It’s me again.
setting sync to vblank off and double side off I have double score…
Cpu: i7 980X
Gpu Quadro 4000
Screen resolution: 1919 x 1122
Overall score: 33.7
Time: 1min 42Sec
Sorry, I was talking about the RAM usage by Blender+System. I think that in any modern 64bit OS the RAM usage during render may be hovering around 6GB. That can cause the system to begin intensively use swap memory and slow down the whole system. This could interfere with the result in computers with less than 8GB of RAM. I guess.
In fact, I noticed that my system (8GB RAM) was using a lot of RAM during rendering when I started to hear the typical sound of my old SATA 7200rpm hard disck writing the swap.
Yes system memory has not to go above the RAM limit, but i was (i am) pretty sure running this test is safe on a system with 4GB…i will make more tests.
Ok, my bad. I corrected the link in the 1st post cause the first time I edited the 1st post, I edited how the link appeared and not the link itself.
My results :
CPU : i7 980x
GPU : GTX 580 3Gb
Blender version: (2, 72, 0)
Revision: b’9e963ae’ , b’2014-10-21’
Platform: b’Windows:32bit’ (it’s 64 bit)
RESULTS
Object mode
Wireframe
-rabbit (4150 polys with 5 levels subsurf*, 4.2mln polys, 8.5mln tris): 46.95 fps
-bolts (8.3mln polys, 16.6mln tris): 26.89 fps
-robot (1.5mln polys*): 59.97 fps
Solid
-rabbit (5 levels subsurf*): 29.98 fps
-bolts (8.3mln polys, 16.6mln tris): 19.85 fps
-robot (1.5mln polys*): 50.35 fps
Material
-robot (225k polys): 28.20 fps
-robot (1.5mln polys*): 4.57 fps
Edit mode
Wireframe
-rabbit (265k polys, 530k tris): 11.87 fps
Hiddenwire
-rabbit (265k polys, 530k tris): 9.98 fps
Solid
-rabbit (265k polys, 530k tris): 6.58 fps
Sculpt mode
Solid matcap
-basemesh (25k polys with 4 levels multires, 6.5mln polys, 13mln tris): 29.58 fps
-basemesh (5 levels multires, 26mln polys, 52mln tris): 14.11 fps
Screen resolution : 1919 x 1148
VBOs: True
Anisotropic filter: 2x
Draw method: AUTOMATIC
Multi sample: NONE
Mipmaps: True
GPU Mipmap: True
Overall Time (to run the benchmark): 3 min 46 sec
I get worse FPS than Manolo76 (that’s OK, I have a bad GT430), but the overall time is similar. Why is that?
OS: Kubuntu 14.04 64bits
CPU: intel i7 3770
GPU: nvidia GT430
Blender version: (2, 72, 2)
Revision: b’4828c6a’ , b’2014-11-13’
Platform: b’Linux:64bit’
RESULTS
Object mode
Wireframe
-rabbit (4150 polys with 5 levels subsurf*, 4.2mln polys, 8.5mln tris): 11.61 fps
-bolts (8.3mln polys, 16.6mln tris): 6.27 fps
-robot (1.5mln polys*): 28.94 fps
Solid
-rabbit (5 levels subsurf*): 4.67 fps
-bolts (8.3mln polys, 16.6mln tris): 2.43 fps
-robot (1.5mln polys*): 12.72 fps
Material
-robot (225k polys): 43.24 fps
-robot (1.5mln polys*): 7.04 fps
Edit mode
Wireframe
-rabbit (265k polys, 530k tris): 14.48 fps
Hiddenwire
-rabbit (265k polys, 530k tris): 10.75 fps
Solid
-rabbit (265k polys, 530k tris): 7.41 fps
Sculpt mode
Solid matcap
-basemesh (25k polys with 4 levels multires, 6.5mln polys, 13mln tris): 4.60 fps
-basemesh (5 levels multires, 26mln polys, 52mln tris): 1.26 fps
Screen resolution : 1919 x 994
VBOs: True
Anisotropic filter: 2x
Draw method: AUTOMATIC
Multi sample: NONE
Mipmaps: True
GPU Mipmap: True
Overall Time (to run the benchmark): 3 min 39 sec
That’s because i put a sort of timeout (approx 10sec) if it takes too long to complete a 360 spin view. In fact the overall time is not so relevant now.
Ok, I see.
Sorry!
CPU: i7 980X
Gpu Quadro 4000
BLENDER VIEWPORT BENCHMARK
Blender version: (2, 72, 0)
Revision: b’73f5a41’ , b’2014-10-15’
Platform: b’Linux:64bit’
RESULTS
Object mode
Wireframe
-rabbit (4150 polys with 5 levels subsurf*, 4.2mln polys, 8.5mln tris): 48.13 fps
-bolts (8.3mln polys, 16.6mln tris): 24.58 fps
-robot (1.5mln polys*): 91.45 fps
Solid
-rabbit (5 levels subsurf*): 34.39 fps
-bolts (8.3mln polys, 16.6mln tris): 17.90 fps
-robot (1.5mln polys*): 50.27 fps
Material
-robot (225k polys): 42.50 fps
-robot (1.5mln polys*): 5.12 fps
Edit mode
Wireframe
-rabbit (265k polys, 530k tris): 11.29 fps
Hiddenwire
-rabbit (265k polys, 530k tris): 8.68 fps
Solid
-rabbit (265k polys, 530k tris): 5.77 fps
Sculpt mode
Solid matcap
-basemesh (25k polys with 4 levels multires, 6.5mln polys, 13mln tris): 46.34 fps
-basemesh (5 levels multires, 26mln polys, 52mln tris): 11.49 fps
Screen resolution : 1919 x 1122
VBOs: True
Anisotropic filter: Off
Draw method: AUTOMATIC
Multi sample: NONE
Mipmaps: True
GPU Mipmap: True
Overall Time (to run the benchmark): 3 min 22 sec
Is screen size relevant to the test results? I feel it should…
Out of sheer curiosity i did compare fresh Gooseberry from builder and a year old build which still was on HD here…
Script needed some corrections - git was not available back then.
Here is Libre Office spreadsheet; if you paste test text (which you can save as a text file instead of taking image) in a new sheet it will ask for a table separator. Colon “:” is the right one.
http://wikisend.com/download/862242/viewport_test.ods
Not sure how long file will last there. Forum does not allow such extensions.
Not sure, but i think it’s not so relevant. If you run the bench reducing drastically the blender window, results don’t change.
At the moment i have no possibility to try different monitors.
Somebody with al least 12GB RAM could try to run this version. Only obj mode, 2 wire, 2 solid, approx 16mln faces.
Thanks!
I had concerns that if window is less than full screen there is less work for the video subsystem and as a result test playback is faster.
If you say window size does not matter, well, it’s fine then.
I thought the same way, but probably screen size does not affect window drawing performance (according to my tests)