Posts by FalconFly

1) Message boards : Number crunching : SETI@Home Wow!-Event 2015 (Message 1707304)
Posted 1 Aug 2015 by Profile FalconFly
Post:
Another year - another Challenge :)
2) Message boards : Number crunching : Odd re-occuring issue with Android device (Message 1566394)
Posted 3 Sep 2014 by Profile FalconFly
Post:
No, BOINC is uninstalled from the device and the log did not reveal anything useful anyway (except giving the reason for the failure : "network activity suspended - user request").

The network suspension always occured some 2 hours after starting computation with no other activity and the device not even being touched.

I do believe it is possible that the device temporarily lost WLAN connection (standard power-saving feature) and thus entered into network suspend mode. However, even when allowed to transmit data via any means (UMTS), this condition never changed.

IMHO has to do with some form of long-standing issue/bug of Android BOINC handling its preferences and reliably detecting device status changes, since it has been with me for so long. One of the earliest issues of that kind I ever observed was that it i.e. detected temperatures being over the threshold but never detected them sinking back again to resume computation).
But all in all that's only my best educated guess.
3) Message boards : Number crunching : Odd re-occuring issue with Android device (Message 1566270)
Posted 2 Sep 2014 by Profile FalconFly
Post:
I ran the 7.4.14 Version, which didn't change anything after its upgrade from 7.3.12 two days ago...

I also observed that the tasks were running normal, so computation was running just fine.
It really only seemed to be the network activity that BOINC somehow switched to "suspended at user request" without me doing anything.

I also did a complete uninstall/reinstall, but the problem returned within a few hours. Resetting the project also didn't help, after initially running fine it as well ran into the very same problem. Rebooting the whole device multiple times also didn't do anything.
(all occured while the device was at the charger for the whole time and temperatures running a single task were well in the green)

While I do recall having similar issues in the past, I could always solve them with any of above means. Sometimes changing preference back-and-forth did the trick. Not this time though, that was persistent with "ready to upload" workunits being stuck indefinitely and no new work being downloaded as the server was waiting for the others to be uploaded 1st (running SIMAP at the end).

During the last 3 weeks crunching SETI WorkUnits (WoW!2014 event), the only issue sometimes was that computation was stopped and failed to resume (for whatever reason, usually overnight with 0 other device activity). But other than that, it was crunching along nicely.

Overall, no big deal but at least on my Device, I noted Android BOINC never really did 100% what it should do once a while (ever since release basically). All of it had to do with it not following my preferences correctly - and lately inaccesible preferences being set somehow, effectively rendering itself useless
As said, my device could be faulty so it may not necessarily be a BOINC issue after all ...
4) Message boards : Number crunching : Odd re-occuring issue with Android device (Message 1566168)
Posted 2 Sep 2014 by Profile FalconFly
Post:
Hm, I checked out that NativeBOINC client - and it demands App rights to access personal Files, Images, Videos and Audio files on the device.

That's a NoGo right there, as BOINC needs none of that obviously...

I guess I'll just leave it be, that's the easiest solution - and it's not like I'm losing a whole lot of computing power because of that anyway ;)

PS.
I do suspect that this has to do with Samsung Software Updates. Ever since they introduced the KNOX package into the updates, it appears everytime BOINC is losing control over its preferences and I ran into similar troubles with it...

Plus, the device always had overheating issues in a few certain applications, so I'm pretty confident the device has issues (I think it's the 1.6GHz version as it doesn't have LTE).
It's gonna be my last Android device anyway.
5) Message boards : Number crunching : Odd re-occuring issue with Android device (Message 1566107)
Posted 2 Sep 2014 by Profile FalconFly
Post:
Yeah, that was my next step as I'm running out of options with the official BOINC client :/
6) Message boards : Number crunching : Odd re-occuring issue with Android device (Message 1566095)
Posted 2 Sep 2014 by Profile FalconFly
Post:
Hmpf,

every now and then I run into trouble with crunching on an Android device (Samsung Galaxy S4).

After installation (or re-installation) of the official BOINC client, initially everything runs fine.

Since I have to limit the device to "use 1 processor" due to overheating issues, I have to enter the preferences at least once.

As soon as I do that, the device of course creates its local set of preferences (visible in the event log).

However, something else changes as well :
It from there on also states "suspended network activity (user request)" which of course I never did and I don't even have that setting.

