engine. You’d think all those mobile development years would have bred some some heterogeneous compute wizards somewhere…

[@flameshana9]

I would argue that Android is a prime example of a failure to use all its cores. My cpu usage goes up to 20% and that's about it. Never seen it go higher. A Windows pc will always be able to make use of the cpu since you can actually multitask. But on Android all that power just sits there idle.

[@stephenhawkins555]

What you are describing is a lot more complicated and a lot less practical than it sounds. There are diminishing returns for increasing parallelism, especially in games, which are always going to be sensitive to processing latency. Additional threads requires additional context switching and synchronization overhead.

[@EraYaN]

@@stephenhawkins555 That is why you don’t split up the same task across all cores necessarily, I’m painfully aware of how that can decrease performance in more latency sensitive tasks. Thing is not everything that needs to be done in a game in a given second has the same latency bounds, plenty of stuff can in fact run on slower cpu threads (think all the actor system stuff in a simulation game for example) and many things will benefit from the extra throughput, since a lot of those are embarrassingly parallel or close to it.

[@zhugedai1279]

It's good to see when someone admits they made a mistake. It makes me feel like I'm paying attention to someone worth my time. Not someone who is always right or knows everything but someone who is listening to others and correcting themselves. It humanizes them and helps me understand their thought process more fully.

[@Wellibob68]

Noticed that your studio shots are a tad more colourful 👍 Better warmer feel/look similar to your old studio/video look. Great video as always.

[@alpha007org]

At first I thought 12gen arch with P+E cores is brilliant. And I thought E cores are good for gaming. After I saw the previous video, I had some questions about methodology. But, well, well, I was wrong as shown in this video. There is still a thought in my mind that scheduler is doing something wrong, but in the light of all this information, I can't "fool" myself anymore. E cores are bad for gaming, there just isn't any way around it.

[@Superiorer]

They arent bad for gaming. They are useless until your cpu utilization of the pcores is 100%. They will help then as can be seen in the win 11 vs win 10 test of shadow of the tomb raider test. When the 2 p cores are hammered and the 8 e cores arent helping (win 11) then the framerate is terrible.

[@vMaxHeadroom]

It all depends on workloads, for gaming, even the 5600 will do a stellar job and no doubt at the lower end the 12400 which will launch soon will also just kill it on gaming workloads. The e-cores do make a difference but as they get optimised, and where it makes sense, like multi-tasking and production workloads, low power task, they will come into there own. Super early days for this hybrid design and I think they will only get better. The P-cores on there own are monsters anyway!

[@alpha007org]

Maybe I wasn't clear enough. I thought at first that E Cores will be a net positive for *GAMING*. But when E cores are in use, ringbus speed is lowered to E cores frequency, and this had negative effect for the performance of P cores. We see this performance deg. when any E cores are enabled. It's small but it's there. Bad for gaming. Multithreading on the other hand, when you need as much CPU power is another topic worthy of discussion. Would it be better if Desktop SKUs had instead of 8 E cores, no E cores but two more P cores. And for mobile parts, that's also another topic. For Mobile, I suspect 2P + 8E will be a popular choice for long battery life with a good punch of those two P cores.

[@saricubra2867]

@@alpha007org An E cores quad core cluster lacks hyperthreading and clockspeed is so low, that already makes it worse than a 7700K in games lmao.

[@monkfishy6348]

I feel like the E cores are such a waste. Games perform better with them entirely disabled, not just for avg fps but low% fps too! Complete waste of die space. They should just lop off the E cores and put in half as many more P cores.

[@billwhoever2830]

What do you think about AMD's promise of +15% gaming performance when 3d cache is introdused on the zen3 cores? Do you expect amd to outperforme the brand new Intel cpus in gaming with something as simple as doubling the cache?

[@damara2268]

Something as simple as doubling the already big cache might help in gaming performance but also might make the price shoot up. :D 3d stacking process is expensive.

[@jondonnelly3]

Nearly all the improvement between 12600k and 12900k is the cache amount, not cores or frequency on gaming. So 15% is possible.

[@marvcyber815]

@@jondonnelly3 Adding to that, the IPC improvements we have seen from HUB with 5775c to 10900k mostly came from cache speed and cache size improvements with some came from the interconnect improvements So if zen 3D can maintain the same cache speed while adding that much cache, i can see that promise can be met

[@bogodoyandex9654]

@@damara2268 you can compare 5700g and 5800x, 5800x have double cacha comapre 5700g .gaming performance have 10% different. but only gaming ,another task, it just about 3-4% different. and ,5800g have higher clock speed than 5700g. so the 3d cache for sure can make zen3 stronger in gaming ,but how much? don't know, maybe 10% maybe 5% maybe like amd say 15% .but amd alway lie in their papagonda ,and for productivity ,no much different. and the price....much expensive. now amd become greedy ,unshame company.

[@marksnow8838]

@@bogodoyandex9654 Greedy? What The F? 5600x offered close to or exceed the $500 10900k in gaming when it was released, people always seems to ignore this plain and obvious fact.................you call that greedy?

[@Eternalduoae]

I appreciate you coming out and discussing the limitations of your previous analysis. I highly value this sort of self-reflective content because it implies (shows) that you guys are continuously re-analysing your prior data to actually be consistent. As a scientist, that'st important to me and, while I may not always agree with your stances on various industry trends, I feel like I can at least trust you to be honest and reflective on your now current opinions.

[@chronossage]

I feel like the whole e-core concept is really for laptops and the only reason it's in desktop processors is so Intel can try to hide their high power consumption.

[@bogodoyandex9654]

this wrong. if the make 2die 8core, and decrease the all core clock speed to 4.0, in this case we can got the multhread performance 16x4 =64 now 12900k is 8x4.9+(8x4.9)x0.38=54.1 so if they make all 16 core bigcore cpu, with only 4ghz .also can easy beat 12900k. and ,when the decrease the clock speed from 4.9 to 4ghz.the power compsution you guess how much reduce? the answer is more than 50%/ 12900k 4.9 power draw is 220-250w. if reduce to 4.5 /3.5 ,the power draw will reduce to to around 150w. if only 4ghz/3ghz, the power draw will even lower to 100w. (12400 6core, 4ghz only 74w) so if they make 16 golden core ,4ghz take out that ecore. the power draw will only about 150-180w.(oveclock to 4.8ghz need more than 300w), and 16core 4ghz, multhread performance can easy crush 12900k and 5950x.

[@davipadilha225]

alder lake's got a good power efficiency, just so you can have an ideia, you can underclock 12900k to 4.4GHz and it will perform the exact same as r9 5950x (stock), intel put the clocks higher so you could show a considerable better performance than zen 3.

[@Superiorer]

@@davipadilha225 THIS. And thats why the 12900k looks so inefficient compared to the i5 and i7. The i9 is clocked very high.

[@Bellissima2k]

@@davipadilha225 Wrong, it's already been shown that if the 12900K was power limited to 140W same as 5950X than the 5950X is 17% faster.

[@mclaughlinhugh]

@@Bellissima2k He's talking about underclocking not power limiting, two very different things. Even applying an offset undervolt with stock clocks helps a lot with these. As someone who works from home with remote desktop, ive seen power draw go from about 25w to 6w average (from 8700k to 12900k) throughout the day and always less than 80w when gaming.

[@Chris-nt1ns]

so a 12700k and 12900k both have 8 P-Cores, does that mean that they should perform almost the same for gaming?

[@Hardwareunboxed]

Similar, the larger L3 cache of the Core i9 helps though.

[@AlexSchendel]

Disclaimer: I do like the idea of space and power efficient E-cores. That said, Buildzoid has noticed that due to the lower core frequency limit of the E-cores, they cause the ring bus frequency to tank when they're enabled. With just P-cores, the ring bus can get to around 5GHz but with E-cores enabled that drops to low 4GHz. I can imagine this could cause the results you saw with the 4+0 config being faster than 4+8 config.

[@YeOldeTraveller]

As usual, great content. I would be curious to see a comparison between 0p 8e and 4p 0e (equal thread count). Adding 2p 0e (to see where that fits) and 0p 4e (scaling). I am interested in what Intel releases next.

[@anthonywong7004]

Thanks Steve! This video really cleared up the argument of E cores vs P cores. It's fascinating to see adding more E cores actually leads to a bit of performance regression, but still with the upcoming advancement in CPUs from AMD and Intel I'm sure we can get much faster stuff when 8 P cores in Alder Lake becomes not enough for gaming. And I'm sure there'll be someone who can make good use of 8 E cores or even more.

[@superneenjaa718]

Nobody is going to program a game for a specific line-up of CPUs. E cores are bad for this type of workload by design. Having less communication channels between cores decreases power consumption. In desktop, E cores are good for rendering type applications and that's it.

[@stuartb3082]

@@superneenjaa718 it’s windows job to schedule which cores are used for for what process not the game.

[@saricubra2867]

The perfomance regression is so insignificant combined with the P-cores, it's like enabling hyperthreading many years ago, then years latter, non hyperthreaded quad cores are bad for games.

[@saricubra2867]

@@superneenjaa718 E cores make the P-cores free from background tasks, they can fully focus on game or other similar latency sensitive applications specially good for multitasking or smoother.

[@superneenjaa718]

@@saricubra2867 it is never going to be like hyper-threading. Powerful cores can handle background tasks faster anyway.

[@pointblank722]

Got 12700K on Asus Strix Z690-a with DDR4. Old one was a 9900K. With 3080Ti at 3440x1440 DCS fps improved so good. Ontinuing tests and will post at my channel within couple of days.

[@cliffordjohnson943]

Thanks for doing honest reviews for the consumer and admitting where you may have made mistakes in testing! Great work man!

[@sermerlin1]

What I'd like to see is "Ecores" being used for windows/windows background programs while "Pcores" used when using primary program like a game. So that discord streaming, browser and other crap hammers the ecores while not disturbing the Pcores which should be only focused on rendering the game. Basically "Ecores" for OS... "Pcores" for rendering/processing data.

