On Intel Raptor Lake, any truth to the rumors that disabling all e-cores hurts single threaded performance of the p-cores??

[G J]

You was told to move the discussion on in the other forum and a day later you start a thread here, do you want confirmation that badly that these so called e-waste cores are holding back your processor as disabling or enabling them is your choice as you said every use case is different but dont bend the truth to push a narrative onto others.

You best be using a the best of the best system to be trying to pushing this ecore thing so hard.

 

[RavenXXX2]

Life must be really tedious to make such a fuss over this, play and enjoy your games and stop FPS watching.

 

[Wolverine2349]

You was told to move the discussion on in the other forum and a day later you start a thread here, do you want confirmation that badly that these so called e-waste cores are holding back your processor as disabling or enabling them is your choice as you said every use case is different but dont bend the truth to push a narrative onto others.

You best be using a setup like a overclocked delided 13900K/KS with direct die custom water cooling @ 5.9 ghz all cores with 8200mhz tuned ram and a 4090 or otherwise you have performance left to gain.

Well I only air cool. Direct die is very difficult.

The start of this thread is because someone over at overclock.net stated that there was weird behavior or performance reduction of even P cores single thread and stated you just need one e-core turned on and I was very worried about that. In reality I was hoping for confirmation that the CPU would perform no different than an 8 core 16 thread CPU with the IPC and perf of Raptor Cove with no e-cores at all than if it has them and you disable them at BIOS level.

Falkentyne at overclock.net stated there were weird issues. It seems they are probably unfounded on WIN10, though Thread Director 2 is exclusive for WIN11 and yes based on this below:

Core i7-12700K review revealing that E-cores hurt P-cores, but without them, everything is only worse

A review of the Core i7-12700K reveals that the presence of E-cores negatively impacts the performance of P-cores, but without them, the overall performance is even worse.

fox-laptop.com

It would seem turning all e-cores off screws with it only on WIN11 because of thread director which fortunately WIN10 is not aware of. As thread director will auto assign 2nd priority to e-core and if no e-cores it has no idea how to assign threads to logical cores.

Though I made this thread to get opinions to be sure.

It does appear based on testing only a WIN11 problem fortunately.

Problem started when Bencher came in and stated nada reason to disable e-cores and that is how the arguments started and they were trying to shove their opinion down my throat as well and started it.

Every use case is different but Bencher keeps saying its beyond illogical to disable them when it is not. You cannot compare that to someone buying a 7950X and disabling 1 CCD because you can buy a 7700X. With Intel no P core only option and not everyone likes the hybrid approach either. And Intel has no 8 P core only CPU base don current arch so you have to buy 13900K or 13700K and disable e-cores to get 8 Intel P cores.

 

[G J]

Well I only air cool. Direct die is very difficult.

Performance is performance regarldess of how small or big.

Well then thead over as the gains from overclocked direct die/water cooling I would guess would net you better gains than arguing if the ecore's have to be on or off.

 

[MartinPrince]

Performance is performance regarldess of how small or big.

And sometimes even big performance drop can be a non issue if FPS is high enough.

Still does not change that overall games do perform better with e-cores off almost all games except Spider Man Remastered.

Yeah maybe it is a non-issue mostly, but disabling e-cores is also a non issue and gives much more thermal and power headroom for higher P core clocks and higher ring clock especially on air cooling.

I did purchase the chip once again as an 8 core chip and wanted the lots of L3 cache and no e-cores.

If Intel had a 10 P core chip I would have purchased that instead. But they do not. Fortunately games do not have any meaningful benefit from any more than 8 cores or even 6 right now unless you run Discord and a bunch of intense background stuff which I do not as I keep my OS clean and lean and minimal. But does not change fact that fewer faster cores are better for gaming than more slower cores unless you go to 4 cores or less in which things change.

I can see it is probably going to be futile to get a simple answer from you for a simple question.

I'll try one more time. Why did the 'proof' you put forward to support your premise actually tell people do the exact opposite of what you are doing?

 

[Bencher]

I can see it is probably going to futile to get a simple answer from you for a simple question.

