Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Performance of Nano versus Atom
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Performance of Nano versus Atom

bitswitchbitswitch Member

A VIA Nano U2250 should be theoretically somewhat slower than an Atom N2800 (at least for multicore), shouldn't it?

I recently got a Kimsufi machine and that statement does not seem to apply, instead it seems to be actually the other way round. The N2800 is not only in synthetic benchmarks noticably slower than the U2250 (sysbench - single core to be fair though) but also in real world scenarios. For example, a configure of Apache took 70 seconds on the U2250 versus 105 seconds on the N2800 (+50% on supposedly faster hardware).

Even if the single core performance of the N2800 is (for whatever reason) is slower than the U2250's I'd still "assume" in a real world case it should be overall faster. Am I measuring wrong, do I have wrong expectations or is there a performance issue?

Both machines are running a 32 bit Debian, the U2250 Wheezy, the N2800 Stretch. The N2800 is using Debian's standard 4.9 kernel instead of Kimsufi's custom 3.x one (though I think that shouldnt matter in this case here).

Comments

  • rm_rm_ IPv6 Advocate, Veteran
    edited May 2017

    I can confirm this, was surprised to find out some time ago that Nano is about 40% faster than N2800 on single-thread tasks. For example generating a DokuWiki page (PHP & Lighttpd), typical result across multiple runs:

          U2250      N2800
    real  0m0.143s   0m0.210s
    user  0m0.024s   0m0.032s
    sys   0m0.024s   0m0.020s
    

    The saving grace with N2800 is that it has 2 cores and HT on top, so if you have multiple visitors accessing your website, those can be handled in parallel, even if each single request is handled slower.

    The reason for faster single-core could be that Nano is an Out-of-order execution CPU (and generally does a whole ton of advanced stuff), while Atom N2800 is a much simpler In-order one "without any instruction reordering, speculative execution or register renaming.".

    Thanked by 1sin
  • is there any proper benchmark?

  • Thanks @rm_, so my findings are not completely off then and somewhat to be expected :).

  • SplitIceSplitIce Member, Host Rep
    edited May 2017

    I find it rather impressive that VIA is able to launch a competitive processor in this day and age given the market cap of Intel (and to a lesser extend AMD).

  • rm_rm_ IPv6 Advocate, Veteran
    edited May 2017

    SplitIce said: in this day

    Via Nano has been launched in 2008. (Atom N2800 in December 2011)

  • tmwctmwc Member

    Off topic: I actually thought you were talking about the text editors Nano and Atom for some reason.

  • @bitswitch said:
    A VIA Nano U2250 should be theoretically somewhat slower than an Atom N2800 (at least for multicore), shouldn't it?

    As you've found out, not necessarily. It comes down to how the chip is designed, and the IPC of the core.

    The VIA Nano should be a full OOO superscaler core more comparable to Intel's and AMD's mainstream cores instead of their in-order mobile Atom cores. It was a modern, low-power x86 chip targeting the embedded market rather then a Intel's take on an ARM clone, which was the appeal of them.

    At some point Intel switched the Atom cores to OOO cores as well, but I don't remember exactly when that happened. It might have been just recently with the 3000 series. I definitely know those aren't in-order anymore.

    Too bad people could never buy the VIA Nanos. I really wanted one at the time.

  • flatland_spider said: As you've found out, not necessarily. It comes down to how the chip is designed, and the IPC of the core.

    I did, yes. My initial assumption was that the Atom would be "somewhat" faster but that seems to apply only to multi-core tasks. An untar, for example, completed faster than on the on the Nano. Well, learned something new :)

Sign In or Register to comment.