DCTF to 70 now fully assigned...
 

Just so everyone knows, all DCTF candidates to 70 are now assigned.

I have adjusted the DCTF assignment page such that if MISFIT or mfloop is the requester and the "pledge" is 70, it is automatically bumped up to 71. This is so no workers are left without work.

[QUOTE=chalsall;409175]It's a bit tricky (many "waves", etc), but if you look at [URL="https://www.gpu72.com/reports/available/"]this report[/URL] and don't see at least a few hundred candidates available in the "75" column, then we're short feeding the P-1'ers.[/QUOTE]

when you say the 75 column do you mean the 67-79M table or the 332.2-332.9M table?

[QUOTE=kladner;409186]Again, many thanks! I did spend a while this morning running chunks of cufftbench.bat on the MSI card. It runs for a while, stops, I start it again, it zips through what it's already done, and then does some more. No error messages. The EVGA card always quit with an 'unknown error."[/QUOTE]

Please excuse the forgoing ignorant drivel. I completely misunderstood what was happening and what was being said.

After the registry change, the MSI card completed:[INDENT]CUDALucas -cufftbench 1024 8192 5
CUDALucas -threadbench 1024 8192 5
CUDALucas -r 0
[/INDENT]I'll run -r 1 overnight. It takes quite a while, doesn't it?

[QUOTE=dragonbud20;409228]when you say the 75 column do you mean the 67-79M table or the 332.2-332.9M table?[/QUOTE]

The former.

[QUOTE=kladner;409249]Please excuse the forgoing ignorant drivel. I completely misunderstood what was happening and what was being said.

After the registry change, the MSI card completed:[INDENT]CUDALucas -cufftbench 1024 8192 5
CUDALucas -threadbench 1024 8192 5
CUDALucas -r 0
[/INDENT]I'll run -r 1 overnight. It takes quite a while, doesn't it?[/QUOTE]

-r 1 completed. Yay!

Woohoo! Glad it worked! I don't think my EVGA 570 ever made it through -r 1.

[QUOTE=wombatman;409277]Woohoo! Glad it worked! I don't think my EVGA 570 ever made it through -r 1.[/QUOTE]

Have you tried reducing the RAM clock?

Hmmm, perhaps not. I'll have to try that out at some point. I was generally content to only have one or two crashes and restarts per LL test, but it's worthwhile to try and do better than that.

I can't say if it is universal, but many consumer Nvidia cards (as opposed to Teslas) have to have the memory slowed by a couple to a few hundred MHz to get through the CuLu self tests. It certainly has been true of all except (possibly) one of the cards I have tested with CuLu.

More "Depth First" LLTF'ing please...
 

Just so everyone knows, we're about a day away from starting to have to release candidates for P-1'ing at 74 bits again instead of the more appropriate 75.

Not the end of the world, of course. But sub-optimal.

I would ask that those who are doing LLTF'ing, please "pledge" to at least 74, or choose the "Let GPU72 Decide" option (which will take candidates to 75 currently). There are many "big guns" who are currently doing "breadth first" (read: TF'ing from 71 to 72); yes, that work has to be done in the future, but we have hungry P-1'ers and LL'ers to feed right now.

Also, those who are currently doing DCTF'ing might consider doing some LLTF'ing for a little while. We're currently about a year or so ahead of even the DC'ing Cat 4 wave so we have lots of time to divert some resources for a while.

But, as always, everyone's free to do whatever they enjoy most; it's your hardware / time / money after all. :smile:

[QUOTE=chalsall;409354]Just so everyone knows, we're about a day away from starting to have to release candidates for P-1'ing at 74 bits again instead of the more appropriate 75.

Not the end of the world, of course. But sub-optimal.

I would ask that those who are doing LLTF'ing, please "pledge" to at least 74, or choose the "Let GPU72 Decide" option (which will take candidates to 75 currently). There are many "big guns" who are currently doing "breadth first" (read: TF'ing from 71 to 72); yes, that work has to be done in the future, but we have hungry P-1'ers and LL'ers to feed right now.[/QUOTE]

I just grabbed 1300 GHzd of LG72D LLTF and it gave me back 34 72->74 assignments in the 72.9M range. Was that intended?