Posts by HAL9000


log in
21) Message boards : Number crunching : Top GPU models list (Message 1632643)
Posted 5 days ago by Profile HAL9000
http://setiathome.berkeley.edu/forum_thread.php?id=75889 in the Questions and Answers forum may be of interest.
Joe

Indeed it is.
I do wonder if providing of list with GFLOPs might be more useful. Somewhat like the CPU list.
22) Message boards : Number crunching : Lost Credit and Tasks? (Message 1632580)
Posted 5 days ago by Profile HAL9000
Thanks guys, I noticed my client had 'Use GPU Never' - which I must have selected recently by accident.

Allowed GPU usage again.

Looking at the App details for your host. It looks like you have not done any SETI@home v7 or AstroPulse v7 on your GPU. AstroPulse v7 was added in October. So your RAC may have been falling for quite some time with your GPU set not to be used.
23) Message boards : Number crunching : Question about AP vs MB Preferences (Message 1632576)
Posted 5 days ago by Profile HAL9000
With the changes to the feeder only sending out a maximum of 7 AP tasks the feeder may have AP work to send out much more frequently. However with AP + other set you will get MB work when the feeder has 0 AP work.
I am ok with going several hours or a few days with no new AP. Even with a queue of 100 AP my fastest GPU takes about 2 days to do them all.

For one of my machines that is AP only & rarely ever has to get switched to my AP+MB venue. Most of the requests of AP come back with 0.
Total number of work requests: 1000 Number of requests gaining work: 132 Number of requests not gaining work: 868 Tasks per request percent and count: Total 0 tasks: 86 %, 868 1 tasks: 9 %, 98 2 tasks: 2 %, 24 3 tasks: 0 %, 8 4 tasks: 0 %, 2 5 tasks: 0 %, 0 6 tasks: 0 %, 0 7 tasks: 0 %, 0 8 tasks: 0 %, 0 9 tasks: 0 %, 0
24) Message boards : Number crunching : Lost Credit and Tasks? (Message 1632572)
Posted 5 days ago by Profile HAL9000
With the current 123057 active users the 5% badge would be granted to those from rank 6152 to 1231. Looking at the Top participants for that area you have just fallen below 6152.
Looking at the external stat site Free-DC your daily output in the past week looks to have dropped a bit. Probably when we were out of AP work recently. Which would reduce your RAC compared to others that say do MB work only.
25) Message boards : Number crunching : Panic Mode On (94) Server Problems? (Message 1632470)
Posted 6 days ago by Profile HAL9000
They really should turn on the ap assimilator v6 too. I still have 6 AP v6 tasks that needs to be assimilated.

I'm sure there are still thousands of AP v6 tasks in need of assimilation.

Edit, added:

Not good:

5471 SETI@home 2015-01-24 16:21:58 Sending scheduler request: To fetch work.
5472 SETI@home 2015-01-24 16:21:58 Requesting new tasks for GPU
5473 SETI@home 2015-01-24 16:22:41 Scheduler request failed: HTTP internal server error

I have 26 APv6 that are ones stuck in limbo & will require the clean up script to be set free.
26) Message boards : Number crunching : Seti@home 7.7.03 OpenCL Intel sah problem (Message 1632468)
Posted 6 days ago by Profile HAL9000
Your Intel GPU driver version of 10.18.10.3939 could also be an issue. It has been found with a few drivers so far that they do not work well with SETI@home applications, or a few other projects as well.
10.18.10.3621 and 10.18.10.4061 are known good drivers at the moment. Recently some discussion about drivers was had here.
27) Message boards : Number crunching : Question about AP vs MB Preferences (Message 1632466)
Posted 6 days ago by Profile HAL9000
Right now, I have my SETI preferences set to AP and MB. I get relatively few APs, of course. From reading the Fora, I get the distinct impression that those who have their preferences set to AP only get a lot more APs than I do.

If I changed my prefs to AP only, no MB, BUT also said "send me anything you got if you don't have the apps (AP) I checked available"

1) would I get APs preferentially (to MB) when both are available?
2) would I still get MBs as before when no APs are available?
3) would I net more APs that way?

Thanks for any info!

You will not really get any more AP than you do now if you do AP + other apps than you would with AP + MB enabled.
If you want to mostly do AP then you would set only AP & manually toggle MB when required. I have one venue set for MB only, one for AP only & one for MB + AP. If I see an AP only machine get down to only a few tasks remaining I switch it over to the MB + AP venue until more become available.

Given that my machines are not super speed demons I rarely have to do anything.
28) Message boards : Number crunching : Top GPU models list (Message 1632385)
Posted 6 days ago by Profile HAL9000
A bit worrying to see the GTX 980 coming in at number 14

Perhaps it takes a while to build up in the ranks?
Also the performance most likely doesn't take into account the number of tasks being run at once. I would imagine those high end cards are running 3+ tasks at once.
29) Message boards : Number crunching : Panic Mode On (94) Server Problems? (Message 1632360)
Posted 6 days ago by Profile HAL9000
Oh, oh, oh. Something is struggling now. Not good, not good at all...

I'm hoping that the purging of the AP workunits that were assimilated yesterday has started.

Edit: My total AP valid has just dropped about 150. So looks like purging is now occurring!

Well, I can't even get to my task pages. Everything task related is not loading for the moment.

Luckily I had mine open & refreshing it only took 4 or 5 min.
30) Message boards : Number crunching : Panic Mode On (94) Server Problems? (Message 1632351)
Posted 6 days ago by Profile HAL9000
Oh, oh, oh. Something is struggling now. Not good, not good at all...

I'm hoping that the purging of the AP workunits that were assimilated yesterday has started.

Edit: My total AP valid has just dropped about 150. So looks like purging is now occurring!
31) Message boards : Number crunching : Top GPU models list (Message 1632342)
Posted 6 days ago by Profile HAL9000
I was checking out the Top GPU models list today & noticed something strange.
The unique name I modified BOINC to display of "ATI Radeon HD 6870 (Barts XT)" for my hosts 5837483 & 7421701 is displayed. Rather than the BOINC default of "AMD Radeon HD 6790/6850/6870 series (Barts)". I had thought this list was generated only using data from the stock apps, but it looks to be including data from Anonymous platform apps as well. Which makes it much more useful.
From what I can tell there may be a requirement of two hosts with the same GPU to have it listed. Like the CPU list requiring at least 10 hosts for the CPU to be listed.
32) Message boards : Number crunching : Move Boinc & Seti to new partition.... (Message 1632258)
Posted 6 days ago by Profile HAL9000
I can guess at the answer, but will ask the question anyway.

Is there any way to move Boinc to another drive/partition without having to finish all tasks, report them, un-install & re-install specifying the new location?

I ask as I recently upgraded my C drive from a HDD to a SDD. I realise it should take many decades for the read/writes from Boinc to degrade my SSD, but as Seti/Boinc do not need the access speeds a SSD gives It would make a little bit of sense to shift them to disk drive and off the solid state drive, and reduce the wear of the SSD as much as possible.

Thoughts, comments and criticisms welcome.

Don't know about Windows, but in Linux I can attach whatever partition I want to to the path of the boinc folder. As long as you move the boinc folder as a unit and change the commandline options of boinc.exe to use the new location as the home for it, you should be able to put it wherever you want. I've even transfered a boinc folder to another cruncher and had him finish some tasks for me before.

Boinc uses relative path and not absolute.

Another option windows provides is with the mklink command. Much link the directory and file linking in *nix systems. So you could move BOINC where you like & then just make a link back to the original path to keep everything everything else happy.
Using something along the lines of:
mklink /D C:\Program Files\BOINC D:\BOINC
mklink /D C:\ProgramData\BOINC D:\BOINCData

Remote paths can also be uses if you wanted to keep your BOINC stuff on a remote machine or network storage device.
33) Message boards : Number crunching : AP V7 (Message 1632232)
Posted 6 days ago by Profile HAL9000
That longer running task finished at 1.465 million seconds (16 days 23 hours, or 407hours), Credit received was over 600 at 606.31,

Claggy

The 2-3 days for the SSE versions blast right along in comparison.
34) Message boards : Number crunching : Win 10 will be a FREE Upgrade to Win 7 & 8 Users!!! (Message 1632231)
Posted 6 days ago by Profile HAL9000
For those that are curious like me, I found that what Microsoft is calling OS as a Service simply means they will push all updates to your computer automatically. No more Service Packs. Possibly no more Patch Tuesday. When an update is ready, it goes out.

-[ snip ]-

Greetings OzzFan,

Are you saying that we will have NO control over updates as we do now? I have Win7 inform me that there are updates and that I will deal with them at my leisure, meaning I don't want the update system rebooting my PC "automagically" without my consent or me having a hand in it.

The update system does not know my password to finish the boot process. At 1am, my PC would be sitting here waiting for it. BOINC would not be running until I get up and realize the M$ sent an update and tried rebooting my PC. BOINC would miss out on hours of crunch time.

I certainly hope that M$ is not planning on taking FULL control of the updates and allows me to continue as I have for the past couple decades.

Keep on BOINCing...! :)

MS specifically stated that updated could be be delayed for those that needed or wished to run longer stability situations. They also stated that they want to have things like security patches & fixes pushed out as soon as they are available. So Patch Tuesday may in fact be going away.
I imagine the auto update service will still be able to be disabled as it can be now. At least in the preview releases of Windows 10 you still can. However, I haven't had a change to install the newest version released this week.
35) Message boards : Number crunching : Win 10 will be a FREE Upgrade to Win 7 & 8 Users!!! (Message 1631953)
Posted 7 days ago by Profile HAL9000
I would think your old LaserJet drivers would still work (I have a 4000N, 8100DN, and DesignJet 600) that are all working in 7-64. The two LaserJets were installed through finding the drivers via Windows Updates, but the DesignJet was tricky. I had to track down the 64-bit win2k drivers for that one.

64-bit drivers for really old hardware are basically universal, though they won't be signed and "trusted" by newer OSes, but they will still work just fine. Regarding the fact that the 4000 and the 8100 both got the drivers via Windows Update, I would have to assume they'll still be there and available for Win 10, as well.

(all three of my printers have JetDirect cards in them, so they are all network printers, and installing them is not specifically straightforward like plugging a USB printer in, or running the 800MB bloatware software suite for modern printers and selecting how the printer is connected.

To manually add a printer: control panel > devices and printers > Add Printer Wizard > add local printer > create a new port > type of port > TCP/IP > enter IP address, and then it should be pretty self-explanatory from that point. I should also note that my printers do use DHCP, but have reserved addresses. If you can't reserve a DHCP address for them, you should configure them to a static address, otherwise they might stop working because their IP changed.)

A driver that has been signed by MS for an older OS will have a compatible signature for all newer OS. Until they removed the ability to do so. We were only submitting our Wind 7/8 drivers for Windows XP digital signatures. Now we just submit for Windows 7 & there are check boxes to include the older OS such as XP & Vista. If a driver has only been signed for one platform you can get the unsigned driver message in the other platform. Such as an x86 signed driver in an x64 OS.

We haven't had any issues with our printer drivers in Windows 10, but they haven't really changed anything in the printing subsystem since some new things added in Windows 8.x. XP, Vista, & 7 drivers should continue to work with any issues. Miscellaneous software, from those 800MB bloatware packages, on the other hand may need an update. I do know the scanner software we provide with our MFPs does not care for Windows 8 or 10. So the overlords at the head office have finally decided to replace it.
36) Message boards : Number crunching : Panic Mode On (94) Server Problems? (Message 1631817)
Posted 7 days ago by Profile HAL9000
...
The Science is compromised by all those crazy computers, filling the database with invalid results, and nobody from the project seems to care one iota about that.
...

Don't worry the instances of bad GPU results validating against themselves to cause a good CPU tasks to be flagged as invalid happens with AP as well.
http://setiathome.berkeley.edu/workunit.php?wuid=1673423338
http://setiathome.berkeley.edu/workunit.php?wuid=1659064648

I wouldn't call those GPU results bad, IMO the WUs simply had signals so close to critical threholding levels that quirks of the AP validator caused the CPU result invalidation.

For wuid=1673423338, the single pulse found by both GPU and CPU processing is well below the 1.01*threshold level so the Validator ignored it. The Rep. pulse found only by the GPUs is probably right at threshold, but I've downloaded the WU and will check that.

