crashing in SL -- are you sure your video card isn't overheating?

Recently my computer started crashing in SL and other heavy OpenGL applications. I blamed it on the applications and the driver but somone eventually suggested that maybe my card was overheating. Duh! Why didn't I think of that? -- I've been using an nVidia geForce 3 Ti500 that has been rock-solid up to this point.

So I popped the case open and found great gobs of dust bunnies that had accumulated on the back of the card. I also noticed that the PCI sound card as up as close to the front of the card as it could get and was crowding the heatsink and fan. I cleaned off all of the dust, and moved the sound card down to the bottom of the tower.

I checked the heatsink fan and found it to still be working.

I noticed that the back of the tower was a bit crowded with wires and furniture, so I tied the wires off to the side and moved some things around to give the hot air more places to escape. Finally I realized that the Dell case had some hidden vents in the front directly under the spot where I had a habbit of putting my bag, so now I place the bag in a different corner.

I haven't crashed since.

This story has a moral: If you start crashing when it appears that nothing else has changed make sure your video card is properly cooled.

Oh come now, out of the 530 some posts of yours here, round about 30% of them blame firewalls, drivers, BIOSes, dirty SL caches and now overtemperature video cards. You continue to make these claims even when a random sample of your postings show that the person you are trying to help often says that the proposed "fault" Is not present at all.

What's next, recommendations for voltage stabilizers and $100 power cords?

I'm not saying that your recommendations are never applicable, but pointing the finger at everything else for the fact that SL is crashtastic is ever so slightly exaggerated. Why not just blame Microsoft as everyone else does?

Incidentally, as I use my machine for production work, it is as vanilla and stable as I can make it. Indeed, XP doesn't crash, photoshop doesn't crash, firefox doesn't crash, picasa doesn't crash. iTunes doesn't crash, SL crashes. And my nVidia 6800 is always at 53C in an ambient temperature of 38C and it both alarms and slows the GPU at 135C (per manufacturer defaults).

I can't say I could really describe Andrew's post as "blame" at all. He was providing an anecdote of a case where the primary caused ended up being poor cooling and circulation.

I can't imagine the time it took to review over five hundred of Andrew's posts, and I hope it's not necessary to review over five hundred of mine in order for my suggestions to be worthy of merit. ;^)

Part of technical troubleshooting is to eliminate the most frequent, most likely, most obvious, and most of all, most easily corrected problems. What would be sillier: going through all the common problems only to discover it's an obscure thing we need to fix; or immediately pulling an engineer from his work to spend days scouring Second Life's code, only to find out the resident just needed to blow dust out of their computer?

The forums don't paint an accurate picture... we receive hundreds of support requests every week, and the overwhelming majority of them are, in fact, solved by the troubleshooting tips at http://secondlife.com/help or the basic questions we first ask ("have you uninstalled/reinstalled", etc.). Hopefully, more than a few people who visit the forums for the first time, with a problem, read the stick at the top of Technical Issues and check those out. Many don't, which is why we still ask.

I say it every time someone mentions "everything else works", but it bears repeating: Second Life really isn't like any other program. The ability to run Photoshop, Firefox, iTunes, or even most current Direct3D games doesn't inherently mean Second Life should run correctly; in fact, the ability to run those programs doesn't even mean you meet the Minimum System Requirements. Non-3D programs don't even use your graphics card; most 3D programs are not OpenGL (and thus don't run into OpenGL driver compatibility issues), and video games and their pre-designed, pre-optimized, professionally-crafted environments don't push a computer to their limits the way Second Life's streaming technology and resident-created object-rich, texture-full environments do.

I don't know as much about the fine details of GPU cores as I might, but there's a chance the "manufacturer details" say to shut down at 135C because that'd be the point beyond which the hardware might be damaged. If that's the case, then there's certainly a point before that where performance and/or stability might be affected. I've gone beyond "manufacturer's spec" before, and still had stability troubles that plagued me until I went a step further... i.e. consistent crashing until I replaced a 400W Antec with a 450W Antec. We're not going to tell people to buy $100 power cables, but I am going to say something when (to provide a case example) they bought a $20 power supply.

Second Life is a program in a continual state of development. As issues arise, we do look into them. If something's our fault, believe me, I'll be the first to say it. But there's no conspiracy. There's no intent to "shift blame" to anyone that isn't us. Troubleshooting is about tracking down symptoms and eliminating potential causes. To do that effectively, the most wildly varying and problematic elements (differences in hardware, drivers, environment, and network connection) need to be examined first.

