Cool - thanks for doing the tests. It's good to hear. If Nvidia is better we might need to retro-fit the K6000 to the new MacPro!

If possible can you post a file with the 100 separate objects. Not sure what I did wrong but I tried a test file and was getting ~2fps with only 60 objects of 250K polys. Thanks!

Cool - thanks for doing the tests. It's good to hear. If Nvidia is better we might need to retro-fit the K6000 to the new MacPro!

I wish you luck in that – the dual AMD cards seem pretty custom to the case. But who knows, maybe Nvidia will release a board that fits into the new MP.

If possible can you post a file with the 100 separate objects. Not sure what I did wrong but I tried a test file and was getting ~2fps with only 60 objects of 250K polys. Thanks!

Here you go. An AMD 5770 in my mid-2008 MP was running it between 120-140ms (7.1-8.3fps). The GT120 in the same system was a bit slower at 190-210ms (4.8-5.3fps). My Win7 system with the FirePro W8000 was running it at 47ms (21fps) in GL2.

This is using point instancing to draw the data, so it's possible that if you have 100 unique 250K quad grids the GPU would run into memory swapping issues and slow down, which instancing avoids.

Cool thanks! That scene is running at ~ 10fps. Good stuff. Houdini 13 sounds very promising. Good to hear about the optimizations.

I also exported the original 25mil as an obj, then imported it and spun it Maya 2013 and Houdini. the results are 0.6fps for Maya and 4.4fps/Houdini. pretty nifty!

Edit: it would be good if someone who is skilled in Maya test it. I probably did something wrong

Edit: I tested this in Maya 2014(downloaded today) and it totally dies. I must be doing something wrong!

twod

Cool - thanks for doing the tests. It's good to hear. If Nvidia is better we might need to retro-fit the K6000 to the new MacPro!

I wish you luck in that – the dual AMD cards seem pretty custom to the case. But who knows, maybe Nvidia will release a board that fits into the new MP.

If possible can you post a file with the 100 separate objects. Not sure what I did wrong but I tried a test file and was getting ~2fps with only 60 objects of 250K polys. Thanks!

Here you go. An AMD 5770 in my mid-2008 MP was running it between 120-140ms (7.1-8.3fps). The GT120 in the same system was a bit slower at 190-210ms (4.8-5.3fps). My Win7 system with the FirePro W8000 was running it at 47ms (21fps) in GL2.

This is using point instancing to draw the data, so it's possible that if you have 100 unique 250K quad grids the GPU would run into memory swapping issues and slow down, which instancing avoids.

Which shading mode did you use to get the 21 fps with the W8000?

misterluqman

twod

Cool - thanks for doing the tests. It's good to hear. If Nvidia is better we might need to retro-fit the K6000 to the new MacPro!

I wish you luck in that – the dual AMD cards seem pretty custom to the case. But who knows, maybe Nvidia will release a board that fits into the new MP.

If possible can you post a file with the 100 separate objects. Not sure what I did wrong but I tried a test file and was getting ~2fps with only 60 objects of 250K polys. Thanks!

Here you go. An AMD 5770 in my mid-2008 MP was running it between 120-140ms (7.1-8.3fps). The GT120 in the same system was a bit slower at 190-210ms (4.8-5.3fps). My Win7 system with the FirePro W8000 was running it at 47ms (21fps) in GL2.

This is using point instancing to draw the data, so it's possible that if you have 100 unique 250K quad grids the GPU would run into memory swapping issues and slow down, which instancing avoids.

Which shading mode did you use to get the 21 fps with the W8000?

Shaded mode only, Headlight, default material, no AA, GL2.