Are all software/scheduling issues with e-cores fixed on Intel 12th to 14th Gen - Willing to give hybrid ach a chance now

[Wolverine2349]

That doesn't make much sense. It's mostly a side-grade. Unless you're doing it just out of curiosity.

I am going to go for more cores and turn off hyper threading.

Plus I am curious.

 

[Space Lynx]

I have heard so many say e-cores are good and help in games yet I also read reports that they cause stuttering in some games like Star Citizen Elden Ring and even Cyberpunk??

Well I will give a pass for Star Citizen as it is a bad buggy beta game, but 2 year old Elden Ring and what Cyberpunk which is supposed to be good with them? That gives me pause as well?

Are the issues popping up just FUD and anti e-core propaganda ( I will admit my own fears and skepticism and preference for more P cores has had me biased against e-cores a lot but I am willing to give them a chance now) and easy work arounds or are they real?

And you you effectively use WIN10 with 12-14th Gen or is WIN11 actually required to not have those issues. Cause I hate WIN11 interface and hearing about the more embedded telemetry harder to turn off than on WIN10.

I would prefer more than 8 P cores on a single ring/CCD, but none exists, so have thought of getting a 14700K or 14900K/138900K and disabling HT and just using e-cores for extra threads and seeing how it goes. Will things work smoother in games than just a good 8 core 16 thread CPU with HT on from either intel or AMD Zen4 and Raptor Cove or newer for today's most threaded games and tomorrow's as well? Or not really?

Yes AMD has Zen 5 coming and supposedly better IPC, but still the dual CCD severe latency issues and Intel is much better in that regard even with e-cores it seems as thread director 2 likely works very well and all e-cores and P cores are on the same ring bus unlike AMD's extra P cores.

Cyberpunk just had an update recently that had to specifically tell e-cores and p-cores what to do. I can tell you not a lot of devs will go out of their way to do that. Total War Warhammer the original one, in custom battle, I would sometimes see 81% e-core usage and almost no P-Core and my fps would dip down too. Most of the time it was allocated to P-Core though. I doubt it ever got patched to be honest, someone else can test it if they want, but I already decided based on that experience I will stick with x3d Ryzen chips.

 

[FoulOnWhite]

In my experience with mine, E-cores or not, stability and ram compatability is the most compelling reasons to stick with or buy intel right now. No constant ageesa fixes, and with mine at least, granite stable. There was no X3D when i built it, and given the choice between AM4 with its xx years old boards or risk a new LGA setup, guess which i picked, glad i did. It is pointless and blinkered to not consider the other option when building new.
Roll on LGA1851/AM5 when i will consider the merits of both and switch or stick with Intel. Also i don not care and never have about power use, i can pay my bills and as long as i can keep the chip within it's temp limit, what does it matter.

 

[Vario]

I am on Windows 10 LTSC , been on my 12900KS going 2 years now, never had any problems with e-cores.

 

[evernessince]

Basically everyone in this thread that has / uses Intel CPUS recommends leaving Ecores on for games, everyone that DOESNT have Intel recommends turning them off. I wonder who knows what they are talking about and who is spewing the usual hatred towards a specific company. Hmm, hard question

Character assassination nonsense based on nothing but an assumption. People can have systems outside of the one listed on their profile. It's pretty silly to assume anyone on a tech enthusiast form would only have a single system.

Hi,
Open box time to run away not buy lol

Yeah, open-box is usually a gamble. Half the time you are the guinea pig used to test open box products.

I am going to go for more cores and turn off hyper threading.

Plus I am curious.

What exactly is your use case outside of gaming? Those extra cores will provide 0 benefit outside of gaming and even if they did, now that you've disabled hyper-threading you are giving up a lot of that performance advantage.

I don't really see the point in buying a slower more power hungry core heavy CPU for gaming and then neutering it's only advantage when you could just go 7800X3D. The latter of which will support multiple future generations of CPUs on the same platform so it'll be easy to upgrade to boot.

In my experience with mine, E-cores or not, stability and ram compatability is the most compelling reasons to stick with or buy intel right now. No constant ageesa fixes, and with mine at least, granite stable. There was no X3D when i built it, and given the choice between AM4 with its xx years old boards or risk a new LGA setup, guess which i picked, glad i did. It is pointless and blinkered to not consider the other option when building new.
Roll on LGA1851/AM5 when i will consider the merits of both and switch or stick with Intel. Also i don not care and never have about power use, i can pay my bills and as long as i can keep the chip within it's temp limit, what does it matter.

RAM stability and compatibility has been good on Ryzen for awhile now. Recent AGESA updates even allow 7000 series CPUs to reach higher memory clocks than Intel, although it is pointless due to the higher latency. Ryzen is a better platform for RAM OC as well. As buildzoid but it, memory stabilitiy above 7200 on Intel can be very tricky as sometimes a group of memory settings will randomly throw a memory related error even if the system has passed extensive stress testing where as if there's a lingering stability issue the memory controller on 7000 series CPUs will tend to let you know consistently and upfront.

It's not like Intel is without it's own need for firmware fixes over the years as well. There have been multiple high priority vulnerability updates that had to be pushed through windows to update Intel CPU micro-code and through BIOS updates. Getting e-cores scheduling right and fixing issues with DRM took an entirely updated windows scheduler and moire updates as well. The short of it is, neither company has been without the need for updates at a level lower than the operating system. That's just the world we live in.

 

[bug]

I am going to go for more cores and turn off hyper threading.

Plus I am curious.

This is all you need to know: https://www.techpowerup.com/review/intel-core-i7-14700k/20.html
But if you're curious, go right ahead.

 

[Vya Domus]

The funny thing is that all the scheduler does is avoid using e-cores for foreground task, that's about it. Games run "fine" with e-cores enabled because the scheduler is actively trying to minimize their use.

I am adamant they have no place in desktop PCs, they're only purpose is to give Intel some wins in multithreaded benchmarks, that's about it.

 

[Psychoholic]

If thread director and the OS scheduler work the way they are supposed to, i think it works really well.

If a foreground task takes more than 8 threads, it uses E-Cores before using hyperthreads on P-Cores (Supposedly E-Cores are faster than 2nd thread on P-Cores)

Intel classifies the performance levels on Alder Lake in the following order:

1. One thread per core on P-cores
2. Only thread on E-cores
3. SMT threads on P-cores

That means the system will load up one thread per P-core and all the E-cores before moving to the hyperthreads on the P-cores.

^^ Above from the AT Arch overview of alder lake.

 

[FoulOnWhite]

The funny thing is that all the scheduler does is avoid using e-cores for foreground task, that's about it. Games run "fine" with e-cores enabled because the scheduler is actively trying to minimize their use.

I am adamant they have no place in desktop PCs, they're only purpose is to give Intel some wins in multithreaded benchmarks, that's about it.

What does it matter, a win is a win is it not, and MP is important to some people, so much that they would go with an intel with E-cores over a Ryzen without them.

 

[ir_cow]

stability and compatibility has been good on Ryzen for awhile now. Recent AGESA updates even allow 7000 series CPUs to reach higher memory clocks than Intel, although it is pointless due to the higher latency. Ryzen is a better platform for RAM OC as well.

Absolutely disagree on this. AMD is a pain to do any memory related tweaking. I really enjoy having to reset CMOS when I put the tFAW to low lol.

 

[evernessince]

What does it matter, a win is a win is it not, and MP is important to some people, so much that they would go with an intel with E-cores over a Ryzen without them.

The OP never stated that it would be a "win", just that they wanted to try it. I'm not really sure what the "win" part would be for a 14700K with HT disabled either. You are getting 13600K level MT performance at that point and 7600 level gaming performance for a price higher than a 7800X3D.

Absolutely disagree on this. AMD is a pain to do any memory related tweaking.

Can't say I've had the same experience. Took my vastly less time to tune in my memory settings as compared to my 5820K. The X99 platform in general was terrible for memory overclocking. Not even 1st gen Ryzen was nearly as bad as that platform.

 

[Wolverine2349]

The OP never stated that it would be a "win", just that they wanted to try it. I'm not really sure what the "win" part would be for a 14700K with HT disabled either. You are getting 13600K level MT performance at that point and 7600 level gaming performance for a price higher than a 7800X3D.

HT are not real cores. If you have enough real cores its not needed

Games perform better with it off and e-cores on per benchmarks

Majority of games do anyways.

Intel is ditching HT with Arrow Lake.

Some have stated HT going away is to be celebrated.

Better latency with it off and not helpful nor needed when you have more than 8 cores on a single ring or CCD

 

[ir_cow]

Can't say I've had the same experience. Took my vastly less time to tune in my memory settings as compared to my 5820K. The X99 platform in general was terrible for memory overclocking. Not even 1st gen Ryzen was nearly as bad as that platform.

Ah the older stuff. Well AMD pretty much didnt have any memory tweaks until Ryzen and the first gen was quite bad in support for different types of RAM ICs. Never had a X99, but the Z87/Z97 ( 4th gen) stuff was great. So was Z490 (10th gen).

Overall Intel slaps AMD year after year in memory.

 

[evernessince]

HT are not real cores. If you have enough real cores its not needed

Games perform better with it off and e-cores on per benchmarks

Majority of games do anyways.

Intel is ditching HT with Arrow Lake.

Some have stated HT going away is to be celebrated.

Better latency with it off and not helpful nor needed when you have more than 8 cores on a single ring or CCD

More than 8 cores in general is not need for current games. That's why I don't see the point in getting a 14700K, at least 10 of those e-cores will go completely unused. You are spending more on a processor who's only benefit is Multi-threaded performance in heavily threaded apps.

Ah the older stuff. Well AMD pretty much didnt have any memory tweaks until Ryzen and the first gen was quite bad in support for different types of RAM ICs. Never had a X99, but the Z87/Z97 ( 4th gen) stuff was great. So was Z490 (10th gen).

Overall Intel slaps AMD year after year in memory.

Intel had it's memory latency figured out far before Ryzen entered the market. That's nothing new and not something they've made great strides in recent years on.

The cache and registers are also part of the meory heirarchy and in regards to the former AMD most certainly slaps Intel. The 7800X3D is the fastest gaming processor on the market for a reason. Turns out a super fast high bandwidth on chip cache is superior both in terms of gaming performance and energy efficiency (which makes sense given having to fetch data from memory is going to be vastly less efficient than from cache).

IMO this is sort of like criticizing the winner of a race's stride. It's like yeah it could use work but at the end of the day they are still the winner and likely have after factors that contributed to that win.

 

[JustBenching]

The funny thing is that all the scheduler does is avoid using e-cores for foreground task, that's about it. Games run "fine" with e-cores enabled because the scheduler is actively trying to minimize their use.

I am adamant they have no place in desktop PCs, they're only purpose is to give Intel some wins in multithreaded benchmarks, that's about it.

Totally false

Do you see the scheduler trying to avoid ecores?

The OP never stated that it would be a "win", just that they wanted to try it. I'm not really sure what the "win" part would be for a 14700K with HT disabled either. You are getting 13600K level MT performance at that point and 7600 level gaming performance for a price higher than a 7800X3D.

Totally false too. You are not getting 13600k levels of MT. HT only exists in the 8pcores, disabling them drops performance by 10-15%. CBR23 is still going to be above 30k, making it still more than 50% faster than the x3d. 50%. With HT off. Enough said.

 

[FoulOnWhite]

The OP never stated that it would be a "win", just that they wanted to try it. I'm not really sure what the "win" part would be for a 14700K with HT disabled either. You are getting 13600K level MT performance at that point and 7600 level gaming performance for a price higher than a 7800X3D.



Can't say I've had the same experience. Took my vastly less time to tune in my memory settings as compared to my 5820K. The X99 platform in general was terrible for memory overclocking. Not even 1st gen Ryzen was nearly as bad as that platform.

Try reading the post i quoted, not the OP, -they're only purpose is to give Intel some wins in multithreaded benchmarks, that's about it.-

 

[JustBenching]

Character assassination nonsense based on nothing but an assumption. People can have systems outside of the one listed on their profile. It's pretty silly to assume anyone on a tech enthusiast form would only have a single system.

It's not an assumption. When someone claims that games don't put any load on ecores (see post below) it's blatantly obvious theyve actually never, ever used a cpu with ecores, cause if they had theyd know that's absolutely wrong. Games constantly put load on ecores, in fact that's the whole point of them, to remove any load from the logical cores (HT).

 

[bug]

The funny thing is that all the scheduler does is avoid using e-cores for foreground task, that's about it. Games run "fine" with e-cores enabled because the scheduler is actively trying to minimize their use.

I am adamant they have no place in desktop PCs, they're only purpose is to give Intel some wins in multithreaded benchmarks, that's about it.

Right... Another AMD user that knows and is adamant about it. This forums is just full of useful and objective advice.

 

[JustBenching]

Right... Another AMD user that knows and is adamant about it. This forums is just full of useful and objective advice.

