Mostly valids on a GTX1060 and running 1X or 2X seems to work. 2X gives me 1.5X more production. Now if the run time were to be shortened life would be better.
Your two examples continue the pattern I think I've seen that the Linux version of GW GPU v1.07 runs far faster than does the Windows version on comparable hardware.
Just in the last couple of days, I've run some v1.07 tasks on my Radeon VII Windows machine, which is a Windows 10 machine with a decent CPU. The elapsed times are pathetic, especially considering how wonderfully productive this same machine is on Gamma-Ray Pulsar tasks of the current flavor here.
I'll post details later--my quorum partners are slow in responding, so at the moment I don't even know whether I got any validations on the 2X or 3X runs, though I did at 1X.
Just in the last couple of days, I've run some v1.07 tasks on my Radeon VII Windows machine, which is a Windows 10 machine with a decent CPU. The elapsed times are pathetic, especially considering how wonderfully productive this same machine is on Gamma-Ray Pulsar tasks of the current flavor here.
I'll post details later--my quorum partners are slow in responding, so at the moment I don't even know whether I got any validations on the 2X or 3X runs, though I did at 1X.
The quorums are still very slow in forming, but as of now I have two validations at 3X, one validation at 2X, and zero pending inconclusives, so it is at least possible that with current drivers and software the Radeon VII may be a viable 3X platform for V1.07 GW.
BUT, for my particular machines--both Windows 10, with the RX 570 PC having a very modern but cheap and low clock-rate hex-core CPU (i5-9400F, Coffee Lake, lousy thermal solution, nominally 2.9 GHz), and the Radeon VII on a previous generation but more performance-oriented quad-core (i5-4690K, Devil's Canyon, excellent thermal solution, nominally 3.5 GHz) the Radeon VII is turning in elapsed times at 1X, 2X, and 3x, essentially the same as the RX 570. As the Radeon VII gets about three times as much Gamma-Ray Pulsar throughput as the RX 570, this makes running the Radeon VII on the current GW application 1.07 seem unwise in my case.
Quite possibly the Linux situation for Radeon VII on GW V1.07 may be vastly more favorable regarding elapsed times--I hope someone will report.
I'd say the jury is out, and will remain out, on validity rate for this configuration. My machine has reported 19 task completions with some each at 1X, 2X, and 3X. Only five quorums have formed, and all those declared my result valid, but we have seen the same machine switch back and forth from all invalid, to all (or nearly so) valid on GW V1.07 GPU tasks. I'm not interested in sacrificing the enormous amount of Gamma-Ray work which would be required to begin to clarify this point on this machine.
Running 2X over the last 24 hrs I have processed a bit over 22 GW tasks on a GTX1060 with only 1 so far declared invalid. That's an invalid rate of 4.5%. That's a big improvement but IMO it is not ready for prime time.
I agree that validation takes a long time. My dedicated machine has process almost 1k GW but only 262 have been validated, no invalids or inconclusives. All the rest are still pending.
Your two examples continue the pattern I think I've seen that the Linux version of GW GPU v1.07 runs far faster than does the Windows version on comparable hardware.
How about the RX 570? I haven't seen any Linux times for it. I could do that if necessary, though it is more convenient on Win7 for me.
How about the RX 570? I haven't seen any Linux times for it.
Several people have already posted their experiences with RX 570s in this very thread. This is a post I made and there have been some details given by others as well. I'm not so sure that Linux behaves any differently from Windows. There are probably lots of back-end changes going on that we aren't aware of. It's difficult to draw any proper conclusions just yet.
Mostly valids on a GTX1060
)
Mostly valids on a GTX1060 and running 1X or 2X seems to work. 2X gives me 1.5X more production. Now if the run time were to be shortened life would be better.
What are your times? I'm
)
What are your times? I'm running a pair of 2080 Supers. Time is 1040 sec. 406 pending, 109 valid, no errors or invalids.
Running 2X ~ 7800 sec.
)
Running 2X ~ 7800 sec.
Betreger and Zalster, Your
)
Betreger and Zalster,
Your two examples continue the pattern I think I've seen that the Linux version of GW GPU v1.07 runs far faster than does the Windows version on comparable hardware.
Just in the last couple of days, I've run some v1.07 tasks on my Radeon VII Windows machine, which is a Windows 10 machine with a decent CPU. The elapsed times are pathetic, especially considering how wonderfully productive this same machine is on Gamma-Ray Pulsar tasks of the current flavor here.
I'll post details later--my quorum partners are slow in responding, so at the moment I don't even know whether I got any validations on the 2X or 3X runs, though I did at 1X.
The invalids continue but at
)
The invalids continue but at a much lower rate the previously. Tomorrow I think I'l go back to running 1X.
archae86 wrote:Just in the
)
The quorums are still very slow in forming, but as of now I have two validations at 3X, one validation at 2X, and zero pending inconclusives, so it is at least possible that with current drivers and software the Radeon VII may be a viable 3X platform for V1.07 GW.
BUT, for my particular machines--both Windows 10, with the RX 570 PC having a very modern but cheap and low clock-rate hex-core CPU (i5-9400F, Coffee Lake, lousy thermal solution, nominally 2.9 GHz), and the Radeon VII on a previous generation but more performance-oriented quad-core (i5-4690K, Devil's Canyon, excellent thermal solution, nominally 3.5 GHz) the Radeon VII is turning in elapsed times at 1X, 2X, and 3x, essentially the same as the RX 570. As the Radeon VII gets about three times as much Gamma-Ray Pulsar throughput as the RX 570, this makes running the Radeon VII on the current GW application 1.07 seem unwise in my case.
Quite possibly the Linux situation for Radeon VII on GW V1.07 may be vastly more favorable regarding elapsed times--I hope someone will report.
I'd say the jury is out, and will remain out, on validity rate for this configuration. My machine has reported 19 task completions with some each at 1X, 2X, and 3X. Only five quorums have formed, and all those declared my result valid, but we have seen the same machine switch back and forth from all invalid, to all (or nearly so) valid on GW V1.07 GPU tasks. I'm not interested in sacrificing the enormous amount of Gamma-Ray work which would be required to begin to clarify this point on this machine.
Running 2X over the last 24
)
Running 2X over the last 24 hrs I have processed a bit over 22 GW tasks on a GTX1060 with only 1 so far declared invalid. That's an invalid rate of 4.5%. That's a big improvement but IMO it is not ready for prime time.
I agree that validation takes
)
I agree that validation takes a long time. My dedicated machine has process almost 1k GW but only 262 have been validated, no invalids or inconclusives. All the rest are still pending.
archae86 wrote:Your two
)
How about the RX 570? I haven't seen any Linux times for it. I could do that if necessary, though it is more convenient on Win7 for me.
Jim1348 wrote:How about the
)
Several people have already posted their experiences with RX 570s in this very thread. This is a post I made and there have been some details given by others as well. I'm not so sure that Linux behaves any differently from Windows. There are probably lots of back-end changes going on that we aren't aware of. It's difficult to draw any proper conclusions just yet.
Cheers,
Gary.