SLI bridge GTX 1070 with GTX 960?

[fjguzzone]

Hello,

Just purchased a GTX 1070 to upgrade my system from the GTX 960.

Was wondering if its possible to run both cards at the same time? Will an SLI bridge even work with this setup? Will this combo make for faster octane rendering through the c4d plugin? Or does it make sense to have the 1070 solo?

Fairly new to building my own computer.

Thanks in advance,

FJG

[mojave]

Octane can work with up to 20 cards within the same machine, and you should see a linear performance increase when adding new hardware. However Octane doesn't support SLI.

Something to take into consideration is that Octane needs to upload a copy of your scene to each card, so the maximum available total memory will correspond to the lowest memory size of all your GPUs. In your case 2GB.

To sum up, you may definitely use both cards, however depending on the scene you are rendering, if it doesn't fit in your GTX960 you will need to disable it so the 8GB of your 1070 will be available.

[fjguzzone]

Thank you @mojave for the reply, very helpful.

A few more follow up questions:
- Is there a way to determine if a scene will fit prior to rendering?
- Does it matter if I have the two cards physically bridged using the SLI bridge that came with the motherboard?

Thanks again in advance,

FJG

[mib2berlin]

- Is there a way to determine if a scene will fit prior to rendering?

Yes but scene must be loaded in to Octane or Plugin.

- Does it matter if I have the two cards physically bridged using the SLI bridge that came with the motherboard?

AFAIK no, you can switch SLI off in Nvidia settings, at least on Windows.

Cheers, mib

[FrankPooleFloating]

SLI only works on identical GPUs, so linking your 1070 and 960 would be about as advantageous as you attaching your microwave to your washing machine.

[acc24ex]

and finally GPU processing is done through the PCI bus, and only for gaming purposes SLI is used .. let's put it that way..

..just reading into it a bit, SLI is apparently faster than the PCI for sharing data between two graphic cards but still not fast enough to share the memory, so even in gaming you will always have the same amount of memory not 2x if you use 2 graphic cards..

- why aren't the hardware manufacturers addressing this issue finally - the biggest bottleneck in computer architecture now is the PCI bus - if you can make it as fast as RAM bandwidth - you could use cpu, ram and gpu work seamlessly, maybe even eliminate VRAM all together

- GPUs would be cheaper if they wouldn't have to use their own memory

- maybe the cost of the high speed pci would be too expensive, or maybe the technology is here, only waiting to make as much from this one to make the switch or "it's working good enough"

..I'm sure in like 5-10 years they will do this for sure