Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

twitter.com

Post reply on HN
Print view View original

401–410 of 529 posts

Re: 16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

#401

Post by ummonk »

cercatrova wrote: post #26

Is anyone else thinking, what the f*ck? Are we in a new era of computing? It certainly feels that way when looking at these desktop class ARM chips, where performance doubled every year or so, just like back in the 80s and 90s.

It would be normal if Moore's law hadn't dropped off in recent years.

Re: 16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

#402

Post by unsigner »

sitkack wrote: post #389
epistasis wrote: post #202

Earlier quoted context omitted.

I'm no expert, but the only big architectural differences are a massively larger decoder and a reorder buffer that's several times as large as x86 designs. If these are actually the reasons for the performance difference, and it's difficult to do these on x86 because of the instruction set, it seems to this amateur that ARM64 really does have an advantage over x86.

The ISA differences between ARM and x86 do not account for the difference in performance, there are multiple factors here (process, ssd, memory bandwidth, cache, thermal reservoir, etc). While this is wonderful for ARM in the now-term, we just moved from walled ISAs to a plurality of ISAs, compute just became a bulk commodity in a way that it could not with an x86 duopoly. Anyone can now take off the shelf RISC-V des…

There’s not a shred of evidence RISC-V can approach the levels of performance discussed in this thread. There’s a lot of “big implementations can potentially do X” hand waving in RISC-V land, and not much real silicon to show for it.

Re: 16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

#403

Post by wmf »

pwinnski wrote: post #322
whalesalad wrote: post #48

I love everything about the move to Apple Silicon with the exception of the decision to put memory on-die or in-package (not sure how it is configured). They call it 'Unified Memory'. It makes a lot of sense but I don't know if they are going to be able to pack enough memory in there. A lot of folks are fixated on CPU performance lately (which is rad) but I think that there is a tendency to ignore memory. I have 32gb…

Putting the memory into the SoC seems to be a lot of how they're getting this amazing performance: the memory bus is now twice as wide as anything else. But yes, it means that Apple is going to have to either really start jamming more and more into there, or develop a two-tiered approach to memory where slower "external" RAM can supplement the faster "internal" RAM.

The memory bus is now the same width as every other laptop.

Re: 16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

#404

Post by jacquesm »

This is a super bad benchmark for many reasons, target architecture is different, underlying cache, storage device, bus width and possible optimizations are all different. Compile time in such a situation means next to nothing without isolating all these factors and what's left is then attributable to the M1 vs Intel difference. That would be a lot of work though and would not make for such a catchy and clickable title.

Re: 16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

#405

Post by unsigner »

twoodfin wrote: post #290
epistasis wrote: post #202

Earlier quoted context omitted.

I'm no expert, but the only big architectural differences are a massively larger decoder and a reorder buffer that's several times as large as x86 designs. If these are actually the reasons for the performance difference, and it's difficult to do these on x86 because of the instruction set, it seems to this amateur that ARM64 really does have an advantage over x86.

Don't forget ARM's more relaxed memory model vs. x86's TSO.

One of the reasons Rosetta 2 works so well is Apple silicon sticks to the more restricted x86 memory model.

Re: 16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

#408

Post by mikechen233 »

Eric_WVGG wrote: post #333
whalesalad wrote: post #48

I love everything about the move to Apple Silicon with the exception of the decision to put memory on-die or in-package (not sure how it is configured). They call it 'Unified Memory'. It makes a lot of sense but I don't know if they are going to be able to pack enough memory in there. A lot of folks are fixated on CPU performance lately (which is rad) but I think that there is a tendency to ignore memory. I have 32gb…

I'm afraid I lost the source, but this morning I was reading about some fairly in-depth Xcode benchmarks, the dev was saying that there was almost no hit in performance when it hit swap, and speculated that Apple might be getting ready to move past the concept of RAM altogether in a few years. Sounds a little bonkers to me, but the bus to the SSD is no joke

Next gen consoles and Nvidia rtx io is kind of doing that with graphics textures. Textures are memory mapped but stored in nvme storage. When GPU reads a texture, it first looks in the ram, if not found in the ram, the controller talks to nvme controllers to load from storage.

If you think about it, gen4 pce nvme can reach 5GB/s doing 16k ramdom reads. The bandwidth is getting close to ddr2/3 territory. And new storage tech like 3dxpoint will have ram like access latency to improve small io perf.

You will always still need ram, but you can be more efficient at how to use the ram.

Re: 16-inch MBP 2x slower than M1 MacBook Air in a real-world Rust compile

#409

Post by unsigner »

bob1029 wrote: post #378
cercatrova wrote: post #26

Is anyone else thinking, what the f*ck? Are we in a new era of computing? It certainly feels that way when looking at these desktop class ARM chips, where performance doubled every year or so, just like back in the 80s and 90s.

I thought this for a few seconds and then realized these sorts of benchmarks are lacking obvious controls (i.e. compiling for different architectures as noted by others). If you want a more realistic sense of how the M1 performs relative to x86 peers in raw, equivalent workloads, there are some cinebench numbers appearing out there: https://hexus.net/tech/news/cpu/146878-apple-m1-cinebench-r2... I am way more interes…

People are celebrating a kickass electric bicycle doubling the range of all the existing ones, and you demand to be shown it scaled up to the same power+price budget as an open pit mining truck, hauling dirt in three shifts. It might or might not happen, but the electric bicycle is still impressive.
Post reply on HN