Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Apple M1 CPU & GPU speed is very disappointing

Status
Not open for further replies.
L

Leifi

macrumors regular
Nov 6, 2021
128
121
cmaier said:
Now, once again, before you get to comment on my links to sources, where are your sources? The forum rules require that when challenged on a factual statement, you post proof. Where is your proof?
Click to expand...

I just noted that your links do not prove that SMT is useless :) Please follow the forum rules and prove your statement below inconclusively... :)

"Hyperthreading is only necessary when you have so poorly designed your instruction dispatch that you can’t keep your ALUs busy. M1 has tremendous IPC. It’s execution units are always busy. Adding hyper threading would slow the cpu down because it would cause unnecessary context switching."

Exactly how much would it be slowed down adding SMT... and why would these extra threads mandate uneccessary context switching :)
 
Last edited:
P

pshufd

macrumors G4
Oct 24, 2013
10,151
14,574
New Hampshire
I have my production setup running on my M1 mini and the M1 has more than enough power to run it. I am running my office stuff on an M1 PRO MacBook Pro. The amount of CPU and GPU horsepower on the M1 is more than enough. It would be enough to run both if it had more RAM and support for three displays. The M1 PRO chip is simply more CPU/GPU than I need but the only way to get a system with 32 GB of RAM. I plan to replace this setup with an M1 mini with 32 GB RAM and support for 3x4k and an M2 Air 16 if and when they come out. I could also replace the mini with a 27" iMac if that comes out first. I'm eager for Apple to fill out their product lines with Apple Silicon.
 
  • Like
Reactions: MarkC426
uller6

uller6

macrumors 65816
May 14, 2010
1,072
1,777
Leifi said:
Single thread netbook native-optimized apps and some occasional video editing is fine use-cases for Apples ARM versions.. But those are all use cases where even a $200 phone will get the job done perfectly well these days. Chess and other high-perf. tasks are a different matter all together.. You cannot hide poor overall pipelining and multithread performance (some lower clocked cores as well) by being fast on single-core benches with these kind of more higher-end workloads.
Click to expand...
Intel tries its hardest to hide overall poor pipelining (stalls) by including hyper threading in their cores. And don’t even get me started on multi threading. The new Alder Lake cores use > 200W when running full out. I like to use a quiet and cool laptop, not a George Forman grill.

Edit: Intel also tried its best to hide poor pipelining by doing crazy preemptive branch prediction/speculative execution...and we all know how well that worked out. Now we have Spectre + Meltdown.
 
Last edited:
  • Like
Reactions: cmaier
L

Leifi

macrumors regular
Nov 6, 2021
128
121
pshufd said:
I have my production setup running on my M1 mini and the M1 has more than enough power to run it. I am running my office stuff on an M1 PRO MacBook Pro. The amount of CPU and GPU horsepower on the M1 is more than enough. It would be enough to run both if it had more RAM and support for three displays. The M1 PRO chip is simply more CPU/GPU than I need but the only way to get a system with 32 GB of RAM. I plan to replace this setup with an M1 mini with 32 GB RAM and support for 3x4k and an M2 Air 16 if and when they come out. I could also replace the mini with a 27" iMac if that comes out first. I'm eager for Apple to fill out their product lines with Apple Silicon.
Click to expand...

So you are happy with subpar chess software performance, compared to cheaper alternatives...

No one would take away your "experiences" from you...

I am sure there are some people also very happy with a dual-core i3 computer that are satisfied with that level of performance as well...I would not question that.

It is just is a question of your expectations and needs. But the discussion of "satisfaction" is subjective... Performance benchmarks like the ones discussed for compiled open-source chess engines, and comparison with recent CPUs from AMD and Intel makes a more interesting topic and can be discussed much more objective. It makes a lot more sense to focus on hard performance facts than to focus on personal experiences and personal bias regarding hw choices you make.
 
  • Like
Reactions: Appletoni
J

januarydrive7

macrumors 6502a
Oct 23, 2020
537
578
The trolling here is top-notch.

Let us turn the tides, a bit, with a benchmark optimized for Apple's architectures. Something like:

C: 
#ifndef __APPLE__
  while (true);

return 0;

should do the trick.

/s
 
  • Haha
  • Like
Reactions: jdb8167 and JMacHack
uller6

uller6

macrumors 65816
May 14, 2010
1,072
1,777
Leifi said:
So you are happy with subpar chess software performance, compared to cheaper alternatives...

No one would take away your "experiences" from you...

I am sure there are some people also very happy with a dual-core i3 computer that are satisfied with that level of performance as well...I would not question that.

It is just is a question of your expectations and needs. But the discussion of "satisfaction" is subjective... Performance benchmarks like the ones discussed for compiled open-source chess engines, and comparison with recent CPUs from AMD and Intel makes a more interesting topic and can be discussed much more objective. It makes a lot more sense to focus on hard performance facts than to focus on personal experiences and personal bias regarding hw choices you make.
Click to expand...
How well does your chess benchmark work on Itanium or UltraSparc CPUs? Do you think the benchmark needs optimization for those designs, or do you think you could just recompile and then run whatever code comes out of your compiler as-is? What if you were to run sparklefish on D-Wave's computer? Do you expect it would run exactly the same way, taking 100% advantage of all available hardware, or do you think the code would need to be optimized on these other platforms?

There's a reason you can't trust any single benchmark to be the arbiter of truth - it's always best to look at a range of synthetic benchmarks that are more/less optimized for the hardware for an accurate comparison among all possible workloads. Even then, benchmarks don't reflect the reality of using the machine daily. My M1 MacBook air is faster than my iMac Pro Xeon CPU every day of the week. And it has no fan. I can't stress this enough. The M1 performance per watt is astronomical. And performance per watt is far more important to me than ultimate performance, no matter the power used.
 