[@flameshana9]

Too bad it currently doesn't work that way. Just having the e-cores on means a loss of performance. It's a shame.

[@alanhonlunli]

I always said big.LITTLE is a waste of time. For some reason, there's always those fanboys that want argue how wonderful it'll be. Stop wasting silicon and stick more P-cores on it instead and call it a day.

[@saricubra2867]

That comment aged like milk, specially the recently announced gaming laptops.

[@alanhonlunli]

@@saricubra2867 funny, because this review shows how big.LITTLE has zero benefits for gaming.

[@saricubra2867]

@@alanhonlunli That nonsense statement would be equivalent like hyperthreading doesn't matter for gamers. Time will prove me right.

[@alanhonlunli]

@@saricubra2867 can't argue with the fanboys.

[@saricubra2867]

@@alanhonlunli I can't argue with AMD shills.

[@wpelfeta]

Thanks Steve, this really helped my understanding of the purpose and effect of E-cores. New tech is always confusing, especially if you only listen to the marketing pitch. These new core types are definitely interesting. Maybe in the future Intel could make a Gaming sku vs a Productivity sku based on additional P-cores vs E-cores.

[@the-wert]

What about power draw?

[@nttea]

Thank you for redoing the test in a more realistic(and more fair to e-cores) test environment! I am dissapointed to see that the e-cores fail to pick up any slack of 4 fully utilized p-cores, i had hoped they would prove at least marginally useful in those situations. Not even 2p+8e look particularly encouraging for e-core gaming. I still believe in hybrid core configurations being the future, with efficient cores being able to support gaming tasks with proper technology and software support.

[@OldManBadly]

I think that the question of big / little cores for windows, especially gaming, is pretty meaningless. if they intention is power saving, it's not really a great model to work from. When you consider that a 5800x (which i am using now) takes quite low power to start with, and that it steps down and takes very little power at idle, it's hard to justify having different cores for different things with the goal of "saving power". It's a better idea for phones and other portable devices, where maintaining battery life is key and any saving is valuable. For a plugged in computer, the energy saving seems pretty insignificant. More importantly, as e-cores are quite a bit less powerful, gaming performance can actually be hurt if the wrong workloads get sent to them. You could actually get a decrease in frame rate as a result. Big little would be meaningful for PC if the little was today's full sized cores, and big was actually something truly impressive and "big". The only way the current small cores work is if they are treated by the OS as extra or bonus cores to do background tasks like tracking the mouse or similar. Letting them run main processes or "helping" the big cores seems to be a losing proposition.

[@canyongoat2096]

Yeah I agree if it had like only 2 or 4 performance cores that could clock up to like 5.6ghz or similar performance with better IPC and they would be only used for quick spikes (launching programs) etc. or rendering while the performance cores would be as good as ryzen 5000 or slightly better then that would make more sense. This design is too dependent on good scheduling and I dont think you can do good scheduling for every single program. If games get better performance, maybe rendering wont use the cores to their full potential or things like that.

[@rna151]

There is the detail that in terms of die space 4 e cores = 1 p core and for multithreaded tasks the 4 e cores pack a lot more punch than the one p core. Games don't scale to all that many cores, not every workload is suitable to being diced up endlessly, and for those there could well be a reasonable argument that 6-8 p cores are all they should need and tasks that can be endlessly subdivided are better served by a heaping helping of e cores than a handful more p cores. I feel like it might be a promising approach. One architecture focused on single-threaded performance, another focused on maximizing potential multi-threaded performance.

[@DeSinc]

> Big little would be meaningful for PC if the little was today's full sized cores, and big was actually something truly impressive and "big". that's kind of what it is though. with the current P core sizes, it's not easy to fit even more than like 10 of them on a single die. that's why they went with 8 cores instead of 10 with 11900k because they were so large, and I suspect the same is true of 12900k despite the smaller 10nm process. the e cores are just to get those large p cores as the main performance-getters while still having the 16 core count. I suspect they were unable to fit 16 p cores in a single die given their size.

[@OldManBadly]

@@rna151 I agree, but it runs into the issue that since the e cores and p cores do not process at the same speed (naturally) and that leads to a lot of problems when you are doing things like gaming, live video / editing, or any other "of the moment" task. If you shuffle work load to the wrong core, you get the results later than expected and you end up with a hitch, a pause, or just overall decreased response. So yes, while there is some benefit on the physical side, the number of use cases of e-cores may be limited, or at least more directed at a certain type of operation. As an example, as I type this reply to you, e-cores may be the best way to go. But since the same workload could be handled basically by my idling p-cores, is there really any need? I can see the value of the power saving in mobile / portable devices. I can see the benefit of shifting say I/O processing to e-cores, but I am just not getting the real direct benefits over just having one more dual threaded P-core in the package instead.

