Intel processors crashing Unreal engine games (and others)

[itsmydamnation]

No one metric in isolation ever matters, people presenting it this way are either dishonest or ignorant.

 

[Abwx]

You don’t have to believe me, the 8C SKU appears have a 170W TDP / 232W PPT.

Of course that no one will believe you, that would amount to something like 3W/mm2, wich is just impossible, a current Zen 4 CCD can get to barely half this value, at 3W/mm2 with the same IHS the die temperature would get over 140°C.

 

[H433x0n]

Of course that no one will believe you, that would amount to something like 3W/mm2, wich is just impossible, a current Zen 4 CCD can get to barely half this value, at 3W/mm2 with the same IHS the die temperature would get over 140°C.

Feel free to look at the shipping manifests for Turin and Granite Ridge.

It’s not on the same node as Zen 4. It’s supposedly on N4X, a node that has the potential to be much leakier. Seems Zen 5 is really leaning into frequency for this generation.

 

[Abwx]

Feel free to look at the shipping manifests for Turin and Granite Ridge.

It’s not on the same node as Zen 4. It’s supposedly on N4X, a node that has the potential to be much leakier. Seems Zen 5 is really leaning into frequency for this generation.

Node has nothing to do in the equation, you could take any node there s not a single one that can sustain such a power density, that s a matter of physics limit of the silicon.

One who knows about this limitation would had understood that something is wrong in the shipping manifest in respect of physics laws, you ll see that it will use the same power as a current 7700X at most.

 

[H433x0n]

Node has nothing to do in the equation, you could take any node there s not a single one that can sustain such a power density, that s a matter of physics limit of the silicon.

One who knows about this limitation would had understood that something is wrong in the shipping manifest in respect of physics laws, you ll see that it will use the same power as a current 7700X at most.

It’s not just Granite Ridge but Turin has a similarly large increase in TDP.

I guess within 4 weeks we’ll see you disagreeing with AMD themselves when Zen 5 is launched.

 

[gdansk]

Turin has how many cores with that 25% more TDP? I can't remember if it was 128 cores or 192 cores total. In either case they feel pressured to piggyback on 500W Granite Rapids thermal solutions.

But back on topic: Intel 13/14900K are still returning incorrect results and crashing the most common game engine without underclocks.

 

[Joe NYC]

Going to be hilarious to watch all of this become a non factor in the next 6 months when suddenly Intel CPUs draws less power than their AMD counterparts.

That is quite vague: Less power for a lot less performance?

I think more interesting comparison is performance at a given power consumption. Do you think Intel will lead? As in:

Zen 5 vs. RPL vs. ARL?
Strix Point vs. MTL?

There is a lot more discussion about absolute performance regardless of power, and not enough about power limited performance.

 

[Joe NYC]

It’s not just Granite Ridge but Turin has a similarly large increase in TDP.

I guess within 4 weeks we’ll see you disagreeing with AMD themselves when Zen 5 is launched.

Turin also has increases in core counts. I am not sure how this ends up as far as power per core, but mentioning only one variable, TDP, is incomplete in light of # of cores changing.

It would be more valid in desktop, where the core counts are not changing.

 

[H433x0n]

Turin also has increases in core counts. I am not sure how this ends up as far as power per core, but mentioning only one variable, TDP, is incomplete in light of # of cores changing.

It would be more valid in desktop, where the core counts are not changing.

96C Genoa v 96C Turin is +25% power draw increase.

I don’t know the nT performance of either Zen 5 or ARL but I know the power specs of both and it doesn’t look like Zen 5 is going to have an efficiency advantage.

 

[Markfw]

96C Genoa v 96C Turin is +25% power draw increase.

And what does that have to do with Intel overclocking their factory desktop cpus !

 

[Joe NYC]

96C Genoa v 96C Turin is +25% power draw increase.

96C Genoa vs. 128C Turin is +25% power draw increase.

 

[gdansk]

It was partially my fault I brought up Turin after he mentioned power draw because I'm pretty sure Intel won't be backing off from power draw there as they need to do elsewhere.

But if Arrow Lake doesn't crash, has an only-P-core SKU with better than 9800X3D gaming performance and lower power consumption I'll buy it even if it is more expensive. I won't even complain, that sounds great.

