Alder Lake-S leaks

[Destination]

Thought the whole point of modern tech was to run cooler, not hotter :/
Hell even Intel themselves in their presentation said Alder Lake runs cooler than last gen!

It is.
They just likely haven't managed it yet.
They'll go for perf, not perf per watt, adding e core to make things not seem just as bad.
If they take the perf crown, they won't care about power usage.

 

[Trickle]

240w for a desktop CPU is obscene.

I don't think they expect everyone to be running at that level, though the info yesterday suggested that pedal to the metal will be the motherboard default now.

I think they want all the benchmarks to be done at this wattage, and thus the 'performance rules of thumb' everyone keeps in their mind to be at maximum. [ie. Ryzen 3000 = 20% better than Ryzen 2000 (ipc+mhz), Icelake 0% ipc over Skylake, 4% higher mhz.]

 

[Dave2150]

240w for a desktop CPU is obscence. IMHO more states / countries should adopt the rules introduced in California.

What about those running ryzen 5000 series with PBO or high manual overclocks (which they'll need to compete with 12900k)? The below shows power draw at the wall, not just CPU, though is a good reminder that Ryzen 5000, when not running at stock, is also very power hungry.

 

[Poneros]

lol

they really tried to gimp AMD as much as they could in those comparisons didn't they

 

[LtMatt]

What about those running ryzen 5000 series with PBO or high manual overclocks (which they'll need to compete with 12900k)? The below shows power draw at the wall, not just CPU, though is a good reminder that Ryzen 5000, when not running at stock, is also very power hungry.

That's running CB23 though.

Gaming power draw with PBO+Curve Optimizer is a lot lower, seeing around 105-120W in Far Cry 6 at 5-5.125Ghz as an example.

It will be interesting to compare power draw under similar gaming loads between the processors.

lol

they really tried to gimp AMD as much as they could in those comparisons didn't they

Good catch.

As I said earlier in the thread, they will want to paint their processors in a good as light as possible, so a 5950X that is not restricted should perform better than what they have shown at least.

 

[Destination]

So 105 watt versus 241 watt for Intel internal figures.
Interesting.

 

[mrk]

https://www.pcgamer.com/amp/intel-c...est-gaming-processor-but-has-to-retest-ryzen/

Intel says it will re-measure all of the games it did before with the new patches for AMD and Win 11 and publish its revised findings but remains confident that they are still ahead.

 

[Destination]

What about those running ryzen 5000 series with PBO or high manual overclocks (which they'll need to compete with 12900k)? The below shows power draw at the wall, not just CPU, though is a good reminder that Ryzen 5000, when not running at stock, is also very power hungry.

It really isn't. It's using less than intel did 2 gens ago, and vastly less than the chip you own. Total system power draw is almost lower than the new 12th gen chip will pill on its own.

 

[TrixP10]

IMHO we have cop26, a push for EV cars etc and on the otherhand 240w+ desktop CPU. If Intel can't better that then maybe they will just be left with their fan boy club. Power matters, especially in DC.

 

[alec]

240w+ desktop CPU

but but but it has the Efficiency cores!

 

[SeeNoWeevil]

Does Windows 10 run purely on the Gracemont cores?

 

[timorous]

https://www.pcgamer.com/amp/intel-c...est-gaming-processor-but-has-to-retest-ryzen/

Intel says it will re-measure all of the games it did before with the new patches for AMD and Win 11 and publish its revised findings but remains confident that they are still ahead.

They likely are still ahead.

The 5900X is around 4-5% faster than the 11900K at 1080p. The 11900K vs 12900K slide shows the 12900K 10-15% faster on average. Given that it seems likely that the margin of victory for the 12900K over the 5900X is going to be around 5-10% as an average depending on the exact testing suite.

If you look at that slide and look at the games with big gains they are the kind of games with a lot of CPU load across many AI agents so it might be that those gains are less a function of the CPU itself and more a function of the DDR5. OTOH the games that don't gain much or even regress may well be ones that do see good gains in a DDR4 + 12900K system with lower latency. It will be very interesting to see how it shakes out and for gamers the choice of DDR5 vs DDR4 may come down to the games played as much as cost.

 

[alec]

Does Windows 10 run purely on the Gracemont cores?

Can disable all E-cores.
You can't disable all P-cores, at least one must remain running. Could probably choke this one core into efficient mode by downclocking/undervolting/power limits?

 

[humbug]

What about those running ryzen 5000 series with PBO or high manual overclocks (which they'll need to compete with 12900k)? The below shows power draw at the wall, not just CPU, though is a good reminder that Ryzen 5000, when not running at stock, is also very power hungry.

Dave, these slides are a nonsense. Made by reviewers who have no clue what they are doing.
No one should be running overclocks like that on Zen 3, it makes little if any performance difference to applications like Cinebench, Handbreak ecte.... and actually reduces gaming performance.

All you do is find the happy medium between a negative curve optimiser and a boost off-set, that way you get a reduced power consumption and a performance boost all round.

I'm getting about +5% all round higher performance and i'm using 110 Watts maximum total package power.

 

[uscool]

What about those running ryzen 5000 series with PBO or high manual overclocks (which they'll need to compete with 12900k)? The below shows power draw at the wall, not just CPU, though is a good reminder that Ryzen 5000, when not running at stock, is also very power hungry.

I turned off using motherboard limits on my 5900x and just used PBO and increasing performance over stock with much lower power usage

 

[humbug]

I turned off using motherboard limits on my 5900x and just used PBO and hardly any performance difference with much lower power usage

Yeah, you're doing it wrong

 

[humbug]

Here.... screen cap during the run.

4.7Ghz all core, 105 Watts Total package power, CPU cores 81 Watts, 74c.

4.9 to 5.0Ghz all core in games.

 

[uscool]

Yeah, you're doing it wrong

what do you mean wrong ? cant find the thread here now but was full thread how to disable limits set to motherboard and keep PBO, I went odd 230w cpu at load to around 100+w and still getting the benefits of using PBO

 

[humbug]

what do you mean wrong ? cant find the thread here now but was full thread how to disable limits set to motherboard and keep PBO, I went odd 230w cpu at load to around 100+w and still getting the benefits of using PBO

Reset everything to stock, test in Cinebench, record your score. go back to BIOS

Keep the limits in place, set PBO to advanced, in curve optimiser set symbol or value to negative, set the value, start with 10, save load back in to windows, run cinebench again, compare scores, you should find that your CPU is running a little cooler, using less power and your scores should be a little higher.

Go back to BIOS, set a boost off-set to +100Mhz, test again in Cinebench, compare results.

Now you know how to play with it, find the combination of a negative curve and a boost off-set, just like you would with overclocking until its not stable anymore, once you have everything just so use it, you may find it randomly crashing, often when its not doing anything at all, or watching Youtube, if so go back in to the BIOS and take 2 or 3 off the negative curve optimiser.

 

[uscool]

Reset everything to stock, test in Cinebench, record your score. go back to BIOS

Keep the limits in place, set PBO to advanced, in curve optimiser set symbol or value to negative, set the value, start with 10, save load back in to windows, run cinebench again, compare scores, you should find that your CPU is running a little cooler, using less power and your scores should be a little higher.

Go back to BIOS, set a boost off-set to +100Mhz, test again in Cinebench, compare results.

Now you know how to play with it, find the combination of a negative curve and a boost off-set, just like you would with overclocking until its not stable anymore, once you have everything just so use it, you may find it randomly crashing, often when its not doing anything at all, or watching Youtube, if so go back in to the BIOS and take 2 or 3 off the negative curve optimiser.

thats what I have done, and settled at -25 on all with +50mhz for the past 7 months