Announcement

Collapse
No announcement yet.

GPU benchmarks

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

  • #16
    Originally posted by Metzger View Post
    Here's my video with two 580's.

    http://vimeo.com/23921219
    so you´re running maya on a bootcamped 2010 macpro? is there a mac version of the 580 around? or do you use a pc version? because I have an 8 core here and would love to use it instead of my pc....
    best regards,
    sacha

    Comment


    • #17
      Originally posted by werticus View Post
      how do you fit 9x gtx 580's into a computer?
      lol! I was wondering the same thing but then I noticed it says (DR) in the chart. So it's a DR setup/test, not all in one computer.

      Comment


      • #18
        yes 9x 580 is in DR mode.
        I have 6x 580 plugged on my workstation , i've hear that 7 or more inside 1 box will slow down the rendering . certainly a motherboard data flow issue.
        GHiOM = Guillaume Gaillard
        freelance 3D artist
        www.ghiom.com

        Comment


        • #19
          two gtx 580 each 3gb: 1:22 to finish 512 samples/pixels - without textures though...
          two gtx 580 + one 5gtx 580 DR: 57,3sec to finish 528 s/p - again no textures
          Last edited by nlo; 08-08-2011, 06:29 AM.
          -
          render forza!

          -----

          Office Le Nomade, Vienna

          web: www.oln.at
          blog: blog.oln.at

          Comment


          • #20
            1x EVGA Classified GTX 590: 1:49.0s
            2x EVGA Classified GTX 590: 0:53.9s

            Click image for larger version

Name:	1x590-270.jpg
Views:	1
Size:	265.9 KB
ID:	844374Click image for larger version

Name:	2x590-270.jpg
Views:	1
Size:	266.0 KB
ID:	844375

            Not in DR - single machine.
            Drivers: 270.61 WHQL
            Last edited by renderfarmer; 22-08-2011, 01:43 PM.

            Comment


            • #21
              2 x Quadro 4000 using drivers 270.61 in DR mode. 5m 4.3s. I will post up 3 x Quadro 4000 if I can get the other working

              Click image for larger version

Name:	RT_2x_Quadro4000.jpg
Views:	1
Size:	343.5 KB
ID:	844384

              Comment


              • #22
                I'm thinking of getting a GTX 590 (or two) but i can;t see any resluts for these cards. Any one knows if there is a reason in praticular?
                Also while we're here.. i'm confused if the card (GTX590) can load a 3 gig scene or is limited to 1.5 gig.
                Thanks

                Comment


                • #23
                  There are a couple of results for GTX 590 (11-th and 12-th place from top). The 590 has two GPUs, each with 1.5 GB of graphics RAM, so no - you can't load a 3 GB scene on it (each GPU must store the scene separately).

                  Best regards,
                  Vlado
                  I only act like I know everything, Rogers.

                  Comment


                  • #24
                    Originally posted by fabriceb View Post
                    I'm thinking of getting a GTX 590 (or two) but i can;t see any resluts for these cards. Any one knows if there is a reason in praticular?
                    Also while we're here.. i'm confused if the card (GTX590) can load a 3 gig scene or is limited to 1.5 gig.
                    Thanks
                    Look down 2 posts for 590 benchmark results including dual. Like Vlado says, 1.5GB max.

                    Comment


                    • #25
                      thanks guys, found it now! ok, the memory thing is clear, now have to figure how to appreciate what kind of scene 1.5 gig of gpu memory can take.
                      Is it effectively the same than for CPUs? ie 1.5 gig really won't go very far in this day and age...

                      Comment


                      • #26
                        From what I remember... 1 mil polygon - 100 mb in memory or so...
                        CGI - Freelancer - Available for work

                        www.dariuszmakowski.com - come and look

                        Comment


                        • #27
                          Hey, does anyone know what are the benchmark for the new :
                          EVGA GeForce GTX 560 Ti 2Win

                          Thanks

                          Stan
                          3LP Team

                          Comment


                          • #28
                            old question, new question...
                            i'm testing new features of vray 2.2 on gpu with 290.36 nvidia drivers
                            very old driver 258.96 seems faster, much more...

                            Comment


                            • #29
                              Yes, unfortunately I'm getting the same results here.

                              Running a GTX580 and a GTX560i together (896 cores) I was getting 1:58 for the benchmark using the 266.66 drivers.

                              After installing Vray 2.2 the update, these drivers would no longer work for RT/GPU (OCL file wouldn't compile). So I went ahead and installed the latest drivers found up on the nVidia site, 285.62. Everything looked good and worked OK (the new 2.2 features are great!), but the rendering speed in GPU suffered greatly.

                              I am now rendering the benchmark (same hardware) at a much slower 2:28, which is like a 12% drop in rendering speed performance! Also note that I remember even faster times with drivers in the sub 250.x versions!

                              This of course is very disconcerting and I hope that we can get that performance back soon. Vlado, is there any hope?

                              Thanks,

                              -Alan

                              Comment


                              • #30
                                Yes, we can do certain optimizations. We are working on that and I hope that we can bring some of the speed back. But in general, I too have noticed that recent drivers tend to be somewhat slower. There is also a bug in the most recent nVidia CUDA/OpenCL compiler that prevented us from using the most optimal code.

                                Best regards,
                                Vlado
                                I only act like I know everything, Rogers.

                                Comment

                                Working...
                                X