We built this new machine in our lab with 6 Radeon7 GPUs. It has been running for a few weeks now, and the machine runs out of tasks everyday reporting that "reached daily quota of 1856 tasks".
https://einsteinathome.org/host/12784895
Is there some way that daily quota can be increased? Or will that happen automatically in time? Thanks!
Cheers, Gaurav
Copyright © 2024 Einstein@Home. All rights reserved.
I see that Boinc reports
)
I see that Boinc reports currently 20 processors on that host. You could edit cc_config.xml and change "ncpus" value to something artificial like 128. That would increase the daily quota.
Richie wrote:I see that Boinc
)
In addition to the above you might need to edit "client_state.h" and change 1000 to something like 10000
I posted a walkthrough to do this for 18.04 over at boinc. They are offline at the present but you can find it where I complain about "where is" 7.15.1
Faking the number of your VII to something like 48 (instead of just 6) was harder to do (for me) as I did not want to join the "seti gpu users group ***" but you could join that group and get access to their "7.15.1" version that will fake the holy s**t out of Einstein's download limit.
[edit] *** Would have to give up my GRCPOOL membership but I have been reading where the gridcoin gods are going to allow mining even if not a member of their "club".
JStateson wrote: #define
)
This limit stops new work requests when the number of tasks already in queue is above the stated threshold.
So it does not need revision to fix the problem raised by the original poster. Only if that person wants a larger queue is it relevant.
For my single Radeon VII system, 1000 GRP tasks is about two days of work, so if the machine in question runs at similar efficiency, 1000 tasks might represent about seven hours of work. Possibly a bigger buffer than that might be desired, although Einstein outages much longer than that are uncommon.
JStateson wrote: [edit] ***
)
Yes that's true but probably won't go live for everyone until January or so depending on how the testing goes. I think it also requires the next version of the wallet software, the current one being 4.0.5.? and the one that will allow that in bulk is the 4.0.6.? version.
The Gridcoin developers have
)
The Gridcoin developers have been actively begging for more teams to participate in the beta testing. I don't think they have had any luck even after posting solicitations all over every project that I participate in.
The open team release is called "Fern" and follows the partial team whitelist "Elizabeth" release I am running. I was one of the testers for the gridpool team requirement relaxation. I never participated before because I didn't want to leave my current team. I haven't seen any schedule about when the next release is going to drop. The developers just say it will be available when it is ready.
Quote:I posted a walkthrough
)
You can always access the BOINC sources at Github.
https://github.com/BOINC/boinc
Quote:Faking the number of
)
You might want to search around the forums for ATI/AMD card users that have had issues with multiple OpenCL drivers installed in a host and the issue that causes of too many physical gpus reported by BOINC as another avenue of spoofing gpu count. That does not involve compiling your own client from source and only needs manipulation of default BOINC files.
Gaurav Khanna wrote:Is there
)
You can cheat by telling BOINC you have lots more CPU cores than you actually have. Here is a thread that gives the details. There doesn't seem to be a problem with making an outrageous claim for the number of CPUs so, if necessary, you can do a lot more than just double the number as was suggested in that example, if doubling is not enough.
Cheers,
Gary.
Don't know about cpus and if
)
Don't know about cpus and if they have a hard limit in BOINC. Would have to crawl through the code and see if it is defined in a header file.
There is a hard limit on the number of gpus supported and that resides in the server code. That is why you haven't seen any spoofed client with more than 64 gpus reported.
This approach appears to have
)
Adjusting the "ncpus" to 128 appears to have done the trick! Thanks.