X86-64, ARMv7 vs new ARMv8 CPU performance


#1

Scaleway

As you have read in the blog about new Scaleway Disruptive ARMv8 Cloud Servers

Scaleway is the first cloud computing provider worldwide to offer a true solution for multi-architecture project with ARMv7, ARMv8 and X86-64 cloud servers.

Has anyone tested those new ARMv8 servers and can tell if there are any difference with the previous ones?

Percentage or comparison of script execution times, etc.?

Update

Here is a post from Hackernews user mtgx

Worth pointing out that these are new ARM processors, potentially 3x faster than the (very) old ARM processors it was using before.

I guess it is a good point to motivate me for a test ride. :grinning:


CPU performance
#2

I did some sysbench CPU test runs (10000 prime numbers). It’s a very basic test, but here are some ballpark figures:

  • x86-64 “Starter” (VC1), Atom Avoton C2750, 10000 primes in 9s on one core
  • ARMv7 “BareMetal” (C1), 10000 primes in 258.5s on one core
  • ARMv8 “Starter” (ARM64), 10000 primes in 6s on one core
  • x86-64 “Intensive”, not tested

This is a very parallelisable task, so e.g. 2 cores took half the time of 1 core. I suspect one of the operations used was particularly bad for the ARMv7 core, since the result is very poor.

Of course, you must consider your specific use case - this is a very simple benchmark. You can test yourself at the low €/hour rate Scaleway charges…


Starter X86-64 vs ARM64-2GB ARMv8
[SOLVED] - What are Scaleway core frequencies
#3

Thanks, looks promising.

I’m currently using only x86-64 “Starter” (VC1) containers. A 30% performance gain is worth testing.


#4

But these aren’t baremetal or dedicated like the C1s right? I mean what’s the purpose of using the ARM VPSes over the regular VPSes unless you’re a developer.


#5

30% increase in performance for the same price seems to be a very very good reason to.

Anyone tested switching from VC to an ARMv8 instance with an image / snapshot from VC yet?


#6

Not baremetal? Good for some but no thanks! :slight_smile:


#7

have a look into the machines they are hosting the new ARMv8 vps on :wink: I think its just not economical beneficial to host baremetal system with 48 cores.


#8

Understandable but still I wouldn’t touch a poorly isolated low performance VPS shared with 6-24 noisy neigbours.
After two decades using dedicated and shared servers I’ve learnt my lesson… :smile:


#9

Since they’re using KVM, it’s actually pretty well isolated. Not like those OpenVZ VPS providers (ugh).

I assume they actually assign/pin dedicated cores to each instance and don’t overload the servers.

Maybe there’ll be some disk contention, dunno what they do there… that’s probably the one place noisy neighbours might be an issue. Then again, the baremetal instances use network drives, which isn’t any better.


#10

I’m afraid you’re right on the network drives issue. :slight_smile: