I realize that there is a creditnew effect working its way through RAC figures these days, but my rigs are off about 40% from where they were about six months ago. While I have reduced WU per gpu down to one on several cards/machines,(because of heat) I can't imagine that it would make that much difference. I've even upgraded two machines recently. All machines have been running fairly steady (when WU availability allows) for several months. I run a stable 266.58, a stable Lunatics x0.37, and a stable BOINC 6.12.33, 24/7 on all machines. I was having a lot of invalids and errors months ago with newer drivers. I since rolled back, and that seems to have calmed down. As I look at my results, they seem to be OK time wise. I'm just concerned that I don't get near the RAC that I used to with better hardware. What am I missing? Where can I look to find any bottlenecks? Thanks for any advice.

Russ

---

Proud member of the half bill club!

I can't say for sure, but I think SETI has been hit by the ghost bug again. I know I have more pending work units than I used to. I think this may be related to the Hurricane Electric issue. It could be a situation where wingman can't upload and report timely but were able to download and crunch the work units. This is just speculation.

---

Are you still running x32f on all your hosts? Get the 0.38 installer and install the x38g Cuda app, it's quite a bit faster, especially on Fermi GPU's,

Claggy

Yep, and I had all kind of DC problems. That's why I'm on stable versions. The same I was on months ago when I had better figures.

---

Proud member of the half bill club!

I have better rac and times with my older driver on the 37 lunatics app and less driver problems. The older and more stable driver and the older more stable lunatics apps is best for me also. I took a 10000 rac hit with the newer drivers and the newer lunatics .38 app.

---

For me i've had no stability problems with the current Lunatics build & the 275.33 drivers. No errors, BSODs or lockups.
RAC is at an all time high.

---

Grant
Darwin NT

Don't know the numbers but going from 2 or 3 WUs to 1 would mean a great deal if I'm not mistaken, the Fermi cards don't work at all up to its potential (is that spelled right?? :) ) with only 1.

I realize that there is a creditnew effect working its way through RAC figures these days, but my rigs are off about 40% from where they were about six months ago. While I have reduced WU per gpu down to one on several cards/machines,(because of heat) I can't imagine that it would make that much difference. I've even upgraded two machines recently. All machines have been running fairly steady (when WU availability allows) for several months. I run a stable 266.58, a stable Lunatics x0.37, and a stable BOINC 6.12.33, 24/7 on all machines. I was having a lot of invalids and errors months ago with newer drivers. I since rolled back, and that seems to have calmed down. As I look at my results, they seem to be OK time wise. I'm just concerned that I don't get near the RAC that I used to with better hardware. What am I missing? Where can I look to find any bottlenecks? Thanks for any advice.

Russ

280.26 has made a big difference to one of my machines and seems stable. The GTX570s were downclocking so that I was seeing GPU usage in the 80% range instead of 99% now - only seen on Vista64. XP 32 and 64 machines always show 99% with both drivers. (all cases with lunatics 38g)

That was starting to make a difference to RAC but at the moment it seems to be generally on the rise on all my machines for some other resaon.


John.

---

GPU Users Group