Of course they are not. The current narrative is that their factory-overlocked 5900HX gaming laptop that pulls 100W+ sustained CPU (basically a desktop 5800X) is 50% faster than the 35W 10-core M1, so "AMD mobile much better than Apple Silicon". Go figure.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Stockfish patch to improve performance on M1 by up to 80%.
- Thread starter Sopel
- Start date
- Sort by reaction score
Of course they are not. The current narrative is that their factory-overlocked 5900HX gaming laptop that pulls 100W+ sustained CPU (basically a desktop 5800X) is 50% faster than the 35W 10-core M1, so "AMD mobile much better than Apple Silicon". Go figure.
You're confusing mobile 8-core 5900hx (35w to 54w) with desktop 12-core 5900x (105w).
Mobile cpu, notebook cpu, desktop cpu. You can compare everything because Apple said:
The M1 chip is using the best cpu.
There's no mobile AMD CPU that pulls 100w sustained. 5900hx is a better binned 5800h that I have here that pulls ~50w sustained on 7nm.
Always the mark of a good faith conversation when one has to make up Apple marketing material to make a point.
According to Notebookcheck, ASUS RoG Duo pulls over 70Watts on CPU alone in Prime95 and can maintain over 4Ghz. This is way over the CPU base spec (3.3Ghz at 45W). The full laptop power running Prime95 is ~ 130W. I am not surprised if one gets good results from this machine when running throughput-oriented AVX2 workloads.
I know it can be difficult to understand something without hands on. However, I have one here. AMD CPU can operate at base clock and optionally at < base clock, boosted clock and overclocked (PBO). A 5800h at base clock maxes out at 54w and low of ~43w running stockfish so inline with their spec sheet cTDP. 5900hx is a hair better binned to run at 3.3GHz vs 3.2GHz base clock at same power consumption. Even with overclocked PBO it's not even close to your 100w+ sustained CPU claim and no one questioned GPU TGP since stockfish doesn't use it nor total system consumption.
https://www.amd.com/en/products/apu/amd-ryzen-7-5800h
I am not doubting any of this. However, the link I provided above clearly shows that Rog Duo maintains over 4Ghz running Prime95, with CPU cluster power consumption over 70 watts and full laptop power of 130W. That is not 3.3Ghz. So if someone running on an overclocked laptop like this gets 20m nodes/second in Stockfish — 80% more than a "normal" 4900H, I am not surprised. These are numbers well expected from Zen3 running at 4.2-4.3Ghz. Just not at 45W.
Numbers are in the notebookcheck article, did you read it at all? This has nothing to do with GPU, Prime95 does not use the GPU.
It‘s not possible to overclock Apples M1 CPUs??
But we can overclock Intel and AMD CPUs.
That‘s very bad for Apple.
If we use Intel or AMD and need more CPU power, we only need the overclock the CPU and we will get much more power, while Apple will be still a lot slower.
No, it's not bad for Apple. I realize it's hard for people like you to understand, but most people who buy PCs don't ever overclock them.
If you want high throughput on numerically intensive CPU workloads, you get a x86 workstation. If you want excellent performance on a range of hybrid CPU workloads with ultra low power consumption, you get the M1. It’s that simple.
Can you please explain why it's "very bad for Apple" that users are unable to overclock M1 CPU's.
You can overclock an M1. You just can’t overclock a Mac. You never could. Not even an Intel Mac. But if you want to take your M1 or Intel processor out of your Mac and put it on another motherboard, have fun.
But if that is “bad for apple” then it always has been. Has nothing to do with M1.
That's not quite true.
It was possible to overclock the Mac Pro 3,1 (2008 model) with an application ZDNet released in 2008 called "ZDNet Clock". It didn't work past 10.7 though and was using 32-bit kernel extensions.
I still have the application on my backup drive and it reads 24 June 2008.
Please see Post #82 for the full Stockfish benchmark comparisons.
Solouki
Edit: See post #82 if you are interested in the Stockfish comparisons of six machines.
Solouki
Edit: See post #82 if you are interested in the Stockfish comparisons of six machines.
Last edited:
Hyperthreading is not engaged unless the workload demands it. The first logical core for each physical core is used first, and then if there's still enough work to schedule to the second logical cores, it will do that. This is primarily because running two threads simultaneously will hurt performance of both threads, making them take longer. For things that are responding to UI input, you'll get better responsiveness if you aren't sharing the core with another thread, and this is one way to help achieve that.
Yes, Apple's scheduler will try to keep the P1 cluster quiet and powered off as often as possible, since it can power gate the whole thing, and avoid paying the battery cost to have it running.