GPU benchmarks

[vlado]

In this case my next GPU upgrade will be when cards will be able to pull times below 1 minute on this test

You can always get two Titans

Best regards,
Vlado

[lukx]

yeah and sell my kidney to pay for electricity bills

[stilgarna]

you want 670 here it is with CUDA engine:
Evga 670 GTX FTW 4GB : 3min6.4s

In this case my next GPU upgrade will be when cards will be able to pull times below 1 minute on this test

Really close to the 680GTX. Thanks !!!

[lukx]

Really close to the 680GTX. Thanks !!!

yep , this is why I decided to pay less and still got almost same power.

[stilgarna]

yep , this is why I decided to pay less and still got almost same power.

Very smart !!! 680gtx is very desapointing...

2x 670 will be slightly slower, will consume slightly more power but you have the option for 4GB per card. 2x EVGA GTX 670 SC 4GB are cheaper than a 690 that is limited to 2GB per GPU.

If you want the maximum computation density, atm nothing is faster than a 690, but you will be limited for tasks that can fit in the 2GB buffer.

I need to make some additional tests to know how many vram fits our needs.

Thank you guys !

[stilgarna]

Finally we are moving for a distant RT server with 2x 580GTX and 1x Titan.

In the future we will change 580GTX for 2x Titan or better
I'll post some results as soon as the machine is ready.


Best regards and thank you for your help.

[3LP]

When you're speaking about distant, do you mean local network but using it as GPU node so you only spawn on it?
Or do you have any other trick?

Stan

[stilgarna]

When you're speaking about distant, do you mean local network but using it as GPU node so you only spawn on it?
Or do you have any other trick?

Stan

Local network with RT service.

1 titan + 2 Gtx 580 ( 3gb ) : 49 sec

Using network, not local. Works great.

In our current project we will massively use RT for lighting. We'll see if we 'll add more RT node like this one.

[stilgarna]

Hello,


I notice that GTX670 and GTX660 ti are very close, based on specification, anyone can test the GTX660 ?

Best regards...

[retardus]

Workstation i920 18mb ram

1:10 Titan+580
2:43 580
1:54 Titan

Nvidia Driver 314.22

[retardus]

GPU distributed rendering

Workstation w/ Titan (6gb) + 580 (3gb)
Workstaion w/ 470 (1.5gb)
Laptop w/ 650M (2gb)

46.6s

[eyepiz]

(4) Evga Titans SC on 3DS Max 2014

28.04 sec

[bear75]

(4) Evga Titans SC on 3DS Max 2014

28.04 sec

[ATTACH=CONFIG]13960[/ATTACH]

hi, are all these cards on your workstation or are you using DR on remote slaves?
what about using an external pci-e case?
and last, max 2014 works correctly with vray 2.4?

[eyepiz]

All cards are on one workstation with a 4-way SLI board...Just needed one beefy power supply.
Max 2014 seems to be working correctly however I did see some issues with gamma output in PNG files. 3DS Max 2014 made some changes in the gamma input and output settings the option are no longer there it's "Auto" now but I think that is having something to do with it. Jpegs and tiff's are fine.

[vlado]

The gamma of PNG files has always been wrong in 3ds Max right from the start until 3ds Max 2013 (the inverse gamma was written into the file) which meant you couldn't open those PNGs properly in any other software that was aware of the gamma. In 3ds Max 2014 this is corrected.

Best regards,
Vlado