GPU3 beta available

Ok guys..... fire it up!

Let us know some ppd numbers!
 
Just started running it and it pulled down a new core(FahCore_15), downloaded a work unit(Project: 10627 (Run 6, Clone 87, Gen 0)) and started churning so its at least working.
 
I gotta try this when I get home from work.

One thing that caught my eye was the requirement for driver version 197.45. Looks like an upgrade will be in order.

There is also a FYI for the ATI folks but not a good one:

While this release is for NVIDIA only to start, we are actively pushing ATI support (with the help of AMD/ATI), although we have no ETA at the moment. However, please do not use this client with an ATI GPU at the moment.
Click to expand...
 
Last edited:
tjmagneto said:
One thing that caught my eye was the requirement for driver version 197.45. Looks like an upgrade will be in order.
Click to expand...

Anyone have any word on new(er) nV drivers and whether or not they are good? Still rocking 196.25 here on both gpu systems.
 
Sweet!!! Thanks for the heads up!

edit: I'll post 9800GT 512mb score with 720Mhz core and 1836Mhz shaders when it finishes. I'm hitting 1 Minute 48 Seconds minimum TPF so far.
 
Last edited:
Project: 10627 anyone know the points on this? trying it on my gtx260. 20 second time frams
 
Last edited:
I'll wait to see what others are seeing. I am currently running the latest drivers on my GTX285's. If PpD is the same or higher, then I'll give it a try. They are pretty stable and have never EUE'd even on the 101x or 105x WUs.
 
zero2dash said:
Anyone have any word on new(er) nV drivers and whether or not they are good? Still rocking 196.25 here on both gpu systems.
Click to expand...

I've been using 197.45 without any problems or performance issues. The FAQ page has been updated for GPU3 and based on reading that it looks like you ought to be ok, They are stating the basic requirement are 195.62 drivers or newer for the GTX 2xx series cards and are recommending the 197.45.

Well I'm running a few frames with project 10627 (R5, C89, G0) right now and I don't have any info on ppd from HFM.NET yet. What I do know so far is that my frame times are 1 minute:21 sec, no other folding apps running.

After doing more research, the 10626/10627 is worth 610 pts.
 
Last edited:
zero2dash said:
Anyone have any word on new(er) nV drivers and whether or not they are good? Still rocking 196.25 here on both gpu systems.
Click to expand...

Only needs CUDA 2.3 which 195.x+ drivers have. I'm using the 197.45's though because I was preparing for the near release of GPU3. I didn't want to upgrade drivers again later if they upped the CUDA requirement.
 
geez we need ppd numbers.

particularly for fermi cards
Posted via [H] Mobile Device
 
zodac said:
We have some. :cool:

http://www.overclock.net/overclock-net-folding-home-team/739885-news-open-beta-nvidia-gpu3.html
Click to expand...

That forum makes me wonder why we don't have all of the cool smiley icons :)

Those ppd numbers don't seem to be as good as we thought. I thought we were expecting 21-23k ppd from a 480 GTX? I suppose that is still a lot, but the heat and power consumption make a Fermi GPU folding rig not for the faint of heart.

Anyone here get some figures yet? Better yet anyone here with water cooling on an overclocked 480 have some figures yet? :D
 
zodac said:
We have some. :cool:

http://www.overclock.net/overclock-net-folding-home-team/739885-news-open-beta-nvidia-gpu3.html
Click to expand...

Thanks for the link.

The FahCore_15 WUs (Project 10626/10627) have:

610pts, a Preferred Deadline of 2 days, and a Final Deadline of 3 days.

GTX 470 (stock): TPF of 42 secs (thanks franz), so initial PPD seems close to 12.5k PPD
GTX 470 (1600 shaders): TPF of 32 secs (thank you Tank), so PPD is 16.5k.

GTX 480 (stock): TPF of 37 secs (thanks Interpolation), PPD is 14.2k.
*Still looking for TPF values for an OC'd GTX 480*
Click to expand...

So, thinking outloud for myself, a GTX275 would sell for say $170. It gets around 9k ppd. For an average of 53 ppd per dollar.

A GTX 470 gets say 15.5k ppd, and sells for $350, and gets around 44 ppd per dollar.

That's a 20% markup over a used card, so that's not bad.

Although, Tigerdirect 12% bing cashback make the 470 about $308. There's also a guy selling one used here for $300 flat. That makes the cards relatively equal.
 
Last edited:
Humm.... 12k is a good bit of ppd, but the power and 100+C.