[@JoeWayne84]

The big little is the way of the future though they will be able to add a lot more of the e cores and they or actually hella impressive seeing that it’s first generations of new archetectures i5 outruns the best gaming cpu of AMD at the moment and it beats it in most workloads if it doesn’t just edge it out it ties it and that’s it’s bottom end the new i9 is way the best option today and I’d bet the next gen i9 is going to show even more gains I think they or gonna let AMD release zen 4 before they go all the way it was odd AMD didn’t release a zen3 + right away to take back the lead if they could of easily tweaked it and got it on par with 12th gen they sure would of. Intel is moving to a ARM style architecture and that’s exciting seeing how it will play out in the years to come if they release a 12 p core 24 e core part times could get really interesting . But of course at first they will have bugs with the scheduler but they will get that worked out and it being Intel it will be way faster than if AMD was the one having to fix all the drivers and scheduler. This guys KZbin channel is a hell of a AMD fan boy channel and he’s been making videos the last month trying to find any trip up in the way the scheduler could effect gaming performance by actually disabling the scheduler and setting it manually the. Clocking the intels down to make them still show they or beating the AMD counterparts like all you got to do is let ‘em run against each other all out and it is what it is. The ryzen 5000 or good cpus though they or pretty much on par with Intels new i5 so shit no one should be upset to have either .

[@SB-pf5rc]

14:13 'it's also worth noting that most background tasks only present a very light cpu load...' Yeah tell that to my 2014 dual core laptop.

[@jasonoutman420]

Intel probably wanted a 10/0 config, but was unable to get power consumption under control

[@h1tzzYT]

which is a shame because that would be kickass combo for basically anything and without hybrid teething issues that windows and games currently have

[@SylphidUndine]

so it really is a stretch to be calling the i9 12900K a 16 core cpu. when really it's an 8 core 16 thread cpu regarding gaming.

[@dragam9534]

I wanna see 8p + 0e vs 8p + 8e in gaming... i feel fairly certain that in 90+ % of cases, you will be better off disabling the e cores entirely for gaming. The way i see it, e cores are just a lame feature made for laptop cpu's that we get stuck with on desktop, cause intel doesn't bother making seperate laptop and desktop cpu configurations, and laptops are their primary focus these days.

[@dondraper4438]

They're not a lame feature and Intel is doubling down next gen with up to 16 e cores. It's just that the architecture is new and software hasn't been properly optimized for it yet. AMD is going with their own P/E cores in the future as well. It's just a much more intelligent way of using die space and has been a thing on phones for the past 6 years~.

[@dragam9534]

It is very clearly a feature made for power savings on mobile devices... it has no place on desktop cpu's, where you just want as much power as possible...

[@dondraper4438]

It's like Hyperthreading with pentium 4. It was a box ticking exercise and was trounced by Athlons. That didn't mean Hyperthreading was a terrible concept or idea though.

[@dragam9534]

It is in no way comparable to hyperthreading...

[@dondraper4438]

@@dragam9534 Comes down to die space man. If Intel has to make 16 big cores, they're going to have to save space some where. With a big.LITTLE approach it allows them more die space for the big cores which means higher single threaded performance. It's not just for power savings.

[@yancgc5098]

I would just disable the E-cores and clock my 12900K to 5.2ghz. If you’re a 12900K owner that wants maximum gaming performance then that’s what you should be doing.

[@vgamedude12]

This ecore stuff and the software bugginess is honestly going to tempt me to just pick up a ryzen if they ever have a good sale. Coupled with cheaper mobos. I really thought I was going to go with a 12600k too.

[@andersjjensen]

The 5800X is only $340 on amazon while the 5600X is $300. So exactly on price parity with the 12700K and 12600K. Between what you're going to save on the motherboard and the better software compatibility I'd personally call that a win for AMD. Not by much though. But still a win.

[@vgamedude12]

@@andersjjensen I've seen the 5800x for 315 dollars on ebay (authorized seller and new) I just still haven't bought yet. I'd need new ram probably as well I only have 16gb of 3000mhz ddr4. Alot to buy considering in a year or two ddr5 will be common

[@andersjjensen]

@@vgamedude12 We're far off DDR5 being useful for gaming or even desktop productivity in general. It takes DDR5 6000 to get latency parity with DDR 3200CL14. And you won't lose much performance from keeping your 3000 until DDR5 makes the price of DDR4 plummet (though that's going to take a while). But honestly, at this point I would wait for Zen 3D. If for nothing else, then just for the price squeeze on vanilla Zen 3 parts (though I expect those to dry up fairly fast as soon as Zen 3D launches).

[@vgamedude12]

@@andersjjensen isn't zen 3d a full year off? Thought we were just getting cache increase refreshes as the next launch? Getting on am4 right now feels bad, but so does waiting. My 6700k getting very long in the tooth.

