Game engine code is nested and tightly looped and extremely latency sensitive. Disabling the e-cores forces additional periodic scheduling of lower priority threads on p-cores, even if only relatively infrequently (by processor cycles and scheduler quanta standards), which leads to expensive context switching, and impacts to L1 code and data cache loading. This leads to an outsized performance impact to game code (and other tightly looped) execution that potentially exceeds the impact of actual execution time of the other unrelated process threads. SMT is designed to reduce this impact by splitting the processor core’s front-end into two logical processors, with separate state, registers etc., which helps to reduce the impact of context switching, but this only goes so far (this, BTW, was the original intent of SMT - latency reduction by reducing cache misses, pipeline bubbles and other latency inducing situations - better execution unit utilization was a secondary benefit to later, wider, SMT enabled processors). With the e-cores enabled, unrelated context switching is significantly eliminated on the p-cores, reducing context switching overhead (as well as execution time). So, while those e-cores might look fairly underutilized (especially given the coarse-grained fidelity of something like Task Manager or HTOP), they’re actually doing valuable work by avoiding latency inducing interruptions to the p-cores. Additionally, the scheduler in current operating systems is configured to prefer physical cores over second SMT logical processors, so even without core-type hinting, general scheduling pattern when running, for example, a game that’s heavily utilizing 8 threads, would be that other periodic scheduling will occur on remaining available physical cores, which in this case would be the available e-cores. If all “first” LP’s of the SMT p-cores are busy, along with the single threaded e-cores, then the scheduler will begin dispatching to the second logical processors of the p-cores. This is done via a hierarchical topology map that includes physical cores and logical core associations, and associations to different portions of memory subsystem hierarchy (L1, L2, L3, NUMA nodes, SMT core map, etc.). Finally, schedulers also perform soft-affinity assignments for threads, where (if possible) a thread is scheduled to run on the same processor core to avoid cache misses and reloading (its even more complicated in full NUMA systems), but if a processor core is available, and the thread is waiting, it will likely get scheduled to run on whatever available core is idle. By keeping the scheduler run queue depth low, by reducing execution resources contention, the potential for scheduler frobbing is substantially reduced. The upshot is that there’s no good reason to disable e-cores, as they’re very unlikely to negatively impact performance. (By the way, the exact same pattern applies to AMD’s asymmetric 3d cache enabled processors (7900/7950x3d), since the second, non 3d cache equipped CCD is effectively equivalent to second-tier e-cores in Intel’s landscape.)
Now that is an excellent explanation. Here's hoping AMD will also implement the same level of scheduler optimization when it comes to their hybrid architecture.
@@haziqsofian Decades of systems software development, drivers, developer tools, etc.. Even spent some time in the semiconductor business. Believe me, I'm a riot at cocktail parties, and my wife, in no way, feels that I have too many computers.
@Garrus Vakarian What exactly do you think is nonsense, as I’d be happy to go as deep as you’d like. OS scheduling behavior? Nature of game engine code? I’d be happy to describe how to use commonly available tools (windbg, various sysintermals tools, etc.) to actually illustrate the behaviors that I described.
From my experience with a 13900k. Playing cyberpunk 2077 with E cores on it introduces a lags\stutters\hiccups randomly specially when driving in the city. Turning off E cores gets rid of these hiccups.
@@sxdattxb Ok now get this, after long time of trying stuff and settings I found out something better than disabling with e cores. Try this instead: 1- enable e corse back and everything back to normal. 2- go to Nvidia control panel and look for manage 3d settings > program settings (NOT global settings), find cyberpunk2077.exe and select it, change "power management mode" from normal to "prefer maximum performance", apply and exit. 3- (This a critical step too) change windows power plan to "Balanced". Been playing for so long and haven't had any issues. Seems like normal power mode in Nvidia downclocks the gpu randomly for cp77, and the last update for cp77 demands windows in balanced plan which could be a devs issue. Anyways try these changes. You should be good
Exactly. What most fail to see is the stutter with enabled e-cores. Same thing I tried in Elden Ring, disabling e-cores removed stutter for me significantly for my 13900k
Could you let us know how disabling the E-cores impacted your thermals and power consumption? It would be interesting to know if this slight drop of performance is compensated by thermal and power consumption gains.
I think everyone should test it on their favorite games and see what’s better. I disabled 4 out of 8 e-cores and hyper threading. Now I my CPU (13700k) can run 5800Mhz all core. Horizon Zero Dawn had less fps , because of missing HT, other games a more.
Turning off the E-cores in games; maybe it was true when the 12th gen began. Like he said in the video, Microsoft and Intel have been working for more than a year on this "thread director/scheduler" and they have improved it for both 12/13 gens.
I think it's mostly for 12th gen, and definitely for laptops or systems with restricted cooling solution. this dude overclocked everything on desktop system, no wonder E-cores aren't a drag if you have both power delivery and coolling system to get it going.
When disabling the E-cores, I suspect that you also lower the total cache size available and it should have an impact. I also wonder how Win11 optimizes the threads distribution in a scenario where all P-cores are assigned to the game and all other "small tasks" like handling the USB devices, mouse, keyboard, sound, etc can be assigned to the E-cores. All those small tasks if assigned to the E-core will free more cache on the P-cores.
That's what I was thinking as well. Instead of backgrounds tasks, services, and apps using up resources from the larger P-Cores, the e-cores are able to handle that giving the CPU the ability to really stretch its legs here and prioritize game scheduling on the larger cores.
When a P-core is interrupted to do another job (example: switching from the game code to the mouse driver code), the P-core must flush a lot of temporary datas like its branching prediction table, the code execution and pre-fetch pipelines, maybe make room on the cache for the new datas, etc. If an E-Core can handle the mouse driver instead of the P-core, there is also a gain here.
@@robertr.1879 Then it begs the question if you like to run a neat ship and not have much background running (however possible this is now with Windows 11 bloatware), then would only 1 E-core be enough, so you can clock P-cores to higher freq?
Do you have a high end AMD GPU on hand to test? Would be interesting to see if the decreased driver overhead would lessen the relative gap between e-cores on vs off.
I don't see how more cores would be a bad thing even if half are smalller. Still something to offload background tasks too. Id guess it is a tdp design issue.
In a perfect world where the operating system and the application know exactly how to handle having 2 core types, the e cores are a great addition. Unfortunately not all software has caught up and that's were some of the draw backs come in.