I'll stick to SMP
 
Currently doing 5800PPD on my GTX260. I'm not very impressed. I'll let it run this unit and then probably switch back to GPU2.
 
testing on a stock 470 on a fresh install of win7 ultimate 64

Turned the fan up to 100% for kicks and giggles; WOW, good thing it never goes that high in gaming lol

EDIT-
First results, getting about 12k ppd by their numbers, FAHmon doesnt show vaules, but TPF's put me inline with 12k
Running Stock with the fan's locked at 68% (according to Nv panel, but GPUz shows 77%, odd) I'm showing 68-72's aas my temps, and fan is a tad more quiet than the fan on my 295's running 100%. I'll try putting a 120mm fan on it tomorrow and OC'ing the shaders, if a 470 is capable of 16k OC'd, thats higher than I could get one of my 295's to OC to, so I actually have a even trade up, but now im DX11 (my 295's sold for the same price I got a 470 for)
 
Last edited:
This would get released when I'm down to one PC
one PC with onboard graphics :(
 
Ok, swing and a miss. Temps leveled out at 66-69, not bad for 12k ppd, and fan was a bit better than the 295, all in all an even swap with better heat and a bit better power draw

but, I lost the rig after 2 hours. I had win7 Beta on there, and I could only use my computer for about an hour before getting a client expired warning, then 30-hour later the rig would just reboot or crash. Now on a fresh install of a real win7 I'm getting the sameshutdown windows but without the authentication warning so I know it's a hardware issue. I'm assuming my motherboard is starting to go, but have no clue really

atleast temps and points were good, too bad I didn't finish the unit tho

so, Fermi yes, computer No
Posted via [H] Mobile Device
 
Also...

[04:00:32] Folding@Home GPU Core - Beta
[04:00:32] Version 1.24 (Mon Feb 9 11:00:12 PST 2009)
 
Yeah, it's been established that it was still running the GPU2 core. Looks like the client is smart enough to figure out not to assign GPU3 work to ATI GPUs ;).
 
I haven't been folding in along time... I was all set to get going again with the new GPU3 client. Now I am just a fool.
 
OhMyGod said:
I haven't been folding in along time... I was all set to get going again with the new GPU3 client. Now I am just a fool.
Click to expand...

What kind of processor are you running? You could always contribute with a cpu client :)

If GPU3 ever works with ATI stuff, i'm curious to see whether there will be any increases in production. It was my understanding that running ATI hardware almost wasn't worth the electricity with GPU2, along with needing a higher cpu overhead.
 
got my 9800GTX+ working on it..... no PPD yet 1966 Shaders, pulling 1:40 frames... my 8800GS should cross over to the new WUs in about 20 minutes....

holding back on my trio of GTX260s as they look like they get less PPD on GPU3 as GPU2..... I'll ride it out for the next few weeks and probably update the 260s in June.... since the newer client will have a better life expectancy vs GPU2 which might end before the year does.
 
Last edited:
jfb9301 said:
yanno, the folding forum link says that this will work on ATI...... even tho VJ says no support
can't remember what it says about command line
Click to expand...

The client will support ATI, but there are no WUs being assigned is my understanding from all that.
 
If GPU3 ever works with ATI stuff, i'm curious to see whether there will be any increases in production. It was my understanding that running ATI hardware almost wasn't worth the electricity with GPU2, along with needing a higher cpu overhead.
Click to expand...
Since ATI GPUs have much higher GFLOPS numbers than NVidia GPUs, if there ever were a client that was well-coded to take advantage of so much parallelism (128-thread SMP client is nothing - the ability to split into 10,000+ threads would be more useful) then ATI would almost instantly blitz past NVidia in terms of folding performance.

The client will support ATI, but there are no WUs being assigned is my understanding from all that.
Click to expand...
In all likelihood it won't be this client that will be giving out ATI WUs. It'll be some client using OpenCL, rather than CUDA.
 
nitrobass24 said:
The client will support ATI, but there are no WUs being assigned is my understanding from all that.
Click to expand...

sorry about post editing my text.....

rum and postings that make sense are mutually exclusive......

so I fixed it to something that did..... only that made your's not quote mine any more.....


/me raises glass for another round....
 
Personally, if you're not using Fermi, I wouldn't mess with GPU3. PPD seems a little lower and it uses more CPU than GPU2 but thats just me and I'm fairly new to all of this.
 
TYoda said:
Personally, if you're not using Fermi, I wouldn't mess with GPU3. PPD seems a little lower and it uses more CPU than GPU2 but thats just me and I'm fairly new to all of this.
Click to expand...

In the longer term, I don't have much choice... I am away from my rigs for up to 6 months at a time, and I won't have access to them... even VNC.....

So I have to sometimes guess what I need to do to ride these time periods out. Updating and crossing my fingers makes a sight bit of sense....

but not to immediate PPD.... more for long term... if they end GPU2 and my rigs are not shifted over to GPU3, I could loose months of production time... and at 40 KPPD.... lost time hurts

 
You must log in or register to reply here.
Back
Top