Hi Folks,

One of my three hosts is (or was) a major-brand laptop (2.0GHz C2D, WinXP Pro). For a number of reasons (other than BOINC), I left it running 24/7. Recently, it...um, well...it stoppped working. Making a long story short, I have good reason to believe it's a system board or CPU failure.

When I started BOINCing about 5 months ago, I added this laptop as a host and have been crunching away with a 75% CPU limit ever since. Perhaps out of ignorance, stupidity, denial, or combination thereof, I ignored the fact that the fan ran a lot and there was a good bit of heat being generated.

So, I guess I'm wondering how likely it is that running BOINC over-stressed the laptop in some way and did it in. Is there any safe and reliable way of knowing how much a laptop can be pushed running BOINC 24/7? I know that I'm leaving out a lot of details and there may be too many variables to give a good answer, but I'm hoping that I can return this host to use (maybe limited to 25% or 50% CPU time) after it's repaired and am just trying to get an idea how/if I should do it.

Thanks for the input.

MarkR

Not BOINC directly, however heat can be a real problem in laptops, and BOINC causes the CPU to generate more heat. If the heatsink and fan are clear it should not be a problem though.

Some other hints, keep the laptop on a hard surface so that air can get to the intake. Raising the back a little can help airflow and often makes for more comfortable typing.

most notebooks are not designed for 24/7 100% - you have to take measures to do that. like putting them on cooling pad, cleaning heat-spreaders often, etc..

you'd better do something to check the temperatures inside and controlling the fan(s). [@JR - do you read me?]

for DELL's there is something @ http://www.diefer.de/i8kfan/ I am using. probably you'll be able to find something like that for your NB..

I'll just add that I found out that by limiting the BOINC's CPU usage to 25% on my laptop, the fan doesn't spin up. So I created a venue just for the laptop, which of course stops crunching when running on batteries.

The problem is that unless the WU is short, it may miss deadlines. So I picked and chose only a handful of projects to run on it which have short WUs: IBERCIVIS (Materiales16 & Materiales24), PrimeGrid (LLR TPS & LLR PPS), QCN, Gerasim and Malaria (Predictor).

HTH

Thanks for your input, everybody. I have my laptop back with a new sys board, CPU, and RAM, although I don't know exactly what the real failure was.

@frankhagen...I have a Dell and am now running the ik8fan utility you suggested. Thanks for the tip!

Is there any documentation available that gives max recommended sustained core temps for different CPUs? My 2GHz C2D is running right 60C +/- 5C with both cores running projects with a 60% limit. Any ideas if that's too hot or if I have plenty of room?

The processor data-sheet have such information, but laptops have poor ventilation due to size constraints and weren't really designed to withstand running constantly at full-throttle. It just so happens that somethings will give, be it the fan bearing, dust and fluff in the heat-sink, etc, way before the CPU itself, but taking it with them when they do fail.

I repeat my advise that you should find a maximum CPU limit that doesn't make the fan spin up to the point of being noticeably audible.

HTH

i hope it was covered by warranty - ?



HTH - the nice thing about it is that you can really see what the different sensors are telling. and f. e. can force fans to stay @ max.



depends on the exact model (even steppings). you can find them @ intel - search for TDP - i remeber posting a link somewhere on the forum (huge doc, but it really covers every aspect).



60 leaves plenty of room. im' running my D820 with T7200 mostly above 70.

but it's not that simple.

1st: what does boinc do if you go for less than 100% - ever checked taskman? 100 - 0 - 100 - 0.... it's time throttleing not going for constant load-limit!

if you wan't to reduce temperatures better control CPU andMB frequencies - how? actually it's built in (called speedstep) you just need a program to control it. you might try speedswitch XP (you'd guess: diefer again )

2nd: depends on the projects you are running. i've seen everything from under 60 up to over 80 with boinc on 100%. so you can't set and go..

Fortunately for me, yes.



This is the confusing part for me...Yes, when I have preferences limiting CPU load, most projects seem to bounce between high and low loads that eventually average out to the set limit. However, Prime Grid doesn't seem to care what that limit is -- it runs at 50% on taskman (that's 100% on its core, right?) no matter what I have the CPU load limit set to -- and my T7200 temp spikes as high as 80C. Is PG not playing nice, or do I misunderstand how the BOINC manager works? BTW, I use the 5.10.30 client on this host and set the limits using local prefs, not through BAM!.

1st: depends on the app - 2nd normally taskman shows you a 2 second snapshot. actually it's really doing time-slices of 100% - which is really bad for speed-stepping cores.



jumping too short again - PG-sieves run cool - it's merely pure integer stuff (which i'd recommend for notebooks - 3x+1, ABC, PGsieve...).
PGwoodall is the pure opposite: doing FFTs is the hardest stuff around. you'll be able to fry eggs on it.

if you really wan't to cruch that stuff on NORMAL NBs you should set speedstep to battery-mode (reducing core and memory speed to 50%) and let boinc run @100%.






nothing wrong with either - allthough i'd update to 5.10.45

frank.

Thanks, Frank. Indeed it is a Woodall WU I have that spikes my CPU temp. I feel very ignorant about all this...I thought one CPU cycle was like any other. There's obviously a whole lot more going on than I ever thought!

With regard to PG, I see that you can pick and choose your work. Would I be "safe" with anything other than Woodall?

Do you know of projects/WUs other than those at PG that are particularly intensive and can make the CPU run hot?

a cycle is a cycle - but depending on the code it's using more or less of that hardware on a chip.




everything that sieves - try sr2sieve. - or go for ABC, or 3x+1. RieselSieve ran cool..




QMC for instance. but it's not only CPU - if you check the sensor readouts (you can now) board and memory can run into problems too. for instance on projects juggling around huge amounts of data.

you'll just have to watch what's happening when you start crunching other stuff..

I have a relatively new laptop and it's running unthrottled at 57 C, guess I'll be good? or is it still a high risk?

Thanks for your input, one and all. I've done some picking and choosing between projects and have things throttled back to 65% CPU usage. It seems to be running comfortably at about 60C with occasional peaks to 72C. I'll keep my fingers crossed...

MarkR