And that's where it's crunching its cache dry, just to never upload a single workunit again and of course not getting new ones.
Since I can't access all of its client preferences, it's usually the end of story.

Does anybody know how to convince Android BOINC that I never suspended network activity ?!

I recently updated to the latest version but that didn't change anything.

PS.
I changed all available preferences around and allowed it to transmit data by all means - but still nothing worked.
Naturally, repeating any upload attempt in the projects page also fails (its not accepted and the attempts don't even show up in the event log).

Generally, I sure wish that the local preferences/config files of BOINC were at least be stored in a user-accessible location, so in case of everything going wrong it could be deleted or manually edited. Having full access to the options in the advanced preferences would also help.

Overall, I periodically run into some preferences-related woes. Sometimes it fails to recognize WLAN on/off states or simply fails to resume after suspending computation for whatever reasons. Generally - it's not running reliably smooth and really never has for me. I do wonder if that's my device acting up or BOINC.
7) Message boards : Number crunching : GPU task limits (Message 1564450)
Posted 29 Aug 2014 by Profile FalconFly
Post:
Agreed, I only took a quick glimpse at running multiple BOINC instanced but refrained from it for the sake of reliability and easy management.

I don't like fiddling around deep into that anymore, I'd rather prefer getting along with the established BOINC standards.
The issue was just new to me as I've never before taken a mixed host so hard onto the tasks limit like I did for the WoW!2014 Race ;)

Now I know what to do and next time, it could indeed already be fixed :)
8) Message boards : Number crunching : GPU task limits (Message 1564180)
Posted 29 Aug 2014 by Profile FalconFly
Post:
The Host limit got me into another issue a few times :

On a system with mixed GPUs and with the Host at the limit for tasks in progress, it consistently favored refilling a slower NVidia card, while letting a 4x as powerful ATI card basically run dry - leaving the NVidia card heavily overcommited on tasks and the fast card eventually run idle.

I had to manually micromanage it via Host venues to setup ATI only tasks, just to refill it occasionally.

I often wished the workunits went into BOINC without being hard-coded to the platform (specific GPU type or CPU) they'll complete on, the coding would then occur when the task is launched. Or something like a local workunit pool like in the old SETI classic days (don't remember the name of the Utility that allowed setting that up). All that would have saved me from a high number of sleepless hours micromanaging frequent over/undercommitments of GPUs.
9) Message boards : Number crunching : Lunatics Windows Installer v0.42 Release Notes (Message 1563212)
Posted 27 Aug 2014 by Profile FalconFly
Post:
Considering at least my AMD R9 290 cards have surprisingly similar runtimes for MB tasks as they do for AP, the credits resulting from that are simply *mind blown*

AMD R9 290
~ 21min for 2 parallel average MB tasks, roughly 6 per hour
~ 31min for 2 parallel average AP tasks, roughly 4 per hour

That's about a 1.5:1 runtime ratio for V6 AP tasks vs. V7 MB tasks under OpenCL (Lun 0.42).
Of course, the credit ratio per hour of work for that is like ~2800Cr AP vs. 510Cr MB or a 5.5 : 1 .

1.5:1 for a 5.5:1 crediting I'll take anyday of course ;)
Even with pretty well-blanked tasks (to a limit), the ratio is still clearly in favor of AP) with the current status quo.

I've generally observed that the NVidia cards comparably scream through Cuda MB tasks like hell while taking their time on AP (although still good) - while AMD cards eat AP tasks for breakfast but really take their time chugging through MB tasks.
The differences seem to decrease only when increasing the number of parallel tasks further towards its optimum.
( I got GTX 780 vs. R9 290 for that comparison, as well as GTX 750Ti vs. HD7850 , and that's what I experienced)

I didn't know that before, hence I was surprised to see things moving that way.
10) Message boards : Number crunching : Fundraising time again...... (Message 1562495)
Posted 26 Aug 2014 by Profile FalconFly
Post:
Looks like I was early for the show ;)

Thank you for your gift to:

SETI@home

We received your gift of $50.00 via Mastercard on 08/21/2014. Your confirmation number is 136354.
11) Message boards : Number crunching : AMD and PSU's (Message 1562146)
Posted 25 Aug 2014 by Profile FalconFly
Post:
Thanks FalconFly for the info

Makes it difficult to decide what to do as i all ready knew that you can overpower your system if you get a power supply to big
Working out how much this dam AMD uses at full load doesn't seem to marry up as a 850 should be enough but that does not seem the case.

Starting to wish i just bought Intel oh well i by AMD for the first time and i'm not happy


The calculator Zalster pointed out is actually pretty good. Maybe a tad on the aggressive side (I'd go for 50W more than it recommends just as a safety buffer and a little overhead room for minor hardware changes).

AMD and intel are completely the same when it comes to having the PSU dimensioned appropriately.

If the Host in your list is the existing config (FX-8350 + GTX 650), it should draw roughly ~240W maximum from the plug (assuming a typical setup, Displays not included).
A good quality 350W PSU would have no issue running it.

The FX-8350 is a 125W TDP CPU (~115W peak actual load).
The vanilla GeForce GTX 650 has a 64W TDP GPU, resulting in approx. 80W peak load (its PCB included and peak load likley won't be achieved outside running FurMark stress test).
Typical AMD AM3+ Motherboards clock in at about 25-35W usage.
For Fans and HDDs and RAM, add some 20W and you should be set (2x HDD, 2 Fans, 2 RAM modules).

----------------------
Now for those ~240W, we'd assume a 100.0% efficiency, which doesn't exist. For any 80+ PSU, add 20% onto that figure and you'd have round about what it would need to draw from the plug. We'd be talking about 290W there. That is absolute peak, worst-case short-term power draw. You'd have to intentionally stress-test everything to get there with dedicated Utilities - but in theory you could.
----------------------

Be aware though that PSUs ranging 350W and lower usually aren't too strong on the 12V rails (they're often more like office PSUs), hence people with strong CPUs and an additional PCIe GPU usually avoid going that low - even if the Wattage would make them seem perfectly fit.
The 400W class PSUs are typically the minimum (I'm aware of) that have the usual strenghts associated with the need of putting some serious loading on the 12V rail.
Despite total draw not being in the optimal efficiency region for most PSUs, you'll likely do make good use of its 12V capabilities already. On the upside with i.e. a 400W or 430W PSU, you gain some headroom for future upgrades or changes to the configuration, plus you have a buffer from component ageing of the PSU itself.
12) Message boards : Number crunching : AMD and PSU's (Message 1562028)
Posted 25 Aug 2014 by Profile FalconFly
Post:
May I chime in :

Since I've been using alot of PSUs, this is my experience...

Very reliable, good experience
----------------------------------------
Seasonic
Corsair
CoolerMaster
Thermaltake
XFX
Enermax

(there are more quality brands of course, above were just the ones I mainly employed)

Things I've learned with PSUs, that just worked well for me over the many years :

- never ever judge a PSU just by its advertised Wattage or go for Max Wattage for the buck. Go for quality instead and pay the price.
Many cheap PSUs don't even reach their advertised Wattage, let alone when being sucked by GPUs from the 12V rails. Some just fail, others actually blow or go up in flames (literally).
In contrast, the quality PSUs deliver what they promise and temporarily can even go further than that without taking damage.
Plus, they can handle load changes by the system much better and additionally filter out electrical line fluctuations far better... even brigde mini power surges without a hitch (all these things can make cheap PSUs go ape immediately, including delivering damaging over-/undervoltages to the system they're supplying).

- don't operate oversized PSUs 24/7 at very low load (i.e. 1000W rated with only 25% load)

Especially high-rated PSUs are fully optimized for certain load distributions, they don't seem to like it being taken to very low parts of the envelope for too long.
So when you look for a PSU to operate a certain system, make sure you know very well what Wattage it actually draws from the line and then aim for the desired PSU size.

- aim for about 60-75% sustained load for 24/7 operations and max. longlivety

While optimium efficiency (acc. 80+ certifications) is usually achieved in the ~80%-85% load region, permanently operating it at these very high loads is still putting alot of stress on the PSU and really puts its quality to a test.
Again, this requires actually knowing what Wattage you're expecting.

- PSUs (even quality ones) are aging

While a brand new quality PSU can deliver its full rated power, you have to know that over time of 24/7 operations, this number is shrinking slowly.
Normally not a hughe factor, but when using or considering to implement old PSUs into a System, it must be taken into account.

