Rocket Lake not looking too good...

[dyonoctis]

Tell me, how exactly does the "power draw" personally affect you?

5600X might be king, but the locked 6-cores don't have much power draw. And by now you should already know that "65W" and "105W" are a load of bollocks over at AMD as well, because TDP means next to nothing about practical power consumption for either product.

And that "power draw" still resulted in a family of chips that ran 10-20C cooler than Ryzens That lower power draw did......what, exactly, for Ryzen 3000 and 5000 owners? Allow us to live life on the edge with a 450W PSU?

Do people get some sort of kick out of bringing up "power draw" every time INTEL shows up in ctrl+F?

It's a niche usecase, but it allowed ITX enthusiast to rock up to16 cores with a modest vrm setup It's just a shame that the cpu die are too small to be efficiently cooled.

That would be nice if x86 could reach a point a NH-L9x65 would be enough, CPU cooling is the biggest hurdle when you try to go small

 

[RandallFlagg]

Right, but I'm a bit confused by this:

When running my 4690k at stock clocks, I'd be hard-pressed for it to cross 75W power consumption, even in p95.
Only when I overclocked it to 4.4GHz and raised the power limit in the BIOS that I was able to exceed its TDP: I was drawing 84W-94W at 4.3GHz and now drawing 98W at 4.4 when set in the BIOS at 100W max.

On Intel, a chip rated at 65W with all stock settings will in fact only consume 65W averaged over a time period called called Tau - which is 28 seconds on non K chips. So while the chip may go up to 125W for 5 -7 seconds with a high all-core boost, it will then drop frequencies down such that its average over 28 seconds is 65W.

You can change those settings on most DIY motherboards to allow it to turbo away infinitely regardless of power by changing Tau and the max power limits (PL1 / PL2).

On TPU, when they review a chip they benchmark it using power limits unlocked and with normal power limits. On the benchmark charts used later for other reviews, they just use their DDR4-3200 standard rig which is the green bar below.

You can see the difference below on *synthetic CPU tests* of a 10700 (nonK).

The brown bar is totally stock with stock RAM. The green, stock with DDR4-3200 (TPUs standard setup). Blue bar is with power limits removed, but no overclocking. Edit: they also do it with a +2.5% BCLK on locked CPUs, not shown here since that is true overclocking.

The difference on CPU synthetics is 11.5%. This appears significant, but IMO it really isn't given what the tests are (very well balanced multi-core loads).

In the real world, running all core at 100% is super rare. I've run the Windows performance monitor for multiple days while working and gaming. What I found was that during peak loads I'll get 1 core maxed out for a few seconds here and there and one or two other cores running 30-40% while the rest are way below 20%. That's peak load, so everything I've seen indicates that single core performance is still king so long as you have 4+ cores.

My meaning is that unlocking power limits does not help single core performance, it's just multi-core that hits power limits and hence mainly synthetic loads where this shows up. That +11.5% on synthetics on the 10700 review as example, translated into a negative 0.2% on 1080P gaming. In other words, games aren't going to max out enough cores for power limit unlocking to make much difference. Almost nothing will really, more of a benchmark pump than truly useful.

 

[Vya Domus]

a company that could buy AMD X Times

Not anymore, AMD is worth over 100 billion right now.

 

[EsaT]

Last time I read, a 100W CPU dissipated 100W of heat.
How are your CPUs dissipating twice the heat as the power consumption? Are you mounting one over the other, or something?

Right, but I'm a bit confused by this:

When running my 4690k at stock clocks, I'd be hard-pressed for it to cross 75W power consumption, even in p95.
Only when I overclocked it to 4.4GHz and raised the power limit in the BIOS that I was able to exceed its TDP: I was drawing 84W-94W at 4.3GHz and now drawing 98W at 4.4 when set in the BIOS at 100W max.

That CPU is from time when Intel had no performance competition and basically only absolute top models ever approached TDP rating.
Myself used to have E0 stepping Q9550 undervolted from stock 1,25V to 1,2V running with 30% extra clocks at 3,7GHz.
And monitoring softwares showed power draw to be IIRC ~80W when running IntelBurnTest/Linpack (even heavier load than Prime95) with easy to keep under control temperatures.

But before Core 2 era with Intel stuck on NetBurst/Pentium4 architecture against AMD's Athlons power draw was cranked to nuts level in "Emergency Editions" to maintain some semblance of performance competivity.
(then AMD went for same with "Faildozers", though actually giving some models honest TDP number)

Now multithreaded load situation is same with Intel unable to match more advanced manufacturing node and high core count of AMDs in performance.
Hence for making CPUs to look better in performance Intel basically blesses mobo makers to disregard marketing TDPs in BIOS, allowing CPU draw lot more for as long as temperature is under limits.
As result power draw of top models goes to landfill fire level under full all core load even without any manual overclocking.
And of course manual user made overlockings to those hyped also on forums clock speeds are same:
https://cdn.mos.cms.futurecdn.net/hhakAayRyXRuoAVDo9PpAn.png

TDP is simply more marketing based number than any absolute engineering based value.
Which can be good in some things, like that big PC brand being able to design only couple different coolers for whole range of CPUs.
But then it's also commonly used fraudulently when struggling against competition in performance.

Do people get some sort of kick out of bringing up "power draw" every time INTEL shows up in ctrl+F?

So while it was perfectly OK for everyone to laugh at AMD's problems with Faildozers, it's forbidden to point problems "big blue" is having?

 

[Dredi]

11375H cinbench R20 score showing up as 690. From a dam laptop. Tiger Lake H looks to be a gaming laptop killer, and this is just the H35, H45 should be coming as well.

Intel Core i7-11375H in purported MSI Stealth 15M laptop shown to be 46% faster in single-core compared to Core i7-10750H, indicative of longer sustained 5 GHz boost

We have received a Cinebench R20 single-core score from what appears to be an MSI Stealth 15M laptop powered by the Core i7-11375H Special Edition Tiger Lake-H35 processor. While we have previously reported about leaked scores for this processor, the new score shows it to be much faster and...

www.notebookcheck.net

View attachment 185349

If you look closely, you'll notice that the 690 score is with 1 core/2 threads. If it was a single thread run, it would note that next to the score. When 5000 series ryzen is run in that manner you get a score of around 790.

 

[Zach_01]

Last time I read, a 100W CPU dissipated 100W of heat.
How are your CPUs dissipating twice the heat as the power consumption? Are you mounting one over the other, or something?

For Ryzen the TDP is the heat dissipated towards the cooler under specific circumstances. It consumes more than TDP and dessipate more than that as heat "escapes" an all directions and not only from its IHS.
So a Ryzen 65W TDP consumes 75~90W depending Gen and model.

Intels on the other side consume exactly the rated TDP but only on low power state. Intel CPUs have 2 power state levels. PL1 and PL2 and they fall into low power on a certain period of time (Tau).
A lot of boards have settings to disable PL1 fallback and keep the CPU to PL2 infinitely if thermal or other internal limits allow it I believe.