LuxMark v2.0 released

Discussion in 'GPGPU Technology & Programming' started by Dade, Jan 30, 2012.

  1. Dade

    Newcomer

    Joined:
    Dec 20, 2009
    Messages:
    206
    Likes Received:
    20
    Introduction

    After a year from the release, LuxMark v1.0 has been widely used as OpenCL benchmark by AnandTech, Tom's Hardware, Vr-zone and other sites. AMD has used LuxMark as one of the 5 GPU computing benchmarks to present the new HD7970.

    LuxMark v2.0 includes SLG2 as rendering engine with Metropolis Light Transport, Multiple Importance Sampling, Image reconstruction done on the GPU, support for multiple OpenCL platforms (i.e. Nvidia users can use Intel or AMD CPU device) and many more new features. The new features rise the complexity of the benchmark of nearly one order of magnitude and it should be able to put some serious stress on the new generation of GPUs. The capability to submit results to a centralise WEB database looks like the most interesting new feature of LuxMark v2.0: http://www.luxrender.net/luxmark

    [​IMG]

    Features

    * Open source
    * 100% OpenCL
    * Cross platform (Linux, MacOS, Windows, etc.)
    * Complex/real world code used as GPU computing benchmark
    * Support for OpenCL GPUs-only, OpenCL CPUs+GPUs, OpenCL CPUs-only benchmarking
    * Support for Interactive mode
    * Easy to run
    * Multiple scenes usable as benchmarks

    New in LuxMark v2.0:

    * Totally new rendering engine based on SLG2 with Metropolis Light Transport, Multiple Importance Sampling, Image reconstruction done on the GPU, ecc;
    * Multi-platform OpenCL support;
    * 2 new benchmark scenes, for a total of 3 benchmarks with a raising complexity (~200,000, ~500,000, ~2,000,000 triangles);
    * Run the benchmark only on the OpenCL devices you selected on the device panel;
    * Added the possibility to submit the scored result to a centralised result database: http://www.luxrender.net/luxmark;

    Download

    Precompiled binaries for major OS are available at http://www.luxrender.net/wiki/LuxMark#Download

    License & Sources

    LuxMark is part of LuxRays source repository and is release under GPL license. The sources are available at http://src.luxrender.net/luxrays

    Result Database http://www.luxrender.net/luxmark

    If you have a valid LuxRender forum account, you can submit your results to the LuxMark result database: http://www.luxrender.net/luxmark. The database offers online real-time browsing, searching, average results, top 20s and many more features.
     
    #1 Dade, Jan 30, 2012
    Last edited by a moderator: Jan 30, 2012
  2. Man from Atlantis

    Regular

    Joined:
    Jul 31, 2010
    Messages:
    732
    Likes Received:
    6
    Q9650@3.00GHz, GTX460@675/1350/3600MHz, 9800GT@550/1375/1800MHz
    ForceWare 290.53, Win 7 x64 SP0


    Code:
              CPU   GTX460   9800GT CPU+GTX460 CPU+GPUs
    
    Simple    1294   2663      554      3592       4143
    Medium     156    390       79       483        450
    Complex     91    185      N/A       248        N/A
    
    Q9650@4.25GHz, GTX460@900/1800/4400MHz, 9800GT@750/1875/2200MHz

    Code:
              CPU   GTX460   9800GT GTX460+9800GT CPU+GTX460
    
    Simple    1796   3532      732       N/A       4912
    Medium     216    519      104       620        670
    Complex    122    246      N/A       N/A        337
    
     
    #2 Man from Atlantis, Jan 30, 2012
    Last edited by a moderator: Jan 30, 2012
  3. fellix

    fellix Hey, You!
    Veteran

    Joined:
    Dec 4, 2004
    Messages:
    3,486
    Likes Received:
    397
    Location:
    Varna, Bulgaria
    Core i7-920 @ 3995MHz
    GeForce GTX 570 @ 825/4200MHz

    ForceWare 290.53
    Intel OCL SDK run-time v1.5
    Windows 7 SP1 x64

    Code:
              CPU   GPU   GPU+CPU
    
    Simple    [B]2723  6207  8540[/B]
    Medium     [B]409   921  1274[/B]
    Complex    [B]278   437   686[/B]
    
     
  4. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,797
    Likes Received:
    2,056
    Location:
    Germany
    Very Nice - just finished with all the beta stuff and now there's an even better scene! :)
     
  5. Jawed

    Legend

    Joined:
    Oct 2, 2004
    Messages:
    10,873
    Likes Received:
    767
    Location:
    London
    Are there any benchmarks for CPUs using "native" code?

    Also, I dare say, this could be interesting in comparing the AMD and Intel OpenCLs on both flavours of CPU.
     
  6. fellix

    fellix Hey, You!
    Veteran

    Joined:
    Dec 4, 2004
    Messages:
    3,486
    Likes Received:
    397
    Location:
    Varna, Bulgaria
    I can't see any native CPU option.
     
  7. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,797
    Likes Received:
    2,056
    Location:
    Germany
    AMDs OpenCL driver is way ahead of Intels in terms of performance - on an Intel CPU...
     
  8. Dade

    Newcomer

    Joined:
    Dec 20, 2009
    Messages:
    206
    Likes Received:
    20
    There was a "CPU native" mode in LM v1.0 however it was dropped for the lack of interest (i.e. OpenCL CPU device was performing quite well so there was a lack of interest in maintaining the CPU native code).

    Intel OpenCL has been a quite solid platform since the first public beta however it is about 10-20% slower than AMD OpenCL on my hardware (i.e. Intel i7 860).

    This is a bit surprising because of their large in-house experience in developing CPU compilers (i.e. ICC).
     
  9. Jawed

    Legend

    Joined:
    Oct 2, 2004
    Messages:
    10,873
    Likes Received:
    767
    Location:
    London
    Ah, in the original post you say "* Support for OpenCL GPUs-only, OpenCL CPUs+GPUs, OpenCL CPUs-only and native CPU-only benchmarking".

    I suppose I could rummage and find out, but is Intel actively improving its OpenCL?

    And, since you're here, is the "Static C++" support that AMD is doing for OpenCL attactive to you for this project? Is it out of bounds, since it would appear to be vendor-specific?
     
  10. Man from Atlantis

    Regular

    Joined:
    Jul 31, 2010
    Messages:
    732
    Likes Received:
    6
    ok i updated first post with 9800GT and oc numbers :).. complex test doesnt work with 512MB VRAM it needs ~530MB ram

    9800GT memory controller usage ~30% @oc
    GTX460 MCU ~27% @oc
     
    #10 Man from Atlantis, Jan 30, 2012
    Last edited by a moderator: Jan 30, 2012
  11. Arnold Beckenbauer

    Veteran

    Joined:
    Oct 11, 2006
    Messages:
    1,411
    Likes Received:
    347
    Location:
    Germany
    Sala:
    OpenCL-GPU (RV770@675): 206
    OpnCL-GPU&CPU (Q6600): 269
    OpenCL-CPU: 155
    Win7 x64 and 64 bit exe.
    Room:
    OpenCL-GPU (RV770@675): 103
    OpenCL-GPU+CPU: 155
    Open-CL-CPU: 87
     
    #11 Arnold Beckenbauer, Jan 30, 2012
    Last edited by a moderator: Jan 30, 2012
  12. Dade

    Newcomer

    Joined:
    Dec 20, 2009
    Messages:
    206
    Likes Received:
    20
    Opss, sorry, it was a left over form LuxMark v1.0 feature list. Fixed.

    If you check their OpenCL forum, it is seems active and it looks like there is some interest. The auto-vectorization option is very interesting (even if somewhat limited at the moment).
    However I'm afraid Intel OpenCL will not become an important software for the users (and for the company) until the support for Intel GPUs ... and this may required some new GPU because I don't even know if their current generation of GPUs is able to support OpenCL.

    I would simply love to try "Static C++" and may be I will but, for instance, in the framework of an open source project, using something available only from a single vendor doesn't look like a good idea. It would be like writing something in CUDA.

    For the same reason, I'm still writing OpenCL 1.0 code (but I'm seriously thinking to switch 1.1 because NVIDIA/Apple are finally offering OpenCL 1.1 compatible drivers). It is a pit but when working with an open standard you are somewhat constrained by slower adopters.
     
  13. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,797
    Likes Received:
    2,056
    Location:
    Germany
  14. Man from Atlantis

    Regular

    Joined:
    Jul 31, 2010
    Messages:
    732
    Likes Received:
    6
  15. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,797
    Likes Received:
    2,056
    Location:
    Germany
    Yep, but for what reason? To make Kepler shine? :) Seriously, though, I've contacted Nvidia weeks ago and they said they'd look into it - I wrote two reminders, to no avail until today.

    You would think they're seriously busy launching Kepler, yet, they aren't.
     
  16. Man from Atlantis

    Regular

    Joined:
    Jul 31, 2010
    Messages:
    732
    Likes Received:
    6
    looking your article, the more complexity the more Tahiti shines compared to Cayman.. medium complexity seems like sweet spot for GF110, however GF114 reacts same for medium or high it's already bottlenecked by internal bandwidth..
     
  17. fellix

    fellix Hey, You!
    Veteran

    Joined:
    Dec 4, 2004
    Messages:
    3,486
    Likes Received:
    397
    Location:
    Varna, Bulgaria
    NV broke something in their ICD code while adopting the OCL 1.1 specs -- probably will be solved within the next few WHQL releases, as far as I've heard, but it is definitely not on their priority list. :???:
     
  18. fellix

    fellix Hey, You!
    Veteran

    Joined:
    Dec 4, 2004
    Messages:
    3,486
    Likes Received:
    397
    Location:
    Varna, Bulgaria
  19. CarstenS

    Veteran Subscriber

    Joined:
    May 31, 2002
    Messages:
    4,797
    Likes Received:
    2,056
    Location:
    Germany
  20. OpenGL guy

    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    2,357
    Likes Received:
    28
    HD7970 @ 1050e/1475m
    Beta drivers ;)

    Code:
              GPU
    simple   16787
    medium    1944
    complex   1109
    
    Edit:
    HD7970 @ 1125e/1575m (max OverDrive)
    Code:
              GPU
    simple   18137
    medium    2085
    complex   1198
    
     
    #20 OpenGL guy, Feb 1, 2012
    Last edited by a moderator: Feb 1, 2012
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...