Thanks to Divide Overflow for pointing out that arkayn's download does not contain the readme file. I was not aware of that.

I cannot edit my previous post to change the download link but please note that the readme file is very useful in giving instructions on correctly installing the application and the use of various parameters and count value. It is recommended that everyone installing the manually downloaded version should read it and have it available to refer to if necessary even if they have used a manually downloaded version before. ATI version 0.23 with the readme file included can be downloaded from here.

It includes both 32-bit and 64-bit ATI v0.23 applications. And yes this is the version that includes the Cypress fix (HD 58xx and HD 5970)

And yes this is the version that includes the Cypress fix (HD 58xx and HD 5970)

thanks.

Thanks to Divide Overflow for pointing out that arkayn's download does not contain the readme file. I was not aware of that.

I cannot edit my previous post to change the download link but please note that the readme file is very useful in giving instructions on correctly installing the application and the use of various parameters and count value. It is recommended that everyone installing the manually downloaded version should read it and have it available to refer to if necessary even if they have used a manually downloaded version before. ATI version 0.23 with the readme file included can be downloaded from here.

It includes both 32-bit and 64-bit ATI v0.23 applications. And yes this is the version that includes the Cypress fix (HD 58xx and HD 5970)

And the fix is in on the individual packages, once I actually read this thread I understood what divide overflow really wanted. But it has instillation instructions now as well.

---

Thanks arkayn for adding the readme.txt.

Thanks also for including your instructions.txt which is a thoughtful and considerate addition which will be very helpful to those who have not installed an anonymous platform application before.

Finally my personal thanks for hosting these applications on your site, I have used it quite a few times.

Well,

after two days of testing various apps values and supposed tweaks at least in my case the apps are not worth having
not only do they crunch each wu a tiny bit slower but way more importantly in my case the wu became unstable and about 10% would not validate which if there had been any gain the crashing wu would negate that gain

so i'm back to everything standard and back to everything being fairly happy most of the time

maybe it comes down to card type or number of cards used or something outlandish but yet again i got no performance boost at all

best regards
Ian

maybe it comes down to card type or number of cards used or something outlandish but yet again i got no performance boost at all

best regards
Ian

running 2 WUs single 5870 shaves 1-4 secs, or ~1500-5800 credits gained daily. not much, but not that long ago I would have been ecstatic to have 5000.

however, the thermal cycling note is what got my attention as a potential long term threat and the reason I run concurrent WUs. I noticed 4 WUs provides exactly the same performance as 2 WUs--no additional boost, so tempted to run 4 to further delay thermal cycle. is anyone running 4 WUs? is that a problem for the project or gpu?

I usually run four in one go on a 5970 whenever I do a WU run, its been no problem in a running sense. As to time, it does take off a second or so due to load and unload reduction, and slightly improve throughput - as you point out, no difference in the running time of the actual WU, the slight saving (same order of magnitude as you give) is better than a pig in a poke as they say :)

Thermal cycling definitely a bonus, anything that reduces constant wide up wind down is never a bad thing in the long run as the wind up phase is always the most "stressful" (as such)

Regards
Zy

I usually run four in one go on a 5970 whenever I do a WU run, its been no problem in a running sense. As to time, it does take off a second or so due to load and unload reduction, and slightly improve throughput - as you point out, no difference in the running time of the actual WU, the slight saving (same order of magnitude as you give) is better than a pig in a poke as they say :)

Thermal cycling definitely a bonus, anything that reduces constant wide up wind down is never a bad thing in the long run as the wind up phase is always the most "stressful" (as such)

Regards
Zy

thanks

Thanks for the info guys, crunching two at a time now, just changed count from 1 to 0.5

---