X1950 Pro AGP at OcUK!!!!!!!!!!!!!

You could look at it another way.. with the GPU client I am crunching these WUs in roughly 12-13 hours. Now, for the sake of argument, lets say it is 13 hours. That's 25.3 points per hour. So in a day I am getting around 609 points. That easily equals the performance of both cores of my X2 4400 @ 2.6GHz on a good day. All this is while it is running at 2D speeds. I could if I wanted run it at 3D speeds and probably get 660 PPD.

Now, lets suppose I wanted to overclock my card a bit, which many X1900 owners probably do, then I could be expecting above 800 PPD - just ask Concorde.

I don't own a C2D so I don't know what the most PPD it is capable of, but it is hard to pass up a guaranteed 600+ PPD.

SiriusB
 
james.miller said:
no, no, that wasnt a dig. it sounds far worse than intended, sorry lol:) im quite interesed in what they can do with gpu folding, it's been on the cards for quite a while now. I have a question though - whats the relationship between gpu folder and cpu usage? at what point would it become more economical to fold purely on the cpu alone?:)
Click to expand...

Sorry for jumping on the offence then. :)

As of now a gpu folds better than a cpu but eats a whole core and kernal res.

I hope they can knock it down to about 25< and enable it as a working service.


I wish you could run the gpu client on nvidia's, got a go7900gtx sitting here unused most of the time. :(
 
nVidia architecture does not lend itself to folding. I have also heard it is much harder to write code for it too cos of their drivers. This may be untrue though.

SiriusB
 
SiriusB said:
I don't own a C2D so I don't know what the most PPD it is capable of, but it is hard to pass up a guaranteed 600+ PPD.

SiriusB
Click to expand...

Depends on Wus. Given a steady stream of 364 pointers on both cores (I wish) I would get about 2,800 for the E6600 (1,400 per core) and about 2000 for the E6300. On a steady stream of 44 pointers we would be looking at 460 and 422 respectively - quite a sizeable gap there. At the moment, it is worth my while running the GPU client on both C2D rigs as te majority of CPU WUs at the moment are the relatively low scoring 149s and 44s. If the quality of CPU WUs gets better, it may be worth my while switching off the GPU client from a ppd point of view.

We should bear in mind that the GPU is estimated to be 35-40 times more efficient than the CPU and as such will probably be doing a lot more science in the time.

Stan :)
 
but dont they only work so well on x1900 series cards because of all the pixel shaders/pipelines they have (48) which helps with processing folding data, the best gforce only has 24, so even if they did add support for gforce cards, the performance would be half that of an x1900, so probably only equal, or maybe even less than folding with the core you will have to stop using to fold on the GPU. an x1900gt has 36 and that gives barely any better performance than an OC'd c2d (correct me if im wrong)
also theres the fact that folding on your GPU will waaay increase your electricity usage, costing you more money, if you pay for it anyways :]
 
on a slight side note i notice a few people on here saying bout getting the small WU's, i dont know whether i have been very lucky or what, but apart from my 1st WU, (396 points) on my home pc ive had 600 pointers every single time since then, thats been 4-5 in a row i think, and at work they all got 600 pointers first, but then i cancelled them as they were going to take weeks to complete and put them on that <5mb and now they get 150-300 pointers every time
 
I am crunching three 600 pointers at home, no idea what my borgs are up to. Smallest WU I have had in the last few months is a 153 pointer. Most have been 600 pointers :D

SiriusB
 
At this moment, I have 20 cores folding - 3 of which are tied up with GPU clients.

Out of the remaining 17 cores; 11 are on 149s, 2 are on 44s, 2 are on 364s, 1 is on a 600 and 1 is on a 476 pointer - not a great average :(

Stan :)
 
You must log in or register to reply here.
Back
Top Bottom