Changes between Version 15 and Version 16 of LimulusBenchmarks


Ignore:
Timestamp:
02/03/16 08:45:56 (9 years ago)
Author:
admin
Comment:

added skylake i5-6500

Legend:

Unmodified
Added
Removed
Modified

  • LimulusBenchmarks

    v15 v16  
    33=== HPL Performance === 
    44 
    5 ==== i5-2400S ==== 
     5==== i5-2400S Sandy Bridge ==== 
    66  * '''200.3 GFLOPS''' N=40220 ([wiki:i5-2400S-Raw-HPL Raw HPL Results])  
    77  * 58% of Peak (3.3GHz * 4 cores * 8 DP FLOPS/cycle) + (2.5Ghz * 12 cores * 8 FLOPS/cycle) = 345.6 GFLOPS Peak 
    8   * Three i5-2400S and one i5-2500K each with 4 MB RAM, GbE, Intel MKL and compilers 
     8  * Three i5-2400S and one i5-2500K each with 4 MB DDR3 RAM, GbE, Intel MKL and compilers 
    99 
    10 ==== i5-3470S ==== 
     10==== i5-3470S Ivybridge ==== 
    1111  * '''256.4 GFLOPS''' N=58800 ([wiki:i5-3740S-Raw-HPL Raw HPL Results])  
    1212  * 69% of Peak (2.9GHz * 16 cores * 8 DP FLOPS/cycle) = 371.2 GFLOPS Peak 
    13   * Four i5-3470S each with 8 MB RAM, GbE, Intel MKL and compilers 
     13  * Four i5-3470S each with 8 MB DDR3 RAM, GbE, Intel MKL and compilers 
    1414 
    15 ==== i5-4570S ==== 
     15==== i5-4570S Haswell ==== 
    1616 
    1717  * '''385.5 GFLOPS''' N=60000 ([wiki:i5-4570S-Raw-HPL Raw HPL Results])  
    1818  * 52% of Peak (2.9GHz * 16 cores * 16 DP FLOPS/cycle) = 742.4 GFLOPS Peak (Note: Haswell is now 16 FLOPS/cycle) 
    19   * Four i5-4570S each with 8 MB RAM, GbE, Intel MKL and compilers 
     19  * Four i5-4570S each with 8 MB DDR3 RAM, GbE, Intel MKL and compilers 
    2020 
    21 ==== i7-4770S ==== 
     21  * '''567.4 GFLOPS''' N=60000 ([wiki:i5-4570S-Raw-HPL Raw HPL Results])  
     22  * 76% of Peak (2.9GHz * 16 cores * 16 DP FLOPS/cycle) = 742.4 GFLOPS Peak (Note: Haswell is now 16 FLOPS/cycle) 
     23  * Four i5-4570S each with 8 MB DDR3 RAM, '''10-GbE''', Intel MKL and compilers 
     24 
     25==== i7-4770S Haswell ==== 
    2226 
    2327  * '''444.8 GFLOPS''' N=86000 ([wiki:i7-4570S-Raw-64G-HPL Raw HPL Results])  
    2428  * 56% of Peak (3.1GHz * 16 cores * 16 DP FLOPS/cycle) = 742.4 GFLOPS Peak (Note: Haswell is now 16 FLOPS/cycle) 
    25   * Four i7-4770S each with 16 MB RAM, GbE, Intel MKL and compilers 
     29  * Four i7-4770S each with 16 MB DDR3 RAM, GbE, Intel MKL and compilers 
    2630 
    2731  * '''498.3 GFLOPS''' N=126000 ([wiki:i5-4570S-Raw-128G-HPL Raw HPL Results])  
    2832  * 62% of Peak (3.1 GHz * 16 cores * 16 DP FLOPS/cycle) = 742.4 GFLOPS Peak (Note: Haswell is now 16 FLOPS/cycle) 
    29   * Four i7-4770S each with 32 MB RAM, GbE, Intel MKL and compilers 
     33  * Four i7-4770S each with 32 MB DDR3 RAM, GbE, Intel MKL and compilers 
    3034 
    3135 
     36==== i5-6500 Skylake ==== 
    3237 
     38  * '''480.2 GFLOPS''' N=86000 ([wiki:i5-6500-Raw-64G-HPL Raw HPL Results])  
     39  * 59% of Peak (3.2GHz * 16 cores * 16 DP FLOPS/cycle) = 819.2 GFLOPS Peak (Note: Skylake is 16 FLOPS/cycle) 
     40  * Four i5-6500 each with 16 MB DDR4 RAM, GbE, Intel MKL and compilers 
    3341 
    3442 
    … …  
    8795 
    8896 
    89 '''These tests are very preliminary using old hardware and untuned software.''' 
     97'''These tests are preliminary using old hardware and untuned software. This approach has been abandoned. The results remain for reference.''' 
    9098 
    9199The first issue is how to add 10GigE (GbE) without a switch using low cost dual port GigE cards. Create a four node loop using [http://www.linuxfoundation.org/collaborate/workgroups/networking/bridge#Bridge_priority Ethernet Bridge]. A simple loop will require a Spanning Tree Protocol, but that cuts a link and introduces one 2-hop route, two 1-hop routes, and three 0-hop routes. If one of the bridges is replaced by a [http://www.linuxfoundation.org/collaborate/workgroups/networking/bonding bonded link] in "mode 0" (round robin) then there are ''effectively'' four 1-hop routes and two 0-hop routes, but no route ''effectively'' takes more than "1-hop" in terms of latency and bandwidth.