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

[AusWolf]

Lol what??

90c is just the temp reading.

Meaning the cooler is Not dissipating well 140w.

If the same cpu is 140w, but the temp is 80c, the cooler is dissipating heat better.

If your Tcase temp is oven hot, give up on PCs and find a different hobby.

It's not jsut the cooler. There's also the die's surface area, contact with the IHS, the IHS itself, and to a minimal extent, the thermal paste. Dissipating 140 W off a large, monolithic Intel chip is not the same as dissipating 140 W off a small AMD chiplet. The same power flowing through a smaller area will naturally make it hotter. You're right about the same CPU situation, but we're not talking about the same CPU here.

 

[ThrashZone]

Hi,
Single thread skylake is all e-threads are
If it weren't for the real p-core frequencies the new design would be laughed out of existence.

 

[P4-630]

Hi,
Single thread skylake is all e-threads are
If it weren't for the real p-core frequencies the new design would be laughed out of existence.

There's a 144-E-core Xeon coming soon.

Intel's 144 core Xeon 'Sierra Forest' chips are coming in the first half of 2024

Intel's roadmap has been confirmed.

www.windowscentral.com

 

[TheoneandonlyMrK]

There's a 144-E-core Xeon coming soon.

Intel's 144 core Xeon 'Sierra Forest' chips are coming in the first half of 2024

Intel's roadmap has been confirmed.

www.windowscentral.com

When you think about it is this that far from Knights ferry, with no SMT or Avx512, it's very similar and that went well.

I would expect this to compete more with arm storage array servers personally then anything AMD has out in that moment.

 

[Deleted member 185158]

It's not jsut the cooler. There's also the die's surface area, contact with the IHS, the IHS itself, and to a minimal extent, the thermal paste. Dissipating 140 W off a large, monolithic Intel chip is not the same as dissipating 140 W off a small AMD chiplet. The same power flowing through a smaller area will naturally make it hotter. You're right about the same CPU situation, but we're not talking about the same CPU here.

140w is 140w. Doesn't matter what CPU it comes out of.

Dissipating it from a smaller surface area is more difficult?? Is that all we are saying here?

 

[Lei]

140w is 140w. Doesn't matter what CPU it comes out of.

Dissipating it from a smaller surface area is more difficult?? Is that all we are saying here?

Giving same heat (watt) to a smaller object of the same material (silicon die), makes it hotter.
Temp rises faster when mass is smaller. Q=mcΔt

 

[GerKNG]

140w is 140w. Doesn't matter what CPU it comes out of.

Dissipating it from a smaller surface area is more difficult?? Is that all we are saying here?

Watt/mm²
a small Zen 4 Die with very high density is basically impossible to cool properly. even direct die cooling is not that great.
a huge 14nm 10900k can run at ~50°C at the same wattage/cooler.

 

[Vario]

Giving same heat (watt) to a smaller object of the same material (silicon die), makes it hotter.
Temp rises faster when mass is smaller. Q=mcΔt

Yep, DeltaT = q/mc

Delta T = change in temperature (degree C)
q= heat (Joules = watts/second)
m = mass (grams)
c = specific heat (a constant for the particular type of material, in this case silicon would be 0.7 Joules/(gram*degreeC)

Solving for DeltaT, we show that as mass increases (denominator becomes smaller) the delta T decreases.

 

[Space Lynx]

e-cores are for synthetic benchmark hard-ons oh and the single game that uses them well, spider man, even though no other game utilizes them. lol

13600k is a good chip though, I somewhat regret selling it, but at end of day for me to enjoy 13600k I would have to disable e-cores, which at the time just seemed like I was wasting my own money since I wasn't using the full chip.

total war warhammer had 81% e-core usage and wasn't using my p-cores in some battle scenes... that is the only game I came across though that didn't use the p-cores right. still it pissed me off.

fuck e-cores.

 

[Deleted member 185158]

Yep, DeltaT = q/mc

Delta T = change in temperature (degree C)
q= heat (Joules = watts/second)
m = mass (grams)
c = specific heat (a constant for the particular type of material, in this case silicon would be 0.7 Joules/(gram*degreeC)

Solving for DeltaT, we show that as mass increases (denominator becomes smaller) the delta T decreases.

So don't cool a 5800x with a Hyper 212.

Gotcha!

 

[JustBenching]

e-cores are for synthetic benchmark hard-ons oh and the single game that uses them well, spider man, even though no other game utilizes them. lol

13600k is a good chip though, I somewhat regret selling it, but at end of day for me to enjoy 13600k I would have to disable e-cores, which at the time just seemed like I was wasting my own money since I wasn't using the full chip.

total war warhammer had 81% e-core usage and wasn't using my p-cores in some battle scenes... that is the only game I came across though that didn't use the p-cores right. still it pissed me off.

fuck e-cores.

Lots of games use ecores actually. Spiderman cyberpunk, warzone and now the last of us. The game chokes on normal 6 core cpus but on a 12600k it plays.great cause of the ecores

 

[evernessince]

Watt/mm²
a small Zen 4 Die with very high density is basically impossible to cool properly. even direct die cooling is not that great.
a huge 14nm 10900k can run at ~50°C at the same wattage/cooler.

Power density is definitely an issue but you picked the absolute worst reasoning. Zen 4, or specifically the X class CPUs, run hot because they are designed to boost until 95c to extract maximum performance, not because a CPU with Zen 4's power density is impossible to cool. If you take a look at non-x Zen 4 CPU reviews, those CPUs are very efficient and easy to cool. For example:

As you can see, even with the stock cooler there's less than 1% of a difference vs a higher end cooler. In addition, temperatures are low both with the stock and more expensive cooler.

 

[JustBenching]

Power density is definitely an issue but you picked the absolute worst reasoning. Zen 4, or specifically the X class CPUs, run hot because they are designed to boost until 95c to extract maximum performance, not because a CPU with Zen 4's power density is impossible to cool. If you take a look at non-x Zen 4 CPU reviews, those CPUs are very efficient and easy to cool. For example:

View attachment 290155

View attachment 290157

As you can see, even with the stock cooler there's less than 1% of a difference vs a higher end cooler. In addition, temperatures are low both with the stock and more expensive cooler.

What? That is just wrong. Zen 4 are not "designed" to boost to 95c. They work just any other cpu ever did in the history of cpus. They boost until they hit either a power or a temperature limit. That's literally how any other cpu works. But because they are actually HARD to cool, they reach a temperature limit before they reach the power limit.

Saying zen 4 is easy to cool is just factually wrong. Their heat transfer is bad both because of ihs and big density.

 

[AusWolf]

What? That is just wrong. Zen 4 are not "designed" to boost to 95c. They work just any other cpu ever did in the history of cpus. They boost until they hit either a power or a temperature limit. That's literally how any other cpu works. But because they are actually HARD to cool, they reach a temperature limit before they reach the power limit.

Saying zen 4 is easy to cool is just factually wrong. Their heat transfer is bad both because of ihs and big density.

That's right. With good enough cooling, even a Zen 4 CPU will hit its power limit before it hits its temperature limit.

 

[evernessince]

What? That is just wrong. Zen 4 are not "designed" to boost to 95c. They work just any other cpu ever did in the history of cpus. They boost until they hit either a power or a temperature limit. That's literally how any other cpu works. But because they are actually HARD to cool, they reach a temperature limit before they reach the power limit.

Not sure if you are aware but 95c is the target temperature of Zen 4, aka the temperature the CPU will advantageously boost to. You can verify this in every review of the product as it will continue to boost until it hits that temperature and will scale frequency with that value in mind.

Historically speaking CPUs relied upon a frequency and voltage table and would throttle back down to a lower frequency state if they hit a thermal limit. Boosting is relatively new to the CPU space, which started around 2009 with Intel turbo boost. To say that CPUs always boosted is patently false and also ignores the difference in how AMD's boosting algorithm or how boosting algorithms work in general.

If you take a look at AMD's boosting algorithm for the 3000 series vs the competing Intel products at the time, AMD is able to update it's boost algorithm with the 3000 series once every 2 ns. If I recall correctly Intel was something like one update every 32nm.

That some Zen 4 processors, specifically the X class, frequently sit at 95c tells us two things. One, AMD's boost algorithm is taking full advantage of thermal headroom. Two, Zen 4 CPUs have a target temp of 95c. Any use of that value to imply that they are hard to cool would be despite the plentiful reviews like ones the graphics above were pulled from, clearly showing that Zen 4 with far less cooling grunt that competing high end Intel processors.

Saying zen 4 is easy to cool is just factually wrong. Their heat transfer is bad both because of ihs and big density.

The IHS narrative is just an assumption without anything to back it up from the reddit community and surely your reddit friends know more than AMD engineers on optimal die thickness.

 

[JustBenching]

Not sure if you are aware but 95c is the target temperature of Zen 4, aka the temperature the CPU will advantageously boost to. You can verify this in every review of the product as it will continue to boost until it hits that temperature and will scale frequency with that value in mind.

Historically speaking CPUs relied upon a frequency and voltage table and would throttle back down to a lower frequency state if they hit a thermal limit. Boosting is relatively new to the CPU space, which started around 2009 with Intel turbo boost. To say that CPUs always boosted is patently false and also ignores the difference in how AMD's boosting algorithm or how boosting algorithms work in general.

If you take a look at AMD's boosting algorithm for the 3000 series vs the competing Intel products at the time, AMD is able to update it's boost algorithm with the 3000 series once every 2 ns. If I recall correctly Intel was something like one update every 32nm.

That some Zen 4 processors, specifically the X class, frequently sit at 95c tells us two things. One, AMD's boost algorithm is taking full advantage of thermal headroom. Two, Zen 4 CPUs have a target temp of 95c. Any use of that value to imply that they are hard to cool would be despite the plentiful reviews like ones the graphics above were pulled from, clearly showing that Zen 4 with far less cooling grunt that competing high end Intel processors.



The IHS narrative is just an assumption without anything to back it up from the reddit community and surely your reddit friends know more than AMD engineers on optimal die thickness.

I'll start from the end.

When I say AMD cpus are hard to cool, I mean that at similar wattages, they need a much beefier cooler. That is a fact - whoever has used both cpus can testify to that. It's not something that is contestable, at all. Put a 7950x and a 13900k at 200 watts with the same cpu cooler, the intel part will have much lower temperatures.

Now onto the first part, saying zen 4 are designed to constantly boost until 95c is basically absurd. If that was true, then if i just put my pc in the freezer zen 4 will hit 9 ghz. It won't. It boosts the same way a 13900k or any other cpu does. But exactly because Intel cpus are easier to cool, they can hit their power limit before their temperature limit. On the other hand, because zen 4 are harder to cool, they will their 95c temperature limit before their power limit.

Put a small sff cooler to a 13900k and it will behave just like a zen 4. It will boost until it reaches 100c. That doesn't mean it was designed to boost to 100c,it means your cooling isn't enough to hit the power limit so you are reaching the thermal limit instead.

 

[evernessince]

Now onto the first part, saying zen 4 are designed to constantly boost until 95c is basically absurd. If that was true, then if i just put my pc in the freezer zen 4 will hit 9 ghz. It won't. It boosts the same way a 13900k or any other cpu does. But exactly because Intel cpus are easier to cool, they can hit their power limit before their temperature limit. On the other hand, because zen 4 are harder to cool, they will their 95c temperature limit before their power limit.

Designed to target a max of 95c, assuming other boost variables allow as I stated earlier. Please read my comment before replying.

As I also pointed out earlier, the non-X and 3D cache zen 4 chips do not hit thermal limits first. I'd also argue that the 13900K and 13700K absolutely are going to hit thermal limits in the vast majority of cases before power limits. You keep saying "Zen 4" as if it's across the board but that's obviously and clearly incorrect. Depending on which Zen 4 CPU you are talking about, the power consumption and cooling requirements change drastically. Even at the same power and performance level, the cooling requirements can change because X class CPUs have a more aggressive voltage curve and the boost is more aggressive than non-x CPUs so really "easier to cool" is an extremely complicated topic.

When I say AMD cpus are hard to cool, I mean that at similar wattages, they need a much beefier cooler. That is a fact - whoever has used both cpus can testify to that. It's not something that is contestable, at all. Put a 7950x and a 13900k at 200 watts with the same cpu cooler, the intel part will have much lower temperatures.

If it's a fact, you should have no issue providing a reliable source. That said I do find the premise that a large number of people have a bunch of Raptor Lake and Zen 4 CPUs laying around and that they decided to power limit them all the same arbitrary wattage and test them all with the same CPU hard to believe. It's a tall task for reviewers to test on a variety of CPUs and coolers in a selection of applications. On top of that, it's not really clear what the metric for "easy to cool" is or what TDP you'd limit each CPU to. Are you going to test every Zen 4 and Raptor Lake CPU with 5 different CPU coolers in at least 12 games at 5 different TDPs? Good luck, see you next year. Given that each CPU has a different voltage curve and power consumption, what TDP do you pick when different CPUs will inherently be more efficient at different wattages. The results will vary wildly depending on the CPUs, coolers, power limit chosen, and games tested among other factors. I could easily create a graph showing Intel or AMD being "easier to cool" just based on CPU selection alone, forget about the other factors.

