There’s a possibility that you’re seeing these results based on the power/heat envelope of the cpu package shifting the clocks based on low. So on a lightly threaded game that’s not pushing the package temperate up it may be able to maintain higher clocks on the P-Cores. But if a game is hitting all cores, it’s resulting in the cpu throttling down. What would have been interesting would have been to have the P cores, E cores, and Ring/Cache locked at specific numbers, including when E cores are disabled. This would isolate the performance issues down to the scheduler and give a better idea of why performance is changing in games with them on or off and lead to better guidance on setup for folks.
i think most shouldnt disable e-cores, since these results very much resemble a best-case lab scenario where background activity is at a minimum. with most people running chrome tabs, light software and maybe even music in the background, the e-cores should help keep the game performance running smoothly without sacrificing p-core performance
I agree, I think the E-Cores are best suited for exactly that stuff, background programms, while the P-Cores can concentrate on the game. I'd assume that newer games and some Windows 11 updates will be much better at directing each core for its designed job.
Assuming they actually are scheduled properly maybe., But those e-cores do not even seem good for background tasks. They seem good as assisting for heavily threaded productivity workloads that can scale to infinite cores. Of course equal amount of P cores would be much much better, but adding double e-cores can trade blows with 1/2 amount of P cores added as double amount of much weaker cores in fully parallelized capable workload. But for workloads using less threads, less P cores with no e-cores way way better. Plus no
I've got a strong theory that context switching is hitting the single thread games when disabling the E cores as less total threads are available for the OS itself. A process lasso or affinity would prove this.
I also strongly suspect this. If E-cores are enabled, stuff like context switching of various background tasks is likely offloaded to them and thus a single-thread game can fully utilize one thread. It's probably due to in-CPU controller of P-cores, as well as core-to-core latency.
I really want someone to test this... E-cores enabled BUT make the game only use P-cores with task manager or Process Lasso. It should be the best scenario in theory.
I'm wondering if we lasso pretty much everything o/s related to e cores, and have p cores solely for specific programs that you want them to be used for. e.g. games, encoding, etc.
Process Lasso also gave some extra performance for my aging AMD Ryzen 2700x. In older games that do not need more than 4/8 threads, restricting them to a single CCX gave extra 3-5% framerate
As far as i can tell, from the time owning an 12900k, the E-Cores are causing more "trouble" than anything else. At least when it comes to gaming. There are only a handful of games that are actually benefitting from them but there are far more games where the E-Cores are straight up causing issues since the scheduling can be all over the joint. In Far Cry 5 for example, despite E-Cores being disabled and E-Cores being enabled are giving fairly similar framerates, there's an awful lot of microstutter going on with the E-Cores enabled because the P-Cores aren't hyperthreading (like they are supposed to do) but instead throwing each of their second thread on the E-Cores. Disabling the E-Cores in this scenario isn't even just about the little performance boost i get there but way more about the overall smoothness since it completely eliminates the absolutely awful and constant microstutter. Other Games i've tested saw an significant performance boost, especially in the min. FPS department and some games were straight up broken with the E-Cores enabled causing all sorts of weirdness. My conclusion when it comes to Alder Lake is that the Hybrid Approach seems to be a decent idea at first but fails miserably in the real world, at least when it comes to gaming. Looking at the 12900k's die, i'd rather have the 8 E-Cores replaced with 2 P-Cores. Not a fan of the Hybrid Architecture to be perfectly honest, not at all.
@@johnm91326 well actually there is and second die would be adjusted accordingly to that config Anyway if u mean for that exact size 4ecores is bigger than 1pcores clear as day in die shot
@@farhannik7039 not really, cpus aren’t just about gaming surprisingly also maybe games will try make better support for e cores if ryzen do that as well
How much of this difference can we attribute to uncore/ring ratio? With E-Cores On, ring stays at a rather low 3,6 Ghz. With E-Cores Off, ring clocks at around 4,6-4,7 Ghz.
@@intetx no it does not. L3 latency goes down by over 30% if you disable the e cores and oc a bit, thats absolutely massive for cache/memory sensitive applications
The most important thing to have been addressed was to tellus how the L3 Cache behaves when E cores are disabled. Do they take with them their respective Cache and block it from the P cores ? Or do the E cores handle their cache over to their big brothers ? If the first case is true , is there a way to reclaim back the cache for the P cores ? Also , you haven't mentioned the benefit in thermals, power consumption and the potential OC headroom the disabling of E cores brings to the table ... All these make it a worthwhile exploration for just gaming i believe ...
If you disable the E-cores, you get ALL the L3 cache available for the P-Cores to use, instead of having to split/share it with the E-cores. This is a very good thing. The only reason to enable E-cores is if you REALLY need that extra bit of multithreading performance in fully-threaded workloads. But the multithreaded performance difference isn't even much between a higher OC headroom with P-cores only (lower core voltage needed for higher clocks, AND higher Uncore OC), vs leaving the E-cores enabled and being limited to lower P-core and Uncore clocks. A lot of sacrifices need to be made to keep the E-cores running, and it's not worth it 95% of the time. This isn't even counting the lack of software optimization/scheduling for E-cores, and how you lose AVX512 support with E-cores enabled.
The temps when enabled or disabled are important too, especially if you have a gaming laptop like a 13th gen i9 that runs hotter than a demon. If disabling the e-cores keeps the temps down significantly, then it would be worth doing.
Really love that you tested SC2 on 12900k. It's so hard to find people still testing for it. As to my guess as to why having E-cores makes this single thread bound game perform better. I wonder if it has something to do with how it access memory. I know upgrading only the memory gives a dramatic boost in performance for SC2.
@@donjuan3296 RAM speed can get you much better results specifically in SC2. You can also preload asset into RAM for SC2 to fix FPS drops associated with grabbing assets from the hard drive. The map is called Unit Preloader in the arcade and it will run though every thing in the game in about a min. Every run past the first time you play the map will preform much better. (You have to do this every time you load the game).
@@MogDog9 Yea, I use unit preloader. I'm not sure about the ram speed tho. I'm able to geat gear 1 4133 Mhz, but 4400 Mhz gear 2 at a bit higher latencies. Don't know what should I do :( . I guess test.
I just don't understand why 12th users even have to play this "game" to see which gives them better performance? I get the feeling AMD will do things differently, for their next gen.
@@fpshooterful well they are mad 5000s of amd is currently vs 10th gen intel and then intel reliese 11th gen and fail in multi task xD and amd win again and now after 3 generations they pass amd litte bit like 1% xD zen 4 will propeply be 5nm and have 3D stacking of L-cash so I think it will be like 30 to 60% ipc improment over 5000x of amd.
That isn't true. My 9700K 5,2ghz runs in CPU Intensive Games like Fortnite, Battlefield or CoD always on 100% and this is a big Bottleneck for my RX 6800. DirectX11 runs smoothly with 60-100%, but with direct12 it a runs always on 100% and stutters really hard with normal temps. Hyperthreading is a solution for massive CPU usages and the gain without hyperthreading is only affected by the old titles.
@@ChrisFreibier well on my 5600x I do not turn smt off becuze some games I play need hyperthreding but back in time when 4c 8t was overkill and 2c 4t 4c 4t was main platform I feel it was too much and I gain some one core performence in some games by turn it off
Performance Benefit of turning off E cores increases when manual tuning , my chip uses around .03v less which in turn is good for another 100mhz at the same voltage . 4.9/5 ring also possible. Will be interesting to see games in 12 months from now if they utilise the new architecture better , alder lake may age very well considering it is already on top with existing games
@@jebo4jc "new Bulldozer" lmao, with 40% higher IPC than 10th generation and like 25% higher singlecore perfomance than a 5950X definetly is far away from a Bulldozer scenario... Unlike Bulldozer, Alder Lake's IPC isn't crap and actually is a real world design.
One thing I admire of steve is how he knows beforehand what is going to happen most likely (with a very high chanse of beeing right) yet he still puts the hours of work to prove things practically... If it was me, my brain would not let me keep going after the first 2 or 3 cases confirmring what I expected. Respect.
But this is precisely WHY he almost always knows the answer before-hand: He's done the hard work of learning precisely how the PC architecture behaves to any and all changes :P
Otherwise known as "being a competent professional" ;) You should have a solid expectation of how things are going to go if you're competent enough in your field, and you should test the outcome nevertheless and be open to the possibility your expectations were not correct (and explore why they were correct or why they were not). It's really something that should go without saying, but it is indeed sadly kinda rare anyway.
Im still on the side of E cores are a waste on desktops. 4 E cores is the die space of another P core... thats where it would have been better used. Basically instead of 8+8 , 10P would have been better I cant speak for laptops as i dont use one, and those e cores probably a huge help for performance while keeping a battery charge.
1p core = 6.5 ecores worth of power draw. with 2 p cores added you would essentually be producing about 30 more watts than a 8+8 configuration... i have a 12700kf and i can EASILY pull 250 watts on this thing custom loop cooled. i dont think having 10 would have made a ton of since. how are you going to cool it when they all spool up?
@@p1st0ls44 If the P Cores are that bad on power then I guess Team Red is the way to go. I mean seriously to your question “how are you gonna cool that when they all spool up?” and the answer is Zen rofl
@@depth386 can you make 5800x, running at all clock speed 5ghz? if can , how much tdp for this clock speed? and if 5800x alll clock speed 5ghz, what performance? compare to 12700k turn of ecore, all clock speed 5ghz ?
Because you get nothing out of 10 p cores vs 8 p cores in gaming and in workloads youll get far less performance while also costing far more in electricity and being incapable of cooling it as i mean who doesnt want a cpu capable of pulling the same power of like a 3080 in a small space.
maybe disabling e-cores also affects how much l-cache you have available that would explain performance boost in CS GO as we saw a fps increase in ryzen 5000 that has a lot of cache.
@@qweasdzxc it is shared but what i'm saying is disabling "core" may also disable some other things like part of cache (or something else related to the core). That must be it otherwise it's impossible to explain the boost in fps. After all we don't really know the inner workings of all core-related stuff inside the cpu.
@@ivey77426 that's not how cache hierarchy works. Disabling a core only disables it's private cache(level 1, 2), the shared cache(level 3) is connected to the ring bus and accessable to all cores regardless.
@@humanbeing9079 But if the E cores need information, they would take a segment of the shared L3, remember that cache is basically a very, very fast small size RAM. If the e-cores are disabled, they can't use L3, therefore P-cores now don't have to share L3 with E-cores, we get a slight IPC increase. Finally, by disabling E-cores, there's more power or thermal headroom for the P-cores, therefore you can get a better overclock and enable AVX-512 that also boosts IPC, like AVX2 did for Haswell and beyond.
I did my own testing similar to what you guys did with an I7 12700K and Nvidia 3080. I only did 1440P, since that is my preferred resolution. My limited testing produced very similar results. I did not manage to get any significant gains or losses in FPS on the games I tested. I tried some CPU heavy games like Satisfactory and Civ 6 (measuring the turn times there). Doom Eternal and Titanfall 2 ran as smooth as ever since they are so well optimized. Cyberpunk 2077 was completely ignorant of the E cores, the results were It was pretty much all the same, all the variations I could see as an amateur were pretty much within the margin of error. I did not try CS:GO. I used a Watt meter, the power draw was indeed less on desktop usage with E cores enabled, but not by all that much. While the difference is noticeable, I thought it would be bigger than what I saw. I won't post numbers, because again, I am an amateur and the power draw test I did was extremely subjective with way too many variables for me to actually handle.
@@evalangley3985 Why would he be? The 12700k wipes the floor with everything. My tests have shown the same that it doesn't matter if e-cores are on or off.
Cheers for all the testing, not bad for a bloody aussie. lol I had trouble listening to your info when you were on screen with that jumper you have on, distracted the hell out of me.😊
I don’t want to be worrying about how to delegate cores, especially when paying so much for a CPU. The processor should be smart enough to get me the best performance for the job at hand.
I just created batch file to run my games with specific CPU affinity, and games are running only on C0, C2, C4, C6, C8, C10, C12, C14. So my games are only using first thread of all P cores. That way I disabled games to use hyperthreading threads and E cores ... works like a charm.
Great benchmarks. I want to throw in that from my point of view it would make more sense to do this comparison with a 12600K. Why? a) Due to pricing I believe the 12600K will have more market share, so that it is more relevant for the average user b) The 8 p cores of the 12900K are (almost) always enough to run the games, so that the e cores are not required to support the gaming performance. The 6 cores of the 12600K may make use of the e cores to support the gaming performance. Therefore it is more likely you will get interesting results with the e cores on the 12600K. I believe this benchmark only shows that some applications are too dumb to not stay away from the e cores while they are not needed. Not really worth the time you put into this.
@@----.__ yes. Still from a scientific point of view the 12600K is more interesting. Again this benchmark mainly shows that some applications are not (yet?) good at assigning tasks to the cores. I think it is more interesting to see what the e cores can do if the p cores are at 100% utilization and this you will see more often with less cores. On the other hand I believe HWU did a similar video on this already where they turned off some cores manually. Anyway I do not believe this videos was worth the hours Steve has spent to do all the benchmarking.
12900k video is more interesting, as you say 8 cores is enough for gaming so only using P-cores could be beneficial, in the case of 12600k it's guaranteed to be better to leave E-cores enabled. But for 12900k it's not as obvious if disabling E-cores would boost performance, thanks to this video the conclusion is it doesn't boost performance enough to make the difference, therefore 12600k surely will be even less benefits.
@@WCIIIReiniger I believe it was certainly worth the time. Just because it doesn't cater to your particular interests, it doesn't mean there aren't tens of thousands of people out there who are interested. Scientifically this test was an excellent examination at the effects of performance and economy cores on the flagship processor.
The point is that background tasks can also be served with more P-cores.(obviously you can't add as many P-cores as you can E-cores) *It's not about the number of cores, it's about the total amount of "work" the CPU can do.* The streaming argument is regurgitated by people who don't understand what's going on, nor how CPU works.
If you stream using Nvenc, which realistically you should as it's by far the best option for your typical 6 Mbit Twitch stream, your CPU isn't affected that much. I've seen about 7% additional CPU usage on an 8 core...
You've already made the first mistake of thinking that streaming is a "background task". It's not, and is much more computationally demanding than an actual background task, such as an idle browser tab, music player or an unused (but running) program sitting in memory. Even video playback (like this YT video) would be better served on a P core than on an E core - even if the E core can handle it.
@@Parky_T lol absolutely 1000 % bullshit, video decode is a exactly one of the task that can be done on the smallest processor capable of doing it, as there are next to no variations in performance requirement. You only need bigger faster cores if you have variation in process difficulty and you need more headroom to catch those...+ smaller cores usually are more efficient as they have less overhead powerconsumption AND if you look at loadcycles as electromagnetic waves in siode the circuit the frequency scaling of the energy density is quadratic....
are people out there really buying the 12900K and disabling what justifies the price over a few FPS? power draw aside why would you buy it and then turn around and make it an 8 core? why not just... buy an 8 core? save those dollars or take the time to do what Steve here said and use process lasso.
@@mrclamg Please enlighten us... To me, it makes some sense on a laptop (when the tech is mature enough and software catches up), not so much for a gaming desktop/workstation...
I would love to see this on the 12600K where the E cores would actually sometimes be needed to catch Windows background activity. Of course, you would want to use process lasso or something to make sure games entirely run on the P cores. Disabling the E cores in the bios seems an overly drastic measure to me.
G'day Steve, Teddy Jr. & Tim, Thanks for all the Hard Work you have done this year to keep us informed of what has been happening in PC Gaming Hardware, I wish you all & your families a Wonderful Christmas🎄 plus a Happy & Safe 2022
Context switching by the OS could be one of the reasons why you are seeing odd results here. If E-cores are disabled, this means that P-cores will be handling context switching. The P-cores can perform this task better than E-cores, therefore performance could improve. If you ran perfmon in the background and captured metrics that are typically used to identify bottlenecks, you should be able to see what's going on here. Also there an historic threshold in Windows that might be influencing things here, if a core hits around 75% usage, Windows sees this as a bottleneck and will start to try and handle the situation.
I think the real question is seeing all this, that do you need a CPU with E-cores at all for gaming? I'm thinking about the 12th gen non-K series coming in mid January, especially the 12600 with P cores only. 200 something dollars vs the 350 12600K?
there is a side effect here - disabling ecores gives a 5% boost *assuming nothing is running in the background* - a few chrome tabs open might easly swing this the other way.
@@HasimFN they still consume energy, generate heat and can contribute to thermal throttle of the main cores - and some workloads which are over-threaded do spill to ecores to their detriment.
For starcraft II I'd guess background tasks are running on the E cores, leaving more power and thermal headroom for the single core running the game than when the background tasks are running on the P cores? But afaik these were controlled, so shouldn't have had any impact.
4:12 STARCRAFT 2 MY FAVOURITE GAME THANKYOU STEVE it's been years since I've seen it on benchmarks. My only theory for increased starcraft performance is the way they implemented multhreaded support possibly accidentally utilising the e cores. Or reducing the usage of background tasks on the single p core being used to run StarCraft
For strictly gaming I'd say more than 8 cores is overkill*... having said that I have a 5900x, the extra cores being needed to run streaming/recording/encoding etc.... *Even with "futureproofing" ... most PC gamers use "futureproofing" as an excuse to buy more than they need. I'm guilty of it myself sometimes, but especially when it comes to my case. I figure I'll get multiple builds in a single case over the course of 6-ish years... but it seems I always buy a new case for a new build anyway. *shrug* Although my latest case upgrade should do me good for 6 years.
Thankfully, thats not how it works. The only situation where performance would be "sapped" from the P-cores, is if any one of these P-cores are being fully loaded at 100% usage, and therefore creating a bottleneck. As long as you never hit or spike 100% usage, any background tasks can actually be handled more efficiently and faster with lower core utilization on the P-cores, instead of the E-cores. Having E-cores just gives you an additional buffer layer for certain non-time-sensitive tasks that makes it less likely you'll cap out the P-core usage. All of this assumes Windows 10/11 has perfect thread scheduling/handling, but that's currently far from the truth. The thread handler still needs optimization because some time-sensitive apps are still being run on the E-cores instead of P-cores like they're supposed to.
