Before Felipe asks his monthly question ... ;)

Looks like we reprocessed just over 50 of the old 2011 tapes in February, and started on the 2010 backlog - I saw 21 of those. We also did 76 brand new 2015 tapes, making 150 in all - rather fewer than we normally get through. Still, it was a short month, and v8 processing is slower (because more precise) than before.

Recorded	TOTAL		Processed with		Processed with 
				SaH v7/8 (since		Sah v6 only
				launch June 2013)	(derived)

2007		 350		   4			 346
2008		 916		 874			  42
2009		 548		 456			  92
2010		 762		 159			 603
2011		1148		1082			  66
2012		 846		 819			  27
2013		 590		 585			   5
2014		 260		 260			 n/a
2015		 292		 292			 n/a

Grand total	5712		4531			1181

Before Felipe asks his monthly question ... ;)

Hello Richard. Just came to see if you post your monthly update. As there is not much official information, your input is really valuable for us crunchers.


Thank you once again to taking the time to update the data history processed for us.

It actually helps a lot on keeping interest on the project

Filipe

---

Back after the outage, but:

Replica seconds behind master 20,549

...is rising! :O

---

Aloha, Uli

See:

Replica seconds behind master 22,349

:(

---

Aloha, Uli

Panic Mode off......replica now 10,261 behind.

---

"Sour Grapes make a bitter Whine." <(0)>

It was behind after an outrage and went further behind for a bit? :-O

That's just typical of what happens these days after an outrage. ;-)

Cheers.

I've noticed a spike in inconclusives, now around 6.4% (it was around 4%).

Seems to be 2 main modes,
1 My GPUs v x86_64-apple-darwin systems. Mostly against opencl_nvidia_mac & opencl_intel_gpu_sah applications.

2 My CPUs v windows-intelx86 CUDA42
Mixed in both 1&2 are several results where the spikes, pulses etc number match, but they've been decided not to be similar enough to validate.


I notice that a lot of the inconclusives I've got, have come from 2-5 WUs that I've done all being done by the same machine.
So the main cause of the sudden high number of inconclusives appears to be that when people get WUs they come as a group, so you get systems checking multiple WUs against each other. Also when a group of WUs are completed, and then come up as inconclusive they are released for re-processing as a bunch. So when they are re-allocated, they all go to the one system.

I expect once this batch of inconclusives clears, my inconclusive rate will drop back down to around 4%, till the next time a batch goes to a machine.


I notice that the splitters still tend to bunch up on particular files.
If each splitter worked on just the 1 file, this clumping of WUs would be less of an issue and the allocation of WUs would tend to be more random; there would be less of 2 machine comparing multiple WUs against each other.

---

Grant
Darwin NT

Both those OS X apps have a tendency to spit out inconclusives. There is a beta OS X CUDA app (http://www.arkayn.us/forum/index.php?topic=191.msg4411;topicseen#new) that fixes a lot of that and is twice as fast. Maybe some of your wingmen will see this message...

The most of the Intel iGPU's just barely work on OS X for this. Some work fine and others just spew inconclusives. The same chip can be fine on one machine and not on another.

Chris

---

The most of the Intel iGPU's just barely work on OS X for this. Some work fine and others just spew inconclusives. The same chip can be fine on one machine and not on another.

I don't think they're much better on Windows machines.
It appears that Intel really need to sort out their drivers; they've given them OpenCL support but it would appear they haven't done much work in actually checking that their output is even remotely accurate.

---

Grant
Darwin NT

Question: since last evening sometime, I noticed a gradual decline in the number of tasks on my 2 machines, which normally have 600 WUs on them at any time. It's not like they have been totally shut out - they have been getting WUs in dribs and drabs, just not enough to keep stable or fill up.

Any idea why this might be happening? Is anyone else noticing similar? All I did was switch from stock to Lunatics 44, but that shouldn't affect requests for new work, should it...or does it give me fewer until the new apps are figured out by the servers on my machines?

Thanks for any advice...

EDIT: I just noticed that at least some of the new v8s d/l for GPU have about 1/10 the GFLOPs in the estimate that they had before. Huh????

---

or does it give me fewer until the new apps are figured out by the servers on my machines?

IIRC until the server sees how long the new apps run it vastly over estimates the run times, hence you will get fewer work units.

---

It will take 10 completed tasks to average and then things will go back to the way they were. Just have to wait for that 11th one.

---

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours

Now that was an Uncalled for Pre-Outrage Panic...

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Now that was an Uncalled for Pre-Outrage Panic...

As an added bonus Einstein also was down at 7 am PDT.

---

Has the resend lost tasks feature been turned off? I have lost one WU while trying to get stock apps to run on NativeBOINC and the servers didn't send it back to me so far...

---

Has the resend lost tasks feature been turned off?

From what I can recall it's been disabled here at Seti@home for some time.
As for other projects, no idea.

---

Grant
Darwin NT

OK, than I know... thanks.

---

Did you notice that the Astropulse science database on marvin is disabled?

9 days without any panic, but now that AP's are being split again, I think it would be a good idea to start the AP assimilators. They've been in state "Not running" for a day or so now. The number of Workunits waiting for assimilation for AP is growing and growing, and with new AP's being split and crunched again, if the AP assimilators aren't running, something will break, sooner or later.

---

Donald
Infernal Optimist / Submariner, retired

And many errors in the splitting process, probably for the very same reason...

Cheers.

Did you notice that the Astropulse science database on marvin is disabled?

No, I forgot to open my eyes that wide :-)
Geeze, help me, I'm blind.
Stoopid me....

Damn, all those AP's NOT being produced, my poor gpu is starving ;-)

Back to Einstein methinks, for now.

P.

---