For wuid=1659064648, the single pulse found by both CPU and GPU is very close to the 1.01*threshold level (87.85995151 for scale 5). The CPU app was made before we added enough digits to the peak_power field in stderr to be sure its reported peak_power=87.86 is above or below that critical level, but the GPU peak_power=87.8589 is just below so the Validator ignored that GPU signal. The fact that it declared the CPU result invalid indicates its peak_power was actually above the critical level so the Validator was comparing 1 signal from the CPU result to 0 signals from the GPUs.

A proposed change to the AP Validator would have given the wuid=1673423338 CPU result credit on a "weakly similar" basis (and the accompanying "valid" indication). For wuid=1659064648 the initial validation would have come out "strongly similar" and the CPU result would have become canonical. I hope the server issues will subside enough so that change can move up the todo list.
Joe

A threshold issue sound much better than one of the alternatives.
I'm sure they want the server issues to be resolved more than we do.
37) Message boards : Number crunching : ATI 7950 and 7870 not getting GPU WU's (Message 1631815)
Posted 7 days ago by Profile HAL9000
The only thing I really see wrong with your app_config.xml is that there is no astropulse_v7.1 application. You wouldn't really need an entry for the MB CPU 7.01 app either, but I would not expect any app_config.xml entries to effect downloaded work. Incorrect entries would just be ignored I would image, or more so that you should bee seeing an error in the start up of BOINC.

In the Event Log are the work fetch requests displaying "Requesting new tasks for CPU & ATI" or just "Requesting new tasks for CPU"?
If the machine is busy working on Einstein GPU work it may be requesting SETI GPU work.
38) Message boards : Number crunching : ATI 7950 and 7870 not getting GPU WU's (Message 1631745)
Posted 7 days ago by Profile HAL9000
I just converted a machine with an ATI HD7950 and HD7870 GPUs over to BOINC but am not receiving any GPU work units. I checked and both are DP-capable so that's not it. I've checked my Seti computing prefs and everything looks fine. The machine in question is M6400T.

The machine is happily crunching E@H GPU WU's so it's not a question of BOINC not seeing the GPUs. I've tried a re-set but it didn't make a difference.

All of my other machines are NVIDIA and are doing fine.

Thanks for any help you can provide.

Regards,
Steve

SETI@home only uses SP so DP isn't ever an issue an issue here.
Have you checked your SETI@home preferences and made sure Use ATI GPU is set to Yes for the venue the machine is in?

EDIT:
Looking at the errors on your host 7481448 it does show that you did at one point have 2 GPU tasks on the 19th, but they were Abandoned. Most likely when you reset the project.
39) Message boards : Number crunching : Panic Mode On (94) Server Problems? (Message 1631741)
Posted 7 days ago by Profile HAL9000
And export stats hasn't been updated since yesterday.

There are several hours before today's daily stats dump is generated.
http://setiathome.berkeley.edu/stats/
The timestamp is UCB local time or UTC -8.
40) Message boards : Number crunching : Panic Mode On (94) Server Problems? (Message 1631659)
Posted 7 days ago by Profile HAL9000
Don't worry the instances of bad GPU results validating against themselves to cause a good CPU tasks to be flagged as invalid happens with AP as well.
http://setiathome.berkeley.edu/workunit.php?wuid=1673423338
http://setiathome.berkeley.edu/workunit.php?wuid=1659064648

Well, I just checked the Validation inconclusives on my Test Mac and All the recent inconclusives are against nVidia Driver 347.09, http://setiathome.berkeley.edu/results.php?hostid=6796479&offset=0&show_names=0&state=3&appid=11

If these Bad Results start validating against my test App I will Not be a Happy Tester.

Those two results are with 4 different drivers.
[2] NVIDIA GeForce GTX 980 (4095MB) driver: 347.25
NVIDIA GeForce GTX 750 Ti (2048MB) driver: 344.75 OpenCL: 1.1
[2] NVIDIA GeForce GTX 770 (2048MB) driver: 331.82
NVIDIA GeForce GTX 750 Ti (2048MB) driver: 347.09 OpenCL: 1.1

They all do use 2721 app, but given I have 2 of these vs 11586 valid the percentage is a mere 0.000172622%.


Previous 20 · Next 20

Copyright © 2015 University of California