I hate to quote user benchmark but man was he right when he was calling most of the internet "amd religious fanatics".

 

[Vya Domus]

What does it matter, a win is a win is it not, and MP is important to some people

It matters because adding X amount of normal cores to reach a certain level of MT performance is not the same as adding an Y amount of e-cores to get to the same level of performance. An e-core has roughly half the performance of a normal core, whenever there is any process/thread where it's individual performance is important rather than the collective set of threads for that process (which is most of them) lands on an e-core it will run abnormally slow, no matter how good the scheduler can get it's impossible to not come short of a system that just has normal cores.

 

[bug]

It matters because adding X amount of normal cores to reach a certain level of MT performance is not the same as adding an Y amount of e-cores to get to the same level of performance. An e-core has roughly half the performance of a normal core, whenever there is any process/thread where it's individual performance is important rather than the collective set of threads for that process (which is most of them) lands on an e-core it will run abnormally slow, no matter how good the scheduler can get it's impossible to not come short of a system that just has normal cores.

Yes and if Achilles gives an advantage to the tortoise, he will never be able to catch it. It's all just a matter of logic, isn't it?

 

[JustBenching]

The 14900k, the 13900ks and the 13900k is sitting at the top of the TPU application benchmark graph. All 3 of these CPUs are at the top, with the 7950x being number 4 (soon number 5, after 900ks launches) So even with the scheduler sending tasks to the ecores or whatever the guy above is claiming, the cpus with ecores are the fastest.

So the conclusion is that even ecores are faster than Zen 4 cores. I have no other explanation but I'm all ears.

 

[bug]

The 14900k, the 13900ks and the 13900k is sitting at the top of the TPU application benchmark graph. All 3 of these CPUs are at the top, with the 7950x being number 4 (soon number 5, after 900ks launches) So even with the scheduler sending tasks to the ecores or whatever the guy above is claiming, the cpus with ecores are the fastest.

So the conclusion is that even ecores are faster than Zen 4 cores. I have no other explanation but I'm all ears.

I don't think the E-cores are faster than a Zen4 core. If you disable all P-cores and and try to go against the same number of Zen4 cores, Intel will lose badly. But the total processing power, with P and E cores pooled together, is indeed higher than whatever AMD can muster. This is proof that Intel was smart to go for perf/sq mm, even if E-cores aren't a win in every single scenario.

 

[nomdeplume]

Some have stated HT going away is to be celebrated.

Admittedly someone with Zero experience modern P/E core. Who also stated future unrealized implementation would be the most important decider if it going away should be celebrated.

More of a reactive statement on my part than one that should taken as directly relevant to gaming. Quite a large range of programs were developed in direct conflict with HT.

 

[Wolverine2349]

The 14900k, the 13900ks and the 13900k is sitting at the top of the TPU application benchmark graph. All 3 of these CPUs are at the top, with the 7950x being number 4 (soon number 5, after 900ks launches) So even with the scheduler sending tasks to the ecores or whatever the guy above is claiming, the cpus with ecores are the fastest.

So the conclusion is that even ecores are faster than Zen 4 cores. I have no other explanation but I'm all ears.

E-cores themselves on an equal e-core vs equal Zen 4 count clock normalized are definitely not faster than Zen 4 cores, but as a whole combined with P cores, the high end Intel CPUs on current generations are faster than 16 core Ryzen 7950X. Though the extra e-cores likely faster than Zen 4 cores maybe?

Partly because Intel Raptor Cove has like 7-10% better IPC than Zen 4. Even Golden Cove has better IPC than Zen 4 and it is 1 year older. That is not great for AMD at all.

Though AMD does use less power at full load even though worse at idle as I just am experiencing assuming you leave all stock settings on.

Admittedly someone with Zero experience modern P/E core. Who also stated future unrealized implementation would be the most important decider if it going away should be celebrated.

More of a reactive statement on my part than one that should taken as directly relevant to gaming. Quite a large range of programs were developed in direct conflict with HT.

Lots of games even with current implementation of HT do better with it off. Problem is it helps when core count is low which is why 4C/8T chips hangs in there and aged much better than 4 core 4 thread counterparts on the older Intel chips. But it still adds latency and is not near a replacement for a real core.

With only 8 cores, HT can be useful especially as games are getting more threaded. But not as useful as more cores. Its like how 6C 6T chips are still much better than 4C 8T chips and 8C 8T chips still much better.