- Have a close look at the PSUs cooling behaviour

Every once in a while, even a quality manufacturer puts out "Silent PSUs" that can cross "the line between silent and stupid".
Means, these PSUs compromise their own cooling in order to remain silent - at nearly all costs (BeQuiet had a few of those in the past).
When you notice that even under heavy load the PSU is barely blowing air and the air it does blow is hot (not just warm), that needs to be taken into account for the overall system cooling.

- System cooling in general

Keep in mind that in difference to the old times, when the PSU was basically entirely responsible for moving the hot air out of the PC case, they cannot handle this job alone anymore - especially when GPUs are involved.
PC Cases that have sufficient fans shoveling cold air in and the warm air out are basically the Alpha and Omega of GPU crunching. The PSU will also thank you that, as this will drastically reduce its operating temperatures.

- periodically clean the PSU (pressure air)

PSUs can collect tons of dust without it being visible, as opposed to other hardware (i.e. Video cards etc) where it's usually very easy to recognize.
Most people would be surprised how much dust you get when you give it a thorough pressure air cleaning...
Although normally well-isolated, PSUs still by far operate the highest capacity and electrical components in a PC - and eventually that does not mix well with dust accumulation.

- Manufacturer recommendations for their Video cards are usually way overpowered

Since they have to guarantee that even cheap PSUs get along with their hardware, do not blindly trust their numbers and massively oversize your PSU.
Research the available tables of CPU and GPUs actual power consumption (TDP is at least a very good hint) and make a realistic estimation instead.
13) Message boards : Number crunching : Lunatics Windows Installer v0.42 Release Notes (Message 1561967)
Posted 25 Aug 2014 by Profile FalconFly
Post:
To FalconFly:

This is a "shot in the dark" but I recently started having driver hang/recover events on my GTX550Ti which had been reliable for the past year. After several days, it stopped completely.

I removed the card and carefully cleaned all the contacts with a soft pencil eraser, reinserted it and booted. It has been running again nicely.

Your mileage may vary, but it is a cheap and sometimes useful technique.

Best wishes.


The card is barely 4 days old, so it's definitely good on that. The board however is likely just too old (and possibly defective). It's the 2nd board of its type and has similar issues (although the 2nd did better, giving me hope initially).

I'll let the GPU run dry on a 2sec delay setting now.

The system is just unstable with it and it makes no sense, I just can't get it to work right.
14) Message boards : Number crunching : Lunatics Windows Installer v0.42 Release Notes (Message 1561848)
Posted 25 Aug 2014 by Profile FalconFly
Post:
Yes, after reading into it I found it and created the Key (sry for that, thought my edit came quick enough ;) ).

However, both extremes didn't work for me :

- removing it (as I read on the other thread) gave me a BSOD
(possibly would have been better to leave it and set it to 0, my mistake I guess)
I'll give a last shot setting the values to 0.

- setting the values to 60 still gives me those driver resets

I really think it's the old motherboard, that's my best explanation at this time.

-- edit --

Just tried it with both values at 0 and still gives the driver error message. I'll just leave it running like that and have an eye on the validating tasks.
Since I'm not sitting in front of it and it's just chugging along only connected via network & VNC, that is weird - but still acceptable for crunching purposes.
With a bit (or alot) of luck, the Workunits just continue in progress now with the 0sec timeout delay, that would be sweet.

Anyway, thanks for your help. It's 4AM in the morning here now and I need some sleep :)
15) Message boards : Number crunching : Lunatics Windows Installer v0.42 Release Notes (Message 1561840)
Posted 25 Aug 2014 by Profile FalconFly
Post:
Oki, found that one.

I'll play around with that one, see if it helps.
16) Message boards : Number crunching : Lunatics Windows Installer v0.42 Release Notes (Message 1561807)
Posted 25 Aug 2014 by Profile FalconFly
Post:
Hm, I'm looking for some help for a troubled host now (this one)...

While running one or two MB tasks/GPU (Cuda 5.0 App selected from Lunatics 0.42), all on default vanilla mbcuda.cfg and only 1 CPU MB Task running parallel, I'm getting frequent random Driver resets (337.88) on the GTX 650Ti.
It's running on the last System I could assemble for the current race.

Notably, the card is running on an old AM2 Motherboard on a 16x PCIe 1.1 Slot.
OS is an otherwise naked Vista32 SP2 (originally SP1 install just updated to SP2, no other patches/updates installed for the sake of getting it running quick).

