The heating season started over here, so I got the machine up again. I retired one of the 6 graphics cards, to avoid repeated reboots. It is 24/7 Einstein now, which means FGRPB1G on 5 GPUs only. RACs are still climbing, it is just short of 1 million now, expecting to double over time.
For technical curiosity, and it seems to be a nice community, and because I likely become super-rich that way, ... er ..., I signed up to gridcoins. That single machine running Einstein@Home only yields a magnitude (which is the gridcoin daily cross-project performance indicator derived from RAC) of 171, which then means that 30 gridcoins are assigned to me per day for that machine. At today's exchange rate this would be about 0.1 dollar/day. Hm.
What was kind of surprising is that even though this is only single machine that is contributing to Einstein, this made it among the top 100 as shown on https://www.boincstats.com/stats/5/user/list/12/0/0 . So, this effort makes a difference and I am happy about that. The setup was not too much of effort, and I now think it was well worth it.
What I have not yet done is to allow for multiple tasks for GPU. This is mostly because I consider the system to be mostly CPU-bound, and it is only one PCI lane per card so I don't want too much communication either. But I should give it a shot in a week or two.
The heating season started over here, so I got the machine up again. I retired one of the 6 graphics cards, to avoid repeated reboots. It is 24/7 Einstein now, which means FGRPB1G on 5 GPUs only. RACs are still climbing, it is just short of 1 million now, expecting to double over time.
For technical curiosity, and it seems to be a nice community, and because I likely become super-rich that way, ... er ..., I signed up to gridcoins. That single machine running Einstein@Home only yields a magnitude (which is the gridcoin daily cross-project performance indicator derived from RAC) of 171, which then means that 30 gridcoins are assigned to me per day for that machine. At today's exchange rate this would be about 0.1 dollar/day. Hm.
What was kind of surprising is that even though this is only single machine that is contributing to Einstein, this made it among the top 100 as shown on https://www.boincstats.com/stats/5/user/list/12/0/0 . So, this effort makes a difference and I am happy about that. The setup was not too much of effort, and I now think it was well worth it.
What I have not yet done is to allow for multiple tasks for GPU. This is mostly because I consider the system to be mostly CPU-bound, and it is only one PCI lane per card so I don't want too much communication either. But I should give it a shot in a week or two.
I can tell you that Gary Roberts and Cecht have both helped me multiple times when I've had problems here too!! Gary is worth his weight in Gold to Einstein!!!
The heating season started over here, so I got the machine up again. I retired one of the 6 graphics cards, to avoid repeated reboots. It is 24/7 Einstein now, which means FGRPB1G on 5 GPUs only. RACs are still climbing, it is just short of 1 million now, expecting to double over time......
I can tell you that Gary Roberts and Cecht have both helped me multiple times when I've had problems here too!! Gary is worth his weight in Gold to Einstein!!!
Yip. The progress report was meant to indicate where all our investment into this thread went. Many thanks+RAC to them, indeed. I very much hope this thread not to end any time soon. There should be more former mining rigs out there who deserve a second life in the sciences. There may even be an opportunity for me to initiate that transformation on another rig, which I will report here. But if there are more examples out there, please drop a note.
The change away from AMDGPU-PRO was on the 20200317. Performance-wise as in compute times I get about the same today (22nd) as about a month ago. Directly prior to the change, compute times were 5-7% longer, which I accredit to the reboot unless someone knows about changes to the project that conincide with the 17th. So, besides some confidence that no difficulties are to be expected, there is nothing fancy to report, really.
I had a Mining Rig setup for testing on Seti@Home. It peaked at 16 gpus before I had to take it down.
Since it was a new rig it had all the riser bugs we experience which cause app pausing etc.
But it does look like a MSI B360-F Pro with an 8c/16t cpu and 16GB of memory can pull any BOINC gpu project that does not require gen3 and the full 16bit data path reliably.
Tom
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
This thread stimulated a
)
The work on this thread had stimulated a feature request for BOINC and the friendly developers have already implemented it: https://github.com/BOINC/boinc/issues/3210
Nice, thank you all!
And Gary helped me out where
)
And Gary helped me out where I was convinced this was a missing feature: https://github.com/BOINC/boinc/issues/3229
The heating season started
)
The heating season started over here, so I got the machine up again. I retired one of the 6 graphics cards, to avoid repeated reboots. It is 24/7 Einstein now, which means FGRPB1G on 5 GPUs only. RACs are still climbing, it is just short of 1 million now, expecting to double over time.
For technical curiosity, and it seems to be a nice community, and because I likely become super-rich that way, ... er ..., I signed up to gridcoins. That single machine running Einstein@Home only yields a magnitude (which is the gridcoin daily cross-project performance indicator derived from RAC) of 171, which then means that 30 gridcoins are assigned to me per day for that machine. At today's exchange rate this would be about 0.1 dollar/day. Hm.
What was kind of surprising is that even though this is only single machine that is contributing to Einstein, this made it among the top 100 as shown on https://www.boincstats.com/stats/5/user/list/12/0/0 . So, this effort makes a difference and I am happy about that. The setup was not too much of effort, and I now think it was well worth it.
What I have not yet done is to allow for multiple tasks for GPU. This is mostly because I consider the system to be mostly CPU-bound, and it is only one PCI lane per card so I don't want too much communication either. But I should give it a shot in a week or two.
steffen_moeller wrote:The
)
I can tell you that Gary Roberts and Cecht have both helped me multiple times when I've had problems here too!! Gary is worth his weight in Gold to Einstein!!!
mikey wrote:steffen_moeller
)
Yip. The progress report was meant to indicate where all our investment into this thread went. Many thanks+RAC to them, indeed. I very much hope this thread not to end any time soon. There should be more former mining rigs out there who deserve a second life in the sciences. There may even be an opportunity for me to initiate that transformation on another rig, which I will report here. But if there are more examples out there, please drop a note.
I have the machine
)
I have the machine (https://einsteinathome.org/de/host/12784717) now running with the ROCm (https://github.com/RadeonOpenCompute/ROCm) driver from AMD. With Ubuntu 18.04 I could directly tap into their apt repository. It was very easy.
The change away from AMDGPU-PRO was on the 20200317. Performance-wise as in compute times I get about the same today (22nd) as about a month ago. Directly prior to the change, compute times were 5-7% longer, which I accredit to the reboot unless someone knows about changes to the project that conincide with the 17th. So, besides some confidence that no difficulties are to be expected, there is nothing fancy to report, really.
I had a Mining Rig setup for
)
I had a Mining Rig setup for testing on Seti@Home. It peaked at 16 gpus before I had to take it down.
Since it was a new rig it had all the riser bugs we experience which cause app pausing etc.
But it does look like a MSI B360-F Pro with an 8c/16t cpu and 16GB of memory can pull any BOINC gpu project that does not require gen3 and the full 16bit data path reliably.
Tom
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)