2 x GPU Options for GUI

[sunsetsothickly]

Hi Guys,

I am limited in the scope of what I can purchase for work - resellers/manufacturers-wise.

I am planning to run 50:50 OSX:Win7 - Win7-64 being exclusively for large datasets in Amira until the 64-bit OSX version comes out later this year.
From experience (A Dell T3600 with a single Quadro 4000), a lot of the CUDA-based filters in Amira cause GPU timeouts in windows if you are running a single card for computing and display.

If a second computer (a MP Hex 3.3) were available and I can convince IT purchasing that the EVGA GTX 680 is legit, what would be the best option for a second GUI GPU to avoid those timeouts? Power is going to be the issue? I assume that the 5770 already in there will take too much power to run at the same time as the 680 without installing a supplemental PSU (an option which got a firm 'no' from IT)?

Is it going to be a GT120? Is there anywhere that I can still buy these new? Has anyone confirmed that a GTX680 + GT120 config on standard power cables is stable and free of smoke?

One more thing - who has experience with 64GB of RAM in the hex core? Is the 48GB limit firm?

Thanks for your help

 

[Tesselator]

For figuring out power you can use this site and add up the cards: http://www.hwcompare.com I think MacPro models can supply a maximum of 150Watts from the PCI buss and 75W from each of the two aux connectors (so 300W in total - max).

 

[MacVidCards]

GT120 is fine power wise

So is 2600XT

The GT120 has some CUDA support but is of a different compute capability which causes problems in some apps, but DaVinci Resolve has relied on it as 2nd GUI card for awhile

 

[ashurao]

… I think MacPro models can supply a maximum of 150Watts from the PCI buss and 75W from each of the two aux connectors (so 300W in total - max).

hello,

are you sure? I thought the PCI Express can provide 75W maximum.

 

[OS6-OSX]

300

 

[666sheep]

300

300 total, 4 slots and 75W per slot.

Even as PCIe 2.0 slot is theoretically capable of delivering 150W, official PCIe 2.0 electromechanical specification does not allow cards to use more than 75W.

 

[Tesselator]

300 total, 4 slots and 75W per slot.

Even as PCIe 2.0 slot is theoretically capable of delivering 150W, official PCIe 2.0 electromechanical specification does not allow cards to use more than 75W.

Really? Wow, Apple's site says 150W on a single PCI connection or 300W total between the four slots. And of course 75W per mother board aux socket. Thus 300W cards can be used in a single slot (max) - as 150W on the card-edge and 75W x2 from the aux sockets.

 

[666sheep]

Really?

Seems so. See page 8 of attached pdf.

Wow, Apple's site says 150W on a single PCI connection

Link? (Sorry, I'm lazy today, and I'm watching "Into The Wild" in the background )

 

[Tesselator]

Link? (Sorry, I'm lazy today, and I'm watching "Into The Wild" in the background )

Me too, too lazy... But it was the in Apple data sheets for the 2006 MP1,1 - I suppose t could have been the MP3,1 though as I've looked that up recently as well. This was about 2 or 3 weeks ago.

 

[sunsetsothickly]

Thanks guys,
I have also heard about the GT 640 as a slot-powered card with a bit more oomph than the 120. Have you heard many reports of it working, MVC? Have you ever flashed one?

 

[ashurao]

Hello,
I have a 640 GT 4GB (Evga) and it works very well.

Despite the weakness of this card ((compare to other cards like the GTX 680), display in ArchiCAD and CINEMA 4D is fluid.

But it is not flashed.

 

[sunsetsothickly]

How about the 1GB GT 120s

Any experience anyone? I am watching MVC's store for a GT120 for a 5,1, but have seen flashed 1GB versions around as well.

 

[Tesselator]

Well the GT120 got a ton of negative user complaints from the game and CG crowd for the 1st 6 months it was out. People still point to it when they wanna bash on Nvidia. <shrug>

Myself I just opted for a GTX 570 and I think the discussion of the GT 640 (above) was interesting. The GT 120? Not so much...

 

[sunsetsothickly]

640 is ordered

Looking at the comparisons between the 640 and the 120, I decided to go for the 640. There was a 2GB low profile Zotac version (didn't want it to occupy two slots given the 680 will do that) which I have seen reports about on reduser.net going for AUD79 locally.

I will install it stealthily and let you all know how it goes...

 

[Tutor]

Looking at the comparisons between the 640 and the 120, I decided to go for the 640. There was a 2GB low profile Zotac version (didn't want it to occupy two slots given the 680 will do that) which I have seen reports about on reduser.net going for AUD79 locally.

I will install it stealthily and let you all know how it goes...

My $109 EVGA GT 640/4G is my interactivity card of choice when using CUDA based Octane renderer for Cinema 4d scenes.