Announcement

Collapse
No announcement yet.

39 System Intel & AMD CPU Benchmarks On Linux

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Michael
    replied
    Originally posted by PFee View Post

    Looking at the results I don't see any Skylake CPUs, the closest I see are Broadwell CPUs (5XXX). I'd like to see Skylake Core i7 vs Xeon, since Dell for example will sell a Precision 3620 with i7-6700 or E3-1245 v5 for the same price. Therefore I'd like to know if Xeon offers anything against Core-i7.

    Previously I've experienced a desktop (Dell Optiplex) SandyBridge unit outperform a significantly more expensive Xeon based Dell R720 unit.
    I don't have any Core i7 Skylakes.... But as pointed out, there are Xeon Skylakes and Pentium Skylakes, etc in thsi comparison.

    Leave a comment:


  • devius
    replied
    Originally posted by PFee View Post
    Looking at the results I don't see any Skylake CPUs...
    You didn't look enough then. All Xeon E3 v5 in that chart are Skylake, as is the Pentium G4400.

    Leave a comment:


  • PFee
    replied
    Originally posted by Michael View Post

    I don't own a i7 6700... Intel never ended up sending out desktop SKL samples to me so the only ones I have are the Xeons or Core i5s I bought.
    Looking at the results I don't see any Skylake CPUs, the closest I see are Broadwell CPUs (5XXX). I'd like to see Skylake Core i7 vs Xeon, since Dell for example will sell a Precision 3620 with i7-6700 or E3-1245 v5 for the same price. Therefore I'd like to know if Xeon offers anything against Core-i7.

    Previously I've experienced a desktop (Dell Optiplex) SandyBridge unit outperform a significantly more expensive Xeon based Dell R720 unit.

    Leave a comment:


  • VikingGe
    replied
    Hoping to add my home desktop (Phenom II x6 1055t) after I get home for comparison along with an i7-2600/i7-2700 so that I can actually figure out whether it's worthwhile to upgrade my Phenom to an FX-83xx or wait for Zen.
    I've done that comparison several times in the past and more often than not, my Phenom 1090t (@3.8/4.1, 3.0 NB) ended up being faster than the stock FX-8350, and not much slower in some other tests. FX will be quite a bit faster than a stock 1055t for sure, but seriously, wait for Zen.

    Leave a comment:


  • sterky
    replied
    So out of curiosity i downloaded parallel version of himeno benchmark from here http://accc.riken.jp/en/supercom/him...pp/#itemid4564 - Fortran 77 + MPI version
    Then unpacked and configured it like that: ./paramset.sh S 1 1 8
    Then compiled it: mpifort himenoBMTxpr.f
    After that ran the code with this command: mpirun -np 8 ./a.out
    So now it was using all 8 cores of my FX-8320
    And the result was this: MFLOPS: 2489.7217444251705 time(s): 60.719013929367065

    Edit: Made a simple script to test it, you do need couple of packages but at least on arch linux it's pretty easy to install them
    http://pastebin.com/CGfZrB4J
    Last edited by sterky; 27 February 2016, 09:42 AM.

    Leave a comment:


  • atomsymbol
    replied
    Originally posted by Kano View Post
    Most likely the Pentium G4400 was benchmarked with cpufreq ondemand vs most others with p-state powersave.
    No.

    Originally posted by Kano View Post
    It seems the Turbo did not kick in for some benchmarks.
    Which benchmark(s) do you mean?

    Originally posted by Kano View Post
    Maybe the threshold to increase frequency is differently defined - or some benchmarks use AVX-512.
    Pentium G4400 supports SSE 4.1/4.2.

    Pentium G4400 does not support AVX nor AVX2.

    None of the tested CPUs support AVX-512.

    Leave a comment:


  • atomsymbol
    replied
    Originally posted by xpander View Post
    FFTE and Himeno benchmarks use single core only, not threaded at all.
    Doesn't it appear strange to you how the human mind can parallelize the code and why the machine cannot do something similar?
    Last edited by atomsymbol; 27 February 2016, 06:30 AM. Reason: Fix grammar

    Leave a comment:


  • Kano
    replied
    Most likely the Pentium G4400 was benchmarked with cpufreq ondemand vs most others with p-state powersave. It seems the Turbo did not kick in for some benchmarks. Maybe the threshold to increase frequency is differently defined - or some benchmarks use AVX-512.
    Last edited by Kano; 27 February 2016, 12:33 AM.

    Leave a comment:


  • xpander
    replied
    Originally posted by devius View Post
    How come the Pentium G4400 has the same performance as a Core i7 4790K in Himeno Benchmark and FFTE?
    FFTE and Himeno benchmarks use single core only, not threaded at all. intel rules in singlecore perf.
    luckily on linux most of the stuff is nicely multithreaded

    Leave a comment:


  • atomsymbol
    replied
    Originally posted by devius View Post
    How come the Pentium G4400 has the same performance as a Core i7 4790K in Himeno Benchmark and FFTE?
    In my opinion, the shortest answer is: Mankind doesn't know how to do auto-parallelization in a compiler yet.

    Leave a comment:

Working...
X