  • Like
Reactions: Homy
L

Leifi

macrumors regular
Nov 6, 2021
128
121
uller6 said:
Intel tries its hardest to hide overall poor pipelining (stalls) by including hyper threading in their cores. And don’t even get me started on multi threading. The new Alder Lake cores use > 200W when running full out. I like to use a quiet and cool laptop, not a George Forman grill.

Edit: Intel also tried its best to hide poor pipelining by doing crazy preemptive branch prediction/speculative execution...and we all know how well that worked out. Now we have Spectre + Meltdown.
Click to expand...

I think you are going slightly off-topic. Firstly you are comparing power-consumption with a desktop CPU. Secondly, if you read this thread most of the benchmarking is comparing Apples M1 with less costly CPUs like the AMD 4800 and 5800 mobile CPUs with 8 fullspeed cores and 16 threads capabilty to boost performance even higher on the 8-cores.. from a chess engine perspective they clearly outperfrom Apple silicon per Watt.

_Stockfish.jpg
 
  • Like
Reactions: Appletoni
J

JMacHack

Suspended
Mar 16, 2017
1,965
2,424
I’ll email the relevant benchmark suites and tell them that they’re all ******** if they don’t include this obscure chess bench.
 
  • Like
Reactions: Appletoni
L

Leifi

macrumors regular
Nov 6, 2021
128
121
Romain_H said:
Oh boy.

For your information: you do not compare CPUs. You compare a CPU version with GPU optimized versions. In other words you claim the M1 is inferior to AMD CPUs which aren‘t even used. ????
Click to expand...
Not really. The comparison is between different laptop-alternatives using the GPU-capabilites available in M1 SoC based computers compared to some other laptops.

And this graph is GPU bound engines, Please look at the previous post for CPU-only comparison.
 
Last edited:
  • Like
Reactions: Appletoni
J

jeanlain

macrumors 68020
Mar 14, 2009
2,463
958
Leifi said:
So bite me.. prove me wrong..No one would be happier than me if Apple silicon performed well for chess. It seems there are some non-technical persons here claiming how great the performance the M1 (or Max) "could be" if only people "cared" about optimizing.. But its all talk no evidence,code samples or "proof".
Click to expand...
That's easy. The M1 performs very well in most native test/apps. If it does not perform well in these chess apps, the most reasonable explanation is that they are not as well optimised.
There is really no reason to suggest that, somehow, the M1 is "bad at chess". Since this is an outlandish hypothesis, it is up to you to demonstrate that the Apple microarchitecture cannot be good at chess.
 
Last edited by a moderator:
  • Like
Reactions: Romain_H
L

Leifi

macrumors regular
Nov 6, 2021
128
121
Romain_H said:
it does not. It does not use GPU on M1 at all. There is no Metal version. That‘s what the discussion is about
Click to expand...
Please go read the referenced links to talkchess.com discussion.. Apples GPU is indeed used (trough OpenCL libraries).. Metal is not possible to use effeciently for these tensor AI backends.
 
  • Haha
  • Like
Reactions: Appletoni and jdb8167
L

Leifi

macrumors regular
Nov 6, 2021
128
121
jeanlain said:
Yeah, and posting some random barplot without any link or description doesn't help their case.
Click to expand...
Links to the talkchess threads have been posted.. Plase read trough the discussion before posting.. I will post links again for your convenience..




 
  • Like
Reactions: Appletoni
L

Leifi

macrumors regular
Nov 6, 2021
128
121
Wando64 said:
I am trying to summarise after 20 pages of arguing and I concluded that if you want to play chess you should buy a chess-board.
For anything else, an M1 based computer will do just fine.

Phew
Click to expand...

Duuh.. thats kind of like saying "if you like photography you should buy a camera instead of a laptop for editing"

A 4800u computer with discrete GPU will accidentally also "do just fine" at anything else, just a huge amount better at Chess, Games, AI, etc. than an M1-based computer priced similar.
 
Last edited:
  • Like
Reactions: Appletoni
L

Leifi

macrumors regular
Nov 6, 2021
128
121
jeanlain said:
... it is up to you to demonstrate that the Apple microarchitecture cannot be good at chess.
Click to expand...

Absurd argument..

It has already been tested that the perfromance is sub par when running the same open-source C-code compiled with state of the art compilers.

Going for the russel teapot argumentation is just laughable.

en.wikipedia.org

Russell's teapot - Wikipedia

en.wikipedia.org en.wikipedia.org

The benchmark is what it is.. If you or anyone claims the speed could be significantly improved (by magic) the burden of proof lies upon you, to prove how much speed can be gained, and how.
 
Last edited:
  • Like
Reactions: Appletoni
J

jeanlain

macrumors 68020
Mar 14, 2009
2,463
958
Leifi said:
It has already been tested that the perfromance is sub par when running the same open-source C-code compiled with state of the art compilers.
Click to expand...
I'm not sure which code you refer to, because the M1 performs very well in many apps and several industry-standard benchmarks. SPEC tests for instance. It's the same code compiled for different architectures.
You find a piece of code on which the M1 is slow and claim that the hardware is inherently slower. Then how do you explain the M1 outstanding performance in other tasks, if it's slower? Magic?

Leifi said:
significantly improved (by magic)
Click to expand...
It seems you don't understand what optimising the code can do.
It can speed things up a lot. For instance, x265 got >50% faster after Apple included some optimisations for the M1.
 
  • Like
Reactions: JMacHack
J

jeanlain

macrumors 68020
Mar 14, 2009
2,463
958
Leifi said:
Links to the talkchess threads have been posted.. Plase read trough the discussion before posting..
Click to expand...
Try to post links to results in the same post. It's not up to us to find them if you want to make your point.
 
Status
Not open for further replies.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom