2.7x Cycles benchmark (Updated BMW)

[l4gerardo]

i don´t know what i'm doing wrong i thought my investment in a very expensive pc will pay of, but sadly my results show quite the opposite. my results:

CPU: Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz (24 CPUs), ~2.0GHz
GPU: NVIDIA Quadro 4000
OS: Windows 7 64bit | Blender 2.74
Time: 4 min 14 seconds (GPU - CUDA)
Time: 4 min 54 sec (CPU 24 threads)

[bb3d]

The Quadro 4000 is technically the same hardware as a GTX 560, an outdated midrange gaming card and therefore not a good choice for GPU rendering anymore.
Your CPU might have 24 (logical) cores, but only at 2 GHz. You could try to improve the CPU performance by using smaller tiles, like 32x32.

[okrasit]

@l4gerardo

Which E5 2600 chips are they (v1/2/3)? I strongly feel that v2&3 should do it in way less than 2minutes.

[Grimm]

Ran another test with 2.75a and fiddled with the tile size:

GPU: MSI GTX980
OS: Linux Mint 17.1
Time: 01:10.82 (GPU 240x540 tile size).

[bliblubli]

i don´t know what i'm doing wrong i thought my investment in a very expensive pc will pay of, but sadly my results show quite the opposite. my results:

CPU: Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz (24 CPUs), ~2.0GHz
GPU: NVIDIA Quadro 4000
OS: Windows 7 64bit | Blender 2.74
Time: 4 min 14 seconds (GPU - CUDA)
Time: 4 min 54 sec (CPU 24 threads)

Quadro cards have disabled cuda cores if I remember well. So they are much more expansive than gamer cards while being much slower for rendering. NVIDIA segment their market, they want you to buy one card pro use and the quadro are limited to only fit for CAD/real-time OpenGL. We have many quaddros at work and they are all slower than CPU rendering.

[mib2berlin]

Hi l4gerardo, for CPU change the tiles to 32x32 or 16x16, for GPU use 240x540.
Do you meant Quadro 4000, K4000 or K4200?
The Quardo 4000 is old, the E5-2620 is very old too, when do you buy this?

Cheers, mib

[Winstonian]

EVGA 04G-P4-2768-KR G-SYNC Support GeForce GTX 760 4GB
i7-4790
16GB RAM at 1866
Windows 8.1 64bit
I'm driving a 2k monitor - not sure if that matters.

GPU Cuda: 2:19.19
CPU: 3:51.12


Standard Blend Settings from file

[JulianS]

i thought my investment in a very expensive pc will pay of,

sadly this never pays of. Except of you are really making money with your work! Its like with expensive cars. At a specific point you pay double the price for 10% more power/speed.
Best way is to buy upper midrange hardware for exsample 2x GTX 970 instead of 1 GTX 980Ti. You can even sell midrange hardware much better then high end.

[BernhardS]

Gigabyte R9 390 "G1 Gaming" 8GB
Xeon 1231 v3
Windows 7 64
GPU: 00:59:13 (960x540)
CPU: 04:30:49 (16x16)

-> Keep in mind, that OpenCl does not support transparent shadows [...] yet, so this result is not comparable to nvidia-cards.



2x MSI GTX 750ti "Twin Frozr" 2GB
AMD PhenomII 1090t
Windows 7 64
GPU: 01:46:84 (256x256)
CPU: 07:27:11 (16x16)

In other scenes, the 2x750ti are just as fast as the R9 390.

[Thomas Berglund]

New BMW scene.

Mac Pro (2013)
CPU: Intel Xeon E5-1680 v2 - 3GHz
GPU: 2x AMD FirePro D700
OS: OS X 10.11 Developer Beta 4 (15A226f)
Time: 2 min 43 seconds (GPU - OpenCL) - 128x128 tiles
Time: 2 min 34 seconds (GPU - OpenCL) - 256x256 tiles
Time: 2 min 12 sec (CPU) - 16x16 tiles

Using blender-2.75-548e650-OSX-10.6-x86_64 from:
http://wiki.blender.org/index.php/De...nCL#AMD_on_OSX

[TheIMP67]

Just starting off in the world of Blender, but saw the instruction to share benchmark results so thought I'd do so - although I don't understand the different tile settings people are talking about, see, told you I was new. I've just pressed F12 to render, once using CPU and once using GPU CUDA enabled.