Responding roughly in order to your special pleading above:

1. Way to condescend to your customers, especially ones who might be nearly as clever as you.
   
2. I can't imagine how long it must take to look at 500 posts either. That's why I wrote a program to do the gross textual analysis and used statistical sampling to chose a subset for qualitative "eyeball" analysis. My subset was sufficient to yield a margin of error of +/- 5%. Total elapsed time, 45 minutes, which would have been shorter if vBulletin was a little bit better at coughing up information. I'd done this analysis a couple of weeks ago when I saw somthing similarly outrageous posted but declined to comment at the time.
   
3. What would be sillier than failing to check if the computer is plugged in before blaming SL for being incredibly buggy? I can think of a few things. The first would be publically posting that your machine went flaky because it never occured to you that dust bunnies accumulate in unfiltered PC cases. The second silly thing might be having failed to notice the numerous posts in the SL Technical Issues forum that have already beaten the concept of system and GPU overheating to death. The third silly thing would be having failed to incorporate cooling issue troubleshooting tips somewhere more conspicuously than in the bottom of a locked filing cabinet stuck in a disused lavatory with a sign on the door saying "Beware of The Leopard".
   
4. You're right, the forums don't paint an accurate picture of the sorts of troubles involved. Being a Liaison or a Live Helper does give a more accurate picture of what breaks and why. Having to apologize for the myriad bugs in SL that have no known workarounds for does also. Fortunately, I'm not being paid by Linden Lab to assert that buggy code is in fact operator error. Having done phone support for any number of SL customer who got tired of the LL "uninstall reinstall / firewall / drivers / bios / works for me/ no response" runaround also gives a person both an appreciation of the real problems encountered by real customers and the deplorable state of LL support. Trying to build things in game and having to texture a cube 3 to 7 times before it "sticks" also increases one's appreciation of how buggy SL is, asking if others are seeing the same behavior only supports that confirmation. Characterizing the slow rez of textures pinning it down to age of texture upload only to be told by a Liaison "textures don't rez slowly" lends yet more support.
   
5. Again, I have to agree with you, SL is indeed different from all other programs as it is far, far more buggy than the programs enumerated above. As long as you insist on convincing yourself that there is something that makes SL fundamentally different than any other program known to humankind the more likely you will be to dismiss bugs instead of attending to them. Put another way, if LL should err on the side of assuming a reported problem is real or spurious, you ought err on the side of real but your carefully articulated stance above shows you'd prefer to think spurious.
   This point even deserves a second paragraph. Here's a funny bug that you may not be aware of, following a teleport, your AV may be invisible to everyone but you. A very odd "fix" is for you to change your active group and suddenly there you are visible. Dustbunnies? Methinks not. What would the troubleshooting guide suggest you do? Relogin. That is the hallmark of buggy programs: if it doesn't work, restart it. And just how much code is there for all of the bugs involved? Well, it is hard to know because we don't have information on server-side code, but there are plenty of demonstrably client-side only bugs. How big is the SL client? 9MB app, 2.5MB core dlls. By contrast are Photoshop (19MB app, 20MB core dlls) and Firefox (6.5MB app, 1MB core dlls), so not only is SL way buggier than applications 10 times its size but buggier than apps one half the size as well. Yes, I know that SL beats the shit out of the GPU, but the failure modes (app crash vs. system/driver crash) are clear and happen with a frequency 100s of times that of the driver.
   
6. You may not know much about GPUs, but if you read as carefully as you defended, you would have noticed that my nVidia slows the GPU at 135C which is but a mere 23% hotter than my nominal operating temperature (which itself is but 5% hotter than than case ambient). You might also have notice that Andrew's dustbunnies were causing other OpenGL programs to fail which your defense ignores.
   
7. Assertions of shared hubris that interefere with recognition of error are most assuredly not accusations of conspiracy. I'll decline comment on what your failing to make that distinction indicates.
   
8. Finally, your customers range from the incedibly savvy to the less so. Being dismissive of them certainly works only to hide real defects from yourselves and doesn't do very well for your perception by them. And in case you are inclined to retort that I'm talking about me, you'd be wrong. Yep, I've helped folks who can't understand why SL runs poorly over dialup, I've also worked on diagnosing your bugs with with your customers who ultimately lost entire inventory folders because the best of LL was capable of only zapping the borked DB entry for the folder. I've also helped plenty of newbies extract clothing from the sales box instead of wearing the box on their head, as well.
   That said, you might care to take a look at a few of my posts in "Technical Issues" before stooping to lecture me on what troubleshooting means.... or not.

Wouldn't the forum moderators recommend that your reply might have been be better handled in private communication?

