• Now's your chance win big! Join our community and get entered to win a RTX 2060 GPU, plus more! Join here.

    Pi Cast Episode 3 streams live on Tuesday, August 4th at 2:30 pm ET (7:30 PM BST). Watch live right here!

    Catch Scharon on the Tom's Hardware Show live on Thursday, August 6th at 2:00 pm ET (7:00 PM BST). Click here!

[SOLVED] Nvlink memory question?

Tad2

Distinguished
Apr 3, 2010
45
0
18,530
0
Ok I would like to know how multigpu memory works with nvlink.
So can you actually stack as many quadro's together as you want and it will just add all the ram together? The website words it as if you can only combine two? Double the ram is still amazing, but I thought I read somewhere that it could do more?
 
Ok I would like to know how multigpu memory works with nvlink.
So can you actually stack as many quadro's together as you want and it will just add all the ram together? The website words it as if you can only combine two? Double the ram is still amazing, but I thought I read somewhere that it could do more?
NVLink supports up to 8 Quadro GPUs (setup in a NVIDIA Quadro Virtual Data Center VM ).
With 8 Quadro RTX 8000 GPUs you could have around 300 megabyte of unified memory space and a petaFLOP of compute power available .
Lets say you want to pair 8 + 8 = 16 Quadro RTX 8000 GPUs you could use NVSwitch to connect them.
These 16 GPUs can be used as a single large-scale accelerator with more than 600 megabyte of unified memory space and a couple of petaFLOPS of compute power.
 
Last edited:
Ok I would like to know how multigpu memory works with nvlink.
So can you actually stack as many quadro's together as you want and it will just add all the ram together? The website words it as if you can only combine two? Double the ram is still amazing, but I thought I read somewhere that it could do more?
NVLink supports up to 8 Quadro GPUs (setup in a NVIDIA Quadro Virtual Data Center VM ).
With 8 Quadro RTX 8000 GPUs you could have around 300 megabyte of unified memory space and a petaFLOP of compute power available .
Lets say you want to pair 8 + 8 = 16 Quadro RTX 8000 GPUs you could use NVSwitch to connect them.
These 16 GPUs can be used as a single large-scale accelerator with more than 600 megabyte of unified memory space and a couple of petaFLOPS of compute power.
 
Last edited:
woooahhhh I didn't know it actually worked.
well that's amazing.
whats' the oldest quadro that supports this?
Quadro GP100 and GV100 Cards...But it gets a little complicated

Quadro GP100 cards came with first generation NVLink and it does not work with newer GPUs.
Quadro GV100 arrived with a second gen NVLink and it might not (fully) work with RTX GPUs.
The RTX 5000 has a smaller NVLink connector that cannot be use with any other GPU series.
The Quadro RTX 6000 NVLink Bridge is designed to work on the Quadro RTX 6000 and RTX 8000 series.
You might be able to also use GeForce RTX, Titan RTX since they use the same bridge as the Quadro RTX 6000 and RTX 8000.
 

ASK THE COMMUNITY

TRENDING THREADS