GeForce GTX280 folds three times faster...

This could be very interesting if true... I've been looking at upgrading my gfx card in the future when I upgrade from my 19" to a 24" widescreen :) Would definately be worth using the GPU client if it only uses one core on a quad.
 
I have an 8800GT in the borg. Now, the borg is running two standard clients as a) they are very stable and can be left alone and b) I dont think an SMP unit would have enough time to complete. The PPD from this thing is of the order of about 300ppw.

I might look into scrapping the two standard clients and running the GPU2 client. It does depend a lot on how robust the client is though.
 
theheyes said:
At the risk of sounding like a points who-ere I'd have to see ppd/wattage figures first.
Click to expand...

ppd for each method was never about which crunched the most information though.

So i doubt itll be any higher than an SMP cleint with a quadcore cpu.
 
ghost101 said:
ppd for each method was never about which crunched the most information though.

So i doubt itll be any higher than an SMP cleint with a quadcore cpu.
Click to expand...

If I'm reading you rightly I think we're on the same side of the argument... All I want to know is whether a card is going to be a good investment. At the moment, for me, the cards (nor the ps3) don't produce enough points to justify the wattage. Of course faster processing does not equal proportionally more points, we already see that with the ps3 client.
 
Do we know if the CPU2 client will require a deticated CPU as the older GPU clients did?

I've not folded on GPU's in a long time so may be way out of date here.
.
 
good stuff. Will this be able to run on multiple GPU's as i have a 8800 ultra for my main display and an 8600 gts for my seconday displays. Also will it would witht he nvidia drivers for the Physx card?
 
KE1HA said:
Do we know if the CPU2 client will require a deticated CPU as the older GPU clients did?

I've not folded on GPU's in a long time so may be way out of date here.
.
Click to expand...

Last I read, yes, it does need a CPU to feed it. Also, current CPUs can't feed the card fast enough. I think Nehalem will be a more suitable option for serious GPU folding with QPI and generally more horsepower. I still suspect it will be cheaper and more productive to just run the SMP client though, but time will tell.

Wizardskills said:
good stuff. Will this be able to run on multiple GPU's as i have a 8800 ultra for my main display and an 8600 gts for my seconday displays. Also will it would witht he nvidia drivers for the Physx card?
Click to expand...

People are experimenting with running multiple AMD cards but I don't think it is supported yet. Also, multiple GPUs are sharing the same bandwidth and the limited pool of CPU time so there are a lot of factors to think about. Welcome to high performance computing. :D

Also, more cards = more watts = more heat. Build and running costs are going to skyrocket. I predict (and I could be wrong, it depends on the points system) that in 18 months time people are still going to be folding on Quad CPUs with Linux VMs, albeit Nehalem variants which look like they are going to be 30-40% faster than the current chips. Only problem with Nehalem is that it's going to mean binning your motherboard and RAM as well. :(
 
theheyes said:
Last I read, yes, it does need a CPU to feed it. Also, current CPUs can't feed the card fast enough. I think Nehalem will be a more suitable option for serious GPU folding with QPI and generally more horsepower. I still suspect it will be cheaper and more productive to just run the SMP client though, but time will tell.



People are experimenting with running multiple AMD cards but I don't think it is supported yet. Also, multiple GPUs are sharing the same bandwidth and the limited pool of CPU time so there are a lot of factors to think about. Welcome to high performance computing. :D

Also, more cards = more watts = more heat. Build and running costs are going to skyrocket. I predict (and I could be wrong, it depends on the points system) that in 18 months time people are still going to be folding on Quad CPUs with Linux VMs, albeit Nehalem variants which look like they are going to be 30-40% faster than the current chips. Only problem with Nehalem is that it's going to mean binning your motherboard and RAM as well. :(
Click to expand...

That is true, but wont i be able to run 2 clients. One for each card?
 
when is this NVIDIA client going to be release?

Also if I have two 8800 GTXs will I be able to run this:

1 x SMP
2 x GPU2

????

think of the Points!!!!
 
You must log in or register to reply here.
Back
Top Bottom