I can say that a while back my video card was indeed overheating. New card, extra fan. Still crash.

Well my SL doesn't even launch. Version 1.4 launches ok, but 1.6 doesn't even show this "Second Life is loading" text. Hdd works for a few secs and then nothing. Someone from Linden told me that i don't have open gl, how i could not have it ? I have newest nvidia drivers and still same error. I have Geforce 2 mx 400 64 ram ;/

My sincere apologies for any disparagement of Andrew Linden above. It turns out that I had made an error. When a fellow player asked to see the approach used in my analaysis, I found that in the intervening weeks since my analysis and posting of the results that I had forgotten the exact target of my anaylsis reported above.

I spotted immediately upon looking at my textbase and queries, that it was indeed Lee Linden not Andrew upon whose postings my analysis was performed. Andrew. please accept my humble apologies for my error in memory.

Every other statement I made above holds true but for the target of analysis and now has the benefit of haing been peer-reviewed.

Thomas, it appears that between revisions 1.4 and 1.6 the use that SL makes of the GPU increased enough to have pushed some systems that were close to the edge of power needs over the edge. I'm sorry that I hadn't caught your post in July.

The redoubtable Catherine Omega explains it quite well in this thread; please let me know if you need further help. Aside from the players posting in that thread I know of one friend who was exhibiting the same symptoms upon release 1.6 who did replace the power supply and now has no problems. Pay particular note to Catherine's explanation that the "ratings" on the power supply are often true but inaccurate depending on how the manufacturer labelled the supply.

And if it isn't obvious, if you aren't handy with a multimeter and don't like electrocution, please find someone who is. Speaking from one personal experience with a power supply case that should have been grounded (earthed) but was not, a few seconds unconciousness is not particularly fun and will definitely get you talked about in the office.

Where do you get off trolling Linden? Here Andrew is trying to help the unwashed masses and you stomp on him for it. His apps started crashing so he pulls the computer open to find that poor air circulation (caused by dust and poor layout) is causing the problem. So he reorders the layout and cleans up. I don't know about you but that seem like a righteous fix. Other applications were effected too so he was able to rule out SL.

Even if it was Lee who posted it, what does it matter; it's a valid fix to a valid problem. This is a common problem; while it would be nice if people would use the forum search, most won't. You were just looking for an opertunity to troll.

If you want to yell at Lee for anything; be it for not asking for for people to post thier log file. Alot can be figured out from the log files.

If LL would work on fixing what generagees most of the errors and warning in the log file i'm sure SL would be more stable.

From what i've been able to find out about your video card it only supports OpenGL 1.2, i'm not totaly sure but i think SL requires OpenGL 1.5. Time to aquire a new video card.

If you could post your "debug_info.log" found in "C:\Documents and Settings\<username>\Application Data\SecondLife\logs\" it would really help.

I was having lots of problems with my computer crashing to reboot. Posted here several times about it - finally found the cause!

I had two little areas in my memory (had 2/256 modules) that were bad. The bad areas were in the first module. Just those two little spots - but if they were hit by something then my computer completely rebooted. We were able to find a program online that checked the memory. Replaced the memory and haven't crashed since (i mean to reboot). Who'd have thunk? *grin*

I like this kind of information, I am glad to read it..

As I know enough to read it, see if it is something that would apply to me, add it to my trouble shooting list, done.

I don't understand how saying, check your firewall, check the temp of your video card, check you memory, etc, is bad as in over 90% of the problems they are the cause..

I am happy to read about this and add to my list of things to check when/if I have problems, as they do nothing to hurt, but help...

Keep telling us about these Andrew, Lee, or any Linden, there are some of us know how to take this as information, as just that information. Thanks

What I want to call Lee out for is steadfastly telling people "relog / reinstall / drivers / bios / works fine here / nope textures don't take minutes to load / nope the asset server didn't corrupt your texture / nope the asset server didn't just decide to make your texture folder incapable of holding anything at all" as if he were a cardboard cutout instead of a diagnostician. I'd like for bug reports to be treated as such instead of "just those stupid fucking users". I'd like for Lee to lose the "nope it's not really bugs, it's just that our app is unique in all the world", I'd like for Mr. Onizuka to not get a Linden slapdown for noting a showstopping bug in 1.7 feature feedback.

But until such a time as the Lindens correct their viewpoints, none of this will happen. If every single bug I ever submitted wasn't closed with "presumed fixed in new release please re-report if it is still broken" when they are all still broken I might be a little more kindly disposed. So yes, a lot could be learned from the log files they don't ask for. If they can't be arsed to read bug reports, