PDA

View Full Version : Tesla vs. Quadro Plex



Dexter2999
03-08-2009, 11:42 PM
Hey, I'm a little slow on some things. Maybe someone here can shed some light on some things for me.

Nvidia has had the Quadro Plex line for almost three years now. The price point has been about $10,000. I guess it is basically an external housing that slaves and sync's two Quadro FX 5400 cards together and you can slave an additional two more in a second housing (for another $10,000)and tie them together into one rackmount-able unit. Pretty neat really.

Now years later they introduce the Tesla supercomputer line. It appears to be four video cards slaved together into a single housing. Again the pricepoint is around $10,000. Yet the Tesla has appears to have four cards to the Quadro Plex's two cards.

So does anyone know if the Quadro Plex is going to be discontinued? Is it going to come down in price?

I read in a post here from a Mac user that the Mac has a limit for power being drawn through multiple video cards. This makes the Quadro Plex seem like an attractive option to accellerate functions on a Mac as it is external with an independent power supply.

If NT rents a Tesla for next years NAB (possible for CORE) it is going to be INSANE!

Anyway, anyone else here have info and or opinions on this stuff?

StereoMike
03-09-2009, 02:23 AM
From what I've seen, 3D Max gets realtime DOF, AO, shadows etc. So they will use GPU pretty good (and Mental Images is onto GPU rendering for quite a while, don't know when it will hit the streets, but it couldn't be so long, maybe the Max stuff is even the MI renderer). Nvidia had something in mind, when they developed those machines.

mike

cresshead
03-09-2009, 02:42 AM
From what I've seen, 3D Max gets realtime DOF, AO, shadows etc. So they will use GPU pretty good (and Mental Images is onto GPU rendering for quite a while, don't know when it will hit the streets, but it couldn't be so long, maybe the Max stuff is even the MI renderer). Nvidia had something in mind, when they developed those machines.

mike


yeah the Tesla's 960 cores...Nvidia making it and Nvidia buying Mental Images looks set to push mental images "mental ray" and so 3dsmax, maya and xsi rendering over to the gpu...also factor in the gpu renderer nivdia had before they bought mental images so they've already been that and done the groundwork.

rendering is moving to GPU as it's developing much faster than cpu's from intel and amd are.

also note the tech heads that were once at sgi moved to Nvidia...

currently i see Nvidia as the competition to intel not AMD.

Dexter2999
03-09-2009, 04:01 PM
Well, I just bring it up because the videos about the TESLA is all about CUDA support and while I didn't know what that meant I remember seeing that LW supports CUDA.

900 cores, wow. Instead of buying/building ten $1000 PC's to render you have a varitable renderfarm right there in the room (or close at hand).

But on a more realistic/affordable level, the idea of a Quadro Plex cage being made available for Mac users who want to take advantage of multiple cards and bypass the system power limitations seems like it could be HUGE. I think the number I read in the post said that the Mac literature said that a maximum of 300w combined voltage through the busses was the limit. That seems like a lot to me but then I have no idea how much voltage these high end cards with built in cooling are pulling for these drastic numbers in multi core GPU set ups.

flakester
03-09-2009, 04:10 PM
~ also note the tech heads that were once at sgi moved to Nvidia...

currently i see Nvidia as the competition to intel not AMD.

Very true statement. From what I hear; NV have apparently released some news about development of an in-house x86 chip.
Though they may have to rely on pushing it out under the licence of one of the firms that they aquired a while ago rather than their own name; can't really see Intel granting a x86 licence to someone that they both hate, and see as a credible threat in this dept.

I could be wrong though.

--
flakester.