True nature of E-cores and how effective are they?

[P4-630]

It's the future and I'm supporting it.

 

[AM4isGOD]

Why such vitriol and hatred of E cores or even whatever Intel is doing by the AMD users? you aren't even using Intel, so what does it matter to you. It seems like just hatred of Intel to me. You chose to buy AMD so why spout so much hate for ADL/E cores.

However much power Intel with E cores uses, you don't have to use it, or cool it. So what if it is inefficient compared to what AMD is producing, you are not going to buy it or use it. I am so sick of hearing the same anti E core crap from near every AMD user on TPU.

My 12700k runs cool, and is as good as near any AMD CPU out right now (AM5 is not yet) The only reason i kinda regret it is i am sick of hearing the same shite every day on TPU.

 

[Deleted member 185158]

I have a 12400F.

No E cores.

I can't tell they are not there. Excellent performance.

Intel wanted to try some innovations and so we have it.

Don't see them useful for my personal use case, but I'm sure someone likes the extra cores.

 

[Mussels]

That approach is the future and AMD will have to adapt.
We only need 6-8 extremely fast cores at ridiculously high frequency and a truckload of small ones to work when is needed (MT).

The concept all p cores is a dead end.

I agree - this is why I like the theory that AMD will mix generations of cores to make E and P cores less costly to implement, and use older stock.
Zen 4 (P cores) + Zen 3 (E cores) for example

If they lock them in the area that Zen 3 has it's best power efficiency (lets say 4.4GHz) they'd make some fantastic E cores for no real investment from AMD - they already a plentiful supply of them, and they're already modular thanks to the CCX designs.

I also agree with ERazer - it all comes down to OS support.
Right now AMD use their chipset driver to set preferred cores, so they have a method of doing this (set all P cores to preferred cores, and most of the problem is solved)

 

[hat]

I don't see why it matters. From where I sit, with my 2600k, the e-cores look pretty good if they are indeed as strong as a Skylake core. I know we've had many iterations of Skylake already, but if we can now get those down into low power "efficiency" cores, that seems pretty good... and they're stronger than my 2600k, anyway. Without knowing much about the design, I suspect they're just bare CPU cores without much of the fancy stuff like AVX support added on...

 

[Mussels]

Why such vitriol and hatred of E cores or even whatever Intel is doing by the AMD users? you aren't even using Intel, so what does it matter to you. It seems like just hatred of Intel to me. You chose to buy AMD so why spout so much hate for ADL/E cores.

However much power Intel with E cores uses, you don't have to use it, or cool it. So what if it is inefficient compared to what AMD is producing, you are not going to buy it or use it. I am so sick of hearing the same anti E core crap from near every AMD user on TPU.

My 12700k runs cool, and is as good as near any AMD CPU out right now (AM5 is not yet) The only reason i kinda regret it is i am sick of hearing the same shite every day on TPU.

Nah the problem is your warped views on what efficiency means
Your 12700k runs cool because you barely use it

If I capped my FPS to 60 my "hot" 5800x would barely hit 40C, but it'd be stupid of me to tell everyone that all 5800x run that cold or that those barely above idle figures were 'efficient' as if that's how the CPU always runs in every task.

I don't see why it matters. From where I sit, with my 2600k, the e-cores look pretty good if they are indeed as strong as a Skylake core. I know we've had many iterations of Skylake already, but if we can now get those down into low power "efficiency" cores, that seems pretty good... and they're stronger than my 2600k, anyway. Without knowing much about the design, I suspect they're just bare CPU cores without much of the fancy stuff like AVX support added on...

The concept is great, i'm all for mixed E and P cores - the problem is that intel are still going balls to the wall with their clock speeds past the efficiency settings entirely.
Intel Core i9-12900K E-Cores Only Performance Review - Power Consumption & Efficiency | TechPowerUp



Single threaded, the E-cores are great - chart topping.
The problem is they're used for multi threaded... and they fall behind.
These cores designed purely for efficiency are less efficient than Zen3, Zen 2, and even intels 9th and 10th gen CPUs.

They've been pushed too high on clocks and voltage to beat AMD in benchmarks, negating their entire purpose for existing.

 

[A Computer Guy]

I agree - this is why I like the theory that AMD will mix generations of cores to make E and P cores less costly to implement, and use older stock.
Zen 4 (P cores) + Zen 3 (E cores) for example

If they lock them in the area that Zen 3 has it's best power efficiency (lets say 4.4GHz) they'd make some fantastic E cores for no real investment from AMD - they already a plentiful supply of them, and they're already modular thanks to the CCX designs.

