Intel's Raptor Lake CPU Appears 20% Faster Than Core I9-12900K on UserBenchmark | Hacker News Reader

Intel's Raptor Lake CPU Appears 20% Faster Than Core I9-12900K on UserBenchmark

tomshardware.com

22 pointsrbanffy17 comments

Here's some obligatory information about the credibility of UserBenchmark (video form only, unfortunately): https://youtube.com/watch?v=RQSBj2LKkWg

UserBenchmarks are completely unreliable.

If tomshardaware are starting to promote that, then it makes them unreliable, too.

And the 12900k scores ~14% higher than an M1 Ultra in single core performance.

Using more than double the power?

A 12900KS can draw up to 250w of power, I think the m1 ultra reaches 60w on just single core benchmarks, so 4.16x

Power/perf efficiency does not scale linearly though, nor is there any guarantee a chip can take 4x the power.

I don't disagree, I was just adding information in support of the other commenter. We have seen hints at what can happen when AMD pulls the limiter.

If the M1 can never reach this amount of power or calculation speed (and most likely never will) and all you need is a really fuckin fast single core (because you can't multithread your stuff), power usage doesn't matter.

Also, even if power usage matters, the optimum point may be shifted towards higher power points on desktop because as long as you're waiting for results, the screen is shining, the disks are spinning, etc., so sacrificing CPU performance for CPU power efficiency may actually worsen energy consumption per unit of work done.

M1 power consumption is more interesting for the mobile use cases. Once you get to high end desktop the question is more "how fast can I make it go" not "how efficient is this".

Not quite. Noise and heat can also be issues. I love my nearly silent computer

For high end desktops there really shouldn't be any trouble cooling CPUs, laptops or consumer mini PCs and whatnot certainly though but that's not where things like 12900K are designed to go. E.g. my GPU pulls north of 400W under load but for a very small fraction of the price of the GPU cooling it still runs whisper quiet. Once you've gained the ability to push 5" of air through a wide heatsink with slow thick fans instead of 0.5" through a narrow heatsink with tiny thin fans dissipating heat can be extremely quiet up past the KW range.

Compare that to a Mac Studio or a Mac Mini. They are small and mostly silent. It’s hard to make them be hearable. Compare that to the tower under my desk, which has a lot going on to make it quieter and, yet, I can still tell when it’s idling and when it’s doing some deep thinking by the sound it makes.

The Mac Studio does in fact dissipate up to 370 Watts quietly via a larger design, 2 very thick fans, and utilizing an external power supply. The Mac Mini uses the laptop/tablet class M1 SoC and does not deliver on HEDT workloads so cooling is unexpectedly an easy task.

The sweet spot in uniqueness with the M1 seems to have been the M1 Ultra mobile case. I've got a 16" Macbook Pro with one and AMD/Intel still haven't quite caught up in the workstation laptop segment without constantly kicking fans on.

Is there a speed/power usage metric that benchmarks can use or just pin to a max power? A cpu boosting and using 300w is a bit much.

PassMark appears to be more reliable. https://www.cpubenchmark.net/

“Appears” is a very particular word with Intel

Reply on news.ycombinator.com