Presently I have a AMD PHENOM II 1090 running @3.4GHZ, water cooled.
The vid card is a GTX 550, which puts out 150,000 WU per day, half the time the machine is running Windows 7 32 bit and the outher half its Windows 7 64 bit.
I narrowed down my search to these two, what do you all think I should go with:

ASUS ENGTX580 DCII/2DIS/1536MD5 @ 1581 GFLOPS/sec or

ASUS EAH6970 DCII/2DI4S/2GD5 @ 2703 GFLOPS/SEC.

Which one would be better and why?

I am also interested in feed back on this. I don not have either one( I am an AMD/ATI guy, having 2 HD5870) but from my own research the 580 is the fastest and its CUDA, its good in its own right but I like OpenCL, which the 6970, not the top card but blazzes in its own right and pretty dern quick. The flops tell the story. I think AMD/ATI is optimal here and at most projects. A 6850 is almost just as good and a little cheaper if $$ is a problem, but the add DDR5 is worth the extra $$ for the 6970.
I myself want to get a few.

---

go with an AMD 6970 you want to run moo!, since it runs faster for rc5-72.

I am also interested in feed back on this. I don not have either one( I am an AMD/ATI guy, having 2 HD5870) but from my own research the 580 is the fastest and its CUDA, its good in its own right but I like OpenCL, which the 6970, not the top card but blazzes in its own right and pretty dern quick. The flops tell the story. I think AMD/ATI is optimal here and at most projects. A 6850 is almost just as good and a little cheaper if $$ is a problem, but the add DDR5 is worth the extra $$ for the 6970.
I myself want to get a few.

I myself got a 6850 right after they came out and I too am disappointed about it's overall performance! I usually buy 5770 cards, they are cheaper but I had some extra money when I bought the 6850 and thought, 'hey higher number equals better card'! Well it is better but not THAT much better! My 5770's are getting around 160k per day rac while the 6850 is only doing around 170k at this point. I did bring the 5770's over here sooner but if the 6850 were that good it should be waaaay ahead and it is not. The 6850 only gave me about 45k or so more on Collatz when both 5770's and 6850 were over there. In short I would NOT buy another 6850 if given the choice. I HAVE bought several 5770's since I bought the 6850 though!

Good to know about the 68XX, One would thunk that the new version of the 58XX would be better. Maybe the 69XX is the second gen and works better, given that the 6990 is a monster, from what I read. That will all pale in comparison when AMD/ATI comes out w/there quad core gpu's in early/mid 2012. Nice gpu farm mikey. I got a small farm going myself. Is 4 considered farming? :)

---

Question? I know there are agp/pci gpu card with 96 cores, so..is it possible and/or fesible, to lets say a MB has 3 pci slots, could you use all 3 gpus to crunch, nice psu of course to feed the pci slots? If you have a single cpu not crunching to feed them, or a dual core would be better, of course, or would the irq get all tangeled up and confused. Im sure the flops..or flop of the gpu would be around the same as the cpu, but it would be like having a tri/quad core so to speak..Just a thought, im sure some on has tried it, and the next question where could it crunch? Thats probly the big henderance other wise that combo would be out there. Just imagine, as in my case, an 5870 and 3 pci gpu's with my I7...Funny. It sure would utilize non pci-e machines.

---

Good to know about the 68XX, One would thunk that the new version of the 58XX would be better. Maybe the 69XX is the second gen and works better, given that the 6990 is a monster, from what I read. That will all pale in comparison when AMD/ATI comes out w/there quad core gpu's in early/mid 2012. Nice gpu farm mikey. I got a small farm going myself. Is 4 considered farming? :)

Thanks! I too started with 1 and have been growing, I do now what it is in gpus though. In cpu's there was an actual line of from 1 to 5 machines was a gardener, 6 to 10 was a something else, then 10 to 15 was something else and then over 15 was a rancher!

Question? I know there are agp/pci gpu card with 96 cores, so..is it possible and/or fesible, to lets say a MB has 3 pci slots, could you use all 3 gpus to crunch, nice psu of course to feed the pci slots? If you have a single cpu not crunching to feed them, or a dual core would be better, of course, or would the irq get all tangeled up and confused. Im sure the flops..or flop of the gpu would be around the same as the cpu, but it would be like having a tri/quad core so to speak..Just a thought, im sure some on has tried it, and the next question where could it crunch? Thats probly the big henderance other wise that combo would be out there. Just imagine, as in my case, an 5870 and 3 pci gpu's with my I7...Funny. It sure would utilize non pci-e machines.

I know John Clark is playing with pci crunching gpu's and has put several in one machine, or is at least trying to. He is on Malaria and Prime Grid if you want to ask him. Here is John on PG:
http://www.primegrid.com/show_user.php?userid=78354

Tell him I sent you and he will help you if he can.

Good to know about the 68XX, One would thunk that the new version of the 58XX would be better.

AMD changed their numbering scheme the 68XX is a replacement for the 57XX NOT the 58XX. The 68XX does not support DP math.

The 6970 IS the replacement for the 58XX.

Friends,

I bought a XFX HD6850 end of August and my average has increased a lot. What is the best core to set in Moo? Anyone know? Appreciate comments on this. I´m only crunching MOO.

Bernt

---

I think there is a "tester", of which I have not used, maybe someone can point you there. I have read in a few posts about it, to find a good core number. But I think it will come down to going through the cores and finding what works fer your set up.

---

Brent to test & find your best core do this:
Exit Boinc manager
From the run command type in:

"C:\Program Files\BOINC\Data\projects\moowrap.net\dnetc518-win32-x86-stream.exe" -bench

Just make sure of the location of your:
\projects\moowrap.net\dnetc518-win32-x86-stream.exe

The program exits after 10 second so be quick to get the correct core

Now change the settings in Moo! Wrapper preferences & restart Boinc

Copy,

Thanks a lot for your advice. Back from military service so I will test it ASAP!

Rgds

Bernt

---

Brent to test & find your best core do this:
Exit Boinc manager
From the run command type in:

"C:\Program Files\BOINC\Data\projects\moowrap.net\dnetc518-win32-x86-stream.exe" -bench

Just make sure of the location of your:
\projects\moowrap.net\dnetc518-win32-x86-stream.exe

The program exits after 10 second so be quick to get the correct core

Now change the settings in Moo! Wrapper preferences & restart Boinc

Copy,

I did run the program with the -Bench parameter and got #0 as a result. Edit of preferences and restart of Boinc. I found out that #0 fluctuates in processtime quite a lot. I have tried with #3 and then I have an average of 4,5 minutes per unit and a total time of 50 to 55 minutes with 12 units. So I stick to core #3 which is a bit slower but more reliable.

Rgds

Bernt

---

I have tried with #3 and then I have an average of 4,5 minutes per unit and a total time of 50 to 55 minutes with 12 units. So I stick to core #3 which is a bit slower but more reliable.

Rgds

Bernt

I have 2 x i7-920's... one HAD 2 x 4770's and the other 2 x 5850's.

The 4770s outperformed the 5850's amazingly, much to my annoyance, especially when even after I tested which core etc, the results were the same.

I have now put 1 x 6950 in place of the 2 x 4770's but I still dont get any of these amazing times as posted above ^^^

4.5 minutes ??!!??!! - Mine are taking 31 to 33 minutes on the 6950 !!! and the times are equally as bad on the 5850's (even though I have read of others doing them in fantastic times on the same cards...

Any help in rectifying this would be greatly appreciated... maybe (well, definitely) even get me to stay at Moo permanently !

Thanks in advance,

Veebee

There is not a card on the Planet that can do the WU in 4.5mins, they would have to be clocked so insanely high that it would instantly melt down. Not even a dual 5970 thrashing away with 4xGPUs could get close to that figure.

He was refering to the average WU which has 12 packets of Stats Units inside it to make up a complete WU. The 4.5 mins was losely refering to an "average" time per stat unit packet as shown inside the stderr file.

The real world reality times for the 6950 are pretty much what you are getting, your card is doing fine. Arguably a little slow maybe, just depends what clocks you are running the GPU at. Would be better to look at the top computer list, and keep scrolling down until you find a PC with a 6950 loaded, and look at its times as a comparitor. They will not be far different from what you are getting at present.

Regards
Zy