I also agree with ERazer - it all comes down to OS support.
Right now AMD use their chipset driver to set preferred cores, so they have a method of doing this (set all P cores to preferred cores, and most of the problem is solved)

I'm hoping in terms of OS support they can add some parameter per thread to alter/configure threads to automatically prefer or land on Best, Medium, or Low performance cores. Keep in mind this is different than Critical, High, Medium, or Low priority threads. I'm talking about CPU affinity per thread not process. It would take some time for software to be updated but once done software can better tune itself to core types (looking at your Corsair you boost killer) and in terms of gaming perhaps offer the end user a way to optimize their game threads without having to use something like process lasso. And then also enable Virtualization guests to translate that to host cores for the win.

 

[nguyen]

Nah the problem is your warped views on what efficiency means - and your rabid fanboyism.
Your 12700k runs cool because you barely use it

If I capped my FPS to 60 my "hot" 5800x would barely hit 40C, but it'd be stupid of me to tell everyone that all 5800x run that cold or that those barely above idle figures were 'efficient' as if that's how the CPU always runs in every task.


The concept is great, i'm all for mixed E and P cores - the problem is that intel are still going balls to the wall with their clock speeds past the efficiency settings entirely.
Intel Core i9-12900K E-Cores Only Performance Review - Power Consumption & Efficiency | TechPowerUp


Single threaded, the E-cores are great - chart topping.
The problem is they're used for multi threaded... and they fall behind.
These cores designed purely for efficiency are less efficient than Zen3, Zen 2, and even intels 9th and 10th gen CPUs.


They've been pushed too high on clocks and voltage to beat AMD in benchmarks, negating their entire purpose for existing.

So with E-cores disabled, 8P cores get 20K scores, that 2500points per P-core
8E cores take up the same die space as 2P cores, let say Intel make 10P cores, that would get them 25000 points, meanwhile 8P+8E get them 27700 points.

There are non-K version for people who prefer better effieciency at stock (much cheaper too), and have no clue how to tune their PC. Only idiots pay more for K version and not tune their PC to how they like it.

 

[lilhasselhoffer]

So...most of the hype is pretty stupid. You'll see interesting perspectives that e-cores are the future, that p-cores are the only thing that matters...and then in the same breathe people might tell you that RISC-V will trounce any x86-64 processor on the planet.

The simplest bit of this is that Intel and AMD both experiment...with some experiments being better than others. Starting off...Itanium. If anyone fanboys for Intel ask how that turned out.
Now, the AMD fanboys are equally as stupid. Ask them how Bulldozer worked, when each processor had to share a part of the pipeline. Read: terrible.


To the things at hand. You have got multiple core CPUs. Each of these cores is scheduled to perform different tasks...based upon software. E cores, or efficiency cores, are stripped down to do less intensive or otherwise lower demand processes. Think background processes with low priority. The catch is that in most cases there's very little reason for home users to have so many processes open and running that they'd see need for E cores.
Is this a tactic for Intel to claim that their processors have the most cores? I can't attest to their board meetings...but Intel used to sell itself on having the highest frequencies and greatest single thread performance. AMD has made huge leaps there...and Intel has made relatively little progress. By cost, throwing twice the cores onto a chip and claiming they are the best bang for the buck is...pretty easy to call a win. Given their history...I think Intel is trying for any win.


Now...the final bit. Is it worth having E cores? There are arguments for it...in the same way that ray tracing was the thing that sold Nvidia cards. In a very limited usage case, you could see relatively large performance gains. So...why do I think it's not ready for prime time? Well, right now if you're a common user a 5600x is a silly value for the money. It has the same core and thread count as a $600 CPU from 2011...with better connectivity, it's up-to-date, and at about 33% of that price (unadjusted for inflation). In more than a decade we've only managed to start using 6 cores fully on the consumer market...up from the 2 that was the style at the time.
If it isn't clear, my opinion is that the E cores lack integral scheduling to take advantage of them in anything but Windows 11, Windows 11 is currently suffering the every other version of Windows sucks, and even if it did you'd not expect things to be so clear.

Personally, I haven't done Intel products for a few years. They are treading water, and I think that the reliance on a software scheduler is going to make it very hard to justify E cores for anyone but power users who may be better off considering their workloads on different hardware. AMD fans should really be getting flashbacks...because this is basically the argument put forward for Bulldozer.
Most of my games and software library runs on 1-3 cores...based upon the windows scheduler. I run one program at high priority, maybe a second at equal, and a bunch of stuff that can be stacked on a nearly hibernative core and still be updated. I have no reason to believe that what Intel is putting out with an inflated core count will serve me better than AMD...and the fact that Intel doesn't seem to be basing their marketing material on that (I see about 5:1 them going after gamers above core count) is a tacit agreement. While my experience is anecdote, you can find your own by simply doing what you do and recording data inside the windows scheduler... If you're genuinely pinging all cores then it's probably worth adding more...assuming that you've got way more processes running than your core count.

 

[R-T-B]

I don't see why it matters. From where I sit, with my 2600k, the e-cores look pretty good if they are indeed as strong as a Skylake core. I know we've had many iterations of Skylake already, but if we can now get those down into low power "efficiency" cores, that seems pretty good... and they're stronger than my 2600k, anyway. Without knowing much about the design, I suspect they're just bare CPU cores without much of the fancy stuff like AVX support added on...

ecores support avx, avx2, etc. The only thing they lack is hyperthreading and avx512

 

[lilhasselhoffer]

View attachment 260079
So with E-cores disabled, 8P cores get 20K scores, that 2500points per P-core
8E cores take up the same die space as 2P cores, let say Intel make 10P cores, that would get them 25000 points, meanwhile 8P+8E get them 27700 points.

There are non-K version for people who prefer better effieciency at stock (much cheaper too), and have no clue how to tune their PC. Only idiots pay more for K version and not tune their PC to how they like it.

So...help me here.

You're responding to someone saying that you are not using your 12700k. Your response is that you disabled a large chunk of the silicon...thus literally disabling the components that this thread is meant to discuss.
You then link to a video that compare a 12900F and 12900k. One of the video's conclusions is that there's a 25% difference in power draw between the F and K SKUs...and the performance difference is 0-4%. So...the result is that you pay more for the k, you have a much higher power draw, and you have a boost that is functionally within the error for the testing methodology to be reasonably chalked-up to regular process variation. You then say you can get the performance of the k to that of the f by disabling the nice shiny new E cores...and experience an uplift by tuning...despite the literal cited video stating more frequency<>better performance???

You have to be trolling...right?


The alternative is that you are endorsing somebody spend extra money to get:
1) Slightly higher clocks at a much higher power draw.
2) No iGPU...so if you want quick sync that's gone. Of course,,,same die but binned so it's just dark silicon.
3) A much higher thermal envelope. 125/241 TDP versus a 65/202 watt TDP.
4) The ability to disable stuff...because there's nothing quite like paying extra for power windows then immediately ripping them out
All of this is "acceptable" because anybody who wants any ease in over/underclocking knows they have to spend more. If they do buy something...maybe for non-clocking reasons...in the k series they're idiots.



I'm really having a hard time swallowing this when the argument should be about the E-cores...which seem to be entirely unused in the one provided example. Of course, you could be a fanboy... You could be stating that somehow E cores aren't doing jack dandy based upon the numbers.

I mean; the 12600k has 6 P cores, is clocked 48% higher than the 12400f, which has 6 P cores, and in quad core calculations somehow only manages 20% faster speeds. If you instead compare octo-core... where the processors both either multi-thread or use E cores the gap becomes 33%...so a 48% clock increase against a chip with 4 more physical cores can still not manage to keep up with the huge increase in clock frequency.
CPU numbers


Consider me skeptical. I bought Sandy Bridge, I avoided Bulldozer. I did this because artificially swelling core counts was stupid. I'm buying Ryzen 3. I'm skipping big.little. It's the same stupidity. You're welcome to continue to feed the Intel machine. I'd prefer to vote with my wallet, and tell Intel to make steps forward rather than invest their money making a bad product that runs really fast. If they could integrate the scheduler as a hardware component this would be a different story...but it seems like Intel is relearning the AMD lesson from Bulldozer, software is king, and hardware without software is blowing money on nothing.





Let me TL;DR this.
My worst boss continues to tell me "It's about how fast we can make the car go." In this case, an analog for how many units of a thing we can make. That's...cool. The problem is that's 1950's thinking.
It's not about how fast the car can go, it should be about how far we can run on a tank of gas. The analog there is that the production of parts has to be metered by how tightly we control inputs, how efficiently we can run, and how we balance resources to be as profitable as possible.

Intel wants to sell more, and use that gas. They're running out 10 miles into a 50 mile race, 10 minutes in. AMD has given up on clocking to the moon, and is less about single thread performance. They make it 30 miles into the race, at 25 minutes. Arm is a diesel vehicle. They theoretically could go the 50 miles, but cannot enter the race because the world isn't yet ready for diesel. AMD is not the solution. It is not finishing the race. It is not the fastest. Thing is, we pick from what we have...and right now AMD is the best option for most. The only way we make Intel better is to vote with our wallets, and force them to either fundamentally redesign their engine (what AMD did with Ryzen), or compete by undercutting on price.

To extend the metaphor just slightly, E cores are making an electric hybrid vehicle. Theoretically they are powerful...but when you have to contain 2 drive trains and the brains to make them work there's precious little that actually makes them work better...even if you feel better.

 

[nguyen]

So...help me here.

You're responding to someone saying that you are not using your 12700k. Your response is that you disabled a large chunk of the silicon...thus literally disabling the components that this thread is meant to discuss.
You then link to a video that compare a 12900F and 12900k. One of the video's conclusions is that there's a 25% difference in power draw between the F and K SKUs...and the performance difference is 0-4%. So...the result is that you pay more for the k, you have a much higher power draw, and you have a boost that is functionally within the error for the testing methodology to be reasonably chalked-up to regular process variation. You then say you can get the performance of the k to that of the f by disabling the nice shiny new E cores...and experience an uplift by tuning...despite the literal cited video stating more frequency<>better performance???

You have to be trolling...right?

The alternative is that you are endorsing somebody spend extra money to get:
1) Slightly higher clocks at a much higher power draw.
2) No iGPU...so if you want quick sync that's gone. Of course,,,same die but binned so it's just dark silicon.
3) A much higher thermal envelope. 125/241 TDP versus a 65/202 watt TDP.
4) The ability to disable stuff...because there's nothing quite like paying extra for power windows then immediately ripping them out
All of this is "acceptable" because anybody who wants any ease in over/underclocking knows they have to spend more. If they do buy something...maybe for non-clocking reasons...in the k series they're idiots.


I'm really having a hard time swallowing this when the argument should be about the E-cores...which seem to be entirely unused in the one provided example. Of course, you could be a fanboy... You could be stating that somehow E cores aren't doing jack dandy based upon the numbers.

I mean; the 12600k has 6 P cores, is clocked 48% higher than the 12400f, which has 6 P cores, and in quad core calculations somehow only manages 20% faster speeds. If you instead compare octo-core... where the processors both either multi-thread or use E cores the gap becomes 33%...so a 48% clock increase against a chip with 4 more physical cores can still not manage to keep up with the huge increase in clock frequency.
CPU numbers


Consider me skeptical. I bought Sandy Bridge, I avoided Bulldozer. I did this because artificially swelling core counts was stupid. I'm buying Ryzen 3. I'm skipping big.little. It's the same stupidity. You're welcome to continue to feed the Intel machine. I'd prefer to vote with my wallet, and tell Intel to make steps forward rather than invest their money making a bad product that runs really fast. If they could integrate the scheduler as a hardware component this would be a different story...but it seems like Intel is relearning the AMD lesson from Bulldozer, software is king, and hardware without software is blowing money on nothing.


Let me TL;DR this.
My worst boss continues to tell me "It's about how fast we can make the car go." In this case, an analog for how many units of a thing we can make. That's...cool. The problem is that's 1950's thinking.
It's not about how fast the car can go, it should be about how far we can run on a tank of gas. The analog there is that the production of parts has to be metered by how tightly we control inputs, how efficiently we can run, and how we balance resources to be as profitable as possible.

Intel wants to sell more, and use that gas. They're running out 10 miles into a 50 mile race, 10 minutes in. AMD has given up on clocking to the moon, and is less about single thread performance. They make it 30 miles into the race, at 25 minutes. Arm is a diesel vehicle. They theoretically could go the 50 miles, but cannot enter the race because the world isn't yet ready for diesel. AMD is not the solution. It is not finishing the race. It is not the fastest. Thing is, we pick from what we have...and right now AMD is the best option for most. The only way we make Intel better is to vote with our wallets, and force them to either fundamentally redesign their engine (what AMD did with Ryzen), or compete by undercutting on price.

To extend the metaphor just slightly, E cores are making an electric hybrid vehicle. Theoretically they are powerful...but when you have to contain 2 drive trains and the brains to make them work there's precious little that actually makes them work better...even if you feel better.

I have no idea what you are arguing about except that you want to argue, I don't have a 12700K

So you are ignored now

 

[tabascosauz]

Sounds like some of y'all need to re-read the first line of the OP:

PLEASE, NO INTEL/AMD FANBOY ANTAGONISM!

What is your opinion on intels motivation for introducing e-cores.

It's about e-cores, not whether your ego is in bed with AMD or Intel. No reason to be playing the fanboy card at all. Lay off the insults or thread is closed.

 

[ExcuseMeWtf]

I have specifically went for i3-12100F to avoid E-cores, because I am still using Windows 10, AFAIK there are no planned improvements for scheduling to account for those in that system.

 

[AM4isGOD]

Whatever the reason for Intel introducing the E cores, they greatly increase the MT performance which is surely a good thing. Imo there is no doubt, when the scheduler is working correctly, they will pay off. As i have said before, which some people ignore, is it is a waste running background tasks on a P core. When you are gaming surely you only want your game running exclusively on your P cores and not all the other background crap.

With a AMD CPU with P cores only, when you are gaming, all the other crap tasks you have running are using your P cores which will surely have a detrimental effect on your games performance.

Admittedly the E cores are probably not being managed perfectly at the moment, but at some point they will be, then when Intel users are gaming, their background crap tasks will run properly on the E cores, whereas AMD users will have all them same tasks eating game performance on the P core only CPU.

 

[ThrashZone]

Hi,
Doubt anyone would run out of gaming resources using an 8 core and I'd probably call this a main stream core count now days.

 

[Easy Rhino]

The global cabal is hell bent on getting off of oil and gas and therefore energy prices are expected to skyrocket. Consumers won't be purchasing that new phone every year or new PC every 3-4 years, and enterprise won't be replacing servers every 4-5 years if their power bills are all double and tripling.

 

[R0H1T]

ecores support avx, avx2, etc. The only thing they lack is hyperthreading and avx512

And they won't have AVX512 anytime soon, unless they follow the AMD approach. Not to mention essentially neutering AVX512 that's actually present on the P cores! Then there's switching the tasks between various cores & of course priority, which is less of a problem if the OS can handle it properly. Right now Intel is just throwing (E)cores for namesakes & ironically to counter AMD with "more cores"

 

[P4-630]

AVX512

Isn't used for gaming anyway.

 

[dgianstefani]

Nah the problem is your warped views on what efficiency means
Your 12700k runs cool because you barely use it

If I capped my FPS to 60 my "hot" 5800x would barely hit 40C, but it'd be stupid of me to tell everyone that all 5800x run that cold or that those barely above idle figures were 'efficient' as if that's how the CPU always runs in every task.



The concept is great, i'm all for mixed E and P cores - the problem is that intel are still going balls to the wall with their clock speeds past the efficiency settings entirely.
Intel Core i9-12900K E-Cores Only Performance Review - Power Consumption & Efficiency | TechPowerUp



Single threaded, the E-cores are great - chart topping.
The problem is they're used for multi threaded... and they fall behind.
These cores designed purely for efficiency are less efficient than Zen3, Zen 2, and even intels 9th and 10th gen CPUs.
View attachment 260076


They've been pushed too high on clocks and voltage to beat AMD in benchmarks, negating their entire purpose for existing.

They're only designed for area efficiency, power efficiency is a second priority.

 

[Ellertis]

I agree - this is why I like the theory that AMD will mix generations of cores to make E and P cores less costly to implement, and use older stock.
Zen 4 (P cores) + Zen 3 (E cores) for example

If they lock them in the area that Zen 3 has it's best power efficiency (lets say 4.4GHz) they'd make some fantastic E cores for no real investment from AMD - they already a plentiful supply of them, and they're already modular thanks to the CCX designs.

I also agree with ERazer - it all comes down to OS support.
Right now AMD use their chipset driver to set preferred cores, so they have a method of doing this (set all P cores to preferred cores, and most of the problem is solved)

The leaks suggest Zen5(P) + zen4c(E)

 

[TheoneandonlyMrK]

They're only designed for area efficiency, power efficiency is a second priority.

And that's why they're shit and implemented shit IMHO.

They're should be no more than 8 E core's.

They should be actually efficient.

They should be doing 95% of all the work.

With only time sensitive apps pushed to P core's.

But they are not , do not.

Why because they're only there to increase the core count number on the spec sheet.

I got only love for efficiency core's, bit this tat ain't that.

Opinion over.

 

[P4-630]

Mobile phones have it at least for a decade already.....
I wouldn't be surprised if the next gen game consoles will also come with "efficiency" and "performance" cores....

 

[dgianstefani]

And that's why they're shit and implemented shit IMHO.

They're should be no more than 8 E core's.

They should be actually efficient.

They should be doing 95% of all the work.

With only time sensitive apps pushed to P core's.

But they are not , do not.

Why because they're only there to increase the core count number on the spec sheet.

I got only love for efficiency core's, bit this tat ain't that.

Opinion over.

Everything on that list can be fixed with kernel scheduling, software optimisation and bios limits. The hardware is fine.

 

[ThrashZone]

Hi,
Yeah the new hedt chip.

Issue is waiting for ms to optimize it
They can't even design a decent start menu