 

[H433x0n]

96C Genoa vs. 128C Turin is +25% power draw increase.

Feel free to bookmark this and come back to this post after Computex if I’m wrong.

 

[H433x0n]

It was partially my fault I brought up Turin after he mentioned power draw because I'm pretty sure Intel won't be backing off from power draw there as they need to do elsewhere.

But if Arrow Lake doesn't crash, has an only-P-core SKU with better than 9800X3D gaming performance and lower power consumption I'll buy it even if it is more expensive. I won't even complain, that sounds great.

ARL-S will have much better efficiency and lower power consumption but the rest of what you wrote isn’t going to happen.

Technically on paper specs, top GNR-AP will use less power than Turin but it’ll get stomped in performance so it won’t matter.

 

[Joe NYC]

Feel free to bookmark this and come back to this post after Computex if I’m wrong.

Where exactly do you think you are right, and I am wrong?

My contention is that there will be more cores in Turin, Turin Dense (vs. Genoa, Bergamo), for that increased power consumption.

Are you claiming that Turin core counts will stay the same, but power will go up?
(just trying to clarify).

 

[Joe NYC]

ARL-S will have much better efficiency and lower power consumption but the rest of what you wrote isn’t going to happen.

Vs. RPL? It's like saying "He is better looking than Quasimodo". It's literally the lowest bar to clear.

The question will be ARL vs. Zen 4, ARL vs. Zen 5?

 

[H433x0n]

Where exactly do you think you are right, and I am wrong?

My contention is that there will be more cores in Turin, Turin Dense (vs. Genoa, Bergamo), for that increased power consumption.

Are you claiming that Turin core counts will stay the same, but power will go up?
(just trying to clarify).

When comparing Genoa v Turin at iso core counts the power consumption goes up. There is also higher core count SKUs where power consumption goes up even further.

 

[Joe NYC]

Turin core counts go up but iso core counts power consumption goes up.

That's within realm of possibilities.

It implies if core counts go up, then iso core power consumption increase (if any) is less than socket power consumption increase.

 

[Kepler_L2]

This isn't unique to Zen5, every new EPYC gen has increased TDPs at iso-core count. 225W for 64 cores on Zen2/Rome, 280W for 64 cores on Zen3/Milan, 360W for 64 cores on Zen4/Genoa.

 

[adroc_thurston]

This isn't unique to Zen5, every new EPYC gen has increased TDPs at iso-core count. 225W for 64 cores on Zen2/Rome, 280W for 64 cores on Zen3/Milan, 360W for 64 cores on Zen4/Genoa.

Robert Dennard himself is dead yet people pretend Dennard scaling is still alive.
Wild.

 

[coercitiv]

Going to be hilarious to watch all of this become a non factor in the next 6 months when suddenly Intel CPUs draws less power than their AMD counterparts.

Do you reckon that Arrow Lake drawing less power will make Raptor Lake stable again?

 

[H433x0n]

Do you reckon that Arrow Lake drawing less power will make Raptor Lake stable again?

No, we need another 22 pages of the same 15 partisan posters keeping the hate train going. Still haven’t managed to beat this topic into the dirt quite yet.

 

[coercitiv]

No, we need another 22 pages of the same 15 partisan posters keeping the hate train going. Still haven’t managed to beat this topic into the dirt quite yet.

Rest assured, we'll get more and more pages until Intel shines light on the issue. All we need is an official statement and a clear path forward for the affected customers, months after the problem was reported in the media. Apparently that is a hard task for one of the biggest tech companies in the world.

 

[CouncilorIrissa]

If only there was a way to prevent these threads from appearing...
Oh right, I know that: shipping CPUs that are actually stable out of the box, having a spec for them and making mobo manufacturers run them in the spec.

 

[igor_kavinski]

If only there was a way to prevent these threads from appearing...
Oh right, I know that: shipping CPUs that are actually stable out of the box, having a spec for them and making mobo manufacturers run them in the spec.

Yeah. Something expected from world class professionals and silicon engineering wizards...

It seems AMD copied them a bit too well. So well that Intel decided to change to not resemble AMD