Intel i7-5960X @ 3.00GHHz and NVidia GTX Titan X

CPU Render 03:08.94
GPU - CUDA 1:53.45

[okrasit]

New BMW scene

OS X 10.11 latest DP & blender-2.75-33bac1f-OSX-10.6-x86_64
R9 290 (256x256 tiles): 04:04.64

[copperplate]

TheIMP67, the tile settings are found in the render tab (the camera icon) under the subheading "Performance". You can change the x and y pixel dimensions of the tiles Blender will use to dice up your scene. Changing these settings will dramatically reduce your render times, you just have to experiment. You can see the different render times that happen with different tile sizes in my renders here.

Generally the larger the tiles, the faster the render, however, this only works up to a certain point. For example, 240 x and 540 y tiles rendered faster than 480 x and 540 y tiles with my GTX 970.

[nearleyg]

Just finished (2 days ago) building a new machine after a while of saving. Really impressed with the multi gpu scaling. No overclock on the gpus yet. I'll do another when they are overclocked.

My time(s):
CPU: Intel i7 5960x 4.8Ghz
GPU: Nvidia GTX 980 ti 2x
OS: Windows 7 64bit
Time: 50.71 sec (GPU - CUDA) tile size 128x128
Time: 34.92 sec (GPU - CUDA_ tile size 256x256
Time: 31.05 sec (GPU - CUDA) tile size 480x270
Time: 32.85 sec (GPU - CUDA) tile size 512x512
Time: 34.48 sec (GPU - CUDA) tile size 480x540
Time: 2 min 04 sec (CPU) tile size 128x128
Time: 1 min 54 sec (CPU) tile size 64x64

[Csokis]

My time(s):
CPU: Intel i7-3770 @ 3.4Ghz
GPU: Nvidia GeForce GTX 750 Ti
OS: Ubuntu Studio Linux 14.04.2 LTS
Time: 03:20:13 (GPU - CUDA)
Time: 04:10:31 (CPU)

[kaneleka]

CPU: Intel Core i5-4690K
GPU: AMD R9 390
OS: Windows 10 Pro
Time: 54 seconds (GPU - OpenCL) tile size 960x540
Time: 5 min 46 sec (CPU)

[johndoe123]

(Times are for new BMW scene)

CPU : Intel i7 4770k
GPU : AMD 7870 OC
OS: Windows 7 64bit Pro

Time: 02:31:66 (GPU-OPENCL) tiles 256x256
Time: 04:01:84 (CPU), tiles 16x16

And a vray version I made for fun xD

bmw_vray.jpg

[daveywaveyHD]

CPU: Intel Core i7-3770k @ 4.2Ghz
GPU: AMD R9 295x2 @ 1.1Ghz
OS: Windows 10 Home
Time: 1:23 min (GPU - OpenCL) - Default tile size (128x128)
Time: 4:53 min (CPU) - Default tile size (128x128)

[joseph raccoon]

Is there a benchmark page for the gooseberry render scene found here https://gooseberry.blender.org/goose...enchmark-file/

BTW my workstation did an hour for that render, So not to bad for 200 dollars of machine I think.

[ClausBohm]

CPU: Intel i7-5820K 3.8Hgz
GPU: Nvidia Quadro K5200
OS: Windows 7 64bit | Blender 2.75a
Time: 2 min 12 seconds (GPU - CUDA)
Time: 3 min 29 sec (CPU 8 threads)

Will try my 2 OC Titan X's in combo with this card later and then post. I am also getting AIO water blocks so I can crank up the titans a bit more. Its more for Octane then Blender (cycles) but nice to know how it works here as well.