Live data from Hacker News

The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

anandtech.com

161–170 of 266 posts

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#161
post #99

Earlier quoted context omitted.

Not sure if it's still the case but RISC-V was beating ARM in code size as well

I find that a bit surprising since my understanding is RISC-V is more minimal (has fewer higher level instructions) (ATM, that will change with extensions) then ARM. This paper refutes your point and is inline with my understanding: https://carrv.github.io/2020/papers/CARRV2020_paper_12_Perot... One counter point is the simpler the ISA the more work the compiler needs to do. So x86 is may be easier to write an optimi…

The paper you linked to is about embedded controller micro-benchmarks being worse by 5-10%; RISC-V's compressed ISA made some tough decisions to improve density on general-purpose code, not just micro-focused on MCUs. On something larger like SPEC, RISC-V is denser than ARM and x86.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#162
post #75

Earlier quoted context omitted.

AMD should see the writing on the wall and start designing ARM CPUs. I think that would be very exciting because they have the engineering expertise and talent necessary to compete against Apple's M1. I don't know of any other company that can do that (besides Intel, but you know...). Additionally, there is a huge hole open right now for ARM on everything else that is not an Apple product. Apple normalized ARM on con…

AMD is doing fine. "The writing on the wall". Zen3 is going to very competitive, their interconnect on chip is great. Sure, add a line for ARM. What's interesting to me, I feel like a lot of this started when intel wouldn't do the chip for the iphone. All of a sudden you had this leading edge mfg pushing arm hard which drove volumes through the roof, and all the copycats (android etc) copied the chip selection, scree…

> I wonder how different the path might have been if intel had come up with something for apple.

They kind of tried and failed. Intel is completely stuck in their "x86 is the only way" culture. x86 just can't compete in the low power space. For phones, the x86 power requirements made it effectively DOA.

Sure, Atom could have competed with arm in some markets (small laptops for example, they had some success in the early netbook days), but Intel managed to screw even that up by deprioritizing Atom development - they thought they could make more money in the server space. Short term, not a bad bet, but it wouldn't surprise me if it set Intel's eventual demise in motion. We're starting to see the signs now.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#163
post #99

Earlier quoted context omitted.

Not sure if it's still the case but RISC-V was beating ARM in code size as well

I find that a bit surprising since my understanding is RISC-V is more minimal (has fewer higher level instructions) (ATM, that will change with extensions) then ARM. This paper refutes your point and is inline with my understanding: https://carrv.github.io/2020/papers/CARRV2020_paper_12_Perot... One counter point is the simpler the ISA the more work the compiler needs to do. So x86 is may be easier to write an optimi…

The paper/presentation you link to is comparing 32-bit ARM with Thumb and RISC-V. This is mostly relevant to embedded processors, and my understanding is ARM and RISC-V is somewhat equal there.

But for 64-bit, ARM has dropped support for Thumb, while RISC-V still supports compressed instructions. In that case, RISC-V is more dense than ARM. I can't find the numbers for that right now, but I read it just yesterday so I'm pretty sure I remember correctly.

Also keep in mind that Thumb is a mode that the ARM CPU has to switch to, while RISC-V compressed instructions can be mixed with normal instructions.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#164
From the comments:

> Each Neoverse N1 core with 1MB L2 is just 1.4mm^2, so 80 of them add up to 112mm^2. The die size is estimated at about 350mm^2, so tiny compared to the total ~1100mm^2 in EPYC 7742.

So performance/area is >3x that of EPYC. Now that is efficiency!

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#165
post #162

Earlier quoted context omitted.

AMD is doing fine. "The writing on the wall". Zen3 is going to very competitive, their interconnect on chip is great. Sure, add a line for ARM. What's interesting to me, I feel like a lot of this started when intel wouldn't do the chip for the iphone. All of a sudden you had this leading edge mfg pushing arm hard which drove volumes through the roof, and all the copycats (android etc) copied the chip selection, scree…

> I wonder how different the path might have been if intel had come up with something for apple. They kind of tried and failed. Intel is completely stuck in their "x86 is the only way" culture. x86 just can't compete in the low power space. For phones, the x86 power requirements made it effectively DOA. Sure, Atom could have competed with arm in some markets (small laptops for example, they had some success in the ea…

It's a bit ironic that they owned StrongARM which of course became XScale and was sold the year before the iPhone appeared! I think some of the StrongARM team founded PASemi and others may have ended up working for Arm in Texas.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#166
post #91

Earlier quoted context omitted.

I think Chromebooks did ARM years before Apple, no?

Yes but they were terrible. Just being ARM isn't some kind of magic dust. It has to actually be fast.

If you look at the historical benchmarks of the non-apple ARM chips, though they're clearly behind apple, they have largely kept pace. Huawei was perhaps even closing the gap before the trade war knocked them out of contention.

I don't think apple was really all that special in this regard. They were ahead of the curve, which is impressive, but the market was created by arm, not just apple. I doubt we'd be seeing this kind of resurgence had others - especially google - not also jumped on the ARM bandwagon. A rising tide lifts all boats; both android and ios likely benefited from each others shared usage of arm because it helped the whole ecosystem.

For some sense of scale: qualcomm's latest chip just released some benchmark numbers, and while they're not all that impressive (https://www.anandtech.com/show/16325/qualcomm-discloses-snap...) it's still good to put "not that impressive" in perspective: they're comparable to i7-6700k in single-threaded workloads, and to an i7-4770k in multi-threaded workloads. The oldest iPhone that runs the same geekbench version I can find is the iPhone 5s, which is a little more than 4x slower in the single-threaded benchmark, and 8x in the multi-threaded. Yet that chip was no slouch!

Obviously, apple's impressive pole position completely hogs the spotlight, but it's worth remembering that the competition isn't actually all that bad. That level of performance would make for an extremely fast chromebook - were it not for the fact that most are cheap cheap cheap and will never use a high-end chip.

Development across the board has been so fast, that if you remember slow ARM chromebooks, I think you attribute that slowness to 2 things other than apple vs everybody else: firstly, they were built to a bargin-basement price, and thus you weren't getting anything near this performance not because it wasn't apple, but simply because they didn't use the right chips in the first place, and secondly, it's been a while, and when things improve that quickly, then a slightly outdated model can easily appear terrible without that being a condemnation of the whole approach.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#168
post #156

Earlier quoted context omitted.

Probably connected: https://www.extremetech.com/electronics/317329-tsmc-will-ope... Not sure if it has to be absolutely top-tier for defense purposes though.

You can go with larger nodes and larger power budgets for avionics and other electronic stuff in battlefield machines. OTOH you'd rather be top-notch in your communication equipment, low-power sensors for reconnaissance, etc. In some areas, having a US-controlled 7 nm fab may matter even for DoD.

Interesting - thanks.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#169

This line is the kicker for me: "Intel’s current Xeon offering simply isn’t competitive in any way or form at this moment in time." The rest of the paragraph twists the knife more, and declares there to be an open market on server hardware for AMD and Ampere to fight it out. With application stacks being what they are and optimizations for Xeon so deeply integrated, it's unlikely that folks running private datacenter…

are there any good articles about what has gone so, so wrong at intel in the last few years?

There's a lot of great stuff at SemiAccurate throughout the years that details what went wrong at Intel and why. Unfortunately, you have to pay for it because Charlie has a practical stranglehold on semiconductor inside information.

And Charlie is a bit of a polarizing figure. He's gotten some things hilariously wrong, but he's also been dead on the money for a lot of issues, and his documentation of what's happening / happened at Intel is pretty much crystal clear.

You could probably browse through all the 10nm articles there and get a good idea of what happened without purchasing a subscription, if you have a decent knowledge of the semiconductor industry.

EDIT: If you read though these articles, even the parts that are available to non-registered users, a pretty clear picture starts to emerge - https://www.semiaccurate.com/tag/10nm/

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#170
post #101

Earlier quoted context omitted.

I'm excited as well. But as an earlier discussion pointed out I'm worried that the transition might impact the openness in PCs, e.g. locked bootloaders etc

There is already a few ARM laptops that are quite open (like the pine64 ones). There's also ARM boards like the Raspberry PI. I hope that if there's more beefy CPUs and SoCs we'll see those poping up in similar devices as well.

Sort of. Even the Pi is not entirely open-source. The Broadcom SOC relies on some closed-source binaries.
Post reply on HN