I've let the System stand completely ilde for an hour and nothing bad happened...
I tortured it with FurMark plus CPU load and nothing bad happened (GPU Temp peaked at 69deg C, so it seems cooling is okay)... That also makes me think the PSU is okay (FurMark is known to create absolute peak GPU power consumption).
PSU is not hughe but a reliable quality Seasonic 330W and the CPU is only a 65W TDP one, HDD is a small 2.5". System should be clearly under 200W peak and I've powered more hungry devices than this with those in the past.

For the sake of trying and ruling it out, I might toss in a brand new Thermaltake 530W tomorrow.

But when I let it run BOINC, I keep getting the NVidia system message that the Video card/Driver stopped responding and was reset.

In effect, that causes the running GPU Tasks to stop and sometimes be placed into the "waiting to run" state (surprisingly they don't error out as I feared). BOINC then simply moves onto the next fresh tasks. The "waiting to run" ones are resumed after others finish, so technically it's still running overall.

Peak temp I observed with GPU-Z during crunching is 59deg C, which should really okay.

For now I'm a bit out of ideas, so any input is appreciated.
If I'm unlucky, the old components I reused just aren't up for the job anymore and I might have to drop the GTX 650Ti again...

As it stands, the System limbs along with that quirk - BUT I'll have to see if those GPU tasks validate (I've now set the cache to 0.1 days, so I don't waste too many Workunits if things go bad)...

--- edit ---
The Error lines from the Stderr output :

Cuda error 'cudaMalloc((void**) &dev_cx_DataArray' in file 'c:/[Projects]/__Sources/sah_v7_opt/Xbranch/client/cuda/cudaAcceleration.cu' in line 328 : all CUDA-capable devices are busy or unavailable.
setiathome_CUDA: CUDA runtime ERROR in device memory allocation, attempt 7 of 6
cudaAcc_free() called...
Exiting...
Preemptively Acknowledging temporary exit -> Exit Status: 0
boinc_exit(): requesting safe worker shutdown ->
boinc_exit(): received safe worker shutdown acknowledge ->
Cuda threadsafe ExitProcess() initiated, rval 0
17) Message boards : Number crunching : Lunatics Windows Installer v0.42 Release Notes (Message 1561295)
Posted 23 Aug 2014 by Profile FalconFly
Post:
Since I ran a few thousand AP Workunits with the latest AP Apps on both NVidia and AMD cards :

I've noted there recently were some tapes producing a very high number of extremely blanked AP Workunits (basically all of them, the whole tape), namely :

ap_06dc08ac
ap_08ja09ad

They got distributed about 2 days ago.

If someone downloaded a whole bunch of those - that would downright double computing time on the GPUs for quite a while... Maybe worth checking out.
I know I took my time fighting through them, and it's not that I'm lacking hardware right now ;)
18) Message boards : Number crunching : SETI@Home Wow!-Event 2014 (Message 1560346)
Posted 21 Aug 2014 by Profile FalconFly
Post:
Makes sense... and it supports the spirit of international cooperation for a common goal across all nation and team borders :)

I'm fine with the way it is btw., it's my 1st time in that challenge anyway...
19) Message boards : Number crunching : SETI@Home Wow!-Event 2014 (Message 1560331)
Posted 21 Aug 2014 by Profile FalconFly
Post:
I was actually also thinking that awards go to Top 3 overall, Top 3 for their nations and maybe Top x (debatable) inside teams - if a minimum number of members participated for nation/team.

The groups I didn't even know what to do with, so I just joined the one already having a member of my team.

It only dawned much later on me that those are the 12 sky constellations used for astrology/horoscopes ;)
(and I thought astrology and astronomy doesn't go together too well *g*)
20) Message boards : Number crunching : Lunatics Windows Installer v0.42 Release Notes (Message 1559777)
Posted 20 Aug 2014 by Profile FalconFly
Post:
My system with basically the same specs, the HD6370 is just a renamed HD5470, had an issue as well. The r2399 is not compatible with some cards with small WG size like these. Even -tune command line does not seem to help. You can have them try the values I was given to see if it does work on their system. However, I expect not.

Either use older GPU app work wait for v7 app if -tune also does not work for them.


Darn, I was fearing something like that... Thanks for the info.

I wonder if using the previous Lunatics v0.41 would help that, which is using older versions...


Next 20


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.