[@andersjjensen]

@@vgamedude12 Zen 3D is expected late January/early February. And it's AM4. Zen 4 is expected Q4 next year (to go up against Raptor Lake which is launching in the same time frame). But yes, upgrading now is a funny time. LGA1700 will be Raptor Lake compatible, and that's it. AM5 has not had it's support life announced yet, so that's an unknown too. That being said: an MSI B550 Bazooka and that $315 5600X is $455 total. Pair it with your existing DDR4 3000 and you still get 95% of the performance in CPU constrained gaming and 100% when the GPU is sweating. It's a cheap stop-gap upgrade.

[@bradmorri]

how do these cpus perform in gaming if only the 8 x E cores are used? As I understand it, the P cores have a ring bus to themselves and the E cores are on a separate ringbus. Intel is a bit hazy on how the two rings communicate with each other. Looks to me like the connection between the two rings maybe getting in the way, or the E cores performance disparity is causing P core stalls as important threads are being held up waiting on Ecores to complete a task. Looks like Intel's "Ryzen 1000" moment.

[@hamzix6599]

I had the same idea, the 12th gen is ryzen 1000 the 13th and 14th gen will be much better

[@lakshaythapliyal7982]

These e cores doesn't makes sense to me in dekstop why not just use them in laptop only to save power , p-cores are meant for gaming and people saying e cores can handle background task than a 16 (p-core ) cpu can also do the same or a 12 core cpu , as steve said p cores in alder Lake are powerfull which is the reason in increased performance .

[@jayhsyn]

It would be nice to see an updated version of this video , to see if the 13th gen is any better/different, as well as the windows 11 scheduler. To see if they’ve fixed these issues

[@ishanawate1849]

For science: Makes sense to see what the performance difference between p cores and e cores on the same cpu at the same frequency is like and also highlighting the power consumption difference between them while you are at it. For gamers and real world usage: While you are presenting the for science testing results maybe keep highlighting that actually in the real world the e cores work in tandem with the p cores, they are friends and actually help the p cores ...so it is never p core vs e core , it is always p cores with e cores. Also Steve maybe don't take a certain stern tone and make statements like: They are so bad for gaming, so much worse than the p cores.. like dude that just sounds like you are bashing on the e cores and saying boohooo e cores so bad.(This is what it sounded like to me in the last one, and though I understood the data, I didn't see why it was presented with a rather cynical/sceptical tone. I mean the e cores were doing what they are designed to do. What did ya expect ??)

[@ole7736]

So basically what SOTT does is the smart thing: do not distribute gaming loads across P and E cores. Though I would have loved to see a 0P core / 8E core config. E cores have lower performance and lack hyper threading, after all.

[@Superiorer]

Not really, there was a huge performance gain by using the e cores, but only on the condition that the p cores are hammered.

[@Azraleee]

Not possible. ADL will not boot with all P-Cores disabled. You could do something using Virtualization, but that would introduce additional variables.

[@ole7736]

@@Superiorer Thats what I meant by basically. Does not apply to all configs.

[@Superiorer]

@@ole7736 Ah right. Thanks I read it again and I was over critical. I think we feel the same about the p and e cores.

[@MrRedRye]

I'm still of the opinion that the E-cores are simply because Intel couldn't match AMD on core count without exceeding realistic power and thermal limits. It certainly makes some sense on laptops and having the E-cores is obviously better than nothing in heavily threaded workloads. It's clear that Intel are pushing their architecture to the limit once again which is resulting in wayward power usage and heat output, while AMD are operating in the comfort zone of their architecture.

[@kiwd-dynamic]

Unfortunately that's a reality nowadays... I believe even Elon Musk said something about it at some point. Energy, power. That's the shit. That's why nVidia cards always have the edge over AMD cards, for example. They don't care much about power consumption. I believe we are reaching a breakpoint and soon we will be needing new, more efficient power sources if society is to advance further.

[@mr.anirbangoswami]

Hi, can we get an updated video in 2024 regarding E cores

[@theodorosk.8932]

"6 cores is all you need" - Tech Deals, December 2021

[@phoenixzappa7366]

4 is all you need

[@newfaith912]

@@phoenixzappa7366 1 core is all u need to run programs

[@phoenixzappa7366]

1/2 a core is better

[@d.sherman8563]

Pretty sure they said 6 core is minimum you should get for a new system.

[@newfaith912]

@@phoenixzappa7366 hahaha how about no core. Just you calculate data and send them to memory :D

[@lexzbuddy]

I love the intro, integrity. Good man. Keep up the good work :)

[@richardbeckenbaugh1805]

Until intel commits to longer platform support, it doesn’t matter what they do. Buying a cpu that the motherboard costs more and only gives a single generation upgrade path is a nonstarter for most enthusiasts. If AMD continues to offer long term support for the AM5 platform, they will continue to win in the enthusiast market. Buying an AMD 470 motherboard allowed me to go from a 1600af to a 5800x. Four years on one motherboard. My four year old RX580 is starting to limit things but we aren’t going to see GPUs at reasonable prices for years. Possibly a decade. The difference between the AMD 5000 series and alder lake is only 5-7% in most workloads. When you can’t buy a GPU to keep up with your brand new cpu, buying a new cpu makes no sense. That is probably why these cpus aren’t selling. The i5 version is the only one selling at anywhere close to decently. The i7and i9 ( particularly the i9) are selling poorly at best. Having to cool a 280 watt processor is a nontrivial exercise. With a peak power of 95 watts, my 5800x maxes out my RX580 just fine. And I’ve already made that investment. I may eventually upgrade to the AM5 platform in a few years but until GPUs become available again, there is little point.

