Multi gpu usage?

Boomer21b

Prominent
Jun 22, 2017
2
0
510
So I'm building a new rig to serve as a workstation as well as a gaming rig. I'm not personally knowledgeable in regards to gpu's or how computers work in general but I want to build my own for the sake of utility. I'm going to be using a i7 7700k, and a gtx 1050ti (for gaming), and my old nvidia tesla c2075 for 3D cad and general number crunching purposes. I'm wondering about how to make sure the right card is employed at the right times. How do I ensure that my cpu makes the most use out of the two cards available to it? Really tired of trying to game on an old crummy laptop running an i3 and 4 gigs of ram when I can justify having a power house workstation pc for data crunching and r&d. Don't want to build two seperate pc's. I just want one that can do high quality gaming, and still serve as a workstation. Tia.
 
Solution
For your purposes I would suggest a Ryzen 1700 over intel 7700.
The Ryzen has double the cores/threads which would be better for yoru applicaitons.
Intel has better single core performance so the intel is like haivng 8 people that can do 115 tasks per second and the Ryzen is like 16 people that can do 100 tasks per second. If the task (like say installing an engine into a car) can only be done by 4 people then the intel is better because each person is faster/more efficient; however in an all hands task (like building a house) the 16 slightly slower people is going to be better.

I also suggest you just get a better GPU and ditch the Tesla card. A GTX 1060 has almost 3x the number of CUDA cores and will crunch the numbers faster.

You build is totally unbalanced. Why do you have so much CPU power, yet a budget GPU? It may create a bottleneck. However, that 1050 is not going to pair well with the GPU since it doesn't offer SLI. You can get the GTX 1080 Ti (maybe that's what you mean?) and just leave the C2075 alone. The GTX 1080 Ti should be powerful enough to handle any game and data crunching as well to offload the CPU. However, the new Ryzen chips can perform multi-threadded task and heavy loads as well. So maybe pairing up a Ryzen 1800x and a GTX 1080 would be your best move.
 

Boomer21b

Prominent
Jun 22, 2017
2
0
510
I'm not really a computer guy. I do r&d and machining/prototyping for firearms manufacturing. I need to be able to run my cad and make animations of prototypes prior to development as well as doing ballistics and trajectories mapping. I know I need a good cpu and the 7700 seems like a good choice and the c2075, though old, does ok so I was just gonna implement that in the new rig. I didn't actually know the 1050 was considered a budget card, if the 1080 is significantly better I'll pick one of those up instead (tbh I just picked a gaming gpu that was on the newer end of the development spectrum kind of at random).
 
Well it goes like this 1050 ($100) > 1060 ($250) > 1070 ($400) > 1080 ($500) > 1080 Ti ($700) > Titan X ($1200)

So you can get a better understanding, you can see where the cards lay. I wouldn't go for a 1070 right now because of the mining craze, they are either out of stock are marked up. So it would make sense to just get a 1080 or 1080 Ti. If money is truly no issue, then of course the Titan X is the preferred choice. Again, if you are doing a lot of R&D work, get the 1800x. It can handle games, and do some very heavy CPU work as well. It outperforms the 7700K in any shape form or fashion.
 
For your purposes I would suggest a Ryzen 1700 over intel 7700.
The Ryzen has double the cores/threads which would be better for yoru applicaitons.
Intel has better single core performance so the intel is like haivng 8 people that can do 115 tasks per second and the Ryzen is like 16 people that can do 100 tasks per second. If the task (like say installing an engine into a car) can only be done by 4 people then the intel is better because each person is faster/more efficient; however in an all hands task (like building a house) the 16 slightly slower people is going to be better.

I also suggest you just get a better GPU and ditch the Tesla card. A GTX 1060 has almost 3x the number of CUDA cores and will crunch the numbers faster.

 
Solution