This is relevant to our interests.

If that's a super computer then what the hell do you call the farms that some of the people around here have?



 
I have posted a few times over the past year about Teslas but nobody seemed to know or heard about their performance in folding. It may be just another nVidia card like the Quadros, which would mean a GTX295 may actually outperform it. It's anyone's guess unless someone has hard numbers we can look at.
 
I have posted a few times over the past year about Teslas but nobody seemed to know or heard about their performance in folding. It may be just another nVidia card like the Quadros, which would mean a GTX295 may actually outperform it. It's anyone's guess unless someone has hard numbers we can look at.

Supposedly they're just 280/285s with a bunch more Vram. I asked around a while back, and the general consensus was that it would perform the same as a 280/285. Since F@H is shader dependent and not memory dependent, I would concur. But then again, no one will know for sure until someone gets their hands on one. I feel like if they're just like the GTX280/285 though, why even make them, and who would purchase them over 295? They may also may have different abilities, such has massive CUDA scaling thorugh SLI, but I'm just guessing now.
 
These are specifically geared towards CAD & other graphics intensive tasks (not gaming). The drivers normally are not tuned towards games and even with a GTX280 GPU & more vRAM, it is likely that these would be an epic fail at most graphic intensive games.

GTX280's are great PC graphics cards, but not so great at CAD for the same reason, lowever vRAM and the drivers aren't geared towards CAD & other graphics intensive tasks.

These cards are geared toward a specific niche group on PC Users and therefore will carry a hefty premium. These would NOT be good for anyone looking to build a farm or game.

 
These are specifically geared towards CAD & other graphics intensive tasks (not gaming). The drivers normally are not tuned towards games and even with a GTX280 GPU & more vRAM, it is likely that these would be an epic fail at most graphic intensive games.

I think you're confusing Tesla for Quadro. Tesla isn't a video card (notice no outputs), it's a dedicated computing card. Quadros are the business level video cards.
 
Meh, I have more computing power sitting on my desk right now...... And I didn't pay anywhere near what they want for those things.
 
Those Telsa's are definitley interesting, but my pockets ain't deep enough to spend about 10 grand on a computer. I guess these cards really separate the "ami tree" (amateur) folders from the professional folders. I'll just stick with my "anti Q" (antique) stuff :D

DCing for the CURE
 
Here's the Tesla C1060 Computing Processor product page. Apparently you can buy these cards today for $1000+. Since it appears to be a GTX 285 with more memory, slower clock speeds, and no video outputs, it doesn't seem worth it. You can get a GTX 285 2GB card for less than half the price, so this is only for people who really need 4GB for CUDA. I think crazjayz is right that there's no benefit for F@H.
 
Here's the Tesla C1060 Computing Processor product page. Apparently you can buy these cards today for $1000+. Since it appears to be a GTX 285 with more memory, slower clock speeds, and no video outputs, it doesn't seem worth it. You can get a GTX 285 2GB card for less than half the price, so this is only for people who really need 4GB for CUDA. I think crazjayz is right that there's no benefit for F@H.
The only thing we don't know is if the "optimized" drivers would eat FAH for breakfast. It's possible.

However, it's also possible FAH wouldn't even run on this, since FAH is looking for a GPU with a video output.
 
The only thing we don't know is if the "optimized" drivers would eat FAH for breakfast.

Well, that just means we got to get to work hacking the Tesla drivers to work on the GTX cards. ;)
 
Well, that just means we got to get to work hacking the Tesla drivers to work on the GTX cards. ;)
That would probably do it :)

Although, AFAIK, the old consumer card -> pro card modding has somewhat diminished in recent years.
 
On the product page it mentions "Single-precision floating point performance - 933; Double precision floating point performance - 78" Those specs are not listed on the GTX series of cards, but they do indeed have those specs for them. I think that was one of the big hurdles to make the GPU client work was to use graphics cards designed for single-precision to do work that technically might require double precision. With a game if the GPU hiccups you might lose a frame. If your running a FEA stress simulation on an aircraft structure you might not have quite as much forgiveness for computational errors.
 
Back
Top