I expect the cases where performance improved in lightly threaded games with E cores enabled is because the P cores are able to boost higher. Intel only allows the maximum clocks on the P cores when most of them are completely powered down. Offloading background work to the E cores likely makes this possible more often. That said, this effect is greatly diminished with manual overclocking. I personally run 56x2 55x4 54x8 with up to -2 ratio via TVB to control temps in heavy loads. E cores are disabled to allow for higher ring bus clock. I also hate the taskbar and start menu changes in Windows 11, which makes it unusable for me. I don't trust Windows 10 to schedule anything correctly with the E cores enabled.
The First Windows 10 Version was Garbage aswell. Microsoft is going to improve it over time as they always do. Also: I have not OCed 12th gen yet but from my experience with AMD 5k Series its usually not worth it since you gain 5% but you need 50% more power. I rather save they power and invest that money into my next CPU. That is especially true since most Hardware (AMD,Intel and Nvidia) is power hungry AF. You do actually save money that you could invest in a faster CPU by not overclocking these days.
E-cores will improve massively in the next generations... that's why i never buy the first generation of something this new (even though we've seen them on arm for years D:)
@@xflamousz no the 13700k will have double the ecores and they should be a little more powerful. Also once 14th gen comes out and they move to 7nm then it’s expected to be about 20 percent faster for the e cores. I’ll wait for the 14700k. By then ddr5 will have been out for 2-3 years and should be much better and cheaper too etc
This year is becoming stranger than fiction! AMD is now the expensive brand with stone cold predictable performance and guaranteed software compatibility while Intel is trying to swoop us over with good pricing and marginally better performance at the trade-off you'd normally associate with a new player in the field.... What the hell even is this?!?
I have this theory that Alder Lake’s e-cores will really start to shine once the system is loaded with background applications (MSI afterburner, RGB software, antivirus, discord, etc…), especially on 1% or 0.1% lows and general frame pacing since ADL’s architecture can handle those apps and a game asynchronously. It would also be interesting to see how average power consumption compares between Intel and AMD under this scenario as well, or if DDR5 makes a difference under those conditions
I wish we would get an 8 core K model without the e cores. For gaming and overclocking they are not necessary and Id much rather see a little more overclocking headroom. Buying CPUs with e cores and disable them is not an option for me because if we buy them, Intel will continue making them.
What about disabling e cores to get cache speed up? With e cores on a 12600k has it's cache limited to 4-4.4 range. But disabling e cores will let you get the cache all the way up to near p core clock.
The e-cores are so fing good that if you're in the market for a good streaming setup on a single PC, you can not only use the iGPU to deliver surprisingly good image through OBS, the 4 ecores on the 12600K can handle x264 medium surprisingly well and the impact on the main game performance is under 1-5% (compared to up to 30% on normal non big.LITTLE cpu arhitectures).
The thing about disabling the e-cores is that you leave a good temp-headroom for overclocking the p-cores alot thus pushing the cpu way more than you can do with e-cores enabled,,,, And then you will see some pretty good uplift in performance. You only disable e-cores if you really want to push the clockspeeds way up
I disabled e cores to improve the temperatures, but is it actually worth it? For example, if i have programs like discord opened in background or even hosting a minecraft server and playing on it. One thing i noticed is that the system performs better or feels smoother somehow.
@@TheBlueVaron Only if you need to push clockspeeds, if you want to have multiple programs running at all times, then NO. But if you want 5,6-5,7 Ghz corespeed for some demanding game to reach best possible fps then yes... Also remember to clock your ram.👌 Kinda goes together😜
@@ghost500e I actually did the opposite, undervolt the cpu as it was getting reasonably hot for what i actually do, and i saw a tutorial who recommended so much turning e cores off for better temps as it doesnt affect in gaming. But, in case i have discord, many browser tabs opened and hosting a minecraft server with 96 mods, then i should use them, right?
This is the test I've been waiting for, thanks guys. Edit: Some strange results but very thorough testing. I wish they gave us a 10P core cpu. I'm actually gonna wait for better ddr5 and 13th gen or equivalent AMD offerings. Currently running a 10850k.
13th gen will only up the e-core count and slightly increase p-core performance. So if you want non-hybrid design go with zen4 I guess. If it's not gonna be highly overpriced of course.
@@thegameoftimeskiller2984 idk about that, haven't really heard about what's coming after that but it'd make sense. If they also upgrade the small cores to have IPC like zen2 or zen3 then it's good!
exactly i have i7 7700 non k ,32gb ram, rtx 3070 FE. CM v650 80 plus gold psu. shud i go with i9 10850k/msi z490 gaming edge or i5 12600k/asus tuf z690 ??? please suggest me.thanks🙂
@@damara2268 i have i7 7700 non k ,32gb ram, rtx 3070 FE. CM v650 80 plus gold psu. shud i go with i9 10850k/msi z490 gaming edge or i5 12600k/asus tuf z690 ??? please suggest me.thanks🙂
I picked up the 12700K's for $300, 12600K's for $230 and of course i had to get the 12900K though, it was $560. These went back up $30-$50. Is $260 worth 4 E-cores. I can't say that it is. Even at MSRP. $200 for 4 E-cores is still too much. I got a couple get by boards. I'd like to see the best DDR4 Z690 boards. If those even exist.
I think Steve is right on the money with the ring bus speed and the associated L3 latency increase. And may I remind you that there are still programs out there that perform better when you disable hyperthreading... HT has been around for 20 years IIRC...
The switching around of the "uncore" (cache) frequency definitely messes with iRacing, possibly other old games. Not a performance hit, but stutter. No ecores - no stutter
It's two Years later at the time of this comment and newer tests have shown that e-cores do help overall, even on 12th gen. I think a lot of it is due to newer games taking them into account and drivers maturing for GPUs which also make better use of them. There are still some games that perform better without them but in general they do help even more now than they did back when this test was done- just check out newer charts and decide if they're worth it on the games you play. One thing I'd like to see- with e-cores disabled, can you overclock the P cores more? and if so, can that overclock outperform having the e-cores for games that do better with them? I have a 12600kf in my secondary system and I might just do a personal test to find out if I find the time.... time is harder and harder to find as I get older though =\
People were saying this in another video I was watching. I plan on disabling the e cores to see if I can get a smoother experience. I’m not really interested in overclocking
Nvidia vs AMD input lag comparison please. Nvidia focuses a lot on calculations and software, whereas AMD RX6000 series were presented as gaming option
The reason some games perform better with the e cores disabled is because the ring clock is higher without the e cores loaded. It goes down to 3.6 GHz with the e cores enabled or under load. Leave the e cores enabled and overclock the ring clock a bit and you get the best of both worlds. 4.2 GHz should be achievable for most chips.
I'd like a review testing how streaming performs on 12th gen using the cpu to encode. Will the scheduler be smart enough to put all the rendering load on the ecores? That would be a game changer for a lot of people.
No, as h264 is way too intensive to be handled just by e cores without drop outs. It also isnt a game changer since a 5950x still suffers, despite having basically 2 independent 5800x's at its disposal, as the encoding Software also utilizes the cache and memory subsystem, and therefore limiting the games acces to that.
I don't think the rendering is done on the E-Cores. The i9-12900K however is really incredible overall for streaming. Check out EposVox's video covering it.
I would expect enabling e-cores to make games run faster (especially old ones, like csgo, which is cpu bound after a certain point) since it allows the OS to offload scheduling, background tasks and other stuff to the e-cores, leaving the p's available for gaming.
Love the interesting look at the new Intel CPU's. Would have liked to see the power usage and temperature differences in the e-core on/off configs, since on your previous videos you mention that you had to use a premium cooler (MSI MEG CoreLiquid S360) to avoid intermittent thermal throttling. Is it possible that e-cores off is using less power (watts) and generating less heat load so that people who haven't sprung for $280 to $300+ AIO's are seeing lower temps when disabling e-cores? I have seen other testing where it seems they are going with 420mm AIO's (and some are recommending custom water cooling) due to thermal load issues. I love the Intel/AMD competition, but, people going with a 12900k need to realize that to get every last bit of performance out of the 12900k, they must be willing to go with an expensive premium case, cooler, memory, etc .... Again, very interesting testing... love the videos!
@@Deviantsoundz A $89.99 Lancool 215 case won't fit the larger coolers required to cool an OC'd 12900k in a 26c room (78f)... Even my Lancool II Mesh Performance case $110 has issues with some coolers over 280mm.and often premium 360 aio's don't fit in cheaper cases, much less a 420mm AIO.'s.. that's how I come up with a premium air flow case being required... especially if using $150 to $300+ AIO's, Are you really going to try to put it in a $60 to $80 non airflow case and expect good performance??? LOL.
@@SmarterGaming plently of sub 100 cases thats fit 360s so idk why you having issues. Even a tiny nr200p can fit an artic 280 aio that performs close to some 360s even my old cheap fara r1 fits a 360.
I’m wondering if the process scheduler is handing non-gaming process threads to the e-cores in the cases of high single thread performance requirements, and it is giving the assigned P-core a boost as it has the core solely for the game process
I feel like this is Hyper threading all over again, purist said "disable HT, for much better fps" now we can hear the same with E cores, while its not really worth doing so. It will take time when games will be made with hybrid architecture in mind, current games still runs lets say in "legacy mode" so its going to be interesting, and no E cores are NOT only for background tasks, intel actually released instructions for software/game developers how to utilize E cores as well in their workloads so its only a worst case scenario for now. Just leave those E cores (unless you have legit issues launching your software/games) and enjoy awesome performance of alderlake :)
12700k owner, I just leave the E cores on. I play at 1440p so not a huge difference. I do not want the hassle of going into the bios and turning them on or off for different tasks.
What about temperatures and power consumption? Can you overclock the P cores more when you disable the E cores? What about other video editing, modeling, etc.? I think there's much more to explore than just gaming when it comes to E-cores.
another great benefits with disabling ecore is if you work on multiple instance of android emulator. at least in my case got massive boost with performance on 20+ instance while its laging a lot and unresponsive with ecore enabled
Does streaming software utilise e-cores or p-cores more? Because if leaving the e-cores enable doesn’t affect most games, then I guess it might be a benefit to streamers if their software is well utilised by e-cores
Love the test. Only really care about performance in VRCHAT tough. If a get one of these I will test in that game and adjust disable/enable accordingly.
Its funny to see that disabling E cores increases performance. Whole point of having E cores was to increase gaming performance! and we see exact opposite! XD!
The whole point of having E-cores was actually for efficiency purposes and nothing else, because Intel famously has efficiency problems compared to AMD's current-gen Ryzen CPUs. E-cores were Intel's bet on catching up to AMD in efficiency, especially under workloads that don't require the power hungry, high heat P-cores. This is especially important for portable battery-powered devices like laptops.
@@GENKI_INU E cores doesn't matter when you are using it in a PC where power constraints are none. Let's hope game companies adopt E-core design for their games in order to improve fps.
@@cyclonous6240 This is why I think E-cores were a mistake for Intel's desktop SKUs. They could've used the die space of the E-cores to make more P-cores, more cache, or even a better IGPU. Remember that even if games start to utilize E-cores, they will only run as fast as the E-cores are, since that's the main bottleneck, not the P-cores.
@@GENKI_INU Exactly right! They played this card in Desktop market to understand how people are going to react. If the people accept this kind of thinking for PCs then they might start earning tremendous profit again as it is always profitable to produce 4 E cores rather than 4 P cores. Desktop market does not need big LITTLE core design.
@@cyclonous6240 I like to imagine how Intel came to the conclusion of E-cores: "Let's put in 2 more cores for the last two slots of the 12900K die to make 10 cores. It will be a great upgrade from the 8 core 11900K." "Nah, that's way too much power draw and heat, we'll look bad against AMD." "Fine then, let's at least put in extra Cache to fill that space." "Nah, that's way too expensive." "Okay, how about extending the IGPU die in there instead?" "Nah, we're releasing our discreet GPUs later this year anyway. We want everyone to buy those." "What if we cram in 8 crappy 6-year-old Skylake cores without Hyperthreading? We can even call the 12900K a 16 core processor!" "You're promoted to Chief Design Engineer."
Maybe leaving E-cores enabled and performing better in CS:GO and Starcraft it has something to do with the AI in the games itself? If both tests were performed offline with bots it could be that the E-cores are still being used to help with the AI?
Real answer is disabling E core lower power consumption and bring down heat considerably and possibly allowing to boost tcors to hire clock. The real question is how do you do it safely where does all that extra power go where does the chip simply not require as much power when the E cars are disabled.
There are some games like far cry 2 3 and blood dragon that straight up wont run if you have more than 18 cores. Disabling some or all the e cores will fix this
Disable hell no, but should compiliers and schedulars get a heall of a lot more intelligent in the era of multi cores to be able to sort threads more intentionally. That being said the e core should be used in gaming for things that are not crucial like backgound IO preloading, playing soundfiles, background saving etc...
Thank you for this analysis. When will we see an high end roundup of z690 motherboard? I'll suggest using a flir thermal camera more than the sensor used in the last video :)
I would love to see how these results are affected in something like a low P core count processor, where the game might utilize more than the P cores available, 12600K might be a good candidate, since some games might scale beyond 6 P cores.
There are situations where the OS has support threats and background processes that are not game related that would compete for P core resources if the E cores were not available. My suspicion is that the games that perform better with E cores enabled are also well threaded and are utilizing many if not all cores up to 6 or 7 or 8 but have to compete with background processes
Disabling e cores increases ring LLC, which improves latency. W11 has problem with e cores off in that it stops understanding how to balance between real core and HT core. So then have to turn off HT. W10 isn't e core aware so it won't have this problem. HT off in addition improves latency, with less thread switch and more L3 allocated to P cores. E cores have become a productivity addon for CPUs with no benefit to gaming. What background tasks...people tend to not have anything running in background anyway. Latency.
Since this is probably the last video before Christmas - have a merry Christmas and happy new year! I hope you guys can take some time off from testing and benchmarking for a bit and enjoy a break! Thanks for a great 2021!
There’s a possibility that you’re seeing these results based on the power/heat envelope of the cpu package shifting the clocks based on low. So on a lightly threaded game that’s not pushing the package temperate up it may be able to maintain higher clocks on the P-Cores. But if a game is hitting all cores, it’s resulting in the cpu throttling down.
What would have been interesting would have been to have the P cores, E cores, and Ring/Cache locked at specific numbers, including when E cores are disabled. This would isolate the performance issues down to the scheduler and give a better idea of why performance is changing in games with them on or off and lead to better guidance on setup for folks.
@Manuel G this is 12900K which is known as running hot and what he says makes a perfect sense!
@@zwer54 *GAMES dont stress the cpu anywhere as hard as making it run near 200w
i think most shouldnt disable e-cores, since these results very much resemble a best-case lab scenario where background activity is at a minimum. with most people running chrome tabs, light software and maybe even music in the background, the e-cores should help keep the game performance running smoothly without sacrificing p-core performance
I agree, I think the E-Cores are best suited for exactly that stuff, background programms, while the P-Cores can concentrate on the game. I'd assume that newer games and some Windows 11 updates will be much better at directing each core for its designed job.
What about streaming?
Assuming they actually are scheduled properly maybe., But those e-cores do not even seem good for background tasks. They seem good as assisting for heavily threaded productivity workloads that can scale to infinite cores. Of course equal amount of P cores would be much much better, but adding double e-cores can trade blows with 1/2 amount of P cores added as double amount of much weaker cores in fully parallelized capable workload. But for workloads using less threads, less P cores with no e-cores way way better. Plus no
@@Wolverine607 windows is getting better and better at utilizing ecores properly
I've got a strong theory that context switching is hitting the single thread games when disabling the E cores as less total threads are available for the OS itself. A process lasso or affinity would prove this.
I also strongly suspect this. If E-cores are enabled, stuff like context switching of various background tasks is likely offloaded to them and thus a single-thread game can fully utilize one thread. It's probably due to in-CPU controller of P-cores, as well as core-to-core latency.
Is that something that could be fixed with a scheduler update?
I really want someone to test this... E-cores enabled BUT make the game only use P-cores with task manager or Process Lasso. It should be the best scenario in theory.
@@UZieLLable Ideally also force all other processes to use the e cores perhaps?
I'm wondering if we lasso pretty much everything o/s related to e cores, and have p cores solely for specific programs that you want them to be used for. e.g. games, encoding, etc.
Process Lasso also gave some extra performance for my aging AMD Ryzen 2700x. In older games that do not need more than 4/8 threads, restricting them to a single CCX gave extra 3-5% framerate
i would love to see this kind of testing on the 12600k
just get a 12600.....
As far as i can tell, from the time owning an 12900k, the E-Cores are causing more "trouble" than anything else. At least when it comes to gaming.
There are only a handful of games that are actually benefitting from them but there are far more games where the E-Cores are straight up causing issues since the scheduling can be all over the joint.
In Far Cry 5 for example, despite E-Cores being disabled and E-Cores being enabled are giving fairly similar framerates, there's an awful lot of microstutter going on with the E-Cores enabled because the P-Cores aren't hyperthreading (like they are supposed to do) but instead throwing each of their second thread on the E-Cores. Disabling the E-Cores in this scenario isn't even just about the little performance boost i get there but way more about the overall smoothness since it completely eliminates the absolutely awful and constant microstutter.
Other Games i've tested saw an significant performance boost, especially in the min. FPS department and some games were straight up broken with the E-Cores enabled causing all sorts of weirdness.
My conclusion when it comes to Alder Lake is that the Hybrid Approach seems to be a decent idea at first but fails miserably in the real world, at least when it comes to gaming. Looking at the 12900k's die, i'd rather have the 8 E-Cores replaced with 2 P-Cores. Not a fan of the Hybrid Architecture to be perfectly honest, not at all.
„Far more with issues” could you name more than just one?
There isn't the die space or thermal headroom to replace the E cores with 2 P cores.
That shows e core is a scam core.. i hope ryzen dont release this type of core at all
@@johnm91326 well actually there is and second die would be adjusted accordingly to that config
Anyway if u mean for that exact size 4ecores is bigger than 1pcores clear as day in die shot
@@farhannik7039 not really, cpus aren’t just about gaming surprisingly also maybe games will try make better support for e cores if ryzen do that as well
How much of this difference can we attribute to uncore/ring ratio? With E-Cores On, ring stays at a rather low 3,6 Ghz. With E-Cores Off, ring clocks at around 4,6-4,7 Ghz.
All of it.
Literally was his conclusion from the video.
I assume some of it may come from wrong sceduling and processes having to wait for the ecores to be finished.
@@intetx no it does not.
L3 latency goes down by over 30% if you disable the e cores and oc a bit, thats absolutely massive for cache/memory sensitive applications
Just OC your ring bus to 4.2-4.4 with E-cores on.
The most important thing to have been addressed was to tellus how the L3 Cache behaves when E cores are disabled. Do they take with them their respective Cache and block it from the P cores ? Or do the E cores handle their cache over to their big brothers ? If the first case is true , is there a way to reclaim back the cache for the P cores ? Also , you haven't mentioned the benefit in thermals, power consumption and the potential OC headroom the disabling of E cores brings to the table ... All these make it a worthwhile exploration for just gaming i believe ...
exactly my question..
No, you still get the full L3 cache when disabling cores in the BIOS. This applies to either P or E cores being disabled.
If you disable the E-cores, you get ALL the L3 cache available for the P-Cores to use, instead of having to split/share it with the E-cores. This is a very good thing.
The only reason to enable E-cores is if you REALLY need that extra bit of multithreading performance in fully-threaded workloads.
But the multithreaded performance difference isn't even much between a higher OC headroom with P-cores only (lower core voltage needed for higher clocks, AND higher Uncore OC), vs leaving the E-cores enabled and being limited to lower P-core and Uncore clocks.
A lot of sacrifices need to be made to keep the E-cores running, and it's not worth it 95% of the time.
This isn't even counting the lack of software optimization/scheduling for E-cores, and how you lose AVX512 support with E-cores enabled.
Sorry I thought it was obvious that you can't disable portions of L3.
@@gamingunboxed5130 its not
The temps when enabled or disabled are important too, especially if you have a gaming laptop like a 13th gen i9 that runs hotter than a demon. If disabling the e-cores keeps the temps down significantly, then it would be worth doing.
Really love that you tested SC2 on 12900k. It's so hard to find people still testing for it. As to my guess as to why having E-cores makes this single thread bound game perform better. I wonder if it has something to do with how it access memory. I know upgrading only the memory gives a dramatic boost in performance for SC2.
Sc2 gang :D
what do you mean by upgrading memory? Like RAM speed, or capacity? I'm trying to increase my 0.1 and 1% lows.
@@donjuan3296 RAM speed can get you much better results specifically in SC2. You can also preload asset into RAM for SC2 to fix FPS drops associated with grabbing assets from the hard drive. The map is called Unit Preloader in the arcade and it will run though every thing in the game in about a min. Every run past the first time you play the map will preform much better. (You have to do this every time you load the game).
@@MogDog9 Yea, I use unit preloader.
I'm not sure about the ram speed tho. I'm able to geat gear 1 4133 Mhz, but 4400 Mhz gear 2 at a bit higher latencies. Don't know what should I do :( . I guess test.
Disabling e-cores is like disabling hyperthreading years ago...yes you could gain a bit of FPS, but most people probably shouldn't do it
Well I disabled my multi thread on my i7-4770k and when I feel it start to bottleneck then I turn it on before it die🪦
I just don't understand why 12th users even have to play this "game" to see which gives them better performance? I get the feeling AMD will do things differently, for their next gen.
@@fpshooterful well they are mad 5000s of amd is currently vs 10th gen intel and then intel reliese 11th gen and fail in multi task xD and amd win again and now after 3 generations they pass amd litte bit like 1% xD zen 4 will propeply be 5nm and have 3D stacking of L-cash so I think it will be like 30 to 60% ipc improment over 5000x of amd.
That isn't true. My 9700K 5,2ghz runs in CPU Intensive Games like Fortnite, Battlefield or CoD always on 100% and this is a big Bottleneck for my RX 6800. DirectX11 runs smoothly with 60-100%, but with direct12 it a runs always on 100% and stutters really hard with normal temps. Hyperthreading is a solution for massive CPU usages and the gain without hyperthreading is only affected by the old titles.
@@ChrisFreibier well on my 5600x I do not turn smt off becuze some games I play need hyperthreding but back in time when 4c 8t was overkill and 2c 4t 4c 4t was main platform I feel it was too much and I gain some one core performence in some games by turn it off
Performance Benefit of turning off E cores increases when manual tuning , my chip uses around .03v less which in turn is good for another 100mhz at the same voltage . 4.9/5 ring also possible. Will be interesting to see games in 12 months from now if they utilise the new architecture better , alder lake may age very well considering it is already on top with existing games
Is this really where we're at? Getting excited over .03v less and an extra 100MHz?
Alder Lake is the new Sandy Bridge or Haswell, of course it will age pretty well.
@@saricubra2867 Alder Lake seems like the new Bulldozer to me
@@jebo4jc "new Bulldozer"
lmao, with 40% higher IPC than 10th generation and like 25% higher singlecore perfomance than a 5950X definetly is far away from a Bulldozer scenario...
Unlike Bulldozer, Alder Lake's IPC isn't crap and actually is a real world design.
@@NovaDoll "barely faster"
That difference on Battlefield 2042 is not barely faster, it's just massive.
One thing I admire of steve is how he knows beforehand what is going to happen most likely (with a very high chanse of beeing right) yet he still puts the hours of work to prove things practically... If it was me, my brain would not let me keep going after the first 2 or 3 cases confirmring what I expected. Respect.
But this is precisely WHY he almost always knows the answer before-hand: He's done the hard work of learning precisely how the PC architecture behaves to any and all changes :P
Otherwise known as "being a competent professional" ;) You should have a solid expectation of how things are going to go if you're competent enough in your field, and you should test the outcome nevertheless and be open to the possibility your expectations were not correct (and explore why they were correct or why they were not). It's really something that should go without saying, but it is indeed sadly kinda rare anyway.
It's almost like video editing doesn't exist. In hindsight meaning he already knows because the intro can be filmed afterwards
You get more done and dive deeper than channels with 10x the staff. Great job! Great dedication. Keep up the amazing work.
@@evalangley3985 ive been out of the loop. what happened with the halo benchmark?
I mean, he immediately revisited halo infinite right away. And probably will again in future videos, especially if a re-bar on/off video is done.
@@Dark_Luminary nothing, he's being a tool.
Im still on the side of E cores are a waste on desktops.
4 E cores is the die space of another P core... thats where it would have been better used.
Basically instead of 8+8 , 10P would have been better
I cant speak for laptops as i dont use one, and those e cores probably a huge help for performance while keeping a battery charge.
1p core = 6.5 ecores worth of power draw. with 2 p cores added you would essentually be producing about 30 more watts than a 8+8 configuration... i have a 12700kf and i can EASILY pull 250 watts on this thing custom loop cooled. i dont think having 10 would have made a ton of since. how are you going to cool it when they all spool up?
@@p1st0ls44 If the P Cores are that bad on power then I guess Team Red is the way to go. I mean seriously to your question “how are you gonna cool that when they all spool up?” and the answer is Zen rofl
@@depth386 can you make 5800x, running at all clock speed 5ghz? if can , how much tdp for this clock speed?
and if 5800x alll clock speed 5ghz, what performance? compare to 12700k turn of ecore, all clock speed 5ghz ?
@@bogodoyandex9654 Relax I’m not shilling, I have a 9900K. Hot boi tho. For anything but tippyty top a red chip at 4.5ish is fine.
Because you get nothing out of 10 p cores vs 8 p cores in gaming and in workloads youll get far less performance while also costing far more in electricity and being incapable of cooling it as i mean who doesnt want a cpu capable of pulling the same power of like a 3080 in a small space.
maybe disabling e-cores also affects how much l-cache you have available that would explain performance boost in CS GO as we saw a fps increase in ryzen 5000 that has a lot of cache.
isn't the cache shared between all cores?
@@qweasdzxc it is shared but what i'm saying is disabling "core" may also disable some other things like part of cache (or something else related to the core). That must be it otherwise it's impossible to explain the boost in fps. After all we don't really know the inner workings of all core-related stuff inside the cpu.
@@ivey77426 that's not how cache hierarchy works.
Disabling a core only disables it's private cache(level 1, 2), the shared cache(level 3) is connected to the ring bus and accessable to all cores regardless.
You're wrong buddy CS:GO isn't made for hyperthreading. Watch 3KliksPhilip.
@@humanbeing9079 But if the E cores need information, they would take a segment of the shared L3, remember that cache is basically a very, very fast small size RAM.
If the e-cores are disabled, they can't use L3, therefore P-cores now don't have to share L3 with E-cores, we get a slight IPC increase.
Finally, by disabling E-cores, there's more power or thermal headroom for the P-cores, therefore you can get a better overclock and enable AVX-512 that also boosts IPC, like AVX2 did for Haswell and beyond.
I did my own testing similar to what you guys did with an I7 12700K and Nvidia 3080. I only did 1440P, since that is my preferred resolution. My limited testing produced very similar results. I did not manage to get any significant gains or losses in FPS on the games I tested.
I tried some CPU heavy games like Satisfactory and Civ 6 (measuring the turn times there). Doom Eternal and Titanfall 2 ran as smooth as ever since they are so well optimized. Cyberpunk 2077 was completely ignorant of the E cores, the results were It was pretty much all the same, all the variations I could see as an amateur were pretty much within the margin of error. I did not try CS:GO.
I used a Watt meter, the power draw was indeed less on desktop usage with E cores enabled, but not by all that much. While the difference is noticeable, I thought it would be bigger than what I saw. I won't post numbers, because again, I am an amateur and the power draw test I did was extremely subjective with way too many variables for me to actually handle.
@@evalangley3985 Why would he be? The 12700k wipes the floor with everything. My tests have shown the same that it doesn't matter if e-cores are on or off.
@@evalangley3985 nah, I just like testing stuff.
Are you running ddr4 or 5?
@@sgtsavage4991 I'm running DDR 5
Cheers for all the testing, not bad for a bloody aussie. lol
I had trouble listening to your info when you were on screen with that jumper you have on, distracted the hell out of me.😊
I would like to see the temps difference with Ecore in and off.
I don’t want to be worrying about how to delegate cores, especially when paying so much for a CPU. The processor should be smart enough to get me the best performance for the job at hand.
Yep. AI should handle all these things & not the user or reviewer spending time on these things
you have way too much faith in intel.
I just created batch file to run my games with specific CPU affinity, and games are running only on C0, C2, C4, C6, C8, C10, C12, C14.
So my games are only using first thread of all P cores. That way I disabled games to use hyperthreading threads and E cores ... works like a charm.
Great benchmarks.
I want to throw in that from my point of view it would make more sense to do this comparison with a 12600K.
Why?
a) Due to pricing I believe the 12600K will have more market share, so that it is more relevant for the average user
b) The 8 p cores of the 12900K are (almost) always enough to run the games, so that the e cores are not required to support the gaming performance. The 6 cores of the 12600K may make use of the e cores to support the gaming performance. Therefore it is more likely you will get interesting results with the e cores on the 12600K.
I believe this benchmark only shows that some applications are too dumb to not stay away from the e cores while they are not needed. Not really worth the time you put into this.
There's more 12900s out there than you think.
@@----.__ yes. Still from a scientific point of view the 12600K is more interesting. Again this benchmark mainly shows that some applications are not (yet?) good at assigning tasks to the cores. I think it is more interesting to see what the e cores can do if the p cores are at 100% utilization and this you will see more often with less cores. On the other hand I believe HWU did a similar video on this already where they turned off some cores manually.
Anyway I do not believe this videos was worth the hours Steve has spent to do all the benchmarking.
12900k video is more interesting, as you say 8 cores is enough for gaming so only using P-cores could be beneficial, in the case of 12600k it's guaranteed to be better to leave E-cores enabled. But for 12900k it's not as obvious if disabling E-cores would boost performance, thanks to this video the conclusion is it doesn't boost performance enough to make the difference, therefore 12600k surely will be even less benefits.
@@WCIIIReiniger I believe it was certainly worth the time. Just because it doesn't cater to your particular interests, it doesn't mean there aren't tens of thousands of people out there who are interested. Scientifically this test was an excellent examination at the effects of performance and economy cores on the flagship processor.
ur camera footage on video looks amazing great work guys!
Thanks for including PUBG in your benchmarks, not many reviewers do these days despite its numbers on steam.
Thank you for including Starcraft II in the new Alder Lake coverage, especially in a taxing 4v4 game
I think it'd be interesting to see how the performance differs between having the e-cores on or off if there is a background process such as streaming
The point is that background tasks can also be served with more P-cores.(obviously you can't add as many P-cores as you can E-cores)
*It's not about the number of cores, it's about the total amount of "work" the CPU can do.* The streaming argument is regurgitated by people who don't understand what's going on, nor how CPU works.
If you stream using Nvenc, which realistically you should as it's by far the best option for your typical 6 Mbit Twitch stream, your CPU isn't affected that much. I've seen about 7% additional CPU usage on an 8 core...
@@humanbeing9079 windows 11 scheduler makes sure that those tasks do get done in the background by E cores
You've already made the first mistake of thinking that streaming is a "background task". It's not, and is much more computationally demanding than an actual background task, such as an idle browser tab, music player or an unused (but running) program sitting in memory. Even video playback (like this YT video) would be better served on a P core than on an E core - even if the E core can handle it.
@@Parky_T lol absolutely 1000 % bullshit, video decode is a exactly one of the task that can be done on the smallest processor capable of doing it, as there are next to no variations in performance requirement. You only need bigger faster cores if you have variation in process difficulty and you need more headroom to catch those...+ smaller cores usually are more efficient as they have less overhead powerconsumption AND if you look at loadcycles as electromagnetic waves in siode the circuit the frequency scaling of the energy density is quadratic....
are people out there really buying the 12900K and disabling what justifies the price over a few FPS? power draw aside why would you buy it and then turn around and make it an 8 core? why not just... buy an 8 core? save those dollars or take the time to do what Steve here said and use process lasso.
You are wrong because you dont understand the benifits of the 12900k. Have another look at the specs.
@@mrclamg 5950x best cpu 16core P
@@mrclamg Please enlighten us... To me, it makes some sense on a laptop (when the tech is mature enough and software catches up), not so much for a gaming desktop/workstation...
I would love to see this on the 12600K where the E cores would actually sometimes be needed to catch Windows background activity. Of course, you would want to use process lasso or something to make sure games entirely run on the P cores. Disabling the E cores in the bios seems an overly drastic measure to me.
G'day Steve, Teddy Jr. & Tim,
Thanks for all the Hard Work you have done this year to keep us informed of what has been happening in PC Gaming Hardware,
I wish you all & your families a Wonderful Christmas🎄 plus a Happy & Safe 2022
Context switching by the OS could be one of the reasons why you are seeing odd results here. If E-cores are disabled, this means that P-cores will be handling context switching. The P-cores can perform this task better than E-cores, therefore performance could improve. If you ran perfmon in the background and captured metrics that are typically used to identify bottlenecks, you should be able to see what's going on here. Also there an historic threshold in Windows that might be influencing things here, if a core hits around 75% usage, Windows sees this as a bottleneck and will start to try and handle the situation.
I think the real question is seeing all this, that do you need a CPU with E-cores at all for gaming? I'm thinking about the 12th gen non-K series coming in mid January, especially the 12600 with P cores only. 200 something dollars vs the 350 12600K?
I think the price difference will be a lot less. around $60 to 70
Straight down the line tech content - no messing. Great stuff.
there is a side effect here - disabling ecores gives a 5% boost *assuming nothing is running in the background* - a few chrome tabs open might easly swing this the other way.
E cores shouldt touch any of your games
@@HasimFN they still consume energy, generate heat and can contribute to thermal throttle of the main cores - and some workloads which are over-threaded do spill to ecores to their detriment.
@@andytroo just turn them off trust me better
For starcraft II I'd guess background tasks are running on the E cores, leaving more power and thermal headroom for the single core running the game than when the background tasks are running on the P cores?
But afaik these were controlled, so shouldn't have had any impact.
Fantastic job as allways. Thanks a lot for this great video 👍
Love you guys!
And a Merry Christmas to you all!!
4:12 STARCRAFT 2 MY FAVOURITE GAME THANKYOU STEVE it's been years since I've seen it on benchmarks. My only theory for increased starcraft performance is the way they implemented multhreaded support possibly accidentally utilising the e cores. Or reducing the usage of background tasks on the single p core being used to run StarCraft
You can use process lasso to make the game choose P-Cores. Also, skip the virtual threads. Pick only physical ones. 0,2,4,6,8, etc
For strictly gaming I'd say more than 8 cores is overkill*... having said that I have a 5900x, the extra cores being needed to run streaming/recording/encoding etc....
*Even with "futureproofing" ... most PC gamers use "futureproofing" as an excuse to buy more than they need. I'm guilty of it myself sometimes, but especially when it comes to my case. I figure I'll get multiple builds in a single case over the course of 6-ish years... but it seems I always buy a new case for a new build anyway. *shrug* Although my latest case upgrade should do me good for 6 years.
E cores might be performing background tasks - when disabled they might be occupying the same P cores as the games, sapping some performance.