I'll try one more time. Why did the 'proof' you put forward to support your premise actually tell people do the exact opposite of what you are doing?

Because all 3 reviews he posted run windows 11 is his answer.

Now why did he post the reviews to support his position when in fact they defeat it is an enigma

 

[Wolverine2349]

I can see it is probably going to be futile to get a simple answer from you for a simple question.

I'll try one more time. Why did the 'proof' you put forward to support your premise actually tell people do the exact opposite of what you are doing?

It does not for most games. Most games actually perform a little better with e-cores off. It is a fact. Now whether significant enough to matter is another debate and depends on individual. And yes there are a few that do better with e-cores on, but the majority in the reviews for games do better with them off. That is a fact.

 

[Bencher]

It does not for most games. Most games actually perform a little better with e-cores off. It is a fact. Now whether significant enough to matter is another debate and depends on individual. And yes there are a few that do better with e-cores on, but the majority in the reviews for games do better with them off. That is a fact.

A fact not supported by any of the links you provided, but a fact nonetheless

 

[eddyr]

Yes good point there, though how come AMD has an 8 P core only Ryzen 7700X and not just 7900X and 7950X if the fabs are cheaper. I mean the lower end AMD SKUs are defective higher core SKUs right??


I mean why doesn't Intel have defective e-core SKUs with 8 working P cores like AMD has defective P cores to have a count down to only 8 P cores.

They both took different approaches, Intel stayed mostly monolithic for now and AMD jumped head first into using chiplets ('gluing' together). Each approach has advantages and disadvantages. If you haven't read up much on it, it would be worth finding a technology deep dive review from a few years ago.
The 7700X uses a full die. 7950X uses 2x 7700X.
The 7600X uses a partially disabled 6 core chip. 7900X uses 2 of these partially disabled chips.


Intel does have defective/binned chips which it will sell as models down the product stack - with the remaining non defective e-cores.
An extreme few dies are going to have all e-cores defective, certainly not enough to support another line, without disabling functioning e-cores dies. But we've established that is a poorer value proposition, who are they going to market and sell these to at volume when the alternative with the e-cores performs better in metrics. A tiny niche 'because I want to' customer base isn't going to do it, especially when the user can disable them in the bios should they care to anyway.

 

[eddyr]

With Intel no P core only option and not everyone likes the hybrid approach either.

Well I feel you may be disappointed going forward as it seems it is here to stay, and the rumours are that AMD may go the same way. They are designing more power efficient compact dies at the sacrifice of peak performance (Zen4c) and launching for EPYC customers, I wouldn't be surprised it makes its way into consumer products in the future.

 

[Wolverine2349]

Well I feel you may be disappointed going forward as it seems it is here to stay, and the rumours are that AMD may go the same way. They are designing more power efficient compact dies at the sacrifice of peak performance (Zen4c) and launching for EPYC customers, I wouldn't be surprised it makes its way into consumer products in the future.

It should be here to stay for laptops and mobile devices It has no place in desktops.

I mean I am sure Intel could make a separate die with 8 P cores only as it would be cheaper than the dies with 8 P cores and 16 e-cores and just selling defective e-core clusters as 13700Ks and defective P and e-cores as 13600Ks.

Why not another die as they in fact had a separate die for the Core i5 12400 which had 6 P cores only.

They in fact had 8 P core dies only for Coffee Lake and Comet Lake.

And they also had 10 P cores with Comet Lake

Why not they make a 10 P core Raptor Lake as well for gamers.

I mean how about an 8 P core only chip with a better ring clock and IMC and lots more L3 cache. That would knock out AMD's X3D chips and they would corner the gamer enthusiast only market as no games get much if any benefit form more than 6-8 cores and games love and love L3 cache and it would mop the floor with even AMD's X3D lineup as Intel has somewhat better IPC in their Raptor Cove cores than AMD and they clock much higher as well on same type of cooling.

And why not more than 8 OP cores. Oh I know some will say Saphire Rapids HEDT is right around the corner for that. Well that is not good for gamers because Saphire Rapids CPUs are going to use a mesh instead of ring which will reduce gaming performance.

Yes no more than 8 strong cores are needed for gaming but some hate the hybrid arch and use their PCs for gaming and also other intense background tasks and could benefit form more than 8 cores without using the big.little approach and sapphire rapids will not be the answer for them on same system because of the mesh topology certainly not if gaming is a big point of their PC without reduced performance.

 

[MartinPrince]

It does not for most games. Most games actually perform a little better with e-cores off. It is a fact. Now whether significant enough to matter is another debate and depends on individual. And yes there are a few that do better with e-cores on, but the majority in the reviews for games do better with them off. That is a fact.

No, no, no! First of all that response is non sequitur because it does not answer the question I asked. That is typically what you will often find politicians doing when they don't want to admit to something.

Secondly the times where you find the few occasions where games slightly under preform when E-cores are on is because HT is also on; turn HT off and you will find 99% of games performing exactly the same or better.

So the conclusion of this thread is that if you are gaming you can leave E-ores on but you should also turn HT off.

 

[Wolverine2349]

No, no, no! First of all that response is non sequitur because it does not answer the question I asked. That is typically what you will often find politicians doing when they don't want to admit to something.

Secondly the times where you find the few occasions where games slightly under preform when E-cores are on is because HT is also on; turn HT off and you will find 99% of games performing exactly the same or better.

So the conclusion of this thread is that if you are gaming you can leave E-ores on but you should also turn HT off.

Most games perform very close either way and very little difference including Spiderman Remastered e-cores on or off. Benchmarks show most either HT on or off better with e-cores off, but only slightly/ Same for games that perform better with e-cores on, only slightly. There is 2 to 3% difference at most either way.= assuming equal clocks speed of the P cores. But faster clock speed P cores and ring/cache e-cores off will beat slower P cores and e-cores on.

Though all else equal with same P core speed and ring/cache speed with ecores on or off almost no difference either way for gaming.

But e-cores off more thermal headroom for air coolers with only 8 cores to cool instead of 16 additional and potential higher P core clocks with air coolers. With excellent cooling can probably do high P core clocks with e-cores on and thus no difference. With air cooling though much much harder if at all possible.

 

[Bencher]

Most games perform very close either way and very little difference including Spiderman Remastered e-cores on or off. Benchmarks show most either HT on or off better with e-cores off, but only slightly/ Same for games that perform better with e-cores on, only slightly. There is 2 to 3% difference at most either way.= assuming equal clocks speed of the P cores. But faster clock speed P cores and ring/cache e-cores off will beat slower P cores and e-cores on.

Though all else equal P core speed and ring speed almost no difference either way.

But e-cores off more thermal headroom for air coolers with only 8 cores to cool instead of 16 additional and potential higher P core clocks with air coolers. With excellent cooling can probably do high P core clocks with e-cores on and thus no difference. With air cooling though much much harder if at all possible.

I have a small single tower air cooler. I have no problem clocking my pcores with ecores on or off. Why would that make a difference? If a game doesn't use ecores, then having them off doesnt give you any thermal headroom

 

[Wolverine2349]

I have a small single tower air cooler. I have no problem clocking my pcores with ecores on or off. Why would that make a difference? If a game doesn't use ecores, then having them off doesnt give you any thermal headroom

If you do a static all core all the time clock rather than dynamic clock, it does matter. For dynamic clock, there are ways to have speeds boost even on lesser cooling as long as other cores are not used at all.

Though I like a static all core all the time clock speed. I am a traditionalist when it comes to overclocking and that is the way I do it. Better latency as well that way.

 

[Tee Hee Johnson]

Can I disable the P cores and simply run on E?

I like these wee chips and don’t want P getting in the way.

 

[Bencher]

Just tested cyberpunk, in the tom's dinner area, heaviest area of the entire game, disabling ecores murders the performance. Drops wattage by around 25-30 watts, but performance takes a huge nosedive.

 

[Gazb]

From what I've heard from the Star Citizen players with the frame rates using 12th and 13th gen Intel, it wasn't so much the raw peak FPS that was the problem, but random hitching, freezes and stutters. All of which went away when you used something like ProcessLasso to force the game to only use the physical P cores.
Obviously, Star Citizen is poorly optimised and has generally low performance, but it is very very much CPU bound and uses lots of threads and that seems to be the scenario that makes the issues with E cores in gaming apparent.

AMD kinda had similar issues with stuff going between chiplets but that was to a much lesser extent, and I believe caused lower fps but stuttering/hitches etc weren't an issue.

Realistically the only way to fix all these kind of issues is with the Operating Systems getting better at knowing where to allocate/schedule process threads onto which type of CPU cores. That seems like a massive ask/task, unless they just add some kind of flag when launching a game/program that require only the P cores to perform properly. Not as involved as something like ProcessLasso as you'd be expecting the OS to know which cores are the P cores etc, and way simpler to implement from an OS stand point, as it doesn't need to know what each specific program should run on, only to look for a flag and then allocate based on that.

 

[Wolverine2349]

Just tested cyberpunk, in the tom's dinner area, heaviest area of the entire game, disabling ecores murders the performance. Drops wattage by around 25-30 watts, but performance takes a huge nosedive.

Somehow I doubt that unless you have a bunch of stuff running in the background like Windows updates and discord and such. Games just do not scale beyond 8 cores and only come close to even maxing out 8 cores in the most toughest detailed city/town areas of games which is fine unless you have stuff using a decent chunk of CPU resources running in the background. What resolution and framerate.

I keep Windows updates disabled and only use manual updates and have gamer mode so AV does not try and update and games come no where close to using 8 cores let alone more in almost all areas except cities/towns of Red Dead 2. And even in the highest cities and towns with lots and lots of AI and NPCs, with 8 cores active, it only comes close to fully maxing them and it never actually does max them out. Comes to like 75 to 80% consistently and maybe 90 to 95% for CPU usage with 8 cores for a brief second if that.

I suppose things could be different once again if Windows updates starts or browser are open with intense web pages or AV initiates update or if you are running 1080P with 3090 or higher or 1080P/1440P with 4090 or higher which would procue insanely high FPS. None of which apply to me with a 3090 at 1440P or a 4090 I am about to switch to at 4K with graphics details near maxed out,

 

[Bencher]

Somehow I doubt that unless you have a bunch of stuff running in the background like Windows updates and discord and such. Games just do not scale beyond 8 cores and only come close to even maxing out 8 cores in the most toughest detailed city/town areas of games which is fine unless you have stuff using a decent chunk of CPU resources running in the background. What resolution and framerate.

I keep Windows updates disabled and only use manual updates and have gamer mode so AV does not try and update and games come no where close to using 8 cores let alone more in almost all areas except cities/towns of Red Dead 2. And even in the highest cities and towns with lots and lots of AI and NPCs, with 8 cores active, it only comes close to fully maxing them and it never actually does max them out. Comes to like 75 to 80% consistently and maybe 90 to 95% for CPU usage with 8 cores for a brief second if that.

I suppose things could be different once again if Windows updates starts or browser are open with intense web pages or AV initiates update or if you are running 1080P with 3090 or higher or 1080P/1440P with 4090 or higher which would procue insanely high FPS. None of which apply to me with a 3090 at 1440P or a 4090 I am about to switch to at 4K with graphics details near maxed out,

Im running windows ghost with absolutely nothing on the background. Windows defender is completely removed.
Here it is, at tom's dinner, the first 25 seconds of the video, in this area only alderlake and RPL can hold 100 fps, everything else nosedives. Here is my run

Cyberpunk 2077 2023 01 24 18 09 36 04

www.youtube.com

And here is the 5800x 3d dropping as low as 45 fps

Cyberpunk 2077 | RTX 4080 | R7 5800X3D | RTX ON/OFF | DLSS | 4K - 1440p - 1080p | Ultra Settings

Buy Gamekeys at CDKeys ►https://www.cdkeys.com/?mw_aref=JansnBenchmarks Please hit the thumbs up button and subscribe if you like what you see.Recorded with ...

youtu.be