I think when they first came out the windows scheduler would like push off gaming stuff to the e cores and not p cores in some cases so it kind of effected performance. But pretty sure that was in windows 10, and not the case anymore
Elden ring is for sure one of the games for me that benefits from disabling e-cores. It was the only solution that somewhat fixed my stuttering, now I can run it with avg 100 fps with unlocked fps mod
Generally you're limited by the CPU temperature if you overclock, so if you turn OFF E-cores, you can overlock P-cores to higher frequency, improving the performance probably more. I wish you would do such test example, by targeting 90 °C cpu temp. If you don't overclock, then ofcourse the above is not important.
as someone who studies IT and CS, it think its pretty obvious why E cores should remain enabled always. E cores were meant to handle the simpler tasks such as the OS background. why would you want to force your P cores to work harder on the OS when they are already working to produce the best results in games? in theory, alleviating pressure on the scheduler is a good thing, but as we've seen in real world benchmarks, it places that theory as a contradiction. never listen to someone who tells you to disable E cores
Even this video disproves your theory. 8:28. Both lows and avg are statistically significantly higher. Your model of how p and e cores work is too simple.
This is true for Elden Ring - it experiences significant frame time issues with e-cores on. Once I shut down all e-cores - the frame time stutter is almost entirely eliminated. It's the only example I can think of though.
Very good video - thank you very much! There is one area that would still be worth to test - and that is laptops. What seem to keep happening on 13th gen Intel CPUs in laptops is that because of the power limit, the E cores take power from the P cores resulting in lower P cores clocks and also stutters. I recently tested Witcher 3 on my 13900hx with E cores off and the stutters seem to have been eliminated. If you have a 13th gen laptop somewhere this would be a very good test as well :). Thank you!
Try and see if u can unlock turbo ratio limits and try to keep your power plan on balanced which doesn't make the e cores run at Max clock and take so much power, just for reference I tried this on my friends laptop and it worked pretty well. Better than disabling e cores for that drop in multi-core performance. As for how to unlock turbo ratio limits just type in youtube how to unlock turbo ratio limits on Intel 12/13th gen cpu
Definitely interesting. I've found with eCores enabled that CS2 and The Finals have worse 1% lows by a considerable margin. I can get CS2 to run on the P Cores but the FInals + anti-cheat doesn't want to budge, even putting Steam on the P Cores.
Wouldn't disabling E-Cores mean that background tasks (Steam, spotify, discord, anticheats, whatever software we all have installed and running in the background while gaming) mean that they wouldn't be running on the E-Cores anymore, but rather on the P-Cores, because that's all that's left, impacting the gaming performance? I was curious because my motherboard (Gigabyte z790UD has a "gaming mode" which disables all E-Cores)
I don't think the gaming mode means that at the moment, but rather optimise compatibility with all games by not having E Cores (some games crash because of ECores for some reason). Hence diabling E Cores is an option as part of gaming mode. In the future maybe E Cores will help generally.
@@desfefeI managed to run Unreal Tournament 2004 (what a masterpiece) 64 bit version on my i7-12700K. It uses one thread, the e-cores aren't a problem, not a single crash and the scheduler can clean a CPU core perfectly for one job without interruptions. The IPC + clocks and smart scheduling help so much for the games from my teenage years. Best i7 of all time (for now,), the 2600K is overrated
@@saricubra2867 You know what I discovered the general incompatibility of ecores and games was down to DRM. Some weird interraction happening causing crashes. It looks like yet another reason to hate DRMs
I have i5 13600f.Running games like Elden ring, Cyberpunk 2077, Remnant 2 with e-cores on causes stutters and lagging.Disabling e-cores makes those games run perfect.
Thank you! (insert michael gif from the office) I always cringe when people swear by it (turning e cores off) and even test new intel chips with e cores disabled especially 13th gen. For example good old gamer tests his 13900k with e cores disabled and in windows 10... -_- Even if e cores had small gaming performance disadvantage i would be against turning them off because that way you disable massive multicore performance headroom for the sake of small gains (probably unnoticeable) on gaming. Techpowerup also did similar test as you and their results are very similar, although they only tested avg fps so i imagine 1% lows would be in similar boat as your tests. This is the video that im going to redirect each time i hear "TuRn OfF e CoReS FoR BeTter PeRfOrMaNcE"
Just found this video off of my i9 14900k, currently tunning it for temps, I settled at a -0.075, runs between 50-60 °C so I'm confortable. Now I'm debating whether I should bump P-Cores beyond 5800 MHz or start bumping up E cores. Mainly gaming PC, since I play at 4K very demanding AAA games I'm going to bump E-cores a bit. Good video man
people associate the degradation/improvement to E-Cores vs P-Cores... but its more likely a matter of how the game is programed to support more cores vs less cores... like inefficiently using resources between cores, that sort of things... meaning if you use only 8 P-Cores and then turn of 2 P-Cores for example, you likely to see similar statistics.
as a current i7 8700f user, what i really dislike about all the current intels is the power consumption/performance gain. the 8700f is a 65watt tdp cpu that, in gaming, almost never draws more then 55watt package power. take the 13400f, that in some games, can draw 105 watts but doesnt give you double the performance of the 8700f.
What I really want to see is what would happen if you disable all of your P-cores and only allowed the system to use the e-cores. How would something like a 13700k or 13900k with all P-cores disabled compare to something like a 10900k?
Just tested this for the first time running RDR2 only on E-cores, worked fine, with some 10 FPS drop. Although I was looking for reducing the heat on my non-K i9 13900 (as 2 P-cores - 3 and 4 - were showing much higher temp than others), and since the current weather in my country is getting over 40C. What I understand (I could be wrong) is that while Windows 11 and its bloatwares are also using mainly P-cores, and then the game also increases the load on P-cores, all the pressures get on P-cores only, and the processor gets hotter, leaving the E-cores to do only minor background tasks. Then I tried running the game on E-cores only, and it reduced the temp on the overall package, for a cost of FPS drop of course, but now fine with the temp.
the idea of eco cores is to use those for back round apps like even running windows or other apps you got going on when gaming and have the game fully use the cores , but in reality the scheduling is a mess and it all gets crossed and screwed up
Sure most games might get a slightly higher frame count with E-Cores but it also causes games to glitch, freeze, and crash. That’s why many gamers disable them
I tested them with armor core 6, the game stutter and dropped to 7xfps when ecore is on . And it drop only 1 frame(or not) when ecore is off. I think it all depend on how the developers code the game , there is no right or wrong to do so, only the one which fit the game/PC you are playing.
The main thing I wanted to know was power consumption. Does the system use less power in gaming with cores on or off. This is critical for laptop gamers. Thanks.
I wonder how much of this is due to a stock unoptimised Windows 11 doing so many stuff in the background that having extra cores for that will actually improve low% FPS. Because some of these result looks like benchmark of games with a debloated windows 10/11 vs stock windows, where low1% are often affected in a similar way.
11 has more stuff control e cores better then 10. so it can some neat party tricks like move aps around between cores on the fly based on what window is active
There is more then your OS doing stuff in the background (discord, web browser, spotify, background services, everything in your toolbar, etc), even the game will have multiple worker threads often set to span the total physical cores and then some other threads which do work on demand which isn't part of the core frame time which having those extra cores which they could get scheduled on is better, similar to SMT/Hyperthreading. If you only leave your best cores avaliable then the low priority work also need to share them with the high priortiy work. Also even if worker threads go onto those other cores it's still can lead to higher throughput depending on the workloads.
It has nothing to do with "bloated" or "debloated" Windows (that's a gimmick 99.9% of times) and "stuff in the background". Disabling e-cores reduces the cache size.
@@ml_serenity L3/LLC should not change size, and fixed partitioning of L3 hasn't been a thing for a long while, however with ecores disabled your not stopping what running so now the pcore L1 and L2 are going to have more threads using them as there is less cores.
My take - if you want the ultimate thermals and are only gaming, you don't need them. Plus, you can likely overclock your P cores higher when the cooler is only dealing with them. 8 cores is plenty for gaming.
as is switched from 7700k to 12600kf the parking cores was very strage to me. HT deactivated and back activated from the Cpu while running was completly new for me. the concept of E cores on the other hand was very logic and it works very well under W11 in games and anything else. all the benchmarks saying the new i5 is 3 times faster than my old i7 and i can feel and see it in every aspect of my system. i think Adler Lake and all whats coming after it in the future also from AMD side will be truly amazing for all of us.
I have an i7-13620H on a mini-pc, but is a laptop CPU. It has 12 P cores and 4 E cores. Obviously, your excellent testing and the comments shows disabling all E cores is not recommended for many or most games and likely not for any other applications. But I have disabled only 1 E core, a 25% reduction in E-resource, which considering the comments about some games showing stutter effects with the E cores, a partial disable may be in order, but I still am wondering if this limited disable is beneficial in performance or temperature. Any answers?
How did you go that high on the p cores? Got the same cpu and cooler and at stock i jump at thermal throttle in cinebench 24. Gaming is fine at 60-70 degrees
More 1080p 'pc gaming' benchmarks? Hilarious that someone would pay for a 13700K and gaming GPU and then run it at 1080p, like they were playing nintendo switch lol.
the e cores mostly there for background tasks, or things like streaming wile gaming. the people that say to disable them are dumb. it may also depend on what OS Win11 handles them MUCH better then 10 as well
The frametime swing back and forth + worse 1% lows is more noticable on CPU intensive title, especially if you have fewer P cores and way more E cores like the i3 1215U. It might have a very high fps but it feels stuttery
I had game mode enabled, I'm not sure if it works as well, on windows 10 as I haven't tried it. I'd imagine performance would be inferior as Intel themselves recommend Windows 11 for 12th & 13th gen.
It's common sense the e cores do something they are designed to do background tasks and reduce heavy loads in games it's the same, if the task at hand is a game the CPU will share loads to all cores it depends on the games.
@@DannyzReviews great idea because it would be interesting to see what we need to use and what is unneeded for gaming since we can save some voltage and heat and productivity doesn't matter to us gamers all we want to know is is it better to turn off hyperthreading and leave ecores on or to turn off the ecores and leave hyper threading on Because having 32 threads on a 13900 K or 14900 K seems a bit excessive for gaming
I'd love the possibility to assign priority to a certain task (like a game), which then exclusively runs on P-Cores. Or if one looks at it from the other side: To tell the windows scheduler that everything windows, driver, updates etc. related is only allowed to run on E-Cores, even if they are fully utilized. That way the P-Cores stay unbothered while working on the big workloads.
Good video. Just asking can you do this in a real world situation. People using muliple monitors and having stuf going on in the backround. Like RUclips, discord, OBS whadever? I never seem to just game I always have something runing in the backround. E cores might help a lot there.
Save power and be more efficient, e.g. an Avast Scan on my windows drive, task manager shows it uses the E-cores mainly not the P-cores. It seems windows background tasks use E-cores, so your P-cores are free for main apps/games etc.
@@radicalturkey No 1% lows show you the lowest but without showing you the ABSOLUTE lowest FPS because if it's loading or something that might be 1 fps but just for 1 instance over the entire play session. If your 1% low is 40fps but there are multiple framedrops per minute at 40fps you're having a terrible experience.
Underclock E cores to get more headroom for P cores on overclock works flawlessly, end up to have E cores on just to streaming in discord for some people sometimes. Have a loot of background task too
Alder Lake makes Sandy Bridge look completely stupid. These perfomance gains are ridiculous by how big they are. I can't wait for an hybrid and powerful 12 core AMD APU with 3D cache and an amazing scheduler like Thread Director.
i really like hybrid architecture and even if performance in games were lower i will never turned it off because it's just a game...and you lose performance in other applications
@@MarcusH... Yes very true. Most games do not have much difference. Even games that benefit have only marginal improvement. But some games that are hurt by them are a mess. Like Star Citizen. The hybrid tech is just such a massive change from the SMP world we have been in for over 20 years which is why I am not a fan of the design.
This is a good test, but doesn’t turning the E cores off allow you to overclock the cache/ring higher?? I don’t believe these are fully pushed to the limit.
@@GuidoDePalma I'll CALL YOU CLUELESS because you ARE. The E-CORES ARE NOT USEFUL when you can get BETTER numbers with a higher RING and MASSIVE DDR5 OC. OTHER more est. channels SHOW that DDR5 OC gets you better numbers and with the E-Cores off you can get better speed and extreme TIMINGS. Also BIG ring improvements- that FAR outweighs the E-Cores.
You have any guidance on how to overclock memory? I’ve been researching about it but nothing that gives me enough info to get my memory time better or overclocked
@@abritabroadinthephilippines if you are talking about his previous video that is in this videos description, that has no guidance on how to do it. It only goes over results.
@@ripvanwinkle3432 that’s what I’ve been running into lol. The issue is some of the really good programs like Mem Tweakit they used are not long available. I’m hoping Danny has some ideas since he had to have learned from somewhere.
Awesome breakdown, and very useful information. What I'm wondering though is, did you test thermals? My thinking here is that with the E-cores disabled, there's less cores actively producing heat on the IHS. If that is the case, it may be worthwhile to test E-cores off/on with the CPU overclock dialed up- to see if the less heat means higher overclocking headroom and perhaps then an additional performance increase in turn. Normalized results versus the temperature should then tell us if there's more room for overclocking on all-P core config versus hybrid.
I did test this and the best my chip was able to get on its p cores with the e-cores disabled was another 100MHz so like maybe another 1-3fps. Power consumption only went down by around 10-20 watts so not a huge difference in heat either.
Damn. Too much to hope for I guess :). Should be interesting to see if this changes at all with the 14th gen, given they are adding multiple e-cores and increasing power draw by about 10-15%. Probably not though, with how close all the benchmarking results were, doesn't seem like the e-core count/use is really going to be a factor for gaming. Thanks for the information! @@DannyzReviews
@@charlesandresen-reed1514 Intel's really gotta jump to a newer more efficient node, when the only adequate cooling solution is a 360mm or undervolting then that's a big problem.
You're not wrong. If there was ever a time for one of the other companies to pick up some market share, it'd be now. They're at the tail-end of diminishing returns. We're now to the point where there isn't enough material surface area to move heat regardless of what thermal mass they are attached to. @@DannyzReviews
People disable them for win10 and less, there are people very resistant to upgrading to 11, and you can't blame them as they managed to make the UI worse again.
Just get the cheap one from thermalright for 15 bux it even comes with paste. The thermal grizzly contact frame is only worth it if youre going to lap the cpu because it combines with an acrylic physical lap guide to ensure you sand the cpu flat and to the right depth
I get your point about the L3 cache, you are losing it when you disable the E cores, I am pondering that too but I am not sure how pinning games to the P cores proves anything? Isn't the Thread Director and scheduler supposed to be doing that anyway? (the odd game withstanding where it needs to be done manually). So the result after pinning would be the same as his "E cores on" result since they are still there taking over background tasks.
still for me off, because I saw sometime mini stutter in FPS game when they are on, I prefer 2 more P cores in places of 8 e-core ... it's just fake result for me, micro stutter is more difficult to saw than a FPS result ( prove nothing in visual confort )
This only applies to desktops. On laptops, it's often the opposite story, especially with modern i9-s with 16 e-cores. The stuttering in most games is insane and 1% lows are really poor. But if you disable half or most e-cores the framerates are as stable as last year's laptops.
u will missing with windowed/borderless game optimized feature built in win11,which make your game run like fullscreen,even in window or borderless mode
On some games, you need to disable e-cores. One of the best example is Total War Three Kingdoms. With e-cores enabled, my CPU temp would jump to 80 degree on main menu. Meanwhile, with e-cores disabled, it goes to 40-50 degree
Wow the amount of weird people on the internet is amazing. If you turn off E-cores micro tasks and background tasks that the thread director relegates to it will be forced to be offloaded to the P-cores hence borrowing some small amount of computing power. This is similar to 7900X3D/7950X3D where the thread director uses the 6/8 cores with more cache for the game running and use the extra 6/8 for other tasks. There is also some dumb answer on reddit saying the E-cores are not real full pledged functional cores lacking instructions etc., which is absurd as there are intel CPU's that are just totally made up of ONLY E-cores on many low end PC's I believe some are called N100/N200/N300 and all of its variants and are full functioning PC's!
When i tried turning off the e-cores on my i7-12700K, windows felt sluggish and the energy efficency was a lot worse and i lost fps on console emulators. Cores and threads always are helpful, this is why i never cared about the Ryzen 7 X3D chips, nowadays a Ryzen 7 1700 is smoother running Nintendo Switch emulation than an i7-7700K, less stuttering.
Could you benchmark this against concurrently gaming/streaming/web browsing on other monitors? Don't know how much e cores may improve that experience but it seems like a very important test case scenario
Thanks for the benchmarks! It was interesting to see! Your "waste of time" was useful for us, so that we do not need to do it :P Also, would be interesting to check the same thing in laptops! As one guy in the comments have said, when it comes to laptops, it may behave differently! Thanks for the video!
Game engine code is nested and tightly looped and extremely latency sensitive. Disabling the e-cores forces additional periodic scheduling of lower priority threads on p-cores, even if only relatively infrequently (by processor cycles and scheduler quanta standards), which leads to expensive context switching, and impacts to L1 code and data cache loading. This leads to an outsized performance impact to game code (and other tightly looped) execution that potentially exceeds the impact of actual execution time of the other unrelated process threads. SMT is designed to reduce this impact by splitting the processor core’s front-end into two logical processors, with separate state, registers etc., which helps to reduce the impact of context switching, but this only goes so far (this, BTW, was the original intent of SMT - latency reduction by reducing cache misses, pipeline bubbles and other latency inducing situations - better execution unit utilization was a secondary benefit to later, wider, SMT enabled processors).
With the e-cores enabled, unrelated context switching is significantly eliminated on the p-cores, reducing context switching overhead (as well as execution time). So, while those e-cores might look fairly underutilized (especially given the coarse-grained fidelity of something like Task Manager or HTOP), they’re actually doing valuable work by avoiding latency inducing interruptions to the p-cores.
Additionally, the scheduler in current operating systems is configured to prefer physical cores over second SMT logical processors, so even without core-type hinting, general scheduling pattern when running, for example, a game that’s heavily utilizing 8 threads, would be that other periodic scheduling will occur on remaining available physical cores, which in this case would be the available e-cores. If all “first” LP’s of the SMT p-cores are busy, along with the single threaded e-cores, then the scheduler will begin dispatching to the second logical processors of the p-cores. This is done via a hierarchical topology map that includes physical cores and logical core associations, and associations to different portions of memory subsystem hierarchy (L1, L2, L3, NUMA nodes, SMT core map, etc.).
Finally, schedulers also perform soft-affinity assignments for threads, where (if possible) a thread is scheduled to run on the same processor core to avoid cache misses and reloading (its even more complicated in full NUMA systems), but if a processor core is available, and the thread is waiting, it will likely get scheduled to run on whatever available core is idle. By keeping the scheduler run queue depth low, by reducing execution resources contention, the potential for scheduler frobbing is substantially reduced.
The upshot is that there’s no good reason to disable e-cores, as they’re very unlikely to negatively impact performance.
(By the way, the exact same pattern applies to AMD’s asymmetric 3d cache enabled processors (7900/7950x3d), since the second, non 3d cache equipped CCD is effectively equivalent to second-tier e-cores in Intel’s landscape.)
Now that is an excellent explanation. Here's hoping AMD will also implement the same level of scheduler optimization when it comes to their hybrid architecture.
That's really detailed explanation, what is your background? I'm curious
Cheesus computer sciencing crust best yt comment ive seen
@@haziqsofian Decades of systems software development, drivers, developer tools, etc.. Even spent some time in the semiconductor business. Believe me, I'm a riot at cocktail parties, and my wife, in no way, feels that I have too many computers.
@Garrus Vakarian What exactly do you think is nonsense, as I’d be happy to go as deep as you’d like. OS scheduling behavior? Nature of game engine code? I’d be happy to describe how to use commonly available tools (windbg, various sysintermals tools, etc.) to actually illustrate the behaviors that I described.
From my experience with a 13900k. Playing cyberpunk 2077 with E cores on it introduces a lags\stutters\hiccups randomly specially when driving in the city. Turning off E cores gets rid of these hiccups.
@@sxdattxb Ok now get this, after long time of trying stuff and settings I found out something better than disabling with e cores. Try this instead:
1- enable e corse back and everything back to normal.
2- go to Nvidia control panel and look for manage 3d settings > program settings (NOT global settings), find cyberpunk2077.exe and select it, change "power management mode" from normal to "prefer maximum performance", apply and exit.
3- (This a critical step too) change windows power plan to "Balanced".
Been playing for so long and haven't had any issues. Seems like normal power mode in Nvidia downclocks the gpu randomly for cp77, and the last update for cp77 demands windows in balanced plan which could be a devs issue. Anyways try these changes. You should be good
Exactly. What most fail to see is the stutter with enabled e-cores. Same thing I tried in Elden Ring, disabling e-cores removed stutter for me significantly for my 13900k
Same
Just bought a 13700K + Z790 Motherboard, so thanks for the heads up👍🏼
Could you let us know how disabling the E-cores impacted your thermals and power consumption? It would be interesting to know if this slight drop of performance is compensated by thermal and power consumption gains.
E cores dont really generate any heat
@@malcomreynolds4103they do
Gosh this video is awsome, quick, educative, straigh-forward. Keep it up man !
I think everyone should test it on their favorite games and see what’s better.
I disabled 4 out of 8 e-cores and hyper threading.
Now I my CPU (13700k) can run 5800Mhz all core.
Horizon Zero Dawn had less fps , because of missing HT, other games a more.
I'm pretty stunned. I always heard you Should turn the E-Cores off. Seems to me that leaving them on is a better bet. If You Are On Windows 11.
Turning off the E-cores in games; maybe it was true when the 12th gen began. Like he said in the video, Microsoft and Intel have been working for more than a year on this "thread director/scheduler" and they have improved it for both 12/13 gens.
@@robertr.1879still needs to be turned off for cs2 in october 2023
I think it's mostly for 12th gen, and definitely for laptops or systems with restricted cooling solution. this dude overclocked everything on desktop system, no wonder E-cores aren't a drag if you have both power delivery and coolling system to get it going.
When disabling the E-cores, I suspect that you also lower the total cache size available and it should have an impact.
I also wonder how Win11 optimizes the threads distribution in a scenario where all P-cores are assigned to the game and all other "small tasks" like handling the USB devices, mouse, keyboard, sound, etc can be assigned to the E-cores. All those small tasks if assigned to the E-core will free more cache on the P-cores.
That's what I was thinking as well. Instead of backgrounds tasks, services, and apps using up resources from the larger P-Cores, the e-cores are able to handle that giving the CPU the ability to really stretch its legs here and prioritize game scheduling on the larger cores.
This is a caveat I haven't thought about before
When a P-core is interrupted to do another job (example: switching from the game code to the mouse driver code), the P-core must flush a lot of temporary datas like its branching prediction table, the code execution and pre-fetch pipelines, maybe make room on the cache for the new datas, etc. If an E-Core can handle the mouse driver instead of the P-core, there is also a gain here.
@@robertr.1879 Then it begs the question if you like to run a neat ship and not have much background running (however possible this is now with Windows 11 bloatware), then would only 1 E-core be enough, so you can clock P-cores to higher freq?
Do you have a high end AMD GPU on hand to test? Would be interesting to see if the decreased driver overhead would lessen the relative gap between e-cores on vs off.
I don't see how more cores would be a bad thing even if half are smalller. Still something to offload background tasks too. Id guess it is a tdp design issue.
In a perfect world where the operating system and the application know exactly how to handle having 2 core types, the e cores are a great addition. Unfortunately not all software has caught up and that's were some of the draw backs come in.
I thought it was because they share cache with the larger cores
@@TropicChristmas the ecores have their own cache that is shared with the all 4 ecores in the cluster.
I think when they first came out the windows scheduler would like push off gaming stuff to the e cores and not p cores in some cases so it kind of effected performance. But pretty sure that was in windows 10, and not the case anymore
@@randomguydoes2901 the most brainded comment in this thread...
Seems that E cores are fast as first gen skyline i7 6700k... Very interesting 🤔
That is... impressive
No, at least the 4 e-cores on my i7-12700K are equivalent to an i7-3770K.
Elden ring is for sure one of the games for me that benefits from disabling e-cores. It was the only solution that somewhat fixed my stuttering, now I can run it with avg 100 fps with unlocked fps mod
BF1 has slight stutter with e-cores on 😢
Generally you're limited by the CPU temperature if you overclock, so if you turn OFF E-cores, you can overlock P-cores to higher frequency, improving the performance probably more. I wish you would do such test example, by targeting 90 °C cpu temp.
If you don't overclock, then ofcourse the above is not important.
Nice presentation, I always thought the 'disable ecores for better performance' was silly
as someone who studies IT and CS, it think its pretty obvious why E cores should remain enabled always. E cores were meant to handle the simpler tasks such as the OS background. why would you want to force your P cores to work harder on the OS when they are already working to produce the best results in games? in theory, alleviating pressure on the scheduler is a good thing, but as we've seen in real world benchmarks, it places that theory as a contradiction. never listen to someone who tells you to disable E cores
Even this video disproves your theory. 8:28. Both lows and avg are statistically significantly higher.
Your model of how p and e cores work is too simple.
@@whatistruth_1 Meanwhile every other graph shows otherwise
@@Guy_that_exists Rift breaker. This also neglects lost OC potential when E cores are enabled
This is true for Elden Ring - it experiences significant frame time issues with e-cores on. Once I shut down all e-cores - the frame time stutter is almost entirely eliminated. It's the only example I can think of though.
Very good video - thank you very much! There is one area that would still be worth to test - and that is laptops. What seem to keep happening on 13th gen Intel CPUs in laptops is that because of the power limit, the E cores take power from the P cores resulting in lower P cores clocks and also stutters. I recently tested Witcher 3 on my 13900hx with E cores off and the stutters seem to have been eliminated. If you have a 13th gen laptop somewhere this would be a very good test as well :). Thank you!
Try and see if u can unlock turbo ratio limits and try to keep your power plan on balanced which doesn't make the e cores run at Max clock and take so much power, just for reference I tried this on my friends laptop and it worked pretty well. Better than disabling e cores for that drop in multi-core performance. As for how to unlock turbo ratio limits just type in youtube how to unlock turbo ratio limits on Intel 12/13th gen cpu
Awesome video man! Your work is greatly appreciated.
Definitely interesting.
I've found with eCores enabled that CS2 and The Finals have worse 1% lows by a considerable margin. I can get CS2 to run on the P Cores but the FInals + anti-cheat doesn't want to budge, even putting Steam on the P Cores.
How did you run the finals only on P-Cores then?
@@charlybe either BIOS Disabled or disabled in Process Lasso.
The finals runs fine on my P cores
Wouldn't disabling E-Cores mean that background tasks (Steam, spotify, discord, anticheats, whatever software we all have installed and running in the background while gaming) mean that they wouldn't be running on the E-Cores anymore, but rather on the P-Cores, because that's all that's left, impacting the gaming performance?
I was curious because my motherboard (Gigabyte z790UD has a "gaming mode" which disables all E-Cores)
I don't think the gaming mode means that at the moment, but rather optimise compatibility with all games by not having E Cores (some games crash because of ECores for some reason). Hence diabling E Cores is an option as part of gaming mode. In the future maybe E Cores will help generally.
@@desfefeI managed to run Unreal Tournament 2004 (what a masterpiece) 64 bit version on my i7-12700K. It uses one thread, the e-cores aren't a problem, not a single crash and the scheduler can clean a CPU core perfectly for one job without interruptions.
The IPC + clocks and smart scheduling help so much for the games from my teenage years. Best i7 of all time (for now,), the 2600K is overrated
@@saricubra2867
You know what I discovered the general incompatibility of ecores and games was down to DRM. Some weird interraction happening causing crashes.
It looks like yet another reason to hate DRMs
I have i5 13600f.Running games like Elden ring, Cyberpunk 2077, Remnant 2 with e-cores on causes stutters and lagging.Disabling e-cores makes those games run perfect.
Thank you! (insert michael gif from the office) I always cringe when people swear by it (turning e cores off) and even test new intel chips with e cores disabled especially 13th gen. For example good old gamer tests his 13900k with e cores disabled and in windows 10... -_- Even if e cores had small gaming performance disadvantage i would be against turning them off because that way you disable massive multicore performance headroom for the sake of small gains (probably unnoticeable) on gaming.
Techpowerup also did similar test as you and their results are very similar, although they only tested avg fps so i imagine 1% lows would be in similar boat as your tests. This is the video that im going to redirect each time i hear "TuRn OfF e CoReS FoR BeTter PeRfOrMaNcE"
Just found this video off of my i9 14900k, currently tunning it for temps, I settled at a -0.075, runs between 50-60 °C so I'm confortable. Now I'm debating whether I should bump P-Cores beyond 5800 MHz or start bumping up E cores. Mainly gaming PC, since I play at 4K very demanding AAA games I'm going to bump E-cores a bit. Good video man
I’m glad I got e cores, I love playing games while watching you tube and having a lot of browsers open. 😂
people associate the degradation/improvement to E-Cores vs P-Cores... but its more likely a matter of how the game is programed to support more cores vs less cores... like inefficiently using resources between cores, that sort of things... meaning if you use only 8 P-Cores and then turn of 2 P-Cores for example, you likely to see similar statistics.
as a current i7 8700f user, what i really dislike about all the current intels is the power consumption/performance gain. the 8700f is a 65watt tdp cpu that, in gaming, almost never draws more then 55watt package power. take the 13400f, that in some games, can draw 105 watts but doesnt give you double the performance of the 8700f.
my 13600k use 66 watts @ hogwarts legacy, just with 0.0800 offset mode (core voltage)
@@nivea878 ^ this. I think those high power usage numbers are from mostly synthetic tests. no?
@@nivea878 Did you limit something in your mobo bios to have your 13600k CPU wattage peg below 70 watts?
@@BravoSixGoingDark -80mv voltage offset, he said in his comment
@@IIHydraII i see.
What I really want to see is what would happen if you disable all of your P-cores and only allowed the system to use the e-cores. How would something like a 13700k or 13900k with all P-cores disabled compare to something like a 10900k?
You can't. E-cores lack the instruction set to run code properly and can only assist P-cores.
Just tested this for the first time running RDR2 only on E-cores, worked fine, with some 10 FPS drop. Although I was looking for reducing the heat on my non-K i9 13900 (as 2 P-cores - 3 and 4 - were showing much higher temp than others), and since the current weather in my country is getting over 40C. What I understand (I could be wrong) is that while Windows 11 and its bloatwares are also using mainly P-cores, and then the game also increases the load on P-cores, all the pressures get on P-cores only, and the processor gets hotter, leaving the E-cores to do only minor background tasks. Then I tried running the game on E-cores only, and it reduced the temp on the overall package, for a cost of FPS drop of course, but now fine with the temp.
@@MaDrungwrong, you can, i testet it
I've tested setting affinities to eCores only and it wasn't pretty.
the idea of eco cores is to use those for back round apps like even running windows or other apps you got going on when gaming and have the game fully use the cores , but in reality the scheduling is a mess and it all gets crossed and screwed up
Sure most games might get a slightly higher frame count with E-Cores but it also causes games to glitch, freeze, and crash. That’s why many gamers disable them
I tested them with armor core 6, the game stutter and dropped to 7xfps when ecore is on . And it drop only 1 frame(or not) when ecore is off. I think it all depend on how the developers code the game , there is no right or wrong to do so, only the one which fit the game/PC you are playing.
The main thing I wanted to know was power consumption. Does the system use less power in gaming with cores on or off. This is critical for laptop gamers. Thanks.
It was very interesting.
I need to enable E-cores seems like.
They are enabled by default 🤔
@@bns6288 I did some oc testing some time ago
hello, how is your processor doing with avx 512 when e cores disabled? do you have the instruction set in cpu z app shown?
I wonder how much of this is due to a stock unoptimised Windows 11 doing so many stuff in the background that having extra cores for that will actually improve low% FPS. Because some of these result looks like benchmark of games with a debloated windows 10/11 vs stock windows, where low1% are often affected in a similar way.
11 has more stuff control e cores better then 10. so it can some neat party tricks like move aps around between cores on the fly based on what window is active
There is more then your OS doing stuff in the background (discord, web browser, spotify, background services, everything in your toolbar, etc), even the game will have multiple worker threads often set to span the total physical cores and then some other threads which do work on demand which isn't part of the core frame time which having those extra cores which they could get scheduled on is better, similar to SMT/Hyperthreading.
If you only leave your best cores avaliable then the low priority work also need to share them with the high priortiy work. Also even if worker threads go onto those other cores it's still can lead to higher throughput depending on the workloads.
It has nothing to do with "bloated" or "debloated" Windows (that's a gimmick 99.9% of times) and "stuff in the background". Disabling e-cores reduces the cache size.
@@ml_serenity L3/LLC should not change size, and fixed partitioning of L3 hasn't been a thing for a long while, however with ecores disabled your not stopping what running so now the pcore L1 and L2 are going to have more threads using them as there is less cores.
@@reductor_ You're wrong.
My take - if you want the ultimate thermals and are only gaming, you don't need them. Plus, you can likely overclock your P cores higher when the cooler is only dealing with them. 8 cores is plenty for gaming.
as is switched from 7700k to 12600kf the parking cores was very strage to me. HT deactivated and back activated from the Cpu while running was completly new for me. the concept of E cores on the other hand was very logic and it works very well under W11 in games and anything else. all the benchmarks saying the new i5 is 3 times faster than my old i7 and i can feel and see it in every aspect of my system. i think Adler Lake and all whats coming after it in the future also from AMD side will be truly amazing for all of us.
I have an i7-13620H on a mini-pc, but is a laptop CPU. It has 12 P cores and 4 E cores. Obviously, your excellent testing and the comments shows disabling all E cores is not recommended for many or most games and likely not for any other applications. But I have disabled only 1 E core, a 25% reduction in E-resource, which considering the comments about some games showing stutter effects with the E cores, a partial disable may be in order, but I still am wondering if this limited disable is beneficial in performance or temperature. Any answers?
Thank you for all that effort and time u put into this.
Very well done.
How did you go that high on the p cores? Got the same cpu and cooler and at stock i jump at thermal throttle in cinebench 24. Gaming is fine at 60-70 degrees
More 1080p 'pc gaming' benchmarks? Hilarious that someone would pay for a 13700K and gaming GPU and then run it at 1080p, like they were playing nintendo switch lol.
Well we are quite a few people that wants overkill fps. And we play in 1080p and even 720p on high end pc's. So it's kind of legit.
Thank you for the great video!
Now it's time to bench with P Cores disabled
what about power consumption with e cores vs without
the e cores mostly there for background tasks, or things like streaming wile gaming. the people that say to disable them are dumb. it may also depend on what OS Win11 handles them MUCH better then 10 as well
The frametime swing back and forth + worse 1% lows is more noticable on CPU intensive title, especially if you have fewer P cores and way more E cores like the i3 1215U.
It might have a very high fps but it feels stuttery
When you were doing the testing did you have gaming mode turned on? Also does the Intel scheduler work with Windows 10?
The Intel scheduler definitely works for windows 10.
I had game mode enabled, I'm not sure if it works as well, on windows 10 as I haven't tried it. I'd imagine performance would be inferior as Intel themselves recommend Windows 11 for 12th & 13th gen.
It's common sense the e cores do something they are designed to do background tasks and reduce heavy loads in games it's the same, if the task at hand is a game the CPU will share loads to all cores it depends on the games.
Hey Danny how about a video of hyperthreading versus ecores? To see whether or not it is more valuable to turn off hyperthreading and leave acores on
I am actually working on that video, should released in the near future.
@@DannyzReviews great idea because it would be interesting to see what we need to use and what is unneeded for gaming since we can save some voltage and heat and productivity doesn't matter to us gamers all we want to know is is it better to turn off hyperthreading and leave ecores on or to turn off the ecores and leave hyper threading on
Because having 32 threads on a 13900 K or 14900 K seems a bit excessive for gaming
@@DannyzReviews this will also be interesting in the sense that the next generation Intel CPUs are not going to have hyperthreading
It was interesting results!
Thanks!
I'd love the possibility to assign priority to a certain task (like a game), which then exclusively runs on P-Cores. Or if one looks at it from the other side: To tell the windows scheduler that everything windows, driver, updates etc. related is only allowed to run on E-Cores, even if they are fully utilized. That way the P-Cores stay unbothered while working on the big workloads.
Same!
Now an E core off and HT off vs HT on test would be epic, since I presume HT is on here?
Pretty interesting stuff. Cool video my friend!
On the test platform, is HT off or on when enable ecores
E cores disabled on the 14900K, all 8 performance cores at 6 GHZ is worth sometimes turning the Ecores off, though I prefer to normally leave them on.
Is there a quick way to toggle between e-cores on/off without going into BIOS every time?
Good video. Just asking can you do this in a real world situation. People using muliple monitors and having stuf going on in the backround. Like RUclips, discord, OBS whadever? I never seem to just game I always have something runing in the backround. E cores might help a lot there.
E cores OC to 4.5? Stock is 4.2, 0.3ghzt isn't gonna make a difference we are talking about 2/3 FPS at max
Save power and be more efficient, e.g. an Avast Scan on my windows drive, task manager shows it uses the E-cores mainly not the P-cores.
It seems windows background tasks use E-cores, so your P-cores are free for main apps/games etc.
E cores helps in increasing 1% low , which is nice to improve gaming experience than just increasing Average FPS
Do you have hyper threading enabled tho? Or did you turn off e-cores and HT?
Did you disable e-cores only on the games tested or system wide?
FPS is not the same as frame stability. People had problems with stability, not FPS numbers.
That's why he showed 1% lows isn't it?
@@radicalturkey No 1% lows show you the lowest but without showing you the ABSOLUTE lowest FPS because if it's loading or something that might be 1 fps but just for 1 instance over the entire play session. If your 1% low is 40fps but there are multiple framedrops per minute at 40fps you're having a terrible experience.
you can also play with just e- cores, i tested it for myself😁
Yes, just tested RDR2 only on E-cores and it reduced the overall CPU temp.
Can't you just OC the P cores even more once you disable E cores ? Like at least 5.8 under TVB with a 60 degrees or so load.
Underclock E cores to get more headroom for P cores on overclock works flawlessly, end up to have E cores on just to streaming in discord for some people sometimes. Have a loot of background task too
What was the power consumption difference?
In gaming barely a difference we're talking maybe 5-10 watts at most.
@@DannyzReviews and temperature?
Good stuff, also didnt know the bit about Windows 11, gonna look into upgrading from 10 if there really is a benefit
Alder Lake makes Sandy Bridge look completely stupid. These perfomance gains are ridiculous by how big they are.
I can't wait for an hybrid and powerful 12 core AMD APU with 3D cache and an amazing scheduler like Thread Director.
i really like hybrid architecture and even if performance in games were lower i will never turned it off because it's just a game...and you lose performance in other applications
Sadly E-cores is a lie. They should be called A-cores: additional cores. A processor with them do not consume much less power. It's the same.
On my gigabyte motherboard, there is as option to disable e core, for gaming ...
I miss something ?
You should make a video over clocking the intel 13700k i7 . With the MSI motherboard.
so no reason to disable E Cores right
Buildzoid will still tell everyone how E cores destroy gaming performance, no matter how many videos proof that this isnt the case haha
@@Waldherz some games are unplayable unless you disable e.cores though, like nonstop stuttering
@@MarcusH... i had not a single game yet until now who stuttered because of ecores.
@@MarcusH... Yes very true. Most games do not have much difference. Even games that benefit have only marginal improvement. But some games that are hurt by them are a mess. Like Star Citizen. The hybrid tech is just such a massive change from the SMP world we have been in for over 20 years which is why I am not a fan of the design.
Did you run these tests on windows 10 or windows 11 ?
+1700mhz to the memory ? Did you check for performance degradation?
And people said that 12 or 16 cores from AMD wouldn't benefit gaming. Having extra cores for background tasks and or execution bursts should help.
sure it helps, for cpu usage
This is a good test, but doesn’t turning the E cores off allow you to overclock the cache/ring higher?? I don’t believe these are fully pushed to the limit.
very little
@@GuidoDePalma WRONG with the E-cores off you really push the RING and DDR5 OC. you are completely clueless
@@cptnsx you can push the rings yes, but the ecores are useful in most games. Do your research before calling someone "clueless"
@@GuidoDePalma I'll CALL YOU CLUELESS because you ARE. The E-CORES ARE NOT USEFUL when you can get BETTER numbers with a higher RING and MASSIVE DDR5 OC. OTHER more est. channels SHOW that DDR5 OC gets you better numbers and with the E-Cores off you can get better speed and extreme TIMINGS. Also BIG ring improvements- that FAR outweighs the E-Cores.
@@cptnsx pointless to continue this conversation. play with ecores off and whatever you like.
You have any guidance on how to overclock memory? I’ve been researching about it but nothing that gives me enough info to get my memory time better or overclocked
Look in the video description buddy there's a video regarding that.
@@abritabroadinthephilippines if you are talking about his previous video that is in this videos description, that has no guidance on how to do it. It only goes over results.
Most are still clueless on ddr5 so you're out of luck there. But there is info out there on ddr4
@@ripvanwinkle3432 that’s what I’ve been running into lol. The issue is some of the really good programs like Mem Tweakit they used are not long available. I’m hoping Danny has some ideas since he had to have learned from somewhere.
Awesome breakdown, and very useful information. What I'm wondering though is, did you test thermals? My thinking here is that with the E-cores disabled, there's less cores actively producing heat on the IHS. If that is the case, it may be worthwhile to test E-cores off/on with the CPU overclock dialed up- to see if the less heat means higher overclocking headroom and perhaps then an additional performance increase in turn. Normalized results versus the temperature should then tell us if there's more room for overclocking on all-P core config versus hybrid.
I did test this and the best my chip was able to get on its p cores with the e-cores disabled was another 100MHz so like maybe another 1-3fps. Power consumption only went down by around 10-20 watts so not a huge difference in heat either.
Damn. Too much to hope for I guess :). Should be interesting to see if this changes at all with the 14th gen, given they are adding multiple e-cores and increasing power draw by about 10-15%. Probably not though, with how close all the benchmarking results were, doesn't seem like the e-core count/use is really going to be a factor for gaming. Thanks for the information! @@DannyzReviews
@@charlesandresen-reed1514 Intel's really gotta jump to a newer more efficient node, when the only adequate cooling solution is a 360mm or undervolting then that's a big problem.
You're not wrong. If there was ever a time for one of the other companies to pick up some market share, it'd be now. They're at the tail-end of diminishing returns. We're now to the point where there isn't enough material surface area to move heat regardless of what thermal mass they are attached to. @@DannyzReviews
People disable them for win10 and less, there are people very resistant to upgrading to 11, and you can't blame them as they managed to make the UI worse again.
Do you think installing a contact frame for a 13600k is worth it? My max gaming temps don't go higher then 68c. I see you did it for your 13700k.
Just get the cheap one from thermalright for 15 bux it even comes with paste.
The thermal grizzly contact frame is only worth it if youre going to lap the cpu because it combines with an acrylic physical lap guide to ensure you sand the cpu flat and to the right depth
It would have been interesting to also see the power consumption.
It's not the E-cores that you tested.
It's an additional L3 cache.
This can be proven by pinning games to p cores with E cores still enabled.
I get your point about the L3 cache, you are losing it when you disable the E cores, I am pondering that too but I am not sure how pinning games to the P cores proves anything? Isn't the Thread Director and scheduler supposed to be doing that anyway? (the odd game withstanding where it needs to be done manually). So the result after pinning would be the same as his "E cores on" result since they are still there taking over background tasks.
@@redrosecounty yes, you have to push background tasks to p cores too. There is no simple way to achieve this.
where is logic,disable e core can improve the gaming performance,ty bro for the info!
i know this is late but i turned them off recently and been getting smoother gaming performance. I am windows 11 12700k
still for me off, because I saw sometime mini stutter in FPS game when they are on, I prefer 2 more P cores in places of 8 e-core ... it's just fake result for me, micro stutter is more difficult to saw than a FPS result ( prove nothing in visual confort )
pes 21 not woking when enable e-core: out game, haiiiiiixs 😂😂😂😂😂😂😂😂
I have a 12500h on my laptop with 4 p cores and 8 ecores does this mean a i5 9th gen with 6 core and 12 threads is better?
no, the 12500h would have 12 total cores with 16 total threads while the 9th gen is only 6 core 12 thread.
This only applies to desktops. On laptops, it's often the opposite story, especially with modern i9-s with 16 e-cores. The stuttering in most games is insane and 1% lows are really poor. But if you disable half or most e-cores the framerates are as stable as last year's laptops.
If ecores only works well with win 11 then its not for me as I plan on sticking with win 10 for years to come.
u will missing with windowed/borderless game optimized feature built in win11,which make your game run like fullscreen,even in window or borderless mode
@@natsu78999 wowie. Better give up more privacy and gargle more MS balls for Windowed mode.
@@Greez1337 Says the guy using a google account to write a comment.
@@Waldherz 🤣👍. Wait until someone tells him about whatsapp and other applications he probably use every day.
On some games, you need to disable e-cores. One of the best example is Total War Three Kingdoms. With e-cores enabled, my CPU temp would jump to 80 degree on main menu. Meanwhile, with e-cores disabled, it goes to 40-50 degree
That's a thermal problem. Not related to the cpu. You shouldn't need to cut your cpu in half just to have lower temps.
E-Cores introduced stuttering in game, had to disable them.
assasyn oddysey shows the complete opposite
Wow the amount of weird people on the internet is amazing. If you turn off E-cores micro tasks and background tasks that the thread director relegates to it will be forced to be offloaded to the P-cores hence borrowing some small amount of computing power. This is similar to 7900X3D/7950X3D where the thread director uses the 6/8 cores with more cache for the game running and use the extra 6/8 for other tasks. There is also some dumb answer on reddit saying the E-cores are not real full pledged functional cores lacking instructions etc., which is absurd as there are intel CPU's that are just totally made up of ONLY E-cores on many low end PC's I believe some are called N100/N200/N300 and all of its variants and are full functioning PC's!
When i tried turning off the e-cores on my i7-12700K, windows felt sluggish and the energy efficency was a lot worse and i lost fps on console emulators.
Cores and threads always are helpful, this is why i never cared about the Ryzen 7 X3D chips, nowadays a Ryzen 7 1700 is smoother running Nintendo Switch emulation than an i7-7700K, less stuttering.
Weird I seen other videos where turning off e cores increases fps.
Nice and very thorough! Very professional presentation as well for such a relatively small channel. Appreciate the work you put into this man!
My clock speed went up by like 0.5ghz when disabling e cores
Could you benchmark this against concurrently gaming/streaming/web browsing on other monitors?
Don't know how much e cores may improve that experience but it seems like a very important test case scenario
Thanks for the benchmarks! It was interesting to see!
Your "waste of time" was useful for us, so that we do not need to do it :P
Also, would be interesting to check the same thing in laptops!
As one guy in the comments have said, when it comes to laptops, it may behave differently!
Thanks for the video!
I only play cs2. Should I disable e cores or HT? Or both? I got the 14700k
@9:00 he has some info
CS2 and GO max out at 16 threads (i think).
@@denvernaicker8250 that was for csgo brother. Cs2 is completely different and horribly optimized 🫠
@@saricubra2867 so disable e cores but keep HT on?
@@taras_fit Maybe. Compare how it performs with both Big-Little enabled and disabled.
Im sure you could just allocated the p cored to the game your running and e cores to stuff runing in the back like live streams...task manager ftw