Hacker News new | ask | show | jobs
by tagrun 1101 days ago
The worse single-core performance is typical for server CPUs which are intentionally underclocked compared to their desktop counterparts for stability that is expected under non-stop server workloads.

Scroll to the right for the corresponding consumer CPU: Core i9-13900K has higher score in almost all rows.

From the article:

> benchmark result shows Apple's M2 Ultra cannot beat Intel's Core i9-13900K in single-thread workloads and even fall behind in multi-core workloads in Geekbench 5.

1 comments

I have no doubt that Apple is capable of delivering a product that can beat Intel's i9-13900K, especially since Apple is using a much more modern 5nm process vs Intel's 10nm process for the i9.

As you point out, server CPUs are intentionally engineered for a target market. Apple's processors are very intentionally engineered as well. When engineering their product, they had a certain market in mind, and design and performance trade-offs were made accordingly. As others have noted, power efficiency seems to be a big priority. The M2 Ultra has a TDP of 60W. The i9-13900K has a base consumption of 125W, and draws up to 253W under stress. So do the math. Intel achieves 32% better single-core performance and 41% better multi-core performance (Cinebench) for 422% of Apple's power consumption. If there's something impressive here, it's that Apple is able to do so much with so little. If Apple wanted to, they could probably conjoin two M2 Ultra's and soundly beat the i9-13900K by a considerable margin and still use about half the same power to do so. The real question is why any consumer would need that much compute, and the target market for such a niche is probably very small which is why Apple didn't do that

Your fantastic statements about the potentials of Apple is bordering religion.

It's not just Apple, any big player can do what you said, but there are connectivity trade-offs in chiplet design. Its been open to any TSMC customer including AMD (which has been doing it well before Apple). Intel has also fabricated its own chiplet CPUs already and is currently sampling them.

> The M2 Ultra has a TDP of 60W. The i9-13900K has a base consumption of 125W, and draws up to 253W under stress.

What of it? RTX 4900 TDP is 600W. Just get a better PSU. It's a desktop computer that is always plugged to the wall, not a laptop on battery. It's irrelevant.

What people care about desktop computer is the performance/price and upgradability/serviceability. That's the math you need to do. In terms of what currently exists in the real world, with Apple, you get a CPU that is weaker than a 13900K or 7950X3D, with a soldered 128GB RAM and a soldered 4TB SSD for about $7000. I can build a stronger computer for less than $1500, and it will be both upgradable and serviceable. But the Apple device isn't going to pay the $5000+ difference with lower electrical bills any time during my lifetime: it will instead become a very expensive e-waste in less than 10 years.

> Your fantastic statements about the potentials of Apple is bordering religion.

Could you elaborate?

> It's not just Apple, any big player can do what you said

You're fighting a strawman. I never said that it was only Apple. My statements compared Apple and Intel because that was the topic of this whole thread. I would be happy to discuss any other major player in the appropriate place.

> What of it? RTX 4900 TDP is 600W. Just get a better PSU. It's a desktop computer that is always plugged to the wall, not a laptop on battery. It's irrelevant.

I don't think you understood my post at all. *My point was that Apple very clearly optimized for power efficiency.* That is what my statements regarding power usage served to illustrate.

In fact, looking into the promotional materials for the M2, there is very explicit verbiage regarding "new levels of power-efficient performance", "industry-leading performance per watt", and "using very little power."[0] This just proves my point that power effiency was Apple's foremost goal with the M2.

My point was that, if Apple had optimized for all out performance regardless of anything else, they could have very well done so. My point was that they didn't do that but that they could if they wanted to.

E.g. as we saw with the M1 Ultra, they can very easily conjoin two M1 Max dies using a "UltraFusion" technique.[1] Apple no doubt could use the same technology to conjoin two M2 Ultras.

[0]: https://www.apple.com/newsroom/2022/06/apple-unveils-m2-with...

[1]: https://www.apple.com/newsroom/2022/03/apple-unveils-m1-ultr...