flat assembler
Message board for the users of flat assembler.
Index
> Unix > M1 processor performance versus x86 |
Author |
|
revolution 08 Dec 2020, 22:25
Link?
|
|||
08 Dec 2020, 22:25 |
|
donn 08 Dec 2020, 23:05
https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested for starters, word's really getting around.
|
|||
08 Dec 2020, 23:05 |
|
Furs 10 Dec 2020, 14:09
What 5nm x86 CPU did they compare with?
|
|||
10 Dec 2020, 14:09 |
|
donn 10 Dec 2020, 23:18
Screenshot from first link. Looks like it beats AMD Ryzen 9 5950X there. I'm told AMD is 7nm, not sure...
More broadly, from here: https://www.anandtech.com/show/16226/apple-silicon-m1-a14-deep-dive Quote: The new CPU core is what Apple claims to be the world’s fastest. This is going to be a centre-point of today’s article as we dive deeper into the microarchitecture of the Firestorm cores, as well look at the performance figures of the very similar Apple A14 SoC.
|
|||||||||||||||||||
10 Dec 2020, 23:18 |
|
revolution 10 Dec 2020, 23:37
Certainly single thread performance is important as a lot of code (and coders) don't know how to use threads.
But it's Apple, so that means no one can buy the chip to make a system. You have to buy the system from Apple, no one else. And the whole thing will be locked down tightly. So I am sceptical. Performance is one thing, being able to use it without asking for Apple's permission is entirely another. |
|||
10 Dec 2020, 23:37 |
|
Melissa 13 Dec 2020, 11:55
Do we have code for geekbench? I heard it favors ram speed.
|
|||
13 Dec 2020, 11:55 |
|
Furs 13 Dec 2020, 15:32
Yeah and 5nm vs 7nm (both TSMC) gives more raw transistors which is why M1 can afford such a large cache.
|
|||
13 Dec 2020, 15:32 |
|
donn 16 Dec 2020, 02:18
I can't see any information on how the benchmark works from their website..
I found some discussion on it here, here however: https://news.ycombinator.com/item?id=20165254 Quote: Geekbench exposes some strange behaviour around the memory allocator under Windows. On systems with more than 8 cores Geekbench spends a significant chunk of time in the memory allocator due to contention. This issue (at least to this degree) isn't present on Linux, so that's why Epyc scores are much higher on Linux than Windows. It's been around a while, maybe some are gleaming how it works indirectly, just by observing results over time. Also, this is exactly what I was looking for, which backs up the 5nm debate: https://news.ycombinator.com/item?id=25257932 Quote: The i7-1165G7 is within 20% on single core performance of the M1. The Ryzen 4800U is within 20% on multi-core performance. Both are sustained ~25W parts similar to the M1. If you turned x64 SMT/AVX2 off, normalized for cores (Intel/AMD 4/8 vs Apple 4+4), on-die cache (Intel/AMD 12M L2+L3 vs Apple 32MB L2+L3) and frequency (Apple 3.2 vs AMD/Intel 4.2/4.7), you'd likely get very close results on 5nm or 7nm-equivalent same process. Zen 3 2666 vs 3200 RAM alone is about a 10% difference. The M1 is 4266 RAM IIRC. |
|||
16 Dec 2020, 02:18 |
|
Melissa 20 Jun 2021, 05:07
Bought macbook, this CPU is seriously fast. Blows away my desktop, blows away. Eg 3900 scimark2 points vs 3050 for my 2700x.
my problem512 mt solution for minut and a half, while 2700x with 16 threads 5 minutes. It's not memory, CPU is beast... |
|||
20 Jun 2021, 05:07 |
|
bitRAKE 20 Jun 2021, 07:35
This blog post is a nice read and the comments list some related patents if one wants to travel down the rabbit hole.
_________________ ¯\(°_o)/¯ “languages are not safe - uses can be” Bjarne Stroustrup |
|||
20 Jun 2021, 07:35 |
|
ford 21 Jun 2021, 14:34
So, I was initially quite happy with M1. One of my employers issued me a Macbook Pro with M1, and I was shocked at just how good it actually seemed to be. It handled seemingly everything I tossed its way, and didn't even get warm. The battery life could sometimes drain if I was using a ton of stuff in Rosetta, but that has since waned after an OS update. Then, I discovered a few things that were disappointing.
First, it has terrible, and I mean dismal, IO. Performance reading/writing to an external drive is atrocious. Performance to/from a web share isn't bad at all, but if you actually want to plug in a large external disk drive via USB, it's really really bad for any large file operations. Second, the number of PCIe lanes must likewise be very low, because this poor performance gets worse if you actually have a large number of devices plugged in. Third, in any seriously large and multithreaded task, my Threadripper rips the M1 apart. Altogether, if you use more than a handful of external devices, you should avoid M1, and if you require tons of threads you should avoid M1. On a high note, most people will never notice the deficiencies that I found as laptops generally are not being used for large parallel workloads or for massive data operations. I am also very excited for the future of M1. The UTM app is available in the Mac App Store for M1 so you do have a good virtual machine system that doesn't cost tons of money, and if I could use UTM on a more powerful Apple chip it could become very attractive. My final note is that I have been disappointed with Apple on the UNIX front. They started out strong with Darwin and frequent releases of their UNIX system. Things were up to date at all times, and Apple had produced a UNIX that was truly worthy of the name. This has slacked off over the years, and many packages in Apple's UNIX system are far behind the stable current releases, and Apple has stopped releasing Darwin images (sources are still available however). |
|||
21 Jun 2021, 14:34 |
|
Melissa 14 Jul 2021, 21:16
Hm I have 100 megabytes/s via netatalk on Linux server (wifi)
haven't tried tried external drive only apple netatalk and that seems fast, I play and install Steam games on network drive |
|||
14 Jul 2021, 21:16 |
|
< Last Thread | Next Thread > |
Forum Rules:
|
Copyright © 1999-2024, Tomasz Grysztar. Also on GitHub, YouTube.
Website powered by rwasa.