[@crunchychips8123]

I think scheduling is generally a mess, we're going to have to wait a couple of years for games which are E/P-aware not to lose a little perf on hybrid architectures. We see these growing pains whenever there's fundamental changes to the CPU like this, HT/SMT used to hurt performance too, and some games didn't even like dual core CPUs when they first hit the market. The one thing this video confirms for me, at least, is that Intel missed the boat not shipping a 4p/4e i3 SKU. If such a part existed, it would be a really interesting budget option. Oh well, maybe with the 13th gen...

[@saricubra2867]

If the Alder Lake i3 gets launched, it will obliterate the 10100, and 3300X on singlethread perfomance.

[@crunchychips8123]

@@saricubra2867 But the i3 announced is 4c/8t, no e-cores. Still not going to be that appealing for budget gamers. They're treading water with the i3s, doing the bare minimum, because an i3 with a mix of e/p could quickly cannibalise the i5 market.

[@saricubra2867]

@@crunchychips8123 4P+4E would be an insane config.

[@pedrosoares7273]

I can't believe you got backlash from that video, it was actually one of the best so far, so informative! Long story short: Games don't use more than 6 cores, cache size makes can make a big difference, P cores are what dictates gaming performance, E cores should only handle background tasks or aid in some multithreaded workloads

[@glittlehoss]

Awesome stuff bro. You go down the best rabbit holes.

[@attq3980]

I still consider 5950x being the best, 12900k is just too much of a beta test, some games don't even work on it.

[@hamzix6599]

any new gen is a beta test ryzen 1000 was a beta test RTX 2000 was a beta test RX5000 was a beta test and so on

[@Davesretrodungeon]

It would be interesting to see what the power draw is when comparing all the different configurations. Sorry if this has been mentioned already but I didn’t see the original video

[@nathangamble125]

I love this. You questioned your conclusions, and checked again, and found out what was really going on. Integrity is a disappointingly rare virtue.

[@Greendude439]

14:31 Was there always a hole punched through the center of that CPU socket?

[@kathleendelcourt8136]

I didn't have a big issue with the previous video using only 4 E-cores as only the 12900K offers more than 4 E-cores. So the "unfair" comparison with older 4c/8t CPUs didn't seem that unfair to me. If you buy a 12600K or a 12700K, which make more sense for a gaming PC than a 12900K, you won't have more than 4 E-Cores anyway.

[@Filotto]

An observation that comes to my mind about cpu load. From your tests it comes that spreading a gaming load past the p cores, to the e cores too will negatively affect performance. It would be interesting in my opinion to see if different levels of "background loads" would affect gaming if the e cores were handling it versus additional p cores doing it. For example, I'm thinking about different quality software streaming encoding in the background while playing the same game whit 2 configurations that could be 6p cores vs 4p+8e cores. Does it makes sense to you? If the encoding load is less latency dependent it should be better spread across 8e cores than it is across the 2p ones, given that the other pcores are fully utilized by the gaming load (might have included some errors in the scheduling logic of cpus but i'm not very knowledgeable about it)

[@dklingen]

Hey Steve, you are such an excellent person and so objective to re-assess your testing - if only more of the world had your approach to honestly and pursuing the truth!

[@Tmanstomp100]

in my mind the E cores seem good for handling windows background tasks or maybe running a stream while the P cores handle the actual game, seems like they would do good in that situation

[@c.f.m.8133]

Thanks for testing, watching different tests i had a feeling that all that Big little for desktop was a marketing bs to look good on paper. The reason for all this mess is just intel being scared that 12 is less than 16. P cores are amazing and we have those trash cores dragging them down and causing different stability and compatibility issues cause of marketing.... Nice one intel, even when your product doesnt suck you still try to crap it the most.

[@hamzix6599]

the new CPUS are fine and very good

[@InfraredVisuals]

Less than 1 minute into the video & the lighting setup of the studio looks way better than previous ones. Keep it going, HU 👍

[@lansiman]

nobody say E cores alone are good for gaming, or designed to be use by itself the previous title was horribly misleading and give off impression that intel claim E cores are good for gaming by itself extreme wording in title should be use responsibly,HWUB don't need bombastic title to get views

[@Hardwareunboxed]

But that wasn't at all what we were claiming. We're saying (and I feel this was very clear) as a backup to the P-cores they suck for gaming as they reduce performance, as proven here. We never said using just E-cores was even remotely realistic, you're completely taking the testing out of context.

