M1 Pro First Impressions: Core Management and CPU Performance • Hacker News Board

Re: M1 Pro First Impressions: Core Management and CPU Performance

#191

Post by rsynnott » Fri, Nov 05, 2021, 2:29 PM UTC

AnthonyMouse wrote: post #53

zibzab wrote: post #20

Earlier quoted context omitted.

Unity is slow on any machine. With that said, there is a huge M1 hype on HN. It reminds me of the 64-bit ARM hype when someone benchmaked encryption (because ARMv8 has AES instructions) and everyone were made to belive that the new iPhone would be 3-4 times faster.

> With that said, there is a huge M1 hype on HN. The main thing driving this is that people are comparing a new M1 to their Intel Mac laptops with processors from ~2016 and then naturally it's a marked improvement because it's being compared to a five year old machine. You compare them to a modern Intel or AMD system and they're competitive. They win some and lose others. Much was made about the M1 Max taking the sin…

> Golden Cove from Intel just took it back.

I mean, yes, somewhat, in (almost comically) high power desktop configurations, depending on which benchmark you believe. This is of cold comfort to anyone who wants a fast laptop, though.

Re: M1 Pro First Impressions: Core Management and CPU Performance

#192

Post by BoorishBears » Fri, Nov 05, 2021, 2:30 PM UTC

kllrnohj wrote: post #188

BoorishBears wrote: post #163

Earlier quoted context omitted.

Lol what is this vaguely accusatory tone? First off the M1 Air has no fans. Second the MBP 16" was the one with the improved cooling, so it wasn't throttled randomly by using the wrong charging port. And finally the whole reason the M1s are so well cooled is there's no 100W SoC, or 200W probably when you consider the additional dedicated GPU in my 16". My M1 has a 10W TDP! 10! I don't think the GPU alone in my 16" co…

> First off the M1 Air has no fans. Your M1 Air doesn't have an M1 Pro or Max, either, which I clearly specified. Why are you talking about M1 power draw in your response when that's not the SoC in question? > Also note: in your other comments you keep mixing TDP and wall power draw, Power in == heat out. CPUs don't do mechanical work, all power is converted to heat. If the 'TDP' is less than peak power draw (and is…

Because the thread was about why even an M1 not pro is beating a 16"?

You're free to bring in your third option which clobbers it even further, but an $800 fanless machine also clobbers it so it's a perfectly valid comparison when trying to counter weird claims that Apple only just now started working on proper cooling...

tl;dr: How can you claim the old 16" is only being beat because Apple just now started to care about cooling when a fanless machine also beats it? -

Also TDP is an advertising figure but when you're talking about orders of magnitude it's a perfectly fine way to talk about parts as long as you stay consistent in TDP vs wall draw

Wall draw to wall draw, my M1 Air draws under 30W and beats a 16" drawing over 100W

TDP to TDP my M1 is 10W to a 16" closer to 100W.

In both cases it should be obvious why cooling is less of an issue...

Re: M1 Pro First Impressions: Core Management and CPU Performance

#193

Post by jimbob21 » Fri, Nov 05, 2021, 2:31 PM UTC

ricardobeat wrote: post #175

capableweb wrote: post #129

Earlier quoted context omitted.

> From recent benchmarks you need to go into $1000+ territory for the CPU alone to get close to M1 multi-core performance I don't think so, quick comparison, with numbers from https://www.cpubenchmark.net/compare/Apple-M1-Pro-10-Core-32... : - Apple M1 Pro 10 Core 3200 MHz - Score: 23730 - AMD Ryzen Threadripper 1950X - Score: 27293 - Price: ~$499 And I just spent 2 minutes digging up those numbers, didn't even look…

The latest Ryzen CPU is the 5950X, costing near 1k, and it’s behind M1 in that same site’s leaderboard: https://www.cpubenchmark.net/singleThread.html All the recent M1 benchmarks from respectable news sites confirm the same for the majority of multi core real workloads. tl;dr you absolutely will not beat M1 Pro with a $1500 PC for work.

The 5950x costs $750, you can buy it on Amazon right now for that. Much less than 1k. Also, you're comparing single thread performance on only one website. In every website it kicks the M1 Pro's ass on multicore, as it should with 16 cores. https://nanoreview.net/en/cpu-compare/apple-m1-pro-vs-amd-ry... It can also beat it by a hair in single core depending on the benchmark.

Don't be disingenuous just to win the argument.

To actually prove your point, you need to look at GPUs. A 3070 is ~$1100, so there we go, already we've broken his $1600 budget, and we have bought none of the other parts (which would total at minimum ~$1000 more).

If you go 5900x plus 3070, you're pretty much right in line with M1 Pro, but it will cost you around $2500 to equal the Macbook pro. So an upcharge of $1000 to $1500 for the price of being a mobile workstation, as well as all the Mac software. Not really unreasonable.

Re: M1 Pro First Impressions: Core Management and CPU Performance

#194

Post by weatherlight » Fri, Nov 05, 2021, 2:33 PM UTC

floatboth wrote: post #149

ReaLNero wrote: post #50

I have an M1 Max. If anyone wants me to benchmark any workflows, I'd be glad to help!

Relatively obscure fun thing: CRC32 checksum instructions. https://github.com/srijs/rust-crc32fast/pull/6 (Get rust nightly via https://rustup.rs , pull the repo, run: cargo bench --features nightly)

(16-inch, 2019) - 2.6 GHz 6-Core Intel Core i7 - 16 GB 2667 MHz DDR4

    test result: ok. 0 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 0.00s

     Running unittests (target/release/deps/bench-b9bcbb53a5ef2c28)

    running 4 tests
    test bench_kilobyte_baseline    ... bench:         310 ns/iter (+/- 70) = 3303 MB/s
    test bench_kilobyte_specialized ... bench:          48 ns/iter (+/- 5) = 21333 MB/s
    test bench_megabyte_baseline    ... bench:     298,273 ns/iter (+/- 42,157) = 3515 MB/s
    test bench_megabyte_specialized ... bench:      45,100 ns/iter (+/- 7,249) = 23250 MB/s

    test result: ok. 0 passed; 0 failed; 0 ignored; 4 measured

-----------------------------------------------------------

(14-inch, 2021) - Apple M1 Max - 64gb

    test result: ok. 0 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 0.00s

     Running unittests (target/release/deps/bench-3eea1c397faf4328)

    running 4 tests
    test bench_kilobyte_baseline    ... bench:         223 ns/iter (+/- 8) = 4591 MB/s
    test bench_kilobyte_specialized ... bench:         100 ns/iter (+/- 0) = 10240 MB/s
    test bench_megabyte_baseline    ... bench:     231,689 ns/iter (+/- 1,885) = 4525 MB/s
    test bench_megabyte_specialized ... bench:     122,382 ns/iter (+/- 4,651) = 8568 MB/s

    test result: ok. 0 passed; 0 failed; 0 ignored; 4 measured

Re: M1 Pro First Impressions: Core Management and CPU Performance

#195

Post by jmnicolas » Fri, Nov 05, 2021, 2:36 PM UTC

tpxl wrote: post #164

jmnicolas wrote: post #128

Earlier quoted context omitted.

> And each time I read a post like this I wonder: what do these people _do_ with these machines? To give you an idea, I don't have Macs, but I have a beefy desktop because I run my own OpenStreetMap server on it. My old machine: i7 3770, 32GB ram, 1 TB sata SSD: 3 days to import Europe's map. The new one: Ryzen 5800, 32GB ram, 1 TB NVME: 10 hours to import Europe's map. I will probably need to upgrade my ram and NVME…

> because I run my own OpenStreetMap server on it. What do you do with it?

I'm mainly generating map tiles for offline use. I don't want to hit OSM official servers, this is too intensive (and they would probably block me).

Re: M1 Pro First Impressions: Core Management and CPU Performance

#196

Post by jquery » Fri, Nov 05, 2021, 2:41 PM UTC

ArgyleSound wrote: post #123

jquery wrote: post #91

Earlier quoted context omitted.

Power usage doesn’t scale linearly, at all. Mobile Ryzens are very competitive with the M1 watt for watt (trading blows with Ryzen winning sometimes). Seems you’ve fallen for Apple marketing again.

I am very Interested in a good mobile Windows/Ryzen laptop with a MacBook Air-comparable form factor, performance, and battery life. Please link if you know of one. All the Intel machines I've looked into have fans and still manage to be 30-50% slower.

Me too. Let me know if you find one. If I was in the market for something like the MacBook Air, I’d definitely get the MacBook Air.

I’m fact I use an M1 iPad right now. It’s great. For browsing Facebook and video calls and doing remote work, it’s perfect. And for serious workloads, those go straight to my desktop workstation (AMD). Best of both worlds.

Re: M1 Pro First Impressions: Core Management and CPU Performance

#197

Post by sydthrowaway » Fri, Nov 05, 2021, 2:44 PM UTC

blitzar wrote: post #94

sydthrowaway wrote: post #18

These laptops are too expensive. What the hell is the markup on these things

You are too poor.

v true.. I am european

Re: M1 Pro First Impressions: Core Management and CPU Performance

#198

Post by BackBlast » Fri, Nov 05, 2021, 2:58 PM UTC

Zababa wrote: post #102

DSingularity wrote: post #63

Earlier quoted context omitted.

They will skip desktop and ship server chips. If the trends hold, they will have TCO savings which no cloud provider can resist.

I know that AWS already have their own ARM chips: https://aws.amazon.com/ec2/graviton/ . I don't know about chips enough to understand how large is the gap between AWS and Apple when it comes to designing chips though.

I run arm servers in production. M1 performance core performance is about 3x graviton 2 and 2.5x ampere ultra in our stack. The efficiency core is about 0.6 graviton and a bit less than half ampere.

Would love to have reasonably costed m1 instances but the value play for us is ampere.

Re: M1 Pro First Impressions: Core Management and CPU Performance

#199

Post by KptMarchewa » Fri, Nov 05, 2021, 3:05 PM UTC

codeisawesome wrote: post #136

KptMarchewa wrote: post #110

Earlier quoted context omitted.

Do you have two 4k displays? If yes, then does running them affects performance significantly? On i9 MBP this was causing extremely severe performance issues, to the point where processor was throttled to 30% performance.

I run two 5Ks and my 2019 Intel feels quite sluggish. I still pay that price for the real estate.

That's what I would love to avoid.

Re: M1 Pro First Impressions: Core Management and CPU Performance

#200

Post by alwillis » Fri, Nov 05, 2021, 3:14 PM UTC

foldr wrote: post #108

jquery wrote: post #91

Earlier quoted context omitted.

Power usage doesn’t scale linearly, at all. Mobile Ryzens are very competitive with the M1 watt for watt (trading blows with Ryzen winning sometimes). Seems you’ve fallen for Apple marketing again.

What Ryzen laptop would you recommend with performance and battery life roughly equivalent to an M1 MacBook Air? A lot of people on HN obsess over benchmarks, but people buy laptops, not CPUs. I haven’t yet seen a product that’s competitive with the M1 MacBook Air in its niche (which to be fair isn’t really a niche - a laptop for what 90% of people use their laptops for).

I haven’t yet seen a product that’s competitive with the M1 MacBook Air in its niche

This is the inconvenient truth of these discussions. If anything, Apple is under playing the bang for the buck and performance per watt of these machines.