Thankfully, thats not how it works. The only situation where performance would be "sapped" from the P-cores, is if any one of these P-cores are being fully loaded at 100% usage, and therefore creating a bottleneck.
As long as you never hit or spike 100% usage, any background tasks can actually be handled more efficiently and faster with lower core utilization on the P-cores, instead of the E-cores. Having E-cores just gives you an additional buffer layer for certain non-time-sensitive tasks that makes it less likely you'll cap out the P-core usage.
All of this assumes Windows 10/11 has perfect thread scheduling/handling, but that's currently far from the truth. The thread handler still needs optimization because some time-sensitive apps are still being run on the E-cores instead of P-cores like they're supposed to.
I expect the cases where performance improved in lightly threaded games with E cores enabled is because the P cores are able to boost higher. Intel only allows the maximum clocks on the P cores when most of them are completely powered down. Offloading background work to the E cores likely makes this possible more often.
That said, this effect is greatly diminished with manual overclocking. I personally run 56x2 55x4 54x8 with up to -2 ratio via TVB to control temps in heavy loads. E cores are disabled to allow for higher ring bus clock.
I also hate the taskbar and start menu changes in Windows 11, which makes it unusable for me. I don't trust Windows 10 to schedule anything correctly with the E cores enabled.
The First Windows 10 Version was Garbage aswell. Microsoft is going to improve it over time as they always do. Also: I have not OCed 12th gen yet but from my experience with AMD 5k Series its usually not worth it since you gain 5% but you need 50% more power. I rather save they power and invest that money into my next CPU. That is especially true since most Hardware (AMD,Intel and Nvidia) is power hungry AF. You do actually save money that you could invest in a faster CPU by not overclocking these days.
E-cores will improve massively in the next generations... that's why i never buy the first generation of something this new (even though we've seen them on arm for years D:)
yes but they'll always be slower than the P cores, because well, that's the point of using them
13th gen has exactly the same e cores...
@@xflamousz no the 13700k will have double the ecores and they should be a little more powerful. Also once 14th gen comes out and they move to 7nm then it’s expected to be about 20 percent faster for the e cores. I’ll wait for the 14700k. By then ddr5 will have been out for 2-3 years and should be much better and cheaper too etc
@@darkhaven9119 it will have more, but the architecture remains gracemont.
@@xflamousz yes but it’ll also be more optimized and they can still add plenty of improvements to it
Nice sweater Steve!
This year is becoming stranger than fiction! AMD is now the expensive brand with stone cold predictable performance and guaranteed software compatibility while Intel is trying to swoop us over with good pricing and marginally better performance at the trade-off you'd normally associate with a new player in the field.... What the hell even is this?!?
I have this theory that Alder Lake’s e-cores will really start to shine once the system is loaded with background applications (MSI afterburner, RGB software, antivirus, discord, etc…), especially on 1% or 0.1% lows and general frame pacing since ADL’s architecture can handle those apps and a game asynchronously. It would also be interesting to see how average power consumption compares between Intel and AMD under this scenario as well, or if DDR5 makes a difference under those conditions
Hardware Unboxed intro music always brings an instant sense of peace and calm... Merry Christmas all!
I wish we would get an 8 core K model without the e cores. For gaming and overclocking they are not necessary and Id much rather see a little more overclocking headroom.
Buying CPUs with e cores and disable them is not an option for me because if we buy them, Intel will continue making them.
What about disabling e cores to get cache speed up? With e cores on a 12600k has it's cache limited to 4-4.4 range. But disabling e cores will let you get the cache all the way up to near p core clock.
The e-cores are so fing good that if you're in the market for a good streaming setup on a single PC, you can not only use the iGPU to deliver surprisingly good image through OBS, the 4 ecores on the 12600K can handle x264 medium surprisingly well and the impact on the main game performance is under 1-5% (compared to up to 30% on normal non big.LITTLE cpu arhitectures).
another great informational video.. thank you.. I know it takes a ton of time to put this together.
Love the sweater! Merry Christmas! Happy Holidays to you and yours Steve! Hope you all have a fantastic holiday season! 😊🎉🎄
Thanks for testing this, very much appreciated! 👍🎄
The thing about disabling the e-cores is that you leave a good temp-headroom for overclocking the p-cores alot thus pushing the cpu way more than you can do with e-cores enabled,,,,
And then you will see some pretty good uplift in performance.
You only disable e-cores if you really want to push the clockspeeds way up
I disabled e cores to improve the temperatures, but is it actually worth it? For example, if i have programs like discord opened in background or even hosting a minecraft server and playing on it. One thing i noticed is that the system performs better or feels smoother somehow.
@@TheBlueVaron
Only if you need to push clockspeeds, if you want to have multiple programs running at all times, then NO.
But if you want 5,6-5,7 Ghz corespeed for some demanding game to reach best possible fps then yes... Also remember to clock your ram.👌 Kinda goes together😜
@@ghost500e I actually did the opposite, undervolt the cpu as it was getting reasonably hot for what i actually do, and i saw a tutorial who recommended so much turning e cores off for better temps as it doesnt affect in gaming. But, in case i have discord, many browser tabs opened and hosting a minecraft server with 96 mods, then i should use them, right?
@@TheBlueVaron for sure👍
Merry Christmas and happy holidays. Thanks for the video. It was a very informative piece and I appreciate all the work that goes into it.
This is the test I've been waiting for, thanks guys.
Edit: Some strange results but very thorough testing. I wish they gave us a 10P core cpu. I'm actually gonna wait for better ddr5 and 13th gen or equivalent AMD offerings. Currently running a 10850k.
13th gen will only up the e-core count and slightly increase p-core performance.
So if you want non-hybrid design go with zen4 I guess. If it's not gonna be highly overpriced of course.
@@damara2268 even after meteor lake intel will pass 8core P with 32 core E
@@thegameoftimeskiller2984 idk about that, haven't really heard about what's coming after that but it'd make sense.
If they also upgrade the small cores to have IPC like zen2 or zen3 then it's good!
exactly i have i7 7700 non k ,32gb ram, rtx 3070 FE. CM v650 80 plus gold psu.
shud i go with i9 10850k/msi z490 gaming edge or i5 12600k/asus tuf z690 ???
please suggest me.thanks🙂
@@damara2268 i have i7 7700 non k ,32gb ram, rtx 3070 FE. CM v650 80 plus gold psu.
shud i go with i9 10850k/msi z490 gaming edge or i5 12600k/asus tuf z690 ???
please suggest me.thanks🙂
Hardware Unboxing team and family 😍, a Merry Christmas and a Happy New Year 💻🖥⌨🖱 And Greetings from Hans of the Dutch Lowlands NL 😷 🖖
I picked up the 12700K's for $300, 12600K's for $230 and of course i had to get the 12900K though, it was $560. These went back up $30-$50. Is $260 worth 4 E-cores. I can't say that it is. Even at MSRP.
$200 for 4 E-cores is still too much. I got a couple get by boards. I'd like to see the best DDR4 Z690 boards. If those even exist.
Isn't it summer in Australia?
Good to see Steve wearing that nice warm sweater on the first full day of summer...
This really seems like something that's down to software optimization that we will likely see improved as alder lake matures on the market I think
I think Steve is right on the money with the ring bus speed and the associated L3 latency increase. And may I remind you that there are still programs out there that perform better when you disable hyperthreading... HT has been around for 20 years IIRC...
The switching around of the "uncore" (cache) frequency definitely messes with iRacing, possibly other old games.
Not a performance hit, but stutter. No ecores - no stutter
thats what i wanted to know. stutters :P
@@CVLova today I stutter without the E cores. All not easy. Too many things can go wrong
That is a very loud Be Quiet sweater. Happy holidays!
It's two Years later at the time of this comment and newer tests have shown that e-cores do help overall, even on 12th gen. I think a lot of it is due to newer games taking them into account and drivers maturing for GPUs which also make better use of them. There are still some games that perform better without them but in general they do help even more now than they did back when this test was done- just check out newer charts and decide if they're worth it on the games you play.
One thing I'd like to see- with e-cores disabled, can you overclock the P cores more? and if so, can that overclock outperform having the e-cores for games that do better with them? I have a 12600kf in my secondary system and I might just do a personal test to find out if I find the time.... time is harder and harder to find as I get older though =\
People were saying this in another video I was watching. I plan on disabling the e cores to see if I can get a smoother experience. I’m not really interested in overclocking
this is exactly what i was looking for thanks for doing this testing
Nvidia vs AMD input lag comparison please. Nvidia focuses a lot on calculations and software, whereas AMD RX6000 series were presented as gaming option
The reason some games perform better with the e cores disabled is because the ring clock is higher without the e cores loaded.
It goes down to 3.6 GHz with the e cores enabled or under load.
Leave the e cores enabled and overclock the ring clock a bit and you get the best of both worlds. 4.2 GHz should be achievable for most chips.
Alder lake provided us with a lot of fantastic "for science" videos from Steve.
I'd like a review testing how streaming performs on 12th gen using the cpu to encode. Will the scheduler be smart enough to put all the rendering load on the ecores? That would be a game changer for a lot of people.