[@kalestu_]

@@Hardwareunboxed well..that's at least partly caused by the more click-baity titles and thumbnails that you've been using for these videos...as someone else said - "useless" in the title of this one and so on - you can't use wording like that on the package and expect people to go in with a neutral mindset

[@Hardwareunboxed]

Watch the content before commenting is a basic expectation. Are they useless for gaming? Ahh yeah pretty much based on the evidence :S

[@kalestu_]

@@Hardwareunboxed don't prime the audience with the title in a way that can lead to understandable misunderstanding of the presented content as it is seen through the lens of that first impression with the title, could also be a basic expectation - (edit: or at least don't be surprised that this misunderstanding or misrepresentation of the presented content happens afterwards)

[@thatguy7595]

E-cores are useless for gaming in the sense that you don't need them if you already have P-cores, but on their own they can game just fine with well over 100 fps in many AAA titles.

[@SaltyMaud]

If Win11 scheduler is smart enough to keep high power tasks on the P-cores and background tasks on the E-cores, that's exactly what I was hoping it would do. For all intents and purposes I don't consider E-cores to have any direct effect on gaming performance, for gaming 12700/12900k are 8(P)-core processors in my eyes. E-cores are there to (hopefully) handle background tasks, improve efficiency in light use/idle and to contribute to the total CPU grunt when a heavily multithreaded workload calls for it. That being said, I still don't believe adding more P-cores would benefit gaming in any but the most fringe cases, if not for simply coming with a larger cache. 8(P)-cores is more than enough for gaming through the perceivable useful lifespan of alder lake CPUs. I might even go as far as to hope they won't start cramming more P-cores in there in hopes that power density doesn't start becoming an issue that necessitates dialing back per (P-)core performance. E-cores on the other hand are great for that, you can cram a lot of them in for additional multithreaded performance with a more modest impact on die size and power budget.

[@SaltyMaud]

@Sunkeny Nice oneliner, but I'm dead serious. Alder lake is great, my only complain is that we should have followed the hybrid core route much earlier. Future is either this, or ARM, time will tell which will prevail in the desktop space, or if both will coexist.

[@Al-USMC-RET]

I think I'll stick with my 5950x where all 16 cores are performing very efficiently at idle and have great performance when in use. I'm getting 5.1 ghz single core and 4.5 ghz all core with pbo. Never had any issues with it and performance is great. With my 6900xt at 4k it doesn't make sense to get one of these new Intel processors.

[@highlanderknight]

Steve, I thought the previous video was fairly clear overall but its still great to follow up and I enjoy them with my morning coffee. The extra information was still a bonus, but then again, you sold me on the AMD 5900x many months ago so....

[@trjozsef]

The E cores are for Windows Defender, Discord, the RGB bloatware and the automatic updates that we are too lazy to disable. The upside is not gaming on these, but them taking the load off the gaming cores.

[@c.f.m.8133]

Nice marketing driven logic, now go tell it to the game. Smth like: Mr battlefield, plz plz, do not touch those ones or u ll get hurt, just use thebig ones, oh u did, oh...

[@CaptainKenway]

Sounds like a you problem. None of that exists on my system, so I certainly don't have to pay Intel money to ameliorate it for me. AVX-512 is a far more useful feature than these useless cuck cores.

[@NOSfusion]

If you watched the video, you’d know how wrong this comment is.

[@andyastrand]

The e cores are for improving the core counts on Intel chips without them melting due to power draw. You’d be better off with a smaller number of extra P cores. For this we’ve opened the mainstream doors to asymmetrical cpu cores and the software issues they are going to cause for many years. Bad move.

[@A.Martin]

in the future we might get more multithreaded games that can take advantage of them better, but for now many games still only use a few cores, where the faster the better.

[@jonasziegler3025]

Im so hyped for the i3 12100f, basically an i5 10400f in multi core, but 40% higher single core and a maximum power limit of 77 instead of 134 watts. A comparison between those two CPUs with CPU utilization would be interesting, is the IPC being properly used or are the 6 cores still better for overall gaming performance? Questions like that are gonna be interesting. At least in Germany a decent B560 board costs about 85€, the i5 about 140€ and let's say a cooler for 10€. The bundle is now at 235€. The i3 should be like 100€, pair that with a 90€ H610 motherboard and we shouldn't need a cooler. We're now at 190€, close to 20% cheaper, let's hope it's as good as I think it will be.

[@nahush6299]

You don't need a separate cooler for i5 10400f either, it runs pretty good with stock cooler

[@jonasziegler3025]

@@nahush6299 Yes, you don't have to, but if you use it for multi core it sips about 100-120 watts, the stock cooler just isn't good enough there anymore. For normal gaming and basic productivity, sure. But I want a cooler, that keeps my temps in check in every task.

[@nahush6299]

@@jonasziegler3025 Fair enough

[@harryofficial3743]

Isnt the e cores for background tasks? Like when ur gaming and have background proccess like discord browser anti virus nd stuff is running so p cores handles the games and e cores handles the bg tasks? So if this is true then the advice to disable e cores shouldnt make sense as prolly ppl are gonna have bg tasks

[@goa141no6]

I want to see E-cores replacing the celeron line up for laptops, 6 ivy ridge (in performance) cores would be a huge winner for Especially with an arc IGPU for light engineering work and medium-light gaming.

[@NeutronicalGaming]

E cores are the only reason the chip makes sense, you get 4 cores for the same DIE AREA as 1 P core, for Half the speed. that's TWICE as efficient per silicon area, you're not going to get a 2x speed up any other way, intel made the right call. putting p cores on there would slow it down for any kind of multithread load, which is every load these days, the games already use the p cores, so the system gets the efficiency cores, intel did the math, it works don't worry about it.

[@c.f.m.8133]

So we should trust intel and not our eyes?

[@retaora4281]

@@c.f.m.8133 Yes because your CSGO will also run on a 6600k.

[@NeutronicalGaming]

@@c.f.m.8133 no, you should trust math, you're getting more processing on this chip than if it were configured another way ( aka all p cores ), since the video does not present us with a cpu with an extra 2 p cores to test the theory out trusting your eyes isn't relevant either, but i assure you the math is solid.

[@CinePhil101]

Maybe on paper. Real life situations are clearly full of scheduling and compatibility problems.

[@c.f.m.8133]

@@NeutronicalGaming Its not about extra, have u seen the video? 4 p cores outperform 4 p cores + 8 e cores. That basicaly screams that e cores just bring latency penalty and 0 benefits.

[@p3u3g3poultree7]

And Gamers Nexus called the 12900k the best CPU for the year and the 11900k "a waste of sand". I'm laughing now. 50 game titles unplayable regardless of windows 10 or windows 11. That alone is a huge fail.

[@p3u3g3poultree7]

@@Justin-rv1mm but it does not run over 100 game titles consistently for a channel called "Gamer's Nexus". Im sorry if you can't play your game on it = waste of money and sand.

[@AgentSmith911]

The more I hear about Intel's E-cores, the more I prefer AMDs simple core/thread architecture.

[@sanjisanpu]

I totally forgot I wanted to write a scathing, nay, eviscerating, I mean critical but constructive comment on the original video. This video basically addressed all the issues I had, great work. And the result is not what I expected. What a time to be alive.

[@kgonepostl]

Thank you for admitting you were wrong, and remedying it. That shows a strong/humble character. I appreciate your hard work as well, thank you.

[@Dazzxp]

The reason being why you see a regression in performance with just P cores and mixed with P and E core is the memory controller changes. With just P cores the memory controller works at Gear 1, with it mixed it drops down to Gear 2 or if you just use E cores then the memory controller still drops to Gear 2.

[@jarifahmed977]

E cores maybe improve Multi-threaded performance. Cause smaller Cores allows to fit more in same space.

[@andersjjensen]

Only certain kinds of multi threaded performance: the kind where the threads don't talk to each other. The core-to-core latency between e cores and p cores is horrible. That's why Cinebench scores are so good while gaming with e cores involved is so bad. Cinebench threads don't talk to each other at all. They just render pixels in each their own little square on the screen. Game threads have to constantly ask each other for state information to make sure the final frame matches the precise point in time it is supposed to reflect.

[@michaelbaldwin5953]

Like i said in the previous Video , e cores on a desktop processor does seem strange , whereas on a laptop they could do a very good impression of an Arm Processor.

[@asteele7931]

An interesting test might be running several background programs while the game is running and see if windows 11 dumps those tasks onto the e cores or p cores.

[@flameshana9]

The issue might be that Windows is already being slowed down just by having the e-cores running. So what could you gain from it when background tasks start going? It might just push your fps down further. I'd say a good test would be to use p-cores only and then start up a bunch of other programs. If they lose performance, measure how much over a long period (say an hour of game streaming while looping a benchmark). Then try it again with all the cores enabled like normal. My guess is the p-core only setup would take a noticeable hit, but it wouldn't be much lower than the limit already imposed by those e-cores.

[@kiwiasian]

It’s good to see follow up content as new information come to light. Good stuff 👍

[@sigy4ever]

Props to you for coming back and relooking over things

[@INSANEDOMINANCE]

Idk how you do it. I just upgraded my 4790k system to a 12900k and its been a nightmare. It made realize how much work you actually have to do. Again, idk how you do it lol. Drivers, compatibility, updates, hardware changes, and making sure everything is correct. I’m glad you enjoy it all, thank you for all of the help/content!

[@drgr33nUK]

The issue is that Windows 10 doesn't use the Intel Thread Director so it just runs the workload on any core regardless of type where Windows 11 does and knows that E cores ain't for heavy workloads. What you're seeing is expected behaviour. If you're using Windows 10 then you should just disable the E-cores as they will create more problems then good.

[@rbroccoli_]

I didn't see the original video but i'm glad you came back to clarify it all and appreciate the work put into the video. Also, i dont think I want to hear anyone else say P-core and E-cores that many times ever again. :)

[@ryanmartie1244]

Very informative video, thanks Steve!

[@H0don]

I wonder if there is a difference in performance between disabling E-cores by software like ProcessLasso and BIOS.