Is it worth to wait Nvidia pascal GPU?

yzhu25

Commendable
Apr 14, 2016
2
0
1,510
Hi, I want to build a good machine for deep learning research or application recently. Basically it will follow the nvidia devbox configuration. But people keep saying that I should wait for Nvidia Pascal GPU. I just want to ask is it worth to wait? and what about the price?

And if I put up a system right now like devbox, will it compatible with the pascal GPU? I mean, do I need to change MOBO later to fit the new GPU? because I don't know what will the new pascal GPU interface is. THanks a lot. Any comments and suggestions will do.
 
Solution
The server based pascal GPUs use an interface called NVlink according to the link below.

http://www.tomshardware.com/news/nvidia-pascal-tesla-p100-gpu,31557.html

These are the heavy hitters. You aren't buying a gpu or a desktop computer. You are investing in a compute appliance. If this is what you want then waiting for NVlink and pascal is probably a prudent expenditure. If you are not "investing" in a compute architecture I think getting a current gen Compute GPU is probably a better move. No wait and you get up and running today rather than waiting 6-18months for pascal to deploy.
All current commercially available graphics cards and any released in the foreseeable future will use the PCIe standard (PCIe x16 specifically), so compatibility is not an issue. I would actually recommend waiting for the new Pascal GPUs though, yeah. When they come out you can either get one of the stronger Pascal cards or get a current gen card cheaper.
 
The server based pascal GPUs use an interface called NVlink according to the link below.

http://www.tomshardware.com/news/nvidia-pascal-tesla-p100-gpu,31557.html

These are the heavy hitters. You aren't buying a gpu or a desktop computer. You are investing in a compute appliance. If this is what you want then waiting for NVlink and pascal is probably a prudent expenditure. If you are not "investing" in a compute architecture I think getting a current gen Compute GPU is probably a better move. No wait and you get up and running today rather than waiting 6-18months for pascal to deploy.
 
Solution