No, as h264 is way too intensive to be handled just by e cores without drop outs. It also isnt a game changer since a 5950x still suffers, despite having basically 2 independent 5800x's at its disposal, as the encoding Software also utilizes the cache and memory subsystem, and therefore limiting the games acces to that.
I don't think the rendering is done on the E-Cores. The i9-12900K however is really incredible overall for streaming. Check out EposVox's video covering it.
@@xflamousz Considering that we have a lot of E cores and video is more about multithreading, that isn't an issue
@@saricubra2867 exept it is.
@@xflamousz Is not, there are a lot of threads, there's Xe encoder too.
I would expect enabling e-cores to make games run faster (especially old ones, like csgo, which is cpu bound after a certain point) since it allows the OS to offload scheduling, background tasks and other stuff to the e-cores, leaving the p's available for gaming.
Nice to see a youtube channal that test cpu bound games like riftbreaker +1
Love the interesting look at the new Intel CPU's. Would have liked to see the power usage and temperature differences in the e-core on/off configs, since on your previous videos you mention that you had to use a premium cooler (MSI MEG CoreLiquid S360) to avoid intermittent thermal throttling. Is it possible that e-cores off is using less power (watts) and generating less heat load so that people who haven't sprung for $280 to $300+ AIO's are seeing lower temps when disabling e-cores? I have seen other testing where it seems they are going with 420mm AIO's (and some are recommending custom water cooling) due to thermal load issues. I love the Intel/AMD competition, but, people going with a 12900k need to realize that to get every last bit of performance out of the 12900k, they must be willing to go with an expensive premium case, cooler, memory, etc .... Again, very interesting testing... love the videos!
Only on heavy consistent loads, also not sure how a premium expensive case made the list of your requirements.
@@Deviantsoundz A $89.99 Lancool 215 case won't fit the larger coolers required to cool an OC'd 12900k in a 26c room (78f)... Even my Lancool II Mesh Performance case $110 has issues with some coolers over 280mm.and often premium 360 aio's don't fit in cheaper cases, much less a 420mm AIO.'s.. that's how I come up with a premium air flow case being required... especially if using $150 to $300+ AIO's, Are you really going to try to put it in a $60 to $80 non airflow case and expect good performance??? LOL.
@@SmarterGaming plently of sub 100 cases thats fit 360s so idk why you having issues. Even a tiny nr200p can fit an artic 280 aio that performs close to some 360s even my old cheap fara r1 fits a 360.
There is also an option to use ScrollLock to park E-cores when you need it. At least Asus EFI offers it.
I’m wondering if the process scheduler is handing non-gaming process threads to the e-cores in the cases of high single thread performance requirements, and it is giving the assigned P-core a boost as it has the core solely for the game process
I feel like this is Hyper threading all over again, purist said "disable HT, for much better fps" now we can hear the same with E cores, while its not really worth doing so. It will take time when games will be made with hybrid architecture in mind, current games still runs lets say in "legacy mode" so its going to be interesting, and no E cores are NOT only for background tasks, intel actually released instructions for software/game developers how to utilize E cores as well in their workloads so its only a worst case scenario for now. Just leave those E cores (unless you have legit issues launching your software/games) and enjoy awesome performance of alderlake :)
They are tons of framerate/frame pacing issues. Nothing awesome about Alderlake performance when the game fluidity performance feels like crap.
@@CinePhil101 speaking from experience?
@@h1tzzYT Sadly yes
@@CinePhil101 huh interesting, what game you were playing?
@@h1tzzYT These days: Quake. You work for Intel or something?
Welcome back to hardware disabled (:
So the findings from a larger game set give a different conclusion compared to the previous video on the same topic 2 weeks ago?
I know this is old, but good info for any WoW players.. turniing E cores off has pretty substantial gains in World of Warcraft (Retail).
Big thanks for the big effort, Steve 👍
I would be quite interested to see the difference in power consumption for scenarios that produced identical FPS numbers.
Yes, for me, power consumption is the most impressive metric.
Love the 20% Club.
Thanks Steve. 🤘😇
Thanks for sharing and thanks for testing the big 3 resolutions.
Once again, thank you for the detailed benchmarks
Awesome sweatshirt you have there 😬
Good video as always, love it 💪🏼
The big thing that I've noticed is less hitching after disabling e cores. Night and day for me (msi z690i board on w11)
12700k owner, I just leave the E cores on. I play at 1440p so not a huge difference. I do not want the hassle of going into the bios and turning them on or off for different tasks.
What about temperatures and power consumption? Can you overclock the P cores more when you disable the E cores? What about other video editing, modeling, etc.? I think there's much more to explore than just gaming when it comes to E-cores.
another great benefits with disabling ecore is if you work on multiple instance of android emulator. at least in my case got massive boost with performance on 20+ instance while its laging a lot and unresponsive with ecore enabled
Does streaming software utilise e-cores or p-cores more? Because if leaving the e-cores enable doesn’t affect most games, then I guess it might be a benefit to streamers if their software is well utilised by e-cores
Love the test. Only really care about performance in VRCHAT tough. If a get one of these I will test in that game and adjust disable/enable accordingly.
Its funny to see that disabling E cores increases performance. Whole point of having E cores was to increase gaming performance! and we see exact opposite! XD!
The whole point of having E-cores was actually for efficiency purposes and nothing else, because Intel famously has efficiency problems compared to AMD's current-gen Ryzen CPUs.
E-cores were Intel's bet on catching up to AMD in efficiency, especially under workloads that don't require the power hungry, high heat P-cores. This is especially important for portable battery-powered devices like laptops.
@@GENKI_INU E cores doesn't matter when you are using it in a PC where power constraints are none. Let's hope game companies adopt E-core design for their games in order to improve fps.
@@cyclonous6240 This is why I think E-cores were a mistake for Intel's desktop SKUs. They could've used the die space of the E-cores to make more P-cores, more cache, or even a better IGPU.
Remember that even if games start to utilize E-cores, they will only run as fast as the E-cores are, since that's the main bottleneck, not the P-cores.
@@GENKI_INU Exactly right!
They played this card in Desktop market to understand how people are going to react. If the people accept this kind of thinking for PCs then they might start earning tremendous profit again as it is always profitable to produce 4 E cores rather than 4 P cores.
Desktop market does not need big LITTLE core design.
@@cyclonous6240 I like to imagine how Intel came to the conclusion of E-cores:
"Let's put in 2 more cores for the last two slots of the 12900K die to make 10 cores. It will be a great upgrade from the 8 core 11900K."
"Nah, that's way too much power draw and heat, we'll look bad against AMD."
"Fine then, let's at least put in extra Cache to fill that space."
"Nah, that's way too expensive."
"Okay, how about extending the IGPU die in there instead?"
"Nah, we're releasing our discreet GPUs later this year anyway. We want everyone to buy those."
"What if we cram in 8 crappy 6-year-old Skylake cores without Hyperthreading? We can even call the 12900K a 16 core processor!"
"You're promoted to Chief Design Engineer."
World of Warcraft Benchmark would be interesting 👍🏻 0:01
I love the Be Quiet holiday sweater!!!
Patiently waiting for the SAMSUNG G80A review 🙂
Maybe leaving E-cores enabled and performing better in CS:GO and Starcraft it has something to do with the AI in the games itself? If both tests were performed offline with bots it could be that the E-cores are still being used to help with the AI?
Real answer is disabling E core lower power consumption and bring down heat considerably and possibly allowing to boost tcors to hire clock. The real question is how do you do it safely where does all that extra power go where does the chip simply not require as much power when the E cars are disabled.
There are some games like far cry 2 3 and blood dragon that straight up wont run if you have more than 18 cores. Disabling some or all the e cores will fix this
Disable hell no, but should compiliers and schedulars get a heall of a lot more intelligent in the era of multi cores to be able to sort threads more intentionally. That being said the e core should be used in gaming for things that are not crucial like backgound IO preloading, playing soundfiles, background saving etc...
Thank you for this analysis. When will we see an high end roundup of z690 motherboard? I'll suggest using a flir thermal camera more than the sensor used in the last video :)
I would love to see how these results are affected in something like a low P core count processor, where the game might utilize more than the P cores available, 12600K might be a good candidate, since some games might scale beyond 6 P cores.
There are situations where the OS has support threats and background processes that are not game related that would compete for P core resources if the E cores were not available. My suspicion is that the games that perform better with E cores enabled are also well threaded and are utilizing many if not all cores up to 6 or 7 or 8 but have to compete with background processes
Disabling e cores increases ring LLC, which improves latency. W11 has problem with e cores off in that it stops understanding how to balance between real core and HT core. So then have to turn off HT. W10 isn't e core aware so it won't have this problem. HT off in addition improves latency, with less thread switch and more L3 allocated to P cores. E cores have become a productivity addon for CPUs with no benefit to gaming. What background tasks...people tend to not have anything running in background anyway. Latency.
Since this is probably the last video before Christmas - have a merry Christmas and happy new year! I hope you guys can take some time off from testing and benchmarking for a bit and enjoy a break! Thanks for a great 2021!