Where are all the Core17 units?

[bollix47]

Click on the Official stats link at the top of a forum page. Scroll down a bit until you see "Client statistics" and click on it.

[PantherX]

I have heard that AMD GPUs are pretty good for mining so it is possible that some AMD Donors are opting to perform mining on their GPUs for sometime. Alternatively, it could be a server glitch or something.

[Grandpa_01]

For Amd on the other hand, was sure there was roughly 1100 active Amd's last week, but now it's dropped to only 326.

Can I ask where you are getting that stat? If true (and I'm not doubting you) that is incredibly low.

You can get the info at the link below
http://fah-web.stanford.edu/cgi-bin/mai ... e=osstats2

[Joe_H]

There is also a Google spreadsheet doc set up by risto that has is updated daily with the information from the official stats Client Statistics. You can find that here - https://docs.google.com/spreadsheet/ccc ... sp=sharing.

[Rattledagger]

There is also a Google spreadsheet doc set up by risto that has is updated daily with the information from the official stats Client Statistics. You can find that here - https://docs.google.com/spreadsheet/ccc ... sp=sharing.

Thanks for the link, going by this Amd has dropped from roughly 5000 to 300 the last 2 months. It looks like the drop started immediately after the announcement of Core17 being promoted to full release. While the stats-page did earlier include for how long a GPU was counted active, now only the 50-days for CPU's is listed so no idea if the AFAIK 2-days for GPU is still in effect or not...

Since the fall seems to be due to Core17, a couple possibilities are:
1: A mis-configured server, forgetting to include results from Amd's. Since the server isn't including #wu's available in server-status it's clearly mis-configured, but maybe it's got more than one wrong configuration.
2: Core17 doesn't work, possible reasons is too old GPU, too old FAH-client, too old drivers, users haven't installed the neccessary OpenCL-part of driver, users runs the oudated OS WinXP that AFAIK doesn't include OpenCL in Amd-drivers any longer.
3: Core17 reportedly is even worse than previous cores on Ati-5xxx-cards, leading users to stop wasting energy on running FAH on such cards.
4: Users found a better usage of their GPU's. I'm not following bitcoin and similar happenings so maybe something was released start of November?

BTW, talking about mis-configured server, "Client statistics by OS" currently shows: "Table last updated at Tue, 31 Dec 2013 19:18:44" and this doesn't make any sence at all, since it's clearly not UTC since this is 2 hours into the future and if FAH haven't suddenly been kicked-out from Stanford and re-located to Russia or something it doesn't fit with server-time either.

[PantherX]

...1: A mis-configured server, forgetting to include results from Amd's. Since the server isn't including #wu's available in server-status it's clearly mis-configured, but maybe it's got more than one wrong configuration....

Possible. It would need someone to check out the Server and the Stats System to see if something is wrong.

...2: Core17 doesn't work, possible reasons is too old GPU, too old FAH-client, too old drivers, users haven't installed the neccessary OpenCL-part of driver, users runs the oudated OS WinXP that AFAIK doesn't include OpenCL in Amd-drivers any longer...

FahCore_17 supports AMD HD 5000 Series and higher so it's a huge range of GPUs. There were occasional reports where we had to inform them that Windows XP is no longer supported by AMD, hence, we too couldn't support it. Windows Vista or higher is required for FahCore_17.

...3: Core17 reportedly is even worse than previous cores on Ati-5xxx-cards, leading users to stop wasting energy on running FAH on such cards...

There were only a handful of reports where the lower-end GPUs from AMD have an issue folding with FahCore_17. In some cases, updating to the latest WHQL Drivers solved the issue and in other cases, it didn't. As of yet, I don't know what the root cause is.

[bcavnaugh]

Anyone know how to fix this?
15:53:28:WU03:FS01:Connecting to assign-GPU.stanford.edu:80
15:53:29:WARNING:WU03:FS01:Failed to get assignment from 'assign-GPU.stanford.edu:80': Failed to connect to assign-GPU.stanford.edu:80: No connection could be made because the target machine actively refused it.
15:53:29:WU03:FS01:Connecting to assign-GPU.stanford.edu:8080
15:53:31:WARNING:WU03:FS01:Failed to get assignment from 'assign-GPU.stanford.edu:8080': Failed to connect to assign-GPU.stanford.edu:8080: No connection could be made because the target machine actively refused it.
15:53:31:ERROR:WU03:FS01:Exception: Could not get an assignment

See Post: viewtopic.php?f=18&t=25495

[7im]

Yes, be patient. PG has already been contacted to fix it.

[bruce]

Fixed.

If you want more details about what happened, see viewtopic.php?f=18&t=25495

[beer]

Since the PG has fixed the issue with getting WU then I am resiveing core 17 again (after around 2 weeks of only resiveing core 15). The best news are that core 17 does not req a CPU core anymore since its uses around 1-2 % CPU on my system (at least this morning when I left home). And I did not do anything special

[PantherX]

You might be the first to report that FahCore_17 isn't using a full CPU while folding on Nvidia GPU. AFAIK, no new version of FahCore_17 has been released.

[beer]

normally core 17 requires a core when folding on my 660 ti so I was surpriced when it did not this morning. It might be a standalone case. The few core 17 I have seen on quadro 2000 has never used a core. But as far as I know then they use another kind of driver than the geforce serie so that was no suprice to me. Should I report back if I see any changes in this behaviour?

[DutchForce]

Which NVidia driver are you using?
If you're using the latest (331.xx serie) driver on your GTX660 Ti, then it could have a low CPU load on Core 17 WUs. There is an issue with OpenCL and Kepler cards using GK-104 chips, GTX770 and lower cards (incl. all 600 series cards) and will cut your folding performance nearly in half. So it's best to go back to 327.23 WHQL or lower drivers for now, untill NVidia will fix this issue hopefully soon. The only Kepler cards not affected by this OpenCL issue are the GTX780, GTX780 Ti and Titan (GK-110 chips).

[heikosch]

Core 17 uses up to 1 CPU core.
But under high CPU load (i.e. F@H SMP or BOINC) I observe an average CPU usage of only 25-50% (of a core).
Base priority of the FahCore_17.exe is set to "higher than normal" using the SetPriority utility program.

GTX 680 and GTX 780 with Geforce 327.23, combined 250k PPD.

Heiko

[beer]

I am using driver 9.18.13.2049.
I cannot say of it is patten or not since it was one unit. I am back on core 15 (project 8018) again (which realy hurt my PPD)