Benchmark Scene 2 (version 0.3.6a or higher is required)

[voltaire585]

Hi Karba, could you let us know whether the difference between Gtx 1070 and 1080 is a permanent thing or whether tuning to Cuda 8 is somehow going to let 1080 score where its price indicates it should.

Im building a 4x box but dont know which cards to put in 1070 or 1080.

[oguzbir]

RT was faster in my test.. Double checked.

[oguzbir]

Tested with New version 3.7
3x1070 OC'ed 4:05

0.3.6a score was 4:17
Great!!

.....EDIT.....
RT Render was faster than the Max Production Render.
RT Render took 3:52

[aranmor]

So it will always use the less memory card that meets the scene requirement for allocation and buffer, but will use the CUDA from all the those that meet said requirement? right? Is there a way for developers to actually sum up all of the physical memories? would SLI help on this matter? why does my 2 970s show less than 2GB available always on the RT status dialog? (feel free not to answer to my noob questions, but if you do.. Thank you)

[deC9r]

RT render is slower than 3ds max production render.

[deC9r]

Gainward Phoenix GTX1070 @ 2101Mhz / 4399Mhz
Rendertime: 10:00
Fstorm 0.3.7 / Driver 375.57

Gainward Phoenix GTX1070 @ 2101Mhz / 4399Mhz
Rendertime: 10:13
Fstorm 0.3.6a / Driver 375.57

[mitviz]

basically it will render with both cards but as long as the scene uses under 4 gigs it will use both cards if the scene used more than 4 gigs then the 970 could not be used

[aranmor]

Can you RT with this configuration to show your memory? I'm trying to understand how an 8gb (1080) and a 4GB (970) are managed by the engine... so far I can tell that only one memory gets allocated by all the dual card benchmarks here but all of them are of equal ammount, I would like to know what happens in this particular case. Thank you

[davichoxsi]

Alienware 17R2 gtx 970M
Render time: 28:09 min
I need de graphics amplifier, someone can recomend a video card that complement this laptop?

[alancodger]

You have to let it to finish the render. Estimation time doesn't take adaptivity effect into account. Render time in your case would be arounf 30-35 mins, I think.

You stand correct.

Render version: 0.3.6a
Config: GTX 780
Render time: 30.32

[SameHere]

FStorm V0.3.6a
3ds max 2017
Driver 10.18.13.6839
GTX780 + GTX 970 = 13:00 min.

[Karba]

Hi Karba, I got very high render time with this benchmark and I don't know why, I mean I can see 960 with less cuda cores (1024) going 25min and so faster than my GTX 780 with 2304 cuda cores and despite different generation got usually a benchmark closer if not faster, so I stopped the render at 25:40 and the image was like this from the last 5 min or so and look pretty decent to me..but still 1084 samples.. normal? Or is actually about 40 min for that spec of mine? ( can't wait for my 1070 to knock my door )

Nvidia drivers 375.63

You have to let it to finish the render. Estimation time doesn't take adaptivity effect into account. Render time in your case would be arounf 30-35 mins, I think.

[alancodger]

Hi Karba, I got very high render time with this benchmark and I don't know why, I mean I can see 960 with less cuda cores (1024) going 25min and so faster than my GTX 780 with 2304 cuda cores and despite different generation got usually a benchmark closer if not faster, so I stopped the render at 25:40 and the image was like this from the last 5 min or so and look pretty decent to me..but still 1084 samples.. normal? Or is actually about 40 min for that spec of mine? ( can't wait for my 1070 to knock my door )

Nvidia drivers 375.63

[jurasiq]

Render version: 0.3.6a
Config: 2 x GTX 650 Ti
Render time: 1h 6m 22s

[MAROMERO]

[Render test 2] NVIDIA GeForce GTX 1080 Founders Edition.
Max Temp 64°C, Driver 375.57 Win7, Render Time: 10:20 Min.
GPU-Z: https://www.techpowerup.com/gpuz/details/muep4

CPU-Z: http://valid.x86.fr/6663ni