Message boards :
Number crunching :
ETA on new non-beta WUs?
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 5 Mar 14 Posts: 16 Credit: 16,903,909 RAC: 0 Level ![]() Scientific publications ![]() ![]()
|
Sorry if this was addressed prior, but I recently started crunching again and was wondering if there is a mention of when we can expect non-beta WUs to return. The beta ones unfortunately failed for me when I tried it a month ago. Thanks |
|
Send message Joined: 13 Dec 17 Posts: 1419 Credit: 9,119,446,190 RAC: 891 Level ![]() Scientific publications ![]() ![]() ![]() ![]()
|
Unknown. Depends on when or whenever they get a new grad student to start a new research sub-project. There are some very occasional acemd3 tasks slipped every few months or so. But never any sustained long runs like years past. |
|
Send message Joined: 5 Mar 14 Posts: 16 Credit: 16,903,909 RAC: 0 Level ![]() Scientific publications ![]() ![]()
|
Ah damn that's a shame. I loved crunching for this project the most too. |
|
Send message Joined: 13 Dec 17 Posts: 1419 Credit: 9,119,446,190 RAC: 891 Level ![]() Scientific publications ![]() ![]() ![]() ![]()
|
Windows hosts have a very difficult time running the ATMbeta tasks it seems. The Linux hosts don't have much or many issues on the other hand. It you read back in the beginning of the main ATM thread, the Windows hosts mainly had issues with Windows inability to release all the requested memory allocations for the tasks and had "not enough memory" errors. Could be resolved by increasing the pagefile size in Windows to around 50-60GB |
|
Send message Joined: 28 Mar 09 Posts: 490 Credit: 11,731,645,728 RAC: 69 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Windows hosts have a very difficult time running the ATMbeta tasks it seems. That's not always the case. Back in the spring, before switching to Linux, I was getting success rate of 90% plus, with Windows 10, which is about the same as Linux. I was and still am running the units two at a time. If one errors out, typically with the NAN error, the other finishes successfully. I don't remember my exact pagefile space size, but it was a pretty large number. Though, Linux does run faster. If they can fix all the bugs, that were discussed ad nauseam, this app would be great. |
|
Send message Joined: 13 Dec 17 Posts: 1419 Credit: 9,119,446,190 RAC: 891 Level ![]() Scientific publications ![]() ![]() ![]() ![]()
|
I was just posting the observation based on the number of forum complaints about the tasks not working on Windows hosts. Many more complaints from Windows users compared to Linux users. Though expected I guess since there are more Windows hosts attached than Linux hosts by a large factor. But a simple look at the top hosts lists in stats shows almost all Linux hosts on the first page. https://www.gpugrid.net/top_hosts.php |
|
Send message Joined: 28 Mar 09 Posts: 490 Credit: 11,731,645,728 RAC: 69 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I was just posting the observation based on the number of forum complaints about the tasks not working on Windows hosts. Good point, I'm just pointing out that there are exceptions to every rule. |
©2025 Universitat Pompeu Fabra