If you don't have the proper data to compare (ie. average power) then the solution is not to compare. The solution isn't to start comparing largely irrelevant numbers, cause whatever conclusions you may reach will also be largely irrelevant. That's common sense right?
It's certainly more meaningful than the peak numbers.
Huh? Are you making stuff up?
I will argue about your approach: you can't compare M4 SoC transistor counts against a random other CPU to assess the CPU perf/transistor. The reasons are obvious and multiple: do Epyc or Xeon CPU have NPU or iGPU or ISP? What about cache size? Why compare a mobile chip against a server chip?
I get all that and I agree we don't currently have all the data. But it seems that people are jumping to the Apple bandwagon even though as you just said, we simply don't have the data. All I'm saying is, with the current incomplete data we do have, the M4 doesn't really look that impressive. It's a 28b transistor chip losing - and not by a small margin, to a 2 half times smaller 3 year old Intel chip, on a workload like GB6 that doesn't even use the full chip. So sure, you can say that not all of those 28b transistors are put to work in GB6 but the same applies to the Intel chip - as you can tell by the scaling of the scores.
Intel has a bigger headache to take care of with LNL. Snaapdragoon!
Why the hell not? Whats wrong with providing extra data?
"Higher core count" is just meaningless. Who cares about how many cores the thing has? If one approach is using 1000 tiny cores and the other approach is using 2 50b transistor cores are you going to argue "oh it's not fair"? Transistor count is what matters when you want to compare arm to x86. Number of cores is as irrelevant as clockspeeds. Doesn't matter.
forums.anandtech.com
M4 is impressive it achieves its single core score at 4.5GHz while being passively cooled.This is something the 12900K cannot do and in single core the 12900k consumes more power while having a higher clock still delivers less 1t performance than M4 in Geekbench.
Funny you say that but the cpu was passively cooled indeed, the fan is literally at 0% until the cpu hits 80c. Which it never does, not running Geekbench 6. Nice try bud 😁
Again, clockspeeds are irrelevant. It can be doing it at 1 ghz or 5 petahz, really doesn't matter.
Get rid of the heatsink. The M4 doesn't have any cooling or a heatsink in the iPad. Run geekbench with no heatsink and no air cooling that would make it really fair.
In the case of checking GB scores, yes they do matter.
One M4 P core + $L2 is about 4.06mm2 and a 12900K P core + $L2 is around 7.46mm2. We don't look at the transistors for the whole of M4 because, the M4 has a massive iGPU relative to the very very tiny iGPU in the 12900K, M4 has a NPU, massive display and media engines etc.
Now, I've heard everything. There is so much wrong with what you've said, that I'll just have to laugh. I wonder what is next
I highly doubt the CPU fans are idle at 0 rpm. From past experience with Intel desktop CPUs (or even any desktop CPU), whenever the computer is turned on, the CPU fans start spinning. It might not spin at its maximum, but it will spin.
It's a small (relatively) single tower air cooler, u12a.
It has nothing to do with intel cpus and the fans are indeed at 0 rpm.
The m4 was tested under ln2 and still lost. So even though it's unfair that I didn't have Ln2 to use as well, it's fine, don't need it.
Facebook
Twitter