Based on the below benchmarks, the E-cores in desktop Alder Lake gain power efficiency until you go well below 1W per core, while the P-cores scale to around 2W.
chipsandcheese.com
No data suggests that Intel's technology could outperform Apple's at any equivalent power level. Efficiency curves change based on power draw but that isn't relevant to this comparison as Apple chips would likely see similar improvements.Based on the below benchmarks, the E-cores in desktop Alder Lake gain power efficiency until you go well below 1W per core, while the P-cores scale to around 2W.
Alder Lake’s Power Efficiency – A Complicated Picture
Reviews across the internet show Alder Lake getting very competitive performance with very high power consumption. For example, Anandtech measured 272 W of package power during a POV-Ray run. Our o…
So even with only 75W of power, you are not hitting the sweet spot with 16 cores. In fact going from 5W per core to 2W/1W on the P-cores/E-cores respectively seems to yield respective improvements of ~50%/~100% in perf/watt. And these are desktop, not mobile benchmarks. So do I think that with a process shrink and enough E-cores on a mobile CPU you can double efficiency for a given performance level? Yes, I don't see why not. In fact as I said, I suspect a 32 E-core chip on Intel 4 could compete quite nicely with the M1 Ultra.
I literally provided data in my post and you ignored it. The question I am addressing is whether Intel could provide equivalent performance to the M1 Ultra at stock settings with equivalent power draw, not whether they can beat the M1 Ultra at every conceivable power level.
How does the data you provided demonstrate that Intel can match the M1 Ultra at its stock 60 watt power level?
Well, take the limit case where Intel has a 60 E-core CPU. According to the efficiency curves in the article, the E-cores at 1W provide around 45% of their performance at 6W, where they can't scale further. (On the 12900K Anandtech saw max power consumption of 48W for all 8 cores at 3.9 GHz: https://www.anandtech.com/show/1704...hybrid-performance-brings-hybrid-complexity/4). We also know that 8 E-cores give around 25% of the total multithreaded performance of the 12900K at stock (https://www.anandtech.com/show/1704...ybrid-performance-brings-hybrid-complexity/10). So 60 E-cores at stock would be ~1.875 of a 12900K. Scale that by 0.45 and you get 0.84 of a 12900K. That's for a desktop CPU on Intel 7. So moving to a laptop CPU on Intel 4, you would only need an extra 20% performance per watt for rough parity with the 12900K in multithreading.
It's a huge assumption to assume linear performance scaling with 60 cores. All kinds of possible bottlenecks could arise. This would also leave single thread performance well below the M1 ultra.
As I said, I see the 32 E-core 8 P-core design as the more likely option, which would maintain a balance of single core/lightly threaded and multicore performance. And my claim is not that this will happen, but that it is possible. That's a pretty low bar to clear.
We'll have to wait until they're done blaming others for their issuesOn a sidenote (and yes it's a joke...) when Tachyum finally releases its long promised "universal" processor all other CPU developers will be caught with their pants down
Either way to bounce back on topic: comparisons between Apple and N other hardware are perfectly fine by me, but it still doesn't change the fact that in reality Apple is more or less competing with itself only. Obviously M1 and one or more succeeding SoCs will have to deal with sw related hickups but eventually Apple will iron them out at some point. I don't see Apple returning to Intel/AMD/NV hw in any foreseeable future rather the contrary.
How do we know what price they'd need to sell them at?
