Vista RC1 and Einstein SR51 Ver 424
12 Sep 2006 14:28:42 UTC
Topic 191820
(moderation:
Since going up from Vista 5384 (Beta 2) to Vista 5600 (RC1) I have encountered multiple 'Client Error' it seems when I close my machine down the current data is somehow corrupted amd a new work unit is uploaded at the start of my next session. This has reulted in the following credit losses in a 3 day period:
Client error Done 16,056.11 60.92 ---
Client error Done 15,788.30 61.01 ---
Client error Done 10,618.33 41.02 ---
Client error Done 10,802.37 41.50 ---
Client error Done 15,557.46 59.49 ---
Client error Done 15,542.32 59.79 ---
Client error Done 25,822.64 99.23 ---
Client error Done 26,127.59 100.90 ---
Client error Done 16,135.65 61.54 ---
Client error Done 20,035.32 75.64 ---
It does not seem to happen with SETI using Setiathome enhanced version 515
Does anyone have any ideas or comments.
The Barsteward
Beer is proof that God loves us and wants us to be happy.
--Benjamin Franklin
Vista RC1 and Einstein SR51 Ver 424
)
- exit code 1073807364 (0x40010004)
This is usually due to a graphics crash. This can happen because something else is using the graphics, or alternatively you're displaying the graphics and kill the graphics window via Task manager or answering 'Yes' to Microsoft's 'this program is not responding, kill it?' question.
Thanks Jord but no to all, I
)
Thanks Jord but no to all, I have been lucky so far and havent had any BSOD's in Vista all shutdowns have been requested and controlled. I have just been playing with it and if I manually close down BOINC before I close down the machine the data is saved correctly. It must be something Vista is doing in its closedown routine that causes the problem in SR51 424, as I said previously Setiathome is not affected.
The Barsteward
Beer is proof that God loves us and wants us to be happy.
--Benjamin Franklin
I've had a similar problem
)
I've had a similar problem running BOINC on a new machine. It seems that if I do not specifically 'CLOSE' the manager fefore closing down, it will not continue the current task when it is started up again. This is appalling, since the talks take several days to complete, and one cannot really leave a machine running 24/7. It seems as though the latest software does not tell BOINC to terminate, OR BOINC does not obey the instruction, OR BOINC is incompatiable with Vista, OR who knows what. The older version on my XP machine suffers no such problems !!!
Sorry for the typos, above.
)
Sorry for the typos, above. Another (new) feature is that the CPU is dual core, and is running two tasks at once. Oddly enough, also, this problem only seems to occur if the power is switched off? I'd experiment further were it not for the fact that it is not pleasant to lose 30 hours of work (lol, though BOINC starts of predicting 68+ hours, and seems unable to predict the future time to finish - just as it couldn't 2 years ago, LOL)
IIRC, there was an issue with
)
IIRC, there was an issue with some versions of Vista where it didn't wait for everything to cleanup properly before going into some shutdown modes, IOW the ones which aren't a full shutdown. I think they call it 'fast shutdown' or something like that.
However I was under the impression they had fixed that before Vista had gone 'gold', but perhaps not.
Have you checked to make sure you have all the current updates for your version?
Alinator
RE: ... though BOINC starts
)
Actually, estimates for run time are supplied by each of the projects and not BOINC. BOINC does have a mechanism for correcting those estimates. It's called duration correction factor (DCF) and it works rather well. The shortcoming is that you need to crunch quite a few (perhaps more than 20) results to give time for BOINC to progressively adjust the factor and get it right for your machine. It is actually quite trivial to circumvent this delay by making a manual adjustment to the DCF (which is stored in your state file). This is not generally recommended unless you know exactly what you are doing and take personal responsibility for any screw-ups :).
Cheers,
Gary.
RE: IIRC, there was an
)
Well, I had two fast reboots (both of them while uninstalling/installing MS's SDK). The first killed two Predictor results that had racked up a combined 70 hours. The second one didn't affect anything.
I think it's sort of hit or miss. There's a registry hack posted somewhere over on the BOINC forums that will increase the delay before shutdown.
Kathryn :o)
Einstein@Home Moderator
RE: The shortcoming is that
)
Not that many. 2 to 4 is the general medium these days. Especially since all results at EAH take long. It will get difficult if results change length in between, but since we still get the big WUs sent and the results of that are sliced off, it doesn't matter here.
RE: Not that many. 2 to 4
)
He's using 5.8.16 so I think he will need that many :).
Cheers,
Gary.
RE: RE: Not that many. 2
)
That does not seem to be my experience. My new Vista machine is now running its 13/14th "Hierarchical all-sky pulsar search" tasks. It has always predicted about 70hrs each!!
That is not that about which I 'complain' :-) They actually take about 27hrs (good). However, the column which displays a running prediction of the time to completion is nonsense right up to completion. For example, if the time so far is 13.5 Hrs, and the %done so far is 50%, the time to completion, which should be 13.5 hours, will diaplay some absurd figure such as 35 hours (just off the top of my head, you understand). This was the case 2 years ago, but it has never been rectified. Do I really care? Of course not, but it is irritating, is it not? In reality TimeToDate/PercentageDone would give the total time; so subtract the TimeToDate and get the remaining time????!!!