BOINC Client 7.8.4 recommended for MAC OS
15 Nov 2017 4:05:46 UTC
Topic 210994
(moderation:
New Recommended Client for MAC OS
Fixes the calculation of GPU memory when running under OS 10.13 High Sierra.
Bill F
In October of 1969 I took an oath to support and defend the Constitution of the United States against all enemies, foreign and domestic;
There was no expiration date.
It's good that 7.8.3 bug has
)
It's good that 7.8.3 bug has been fixed quite quickly. Just be a bit careful in using the description, "New Recommended Client".
On the BOINC website the 7.8.x series are described as test versions which means that bugs are still quite possible. On past behaviour, when the Devs are hopeful that the bugs are mostly eliminated, they tend to talk about "Release Candidates". It can still be a couple more versions after that before a Release Candidate is actually released as the recommended version. I don't think we are anywhere near that stage yet.
If people want to help with the testing of potentially lower quality versions, that's great. If people desiring a set and forget experience don't understand the risks, it could be a bad experience for them. We should just make that clear so that people understand what they're getting into.
Cheers,
Gary.
I based the use of
)
I based the use of "Recommended" from the fact that 7.8.4 is the Recommended version on the Berkeley download version page at
https://boinc.berkeley.edu/download_all.php
Thanks
Bill F
In October of 1969 I took an oath to support and defend the Constitution of the United States against all enemies, foreign and domestic;
There was no expiration date.
Sure. I tend to look at the
)
Sure.
I tend to look at the detail given in the BOINC client forum where they give lists/descriptions of what bugs have been fixed and what else may have changed. The current thread is here. The last entry is for 7.8.3 and there doesn't seem to be a mention of 7.8.4 yet. Each of the entries has specific warnings about the test nature of these "development versions" and a disclaimer to that effect is given. I just wanted people to understand that (particularly with BOINC) the latest may not always be the best.
In this particular case however, for MAC OS users, it seems a good idea to upgrade if you're already on 7.8.3 and I should have made that clearer. I just wanted to warn users in general that if you're not having any problems, the devil you know may be better than the devil you don't :-).
Cheers,
Gary.
This shows the difference in
)
This shows the difference in what my NVIDIA GTX960 is reporting for memory use from 7.6.44 to 7.8.4 on macOS Sierra. On the "Last 7 Days" chart, note the memory was completely saturated on the left before the break, followed by the jagged usage around 2/3 of capacity after that.
If memory use has dropped
)
If memory use has dropped significantly without impacting on output, wouldn't that be a good thing? It's not clear if you think this is good or bad. It's also not clear to me why BOINC (rather than the app which occupies memory) would have such a dramatic effect on memory use. Do you have any theories for that? Did you happen to change anything else (task concurrency?) at the time of the BOINC upgrade?
The only thing that springs to mind for me is the particular tasks running before and after the BOINC upgrade. However it's a bit of a stetch to imagine that the last tasks under 7.6.44 might have been the high frequency end of the previous data file and after the upgrade the tasks were then the low frequency end of a new data file. You would expect memory use to have risen back to 100% by now in that case.
You really need to complete the story by telling us exactly what has happened to task crunch times, if anything.
Cheers,
Gary.
Gary Thank you ... you asked
)
Gary
Thank you ... you asked all the questions I had thought of and some more that I had missed !!!
Bill F
In October of 1969 I took an oath to support and defend the Constitution of the United States against all enemies, foreign and domestic;
There was no expiration date.
Sorry, I assumed less memory
)
Sorry, I assumed less memory usage was a good thing. The BOINC version was the only difference before and after. I upgraded to 7.8.3, and like many here, experienced 100% dump rate and deprecated to 7.6.44 again.
But the upgrade to 7.8.4 brought the memory use down a lot (leaving more for display I guess, and maybe less memory bandwidth saturation). No dramatic effect on my crunch time. It dropped from about 1hr 4 minutes to 1 hour 1 minute for the NVIDIA GTX 960 and from about 43 to 39 minutes for each of the three units crunching concurrently on my R9 280X.
You are right, though, that some GPU work units use more memory than others, but this was clearly an improvement for GPU memory usage, even though it didn't translate into significant speed increases.
But at least it's not dumping all my GPU units in the trash.
I apologize if I am still not answering your question. I am not as familiar with what happens under the hood with these apps as some on these forums, though I am trying to figure it out. Particularly the constraints.
Yes, indeed, it sounds like a
)
Yes, indeed, it sounds like a very good thing with both memory use and crunch time heading in the right direction. I still can't understand why this should be just attributable to BOINC version only. I'm not questioning what you're reporting, just expressing surprise that a BOINC version change did it. Since you're comparing 7.8.4 against 7.6.44, it shouldn't be anything to do with the memory reporting bug in 7.8.3.
Cheers,
Gary.
I thought it was kind of
)
I thought it was kind of weird too, which is why I reported it. I chalked it up to the way BOINC allocated the memory to the GPU apps and left it at that, but didn't realize this would be a surprise.
I recall that memory chart being a solid blue wall across that 7 day chart, which would be a long stretch for a batch of work units causing it. I can't think of any other configuration changes I made that would affect it. I wonder if there was an OS update somewhere along the way that I forgot about that could've changed something too.