Best Practices For Building A Multiple GPU System

[glimpse]

=DDD that was a post with a lot of guesses..What I meant to say (let's try to get things short).

If the thread is called "Best Practices For Building A Multiple GPU System" - stand for it, "clean the room" so the thread would be worth it's name...&/or create other one to spread "rumors" around.

if this is open forum it would be cool to find relevant information on the topic..otherwise we will see inflation of this..as value of these threads are going to drop..

but that's just my oppinion.

[Tutor]

=DDD that was a post with a lot of guesses..What I meant to say (let's try to get things short).

If the thread is called "Best Practices For Building A Multiple GPU System" - stand for it, "clean the room" so the thread would be worth it's name...&/or create other one to spread "rumors" around.

if this is open forum it would be cool to find relevant information on the topic..otherwise we will see inflation of this..as value of these threads are going to drop..

but that's just my oppinion.

Glimpse,

I greatly appreciate your expressing your opinions. So, thanks for this suggestion. However, please keep in mind that I was responding, in relevant part, to Notiusweb whose last post stated, in part, "The big question in my mind was, is Pascal 'ready' or is it 'delayed', because it had been touted as this amazing successor to the previous generation consumer level GPU and now it only got mentioned as part of the enterprise products. I wonder why they didn't even make mention of an upcoming consumer level GPU. I say this as someone who will buy the next GPU... I'm all for AI and the coming singularity, but hurry up...too slow! [Emphasis added.] For 38 years, I've research issues and gave paid advice to businesses that sometimes involved my giving my best guesses of potential outcomes in situations where there was no exactly matching extant outcomes. So, I now find it extremely hard not to provide free information to my distant cousins/friends here, no whether what it involves nor what the thread is entitled. So, please try to accommodate my idiosyncratic behavior. I'll work on controlling it, but I can't guarantee you that I won't have relapses.

[Notiusweb]

Re: Best Practices For Building A Multiple GPU System
Post by Tutor » Wed Apr 06, 2016 4:51 pm
Glimpse,

I greatly appreciate your expressing your opinions. So, thanks for this suggestion. However, please keep in mind that I was responding, in relevant part, to Notiusweb

Haven't you learned by now? NEVER EVER respond to Notiusweb.
#NeverNotius

[Notiusweb]

I put up a new high score for bench at 1260. I saw some financial predictions coming from analysts saying the stock could go higher if there was a merger with a networked node and if regulations loosen up on the max # of GPU.

[Seekerfinder]

I put up a new high score for bench at 1260. I saw some financial predictions coming from analysts saying the stock could go higher if there was a merger with a networked node and if regulations loosen up on the max # of GPU.

[Notiusweb]

Tutor et al,

What are your thoughts on SSD being used in available PCIE express lanes as far as Best Practices for a multi GPU rig, as in does it work well, or is it a burden on the rig's ability to perform GPU functions. Or if you tried, what have you found works well or what doesn't work well?

[Seekerfinder]

Tutor et al,

What are your thoughts on SSD being used in available PCIE express lanes as far as Best Practices for a multi GPU rig, as in does it work well, or is it a burden on the rig's ability to perform GPU functions. Or if you tried, what have you found works well or what doesn't work well?

Hi Notius,
That's a prevalent question. I think the crux is the statement "available PCIe lanes" - how do you know which / how many is available? There's a lot of info on PCISIG's website and they now have an M.2 specification for PCIe as well. The problem is that it's Greek for most of us (unless you're Greek of course).

The problem is that without advanced diagnostics, all we will know is if it works or if it doesn't in scenario n with a myriad hardware and software configuration combinations to troubleshoot. Another factor is that we have traditionally gotten precious little from Otoy in relation to questions like these so it's trial and error. I would love to see a little app that maps out the PCIe lane usage (perhaps there is there such a thing?), but until then our take on this is to build dedicated render systems and not systems that try to be all things to all men. This could be a challenge to single artists that a single machine for both production work and rendering, however in that case I would recommend solving the storage problem in another way since there are many options that do not 'threaten' those precious PCIe lanes.

I would also love to hear the opinion of others on this matter though.

Best,
Seeker

[glimpse]

Guys, I'm wondering, how many GPUs You managed to plug most on single PSU? (yeah, I'm aware of Watage), but.. what is the highest number of GPUs You manage to connect? =)

[itou31]

on my side : 5 (2 Titan black and 3 780Ti) on a 1600W PSU LEPA

[glimpse]

on my side : 5 (2 Titan black and 3 780Ti) on a 1600W PSU LEPA

Ghm, noOne tried to plug more? 1600W PSU seems to be an overkill for 5GPUs sipping under 1000W in total. I've heard some stability issues with more than 5GPUs & I'm curious if that has anything to dow with reality..