Are you guilty of reporting 'HOT' bugs on Gentoo Bugzilla?

dufeu · l33t Joined: 30 Aug 2002 Posts: 924 Location: US-FL-EST

Before anyone gets heated up, the Subject is a flaming play on words. This post is about 'heat' as in CPU heat causing misleading problems to be reported to bugzilla.

I've posted an enhancement request to bugzilla asking that the x86 Installation Guide be updated. This update is to mention potential 'heat' problems. You can reference the bug here https://bugs.gentoo.org/show_bug.cgi?id=16933

Basically, I found my emerges of gcc kept failing at random places in the java setion. Since I've read a fair number of complaints about gcc failing (especially in the java section), I thought I'd share my findings regarding the real cause of my problems.

What I discovered what that my average operating temperature was 67C. This, my friends, is hot. And this was with one of the high end Themaltake after-market coolers. But first, let's back up a little and describe the problem system.

This machine is an Athlon-XP 1900+ running 512meg of RAM. It includes a high speed (52x) cdrom drive, a 42x10x40x cd-r/rw drive and 6 hard drives ranging from 27 gigs to 100 gigs (total 250gigs). Over time, a layer of dust had built up inside the case especially in the CPU fan area.

After being sufficiently beat in the head by my system freezing randomly and the random crashes of various ebuilds and especially the gcc ebuild, I realized I might be having heat problems. Sure enough, as I noted above, I was running at 67C.

Vacumming the case and especially the CPU fan, brought the 'at rest' CPU temp down to 55C. This is a reduction of a full 12C.

My tip is that if your having random problems with ebuilds, check and see just how hot you're running. Take my 67C before vacumming and add the 10C rise caused by compiling significantly sized ebuilds and you get 77C. This is well into the unreliability zone for the average modern high performance CPU.

:-)

_________________
People whom think M$ is mediocre, don't know the half of it.

AlterEgo · Veteran Joined: 25 Apr 2002 Posts: 1619

Thanks for sharing that tip.
I think vacuuming your case(fans) is kinda dangerous: risk of damaging parts, sucking up cables, static electricity.
I use a can of compressed air to clean my PCs, twice a year.
And it does save me some 5 degrees in temp.

Ari Rahikkala · Guru Joined: 02 Oct 2002 Posts: 370 Location: Finland

Perhaps a link to the sig11 FAQ ( http://www.bitwizard.nl/sig11/ ) would be in order... that's generally the page where you should direct anyone who's suffering from crashes in software known to be quite stable.
_________________
<laurentius> gentoo linux?
<ari> Yesh.
<laurentius> they look horny

mgirard · n00b Joined: 30 Jul 2002 Posts: 64

dufeu · l33t Joined: 30 Aug 2002 Posts: 924 Location: US-FL-EST

Yes, I didn't mean a regular vaccum cleaner! I bad. ;-)

BTW, when I really, really want to clean stuff off, I usually use compressed air. But in this case, my little stinger vac with the small places attachments was more handy and is what I used.

I don't know how closely any of you follow Gentoo Bugzilla, but I know I've seen a fair number of posts specifically for emerging GCC and moreover getting the broken emerges at random spots in the 'java' section. After watching my temp monitor during a GCC compile, I can tell you the java section is particularly intense for the CPU. But the suggestion that the problem might be heat related is rarely made by those trying to help solve the problem.

I feel that it would help if we all raised the general awareness of 'heat' as an issue. In hind sight, I think heat is the root cause of a significant number of bugzilla posts. Not a large percentage, but definitely there.

:-)

Trumpcard · n00b Joined: 27 May 2002 Posts: 36

I've found that I have to remove my case covers to be able to do lengthy compiles. I use Artic Silver and a heavy duty Antec fan that sounds like a friggin Galaxy taking off , and I would still encounter lockups. Once I took my case cover off, all was well. This is an AMD XP-2100 chip. My gut feeling is that AMD's just don't handle heat as well as Intel chips, in fact I've heard alot of people mention that fact, and it exposes itself in gentoo because of the taxing nature of the loads of compiles that are performed.

Its a brand new case/CPU/fan in my case, so dont think dust is an issue.

Gnufsh · Guru Joined: 28 Dec 2002 Posts: 400 Location: Portland, OR

That means you need better airflow through your case, I believe. The best HSF in the world is useless if you stick it in an insulated box with no airflow.

carambola5 · Apprentice Joined: 10 Jul 2002 Posts: 214

Be especially wary of this problem with laptops.

I was in one of my professor's offices bootstrapping when I realized that the one side of the laptop was quite warm. Opening up the temperature file in the ACPI directory told me that I was running at 68C. Crap! And this thing was days old at the time. So I left for a cooler area in the building (the lobby) and got the temp down to 57C.

So don't forget that ambient temperature significantly affects CPU temp.