 

[AusWolf]

Not sure if you are aware but 95c is the target temperature of Zen 4, aka the temperature the CPU will advantageously boost to. You can verify this in every review of the product as it will continue to boost until it hits that temperature and will scale frequency with that value in mind.

With the example of the 7700X, it has a thermal limit of 95 °C, a power limit of 142 W, and a boost/PBO limit of 5.55 GHz. Whichever hits first will stop the CPU from boosting further. With a 280 mm AIO, I hit the frequency limit of 5.55 GHz in Cinebench single-core, but I hit the power limit of 142 W in multi-core, while the CPU cores work at 91-92 °C max. I never reach 95 °C on it, ever (or maybe I will in the summer, we'll see).

Zen 4 does not boost to 95 °C if your cooler allows it to hit its power target, the same way a 13900K won't boost to 100 °C if it can reach its max boost while staying under the temperature limit. Simple as. The fact that 95 °C on Zen 4 is completely safe is a different story.

 

[Space Lynx]

Lots of games use ecores actually. Spiderman cyberpunk, warzone and now the last of us. The game chokes on normal 6 core cpus but on a 12600k it plays.great cause of the ecores

well I don't like losing fps in total war warhammer battles because the scheduler can't decide which is the p-core or not, so no Intel for me.

 

[JustBenching]

If it's a fact, you should have no issue providing a reliable source. That said I do find the premise that a large number of people have a bunch of Raptor Lake and Zen 4 CPUs laying around and that they decided to power limit them all the same arbitrary wattage and test them all with the same CPU hard to believe. It's a tall task for reviewers to test on a variety of CPUs and coolers in a selection of applications. On top of that, it's not really clear what the metric for "easy to cool" is or what TDP you'd limit each CPU to. Are you going to test every Zen 4 and Raptor Lake CPU with 5 different CPU coolers in at least 12 games at 5 different TDPs? Good luck, see you next year. Given that each CPU has a different voltage curve and power consumption, what TDP do you pick when different CPUs will inherently be more efficient at different wattages. The results will vary wildly depending on the CPUs, coolers, power limit chosen, and games tested among other factors. I could easily create a graph showing Intel or AMD being "easier to cool" just based on CPU selection alone, forget about the other factors.

Sure, you can check TPUs CPU cooler reviews. Granted, they are testing Zen 3xxx against cometlake, but things havent changed. You can see coolers that absolutely work fine on Intel - fail on AMD.

I can also post you my results with a u12a on a 13900k. It can actually cool 330 watts without thermal throttling. It can't do that on any zen 4 chip. I don't know why you keep insisting on this, it's beyond a reasonable doubt the case., youll see up to 20c difference between an intel and an amd on same wattages.

Designed to target a max of 95c, assuming other boost variables allow as I stated earlier. Please read my comment before replying.

Well then okay - and intel cpus are designed to hit 100c.. What does that even mean lol

well I don't like losing fps in total war warhammer battles because the scheduler can't decide which is the p-core or not, so no Intel for me.

You can press your scroll lock key on your keyboard and it parks the ecores. Happy gaming

With the example of the 7700X, it has a thermal limit of 95 °C, a power limit of 142 W, and a boost/PBO limit of 5.55 GHz. Whichever hits first will stop the CPU from boosting further. With a 280 mm AIO, I hit the frequency limit of 5.55 GHz in Cinebench single-core, but I hit the power limit of 142 W in multi-core, while the CPU cores work at 91-92 °C max. I never reach 95 °C on it, ever (or maybe I will in the summer, we'll see).

Zen 4 does not boost to 95 °C if your cooler allows it to hit its power target, the same way a 13900K won't boost to 100 °C if it can reach its max boost while staying under the temperature limit. Simple as. The fact that 95 °C on Zen 4 is completely safe is a different story.

Exactly. And the fact that you hit 90+c at 142w is a testament to what im saying. They are REALLY hard to cool. An intel CPU with a 240aio would be happily sitting at 70 or less at that wattage. Heck, mine does on an aircooler

As a proof of concept,, this is a u12a on a 13900k at 260 watts.

 

[oxrufiioxo]

Man I can't believe this thread is still alive

At this point if you're ok with E cores there is an option for you if you don't like E cores there is an option for you.

At this point everyone knows why intel cpu have e cores and for the most part they work quite well better even than amds implementation of 3D V$ on the 7950X it would seem but maybe about the same if we consider 12th gens launch vs 7000X3D somtimes it works well other times youre disabling stuff to get it working properly.

I personally wouldn't buy a cpu with E cores but I also wouldn't buy a 3D V$ cpu that only has it on one CCD life goes on.

 

[AusWolf]

Exactly. And the fact that you hit 90+c at 142w is a testament to what im saying. They are REALLY hard to cool. An intel CPU with a 240aio would be happily sitting at 70 or less at that wattage. Heck, mine does on an aircooler

As a proof of concept,, this is a u12a on a 13900k at 260 watts.

Yeah, and it also proves that Zen 4 doesn't necessarily hit 95 °C if it doesn't have to. I can hit the 142 W power limit on my 7700X without ever reaching 95 °C under any load. The fact that you don't need a 280 mm AIO to do the same with an Intel CPU is just the icing on the cake.

 

[JustBenching]

Yeah, and it also proves that Zen 4 doesn't necessarily hit 95 °C if it doesn't have to. I can hit the 142 W power limit on my 7700X without ever reaching 95 °C under any load. The fact that you don't need a 280 mm AIO to do the same with an Intel CPU is just the icing on the cake.

People usually don't understand what easier to cool means. 13900k is much easier to cool than a 12900k. I even know exactly how much easier. My 12900k taps out at 280 watts roundabout, hitting 95c. My 13900k hits 95c at 330 watts under the same cooler.

I don't think you can cool a zen 4 part at 330 watts unless you use something exotic like moras on the balcony.

 

[evernessince]

With the example of the 7700X, it has a thermal limit of 95 °C, a power limit of 142 W, and a boost/PBO limit of 5.55 GHz. Whichever hits first will stop the CPU from boosting further. With a 280 mm AIO, I hit the frequency limit of 5.55 GHz in Cinebench single-core, but I hit the power limit of 142 W in multi-core, while the CPU cores work at 91-92 °C max. I never reach 95 °C on it, ever (or maybe I will in the summer, we'll see).

Zen 4 does not boost to 95 °C if your cooler allows it to hit its power target, the same way a 13900K won't boost to 100 °C if it can reach its max boost while staying under the temperature limit. Simple as. The fact that 95 °C on Zen 4 is completely safe is a different story.

None of which contradicts what I said. As I said twice "Designed to target a max of 95c, assuming other boost variables allow as I stated earlier".

Sure, you can check TPUs CPU cooler reviews. Granted, they are testing Zen 3xxx against cometlake, but things havent changed. You can see coolers that absolutely work fine on Intel - fail on AMD.

They also are CPU cooler reviews and not a analysis of how easy a given chip is to cool. None of the prior metentioned variables or questions I mentioned earlier are answered by CPU cooler reviews

I can also post you my results with a u12a on a 13900k. It can actually cool 330 watts without thermal throttling. It can't do that on any zen 4 chip. I don't know why you keep insisting on this, it's beyond a reasonable doubt the case., youll see up to 20c difference between an intel and an amd on same wattages.

What are you talking about, the u12a can adequately cool every Zen 4 chip. To say is can't do it for ANY zen 4 chip is obviously false.

People usually don't understand what easier to cool means.

People, specifically you, don't seem to understand that there's no precise definition of what "easier to cool" means. There's no official metric to define your subjective opinion.

 

[JustBenching]

None of which contradicts what I said. As I said twice "Designed to target a max of 95c, assuming other boost variables allow as I stated earlier".

Well if you mean the same way Intel is designed to hit 100c, sure. But that's my point all along, zen 4 works exactly like any other cpu

What are you talking about, the u12a can adequately cool every Zen 4 chip. To say is can't do it for ANY zen 4 chip is obviously false.

Of course it can cool every zen 4 chip, can it cool 330 watts on a zen 4 chip though? Seeing how Auswolf is already at 90+C with a 280 AIO at 140 watts, id probably say no?

People, specifically you, don't seem to understand that there's no precice defintion of what "easier to cool" means. There's no official metric to define your subjective opinion.

Okay, let me put a different way. Intel has much better heat transfer than AMD. Those were exactly my words on the previous page.