Folding Forum

It's a sign of the times I guess. Though I know we've had some CPU projects with big base credits for a while now, these new GPU project are upping the bar even more. Unfortunately for me the one I picked up wasn't going to meet the deadline, so I had to dump it. But I imagine some of the new cards are going to chew through these monsters fairly well.

18036-180432

Base credits of up to 420k. That's a lot of points.




It looks like I'm pretty much shut down on GPU folding. But since prices are finally dropping I'll hopefully be able to pick something up without breaking the bank.

They were launched during the Forum downtime, basically under the radar. They're probably missing the feedback that your species of GPU is not suitable for these monsters.
I was actually wondering when we would see monster work units for the GPUs having had them for CPUs for quite a while now.

So far I've had up to 18041, that took about 7.1/4 hours on my system. I'm still looking out for the monster 18042.
The flipside is that these bigger work units produce much lower PPD on my system. I normally float 2.5M-3.2M, I think 18041 produced about 2.3M PPD.

Have a 18402 running right now on a 2070 Super (Ubuntu 18.04.5, NVIDIA 510.47.03) - the estimated TPF was 00:04:51 and 880,000 points/2.6MPPD running at 125/240W power limit.

Checked HfM.net and already had one two days ago on another 2070 Super at a 170/240W PL which ran at 00:04:25 TPF, 927,000 Points or 3.02MPPD so a little shy of the 3.2MPPD I usually get on the 2070S running at this power level and within variance of the normal spread.

aetch wrote: ↑Mon Apr 18, 2022 7:13 pm They were launched during the Forum downtime, basically under the radar. They're probably missing the feedback that your species of GPU is not suitable for these monsters.
I was actually wondering when we would see monster work units for the GPUs having had them for CPUs for quite a while now.

So far I've had up to 18041, that took about 7.1/4 hours on my system. I'm still looking out for the monster 18042.
The flipside is that these bigger work units produce much lower PPD on my system. I normally float 2.5M-3.2M, I think 18041 produced about 2.3M PPD.

I'm not sure why it took me so long to pick one up then, maybe just the luck of the draw. I guess there is no real plan to classify many of the AMD GPU's or iGPU's, but it seems to me to be a waste of resources really. Even a lot of the old slower cards could easily fold the lower priority stuff with generous deadlines, while the quicker more modern GPU's could fold whatever comes along.

IIRC I got some of the large (for my core count) CPU work units, but they weren't as bad as I was expecting. But with limited cores the really big ones went to you guys with the power processors.

But to put it in context, with slower gear like I have, most work units are looked at in terms of days (or at least big chunks of days) needed vs hours. I guess I'm so used to it that a CPU work unit that would take 2-3 days of 23/7 wouldn't really bother me at all.

gordonbb wrote: ↑Wed Apr 20, 2022 4:49 pm Have a 18402 running right now on a 2070 Super (Ubuntu 18.04.5, NVIDIA 510.47.03) - the estimated TPF was 00:04:51 and 880,000 points/2.6MPPD running at 125/240W power limit.

Checked HfM.net and already had one two days ago on another 2070 Super at a 170/240W PL which ran at 00:04:25 TPF, 927,000 Points or 3.02MPPD so a little shy of the 3.2MPPD I usually get on the 2070S running at this power level and within variance of the normal spread.

It's not really shocking that there is variance even with the really big WU's. Feedback in the beta section seems to be almost none, so I doubt much is getting adjusted unless there are problems with the WU's stability.



At least GPU prices are dropping quickly.