Trumpcard · n00b Joined: 27 May 2002 Posts: 36

I'd dont see how it could be airflow, this case has 2 internal fans in addition to the CPU fan. It moves plenty of air. The lockups occur at relatively low temperatures (relatively low in the grand scheme of things, usually between 40 and 60 C. It finally locked up for it's last time last night..

I've had this problem with more than one AMD, I never had these problems with Intel.. Im starting to see why AMD is the cheaper solution. Im at the point now that Im going to switch back to Intels and not look back I think. I was reading an article not long ago with they were discussing how AMD's qualitity control has gone through the floor, and I'm really starting to believe it. Not trying to start a flame war of AMD vs. Intel, but the new AMD's have the worst heat issues Ive ever seen since the old Cyrix M processors.

Oh well, I'm sure your mileage will vary, but I can say that I'm disgusted with these disposable processors Ive been getting for the past year...
This will be the 3rd time I've had to redo the hardware on my gentoo box in the past year due to hardware defects. In comparison, I just pulled one of my production servers done thats been in service for 3 years for the 1st time for maintence. PII450, 256MegRAM, and its never had issue one (along with all the other intel servers, amd's on the other hand end up needing motherboards/memory/cpu's much more frequently)

Craigo · Posted: Fri Mar 21, 2003 5:58 pm Post subject:

Nice to hear some stories about vaccuming dusts, using compressed air to clear out all sorts etc.

Let me contribute a small story here. We got a dual processor alpha box at work and around 2 gig of ram. And the machine was stored in one of the machine room. And it has never been moved since...... possibly ever.

We had a problem with the machine recently which took around half a day to find (while contacting HP support which to my surprise are really clued up staff!). We decided to open up the case to find serial codes etc and there was tons.. yes.. TONS of dusts and dirt in there. Comparing to your stories, this is one bad ass tons of it.

We had a phone call from HP that they sent out their engineer out to fix it asap. We knew we had to clear up all this dust asap. A quick trip to one of the lab to use their air compressors. I have never seen so many dusts flying out in my life. (who knows, we could have create a dust puppy in there ;)

To most people on this topic, I do know that various heatsinks are really awkward especially AMD's heatsink they give out along with XP cpu. Various reports I've found out reported excatly what dufeu but with dusts between the CPU and the heatsink.

Trumpcard, I kinda have to agree with you there about the heat issue that AMD has but I have never ran into this problem at all. Just probably that I pick the best heatsinks etc. We ran into a interesting discussion that Intel are far better in the server market than AMD. However, the irony is that we are running AMDs on server/desktops because of the cheap pricing.

If Intel brings their pricing down around AMD's level, then people can have a better decision to pick whatever they like.

During my decision to pick a laptop, I just went for Intel instead of AMD basically on the heat issues and the support of Speedstep in Linux (which works really well).

-/Craigo/-

konqueror · Tux's lil' helper Joined: 16 Mar 2003 Posts: 76

hi, sorry if i'm slightly OT, but, could you guys give me some good pointers for getting your small vacuums? thanks.

Trumpcard · n00b Joined: 27 May 2002 Posts: 36

I've had a few AMD based desktops that have been relatively stable, but I stress my servers MUCH more than my desktops.. Frequent compiles mostly, and that works the memory and CPU exhaustively. . Overall, AMD's are fine for a desktop solution, you'll have to excuse my tone in the earlier message, but I had 2 server crashes in one day. The intel one was a man made problem, the AMD server on the other hand is a repeating problem. I've had problems so far with XP1800's, 2000's, and this one a 2200. I just got the box booting again by replacing the CPU with a 2000+ I had stored away for a future project, so I'm mildly calmed down.. lol... Nevertheless, as soon as the Dell PowerEdge gets here, this servers MB+CPU are being regulated to a desktop machine.. I just hope I can RMA this chip.

Fan wise, I've typically used very good ones too. I tend to get Volcano's and Heavy duty Antec fans, and use thermal compound on them as well, so I'm taking the time to make sure I have a safe cooling solution, but Im starting to think nothing short of injecting liquid nitrogen at this point is going to help AMD to made a stable server. In all fairness I haven't tried the MP chips.

Thanks for being patient with my ravings!

Craigo · Posted: Sun Mar 23, 2003 12:44 am Post subject:

Trumpcard,

Since you are running XP cpus for your servers.. I can see the point why they crash during your stress testing on those

Slightly different topic here, I just have been looking at colo boxes and most of them are only AMD durons but the rest are Intel celeron/standard/xeon. I think that just rounds up that Intel are still leading the server market?

Though, I haven't really seen any AMD MPs CPU around ;/

-/Craigo/-

tempy · n00b Joined: 21 Mar 2003 Posts: 15

I cleaned out my fans after reading this, and even though the dust didn't look particulary bad my CPU is running 10 C cooler now.

Hmm... I'd be interested in hearing how well the Crusoe and C3 hold up under stress testing. They run pretty cool compared to both major players.