GPU Assignment Question

[Nert]

I'm not sure where I should post this question. I hope it's o.k. to put it here.

My question is about how work units are assigned to my GTX 750-TI. Recently, I have noticed that I get "Server did not assign work unit" errors whenever I connect to 171.67.108.52 . I understand from another post in this forum that this server has been changed to the "advanced" client. The behavior that I see when looking in my log is that my client will try 171.67.108.52 some number of times (the number of times seems to vary) before eventually trying some other server where it succeeds in getting a work unit assigned.

I understand that I can change my client type to advanced to get work units from 171.67.108.52. However, the description for client-type = advanced in the software section says that this client preference requests "...late stage beta work units if available." I don't really want to work on beta work units (late stage or not). I know they're important, but I'd rather work on non-beta projects.

The GPU eventually gets work units assigned, but there are delays and the process seems inefficient. I'd rather keep my processors working and not waiting if possible. I'm fine with continuing as is, but thought I'd mention this and possibly get some insight from more knowledgeable folks here. I can post my log if that helps. I was going to post it with this message, but It's too long to meet posting requirements. Please let me know what parts I should post if the log provides needed information.

Thanks.

[Joe_H]

First to answer your last question, the important parts of the log to post are the beginning section that lists the client version information, system info and then the configuration settings. If you change the configuration later on, include the log section that shows the current settings. Then portions of the log that show the errors or particular problem you are posting about should be included. You usually can edit out most of the repetitive messages about completing a frame of the WU's as long as you leave enough to give an indication of processing time for each.

As for the assignment issues, they are related to the model of GPU you are using. It is based on the Maxwell chipset and the drivers from nVidia do not yet properly support all OpenCL functions needed by Core_17. Older Core_15 projects use CUDA and that does work okay on the GTX 750ti. PG made some assignment setting adjustments so Maxwell based cards should be sent to servers with Core_15 projects, I don't know if additional tweaks are possible to get your system assigned more quickly. Changing your setting to Advanced won't help. Eventually either nVidia will release drivers that fully support the OpenCL functions needed on Maxwell based GPU's, or a version of Core_17 may be released that has a work around for the issue.

If you are interested in progress on Core_17, you can read the Beta forum where this is discussed by some beta testers. Otherwise PG will probably announce if a fix is released either here or in the blog.

[jadeshi]

The work server in question is currently running projects on full FAH, not advanced. I made an announcement a while back when I made the change. Apologies if you missed it.

[Nert]

Thanks to both Joe_H and Jadeshi for the quick response. Several days ago I had set the advanced flag in trying to resolve this problem. After a day or so, I deleted it (thinking that would clear the flag). I haven't rebooted my system for quite a while. When I looked at the configuration portion of the log as described by Joe_H it still showed the client type as advanced. After a system reboot, the client type flag is no longer set to advanced and my problem seems to be resolved. I just received a unit from 171.67.108.52. Everything seems good right now

[Nert]

Just a quick follow up on this with even better news. I am now receiving and folding x'17 units on my
750 TI. I don't know if there's been an official announcement or not, but it looks like I've completed two of them and the card is working on the third.