Yes, thanks saw that. The heart of the question was, which one is showing on the RS pages or is it something in the middle? If you post a copy of the RS records, will have matching set of local records, maybe your BOINCtasks history if you run that] and what shows on the RS.

TIA

--//--

Yes indeed... Fred's the man in the BOINC single and multi-client management department.

But to the point, reading that, the Result Status page on the website shows 0.02 hours for the top record where 9 minutes 44 sec of Elapsed was incurred (0.16 hours). Let me chew on that for a little [the last of the 5 looks dandy though].

--//--

The object of introducing Elapsed was that the GPU has/(had?) no time concept. Correct me if I'm wrong, but would you not want to get the GPU crunch time recorded, be it when needing many CPU cycles or little CPU cycles to support the GPU task. More graphic... if the HCC task needs 120 seconds worth of CPU cycles, but 60 minutes processing cycles on the GPU, I'd think that 60 minutes needs including, and that is what's in the 9 minutes 44 seconds elapsed of the first record

Except for the bottom one (sorry I meant record 6, not 5) the others do not jive with my concept of what would be proper, and to understand where I'm coming from, this is what I'm looking at - 00:09:44 (00:01:18) . What with the 8:26 minutes gap? The accounting of that the part is what I'm after.

Elapsed time I've never seen going down btw, lest a job is unloaded and restarted. Elapsed is the all encompassing time the job was allowed to run.

What I've starting to gather [actually always understood] is, that at other grids no one is interested in the time keeping history, just the balloon credits per hour because they run 10 times faster [Not to be wrongly taken, 10x faster rightly gives 10x credit for same work unit]. At WCG the time recording factor is of importance... we're unique in that in BOINC world.

Hope I'm clearer now.

--//--

Time to get less unique on that point (but stay unique in reliability and valuable projects).

IMHO, it's ridiculous how important run time at WCG is appraised (look at the badges, based on run time, not on credits).
People are encouraged to keep their old PII's and PIII's alive, sponsoring the utility companies.

The only thing that should count is that the work should done fast with low costs.

To our topic: work done in elapsed time that counts.

IMO GPU-time is not important, because even when it's using only 30% of its capacity or idling like during the start and finish periods of these beta's, the GPU will not do another task.
The only what can be done is to make the GPU more efficient or perhaps by running 2 tasks concurrently on the GPU, if capable.
Supporting so many WCG-projects, you need a bit complex app_info.xml for that.

I'll do another test (if the beta's come again), with all threads busy, but the beta task priority set to normal.

It seems a good idea that there always be announcements (in the forum) for any and all (planned or unplanned) releases of beta_GPU-WUs. I'm not aware that there was one such for 'round 4'.
;