GPU to 72 status...

[Mark Rose]

Originally Posted by LaurV

On the dark side my Titan let the magic smoke out today. It was DC-ing Manfred's assignment in 332M (we did it in parallel to check the residues, and coordinate on the PM on this forum). It was close to 150M iterations (we had a match to 145M, he is a bit more advanced, as he is the assignee, I periodically had to stop and switch to other job, to let him lead the way). Now, I will have to look for some power IC which exploded (that is Q16 and L12 coil, for who knows the PCB layout of the EVGA Titan classical design).

:(

[Mark Rose]

From what I understand, What Makes Sense honours bit level settings and Let GPU72 Decide honours the work type.

I like the idea of a "I Don't Care" option, where GPU72 has complete control.

[chalsall]

Originally Posted by LaurV

In fact, what is the actual, practical, (as opposed to theoretical) difference between WMS and LGPU72D? (on the server side, i.e. what the server does for any of the requests?).

The difference internally is WMS honours the Pledge level (e.g. 73), while LGPU72D sets it (usually to 75, sometimes to 74).

[Gordon]

Originally Posted by chalsall

For a steady state, about 40.

Good job I'm still doing 48 a day then.

[Uncwilly]

Originally Posted by chalsall

George asked me to bring in those candidates in the 332M range which are not yet appropriately TF'ed. I brought in approximately 3,500 candidates between 332192831 and 332.6M which had not been assigned to registered users, and which were TF'ed to less than 78 bits.

Can you have "Fetch Spider" acquire all expos above 332792600 (up to 333M) that are at 72?
That, with what I have queued, will finish off the bit level from 332192831 out to 333M.

[casmith789]

Not that I am doing much, but I have switched to DCTF once my current assignments complete.

What option do I choose to let GPU72 decide whether to do LL TF or DC TF? I can't figure it out.

edit: on misfit

[chalsall]

Originally Posted by Uncwilly

Can you have "Fetch Spider" acquire all expos above 332792600 (up to 333M) that are at 72?
That, with what I have queued, will finish off the bit level from 332192831 out to 333M.

Done. Actually, I've brought in everything below 78 bits not already assigned.

[chalsall]

Originally Posted by casmith789

Not that I am doing much, but I have switched to DCTF once my current assignments complete.

Thanks

Originally Posted by casmith789

What option do I choose to let GPU72 decide whether to do LL TF or DC TF? I can't figure it out. edit: on misfit

For legacy reasons, there isn't a way to have GPU72 assign work types different than what is requested (i.e. LLTF or DCTF). This is probably something which should be retired -- my original thinking was that some definitely wanted to do LLTF, while others definitely wanted to do DCTF.

When I have some cycles I'll update the code such that the "Let GPU72 decide" LLTF requests will assign DCTF if required. Anyone who definitely wants to do LLTF please choose any other option (including "What Makes Sense"), preferably to 75 if you're willing to go that far.

[chalsall]

Originally Posted by Gordon

Good job I'm still doing 48 a day then.

I should have been clearer. When I said "about 40", I meant about 40 more than we were doing.

[Uncwilly]

Originally Posted by chalsall

Done. Actually, I've brought in everything below 78 bits not already assigned.

What I was asking for was above your original 332.6M cut-off.

[chalsall]

Originally Posted by Uncwilly

What I was asking for was above your original 332.6M cut-off.

And that's what I did, but not just those at 72 bits.