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

[Vayra86]

LOL yeah let play pretend that Intel non-K series don't exist , as those non-K model require zero tweak to achive great performance and efficiency, and then some people with 5800X come in and say they can tweak their 5800X to achieve better efficiency

If you don't know how to make your equipments work as you intended, the problem is on you.

The point from your own review was that you needed to adjust stock boost behaviour and manually limit the CPU to 65W out of the box, buddy. And we're full circle now, thanks.

Just so we don't get things mixed up, here:

"When using the RM1 box cooler without any power limits, the 12700 was thermally limited to a score of 19714 points. That's an 8% reduction when compared to what we saw with the Corsair H170i.

With the 65w spec enforced, the score dropped to 16017 points, which is a similar level of performance to that of the Ryzen 7 5800X and Core i9-10900K. A very respectable result given how little power the 12700 is using here."

See? No discussion about efficiency being possibly great on ADL, nobody is fighting that battle. But it does require a tweak.

 

[nguyen]

The point from your own review was that you needed to adjust stock boost behaviour and manually limit the CPU to 65W out of the box, buddy. And we're full circle now, thanks.

Just so we don't get things mixed up, here:

"When using the RM1 box cooler without any power limits, the 12700 was thermally limited to a score of 19714 points. That's an 8% reduction when compared to what we saw with the Corsair H170i.

With the 65w spec enforced, the score dropped to 16017 points, which is a similar level of performance to that of the Ryzen 7 5800X and Core i9-10900K. A very respectable result given how little power the 12700 is using here."

See? No discussion about efficiency being possibly great on ADL, nobody is fighting that battle. But it does require a tweak.

How do you know that 65W spec enforced is not the default setting on every motherboard? guessing?

 

[Vayra86]

How do you know that 65W spec enforced is not the default setting on every motherboard? guessing?

Again from techspot:

"Non-K parts on the other hand, like the Core i7-12700 which are technically 65W parts -- but also not really -- still suffer from a loosely defined spec, which can see them clock as low as 2.1 GHz on the P-cores for AVX workloads, though without power limits should maintain an all-core frequency of 4.5 GHz, which is a 114% increase over the base frequency."

This is the point. We've seen it before with Intel and motherboard vendors where loosely defined specs led to unwanted behaviour. Back then it was about Z-boards. Now we're talking about non-OC boards suffering that situation.

PSA: Don't Buy This Asrock Motherboard

In testing new Intel B660 motherboards for an upcoming VRM roundup, we felt compelled to stop and look at the Asrock B660M-HDV, only because Asrock is at...

www.techspot.com

 

[nguyen]

Again from techspot:

"Non-K parts on the other hand, like the Core i7-12700 which are technically 65W parts -- but also not really -- still suffer from a loosely defined spec, which can see them clock as low as 2.1 GHz on the P-cores for AVX workloads, though without power limits should maintain an all-core frequency of 4.5 GHz, which is a 114% increase over the base frequency."

This is the point. We've seen it before with Intel and motherboard vendors where loosely defined specs led to unwanted behaviour. Back then it was about Z-boards. Now we're talking about non-OC boards suffering that situation.

PSA: Don't Buy This Asrock Motherboard

In testing new Intel B660 motherboards for an upcoming VRM roundup, we felt compelled to stop and look at the Asrock B660M-HDV, only because Asrock is at...

www.techspot.com

Yawn...some boards enforce power limit by default and some don't, so again if you don't know how to make your equipment works as intented, the problem is on you. I'm not going to be hypocritical and talk like the average PC DIYers are so clueless they don't know what TDP their CPUs are running at...or that they can't read reviews before buying something

 

[dragontamer5788]

This is the trouble i've been having discussing this on the forums (and keep getting called a fanboy, a shill, etc etc)
If you need to change almost every aspect of the stock behaviour, that's not some elitist awesome superduper thing to brag about - it's a sign the product is bad.


No one argues intel don't have great single threaded performance - the problem is that they're making you pay for more and more terrible E-cores to get it, and then to sustain it you need to disable the E-cores to keep the TDP down...

Not necessarily?

The P vs E core thing, much like big.LITTLE for the ARM world, has physics-level principles that make sense. The software doesn't quite exist yet to fully take advantage of P vs E cores, meaning its now the user's job to make up the difference. Windows always had good, low-level, controls for which cores get affinity to which processes/threads. Yes, it takes manual effort, but an advanced user can say "these threads use E-cores, those threads use P-cores".

For pure single-threaded performance, you're likely correct that disabling the E-cores is best.

However, for multi-threaded performance, as well as power-efficient background processes, the E-cores will be best. We know this because its already true in the big.LITTLE ARM world.

Intel tried to make automagical thread configuration, and so did Microsoft Windows. Alas, there's only so much "auto-configuration" that you can do. At the end of the day, its the programmer/developer's responsibility to set sane defaults (and barring that, it becomes the IT Administrator's job to set up the threads/processes correctly). Since the typical modern computer user doesn't have an IT degree or understanding, they will inevitably set these knobs to an inefficient setting.

I don't think that makes this whole exercise useless however. Its just the nature of a 1st generation product of a very complicated scheduling/OS level problem. The real issue is that Microsoft / Intel needed to be working on this 10-years ago when ARM made their first big.LITTLE chips.

Especially because of Apple's advancements with M1 and M2, with 2-day battery lives and whatnot. People are beginning to notice the poor power-efficiency of Intel machines (and AMD) over Apple these days. It will be worthwhile to perform a large, optimization pass over the software run on laptops to allow for multi-day computing on one battery charge... it hasn't happened yet though. First, the hardware needs to come out (success: P vs E cores), then the software will change. There's no other way for this to move forward.

 

[Mikael Andersson]

Anecdotal but my 12900K system is the smoothest and most responsive system I have ever had, this includes my previous 10900K and 3900X machines, So i guess the E-Cores are doing a good job at background tasks.

That's the advantage of having sixteen cores and twenty-four threads.

 

[dirtyferret]

This is the trouble i've been having discussing this on the forums (and keep getting called a fanboy, a shill, etc etc)

I would never call you a fanboy or shill; I would simply say you are best with a little white wine, shallots, and frites. I will say reading the whole discussion, it's hard for people to understand some of the points made

If you need to change almost every aspect of the stock behaviour, that's not some elitist awesome superduper thing to brag about - it's a sign the product is bad.

what stock behavior? Are we talking about, some bad mobos from asrock or do all stock settings need to be changed and for what user purpose? Real world use or high benchmark use?

The point is the massive gap between advertised TDP of 65W and peak of 202, or higher in the upcoming gen, and how this affects stock settings and user experience. The gap on competition is smaller and the clock behaviour is less bursty as a result, which also affects cooler requirements.

I'm looking at two 65w CPUs (Intel 12400 and 5600x) so I'm not playing favorites here. Both are similar in gaming power used while playing Cyberpunk and temp wise they are similar while using blender, probably a worst case scenario for most people. We (as educated consumers) know the TDP recommendations by Intel and AMD have been pure BS for years so why is that surprising now?

This is the point. We've seen it before with Intel and motherboard vendors where loosely defined specs led to unwanted behaviour. Back then it was about Z-boards. Now we're talking about non-OC boards suffering that situation.

Is this an asrock issue or an issue with all Intel z motherboards?

 

[Vayra86]

I would never call you a fanboy or shill; I would simply say you are best with a little white wine, shallots, and frites. I will say reading the whole discussion, it's hard for people to understand some of the points made

what stock behavior? Are we talking about, some bad mobos from asrock or do all stock settings need to be changed and for what user purpose? Real world use or high benchmark use?

I'm looking at two 65w CPUs (Intel 12400 and 5600x) so I'm not playing favorites here. Both are similar in gaming power used while playing Cyberpunk and temp wise they are similar while using blender, probably a worst case scenario for most people. We (as educated consumers) know the TDP recommendations by Intel and AMD have been pure BS for years so why is that surprising now?

Is this an asrock issue or an issue with all Intel z motherboards?

You're looking at the wrong CPUs, the above ten+ posts are about the higher end non K's. Note how the temp graph is sharply going up from 12600k onwards. We know the lower and mid range is cool because the core count is lower and so is the peak frequency, by up to a full Ghz even.

As for the TDP values being BS, this is untrue. The underlying principles have been changing the past generations, curiously in the same pace as Intel was losing its leadership.

Is this an asrock issue or an issue with all Intel z motherboards?

I think this is an issue that can exist in the first place because Intel has defined the spec in a silly way, and again, this isn't about Z boards where some tweaking can be expected from users, but about non overclockable chipsets and parts.

I had the same criticism on AMD's approach in the past. Being loose with what MB vendors can do, is always troublesome to the end users. It really only serves to mislead customers.

Yawn...some boards enforce power limit by default and some don't, so again if you don't know how to make your equipment works as intented, the problem is on you. I'm not going to be hypocritical and talk like the average PC DIYers are so clueless they don't know what TDP their CPUs are running at...or that they can't read reviews before buying something

History is full of examples of just that, and really, the average PC DIY job is a complete learning process from beginning to end, for most average PC DIY'ers.

So yes, some clarity would certainly be nice.

 

[AleXXX666]

Anecdotal but my 12900K system is the smoothest and most responsive system I have ever had, this includes my previous 10900K and 3900X machines, So i guess the E-Cores are doing a good job at background tasks.

how about cooling 3900X vs 12900K, for example. I've had 3900X cooled EZ by a simple Dark Rock cooler (even not pro lol). Now let's put it on this 12900K "hot cupcake" LMFAO
So yes, definitely doubling 10900K hypothetically or 3900X would give you same responsiveness. You don't consider 12900K way more power hungry and hot lol.

 

[Mussels]

how about cooling 3900X vs 12900K, for example. I've had 3900X cooled EZ by a simple Dark Rock cooler (even not pro lol). Now let's put it on this 12900K "hot cupcake" LMFAO
So yes, definitely doubling 10900K hypothetically or 3900X would give you same responsiveness. You don't consider 12900K way more power hungry and hot lol.

people mistake the reported temperatures for heat output

A 5800x can run off a 120mm cooler at 80c, and run that same 80c on a custom loop - that's hot therefore its outputting lots of heat! (100-140W range, PBO depending)
But then a 350W CPU that reports 60C because it's easier to transfer heat from, that ones 'cold' and therefore good

These are the same people that think a fan cools a room, and it's very frustrating the lack of common sicentific knowledge out there at times

I've had my 5800x, 5800x3D and 3700x all run off my dark rock slim at various times and honestly, i'd just have bought more of them if I wasn't sponsored the Alphacool parts - the temperature difference is small enough that the heat from the 3090 likely makes the CPU hotter than if it was on that small air cooler

 

[AusWolf]

people mistake the reported temperatures for heat output

A 5800x can run off a 120mm cooler at 80c, and run that same 80c on a custom loop - that's hot therefore its outputting lots of heat! (100-140W range, PBO depending)
But then a 350W CPU that reports 60C because it's easier to transfer heat from, that ones 'cold' and therefore good

These are the same people that think a fan cools a room, and it's very frustrating the lack of common sicentific knowledge out there at times

I've had my 5800x, 5800x3D and 3700x all run off my dark rock slim at various times and honestly, i'd just have bought more of them if I wasn't sponsored the Alphacool parts - the temperature difference is small enough that the heat from the 3090 likely makes the CPU hotter than if it was on that small air cooler

There's also heat transfer. Just one example: My be quiet! Silent Loop 2 dumped a lot of heat into the environment with the i7-11700 at 90 °C, but it's cold to the touch with the R7 7700X at 90 °C. Note that both CPUs report the same core temperature. Both CPUs consume around the 140 W mark when fully stressed. The difference is how much of that heat is transferred to the cooler, and from the cooler to the atmosphere.

 

[JustBenching]

Intel is much easier to cool cause of better heat transfer. How is that still disputable, lol

There's also heat transfer. Just one example: My be quiet! Silent Loop 2 dumped a lot of heat into the environment with the i7-11700 at 90 °C, but it's cold to the touch with the R7 7700X at 90 °C. Note that both CPUs report the same core temperature. Both CPUs consume around the 140 W mark when fully stressed. The difference is how much of that heat is transferred to the cooler, and from the cooler to the atmosphere.

That is literally impossible. If both cpus consume 140w then they both produce the same amount of heat.

 

[unwind-protect]

That is literally impossible. If both cpus consume 140w then they both produce the same amount of heat.

Unless you overclock them to the point where a significant part of the energy output is light...

 

[AusWolf]

That is literally impossible. If both cpus consume 140w then they both produce the same amount of heat.

Feel free to drop by and see for yourself. I also wouldn't believe it if I didn't experience it first-hand. As to why it happens, my only guess is that smaller chiplets are worse at transferring heat to the IHS and cooler compared to larger, monolithic dies.

Another weird difference is that the 7700X cools back to idle temp almost straight after the work is finished while the 11700 needs a minute or two (the warmer liquid within the cooler needs to cool down first, I guess).

 

[JustBenching]

Feel free to drop by and see for yourself. I also wouldn't believe it if I didn't experience it first-hand. As to why it happens, my only guess is that smaller chiplets are worse at transferring heat to the IHS and cooler compared to larger, monolithic dies.

Another weird difference is that the 7700X cools back to idle temp almost straight after the work is finished while the 11700 needs a minute or two (the warmer liquid within the cooler needs to cool down first, I guess).

Your guess is wrong. Yes, smaller chiplets are worse at heat transfer, but the heat doesn't remain on the chip, if it did then your temps wouldn't be similar as youve previously mentioned. Worse heat transfer means that the CPU will be in higher temperature in order to transfer the same amount of wattage to the cooler compared to a CPU that has better heat transfer.

The only explanation assuming what you are saying is true is that your motherboard misreports wattages.

 

[AusWolf]

The only explanation assuming what you are saying is true is that your motherboard misreports wawattages.

Maybe. With the classic Ryzen "power reporting deviation" value gone from AM5's sensors, and I'm not sure how it works on Intel, the only way to check would be to measure directly at the EPS connectors, which I can't do, unfortunately.

 

[Deleted member 185158]

Feel free to drop by and see for yourself. I also wouldn't believe it if I didn't experience it first-hand. As to why it happens, my only guess is that smaller chiplets are worse at transferring heat to the IHS and cooler compared to larger, monolithic dies.

Another weird difference is that the 7700X cools back to idle temp almost straight after the work is finished while the 11700 needs a minute or two (the warmer liquid within the cooler needs to cool down first, I guess).

These are vastly different processors to compare.

Time to heat or cool is the thermal gradient.

I guess depending on where you measure, and or where the sensor location happens to be, could make a difference.

Package or core temps??

 

[AusWolf]

These are vastly different processors to compare.

Time to heat or cool is the thermal gradient.

I guess depending on where you measure, and or where the sensor location happens to be, could make a difference.

Package or core temps??

Package. Sensor location is a good point, I never thought about that as a variable.

 

[Deleted member 185158]

Package. Sensor location is a good point, I never thought about that as a variable.

Package temp could be motherboard socket vs IHS plate temp, which either would be an off die temp reading.

IHS plates are a lot different too. So the gradient would differ. Speaking of course surface area.. time of dissipation and gradient between idle and max temp.

Either way, 140w would be low for a 14nm Intel chip at load depending on clocks and vcore. But perfectly capable of 300w dissipation at its peak frequencies.

 

[AleXXX666]

people mistake the reported temperatures for heat output

A 5800x can run off a 120mm cooler at 80c, and run that same 80c on a custom loop - that's hot therefore its outputting lots of heat! (100-140W range, PBO depending)
But then a 350W CPU that reports 60C because it's easier to transfer heat from, that ones 'cold' and therefore good

These are the same people that think a fan cools a room, and it's very frustrating the lack of common sicentific knowledge out there at times

I've had my 5800x, 5800x3D and 3700x all run off my dark rock slim at various times and honestly, i'd just have bought more of them if I wasn't sponsored the Alphacool parts - the temperature difference is small enough that the heat from the 3090 likely makes the CPU hotter than if it was on that small air cooler

3700X is not so "hot" CPU. 5800X, especially it's "overpriced" variant, is.

Fan doesn't cool a room, but it's comfortable to sit with fan blowing at you room air fast enough to you feel some "chill".

 

[Mussels]

That is literally impossible. If both cpus consume 140w then they both produce the same amount of heat.

You're correct - I believe the unspoken part there was that the AMD CPU uses less wattage, so it produces less heat. He specifically said "at 90C, it outputs less heat" which is correct.

3700X is not so "hot" CPU. 5800X, especially it's "overpriced" variant, is.

Fan doesn't cool a room, but it's comfortable to sit with fan blowing at you room air fast enough to you feel some "chill".

3700x is a 65W-88W CPU split over two core CPU dies, 5800x and 5800x3D are 105-142W over a single die.

That's about as extreme a comparison as you can make, and a ridiculous one when the 5800x3D is going to use less power gaming than the 3700x due to it being massively more efficient.

Also, you proved my point that you have no idea how cooling works. A fan does nothing except move air, it does not lower temperatures of the air it moves. A fan blowing the hot air in my oven provides zero "chill"

 

[Lei]

Also, you proved my point that you have no idea how cooling works. A fan does nothing except move air, it does not lower temperatures of the air it moves. A fan blowing the hot air in my oven provides zero "chill"

That's why we can't use fan to cool ourselves if ambient temp is near 40°C

Fan can only move the hot air away from your skin, not blowing air which hypothetically would be cooler than ambient.
In other words: your skin makes the air around you warmer than ambient. Fan kicks it away and brings you down to ambient again.

 

[lemonadesoda]

Fan doesn't cool a room, but it's comfortable to sit with fan blowing at you room air fast enough to you feel some "chill".

Correct

A fan does nothing except move air, it does not lower temperatures of the air it moves.

Correct

But moving air does cool an object more quickly than stationary air/convection. When air < object ºC

Just as moving air in an oven does heat an object more quickly than stationary air/convection. When air > object ºC

 

[Deleted member 185158]

You're correct - I believe the unspoken part there was that the AMD CPU uses less wattage, so it produces less heat. He specifically said "at 90C, it outputs less heat" which is correct.


3700x is a 65W-88W CPU split over two core CPU dies, 5800x and 5800x3D are 105-142W over a single die.

That's about as extreme a comparison as you can make, and a ridiculous one when the 5800x3D is going to use less power gaming than the 3700x due to it being massively more efficient.

Also, you proved my point that you have no idea how cooling works. A fan does nothing except move air, it does not lower temperatures of the air it moves. A fan blowing the hot air in my oven provides zero "chill"

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.

 

[ThrashZone]

Hi,
The objective of moving even hot air is to try and stay close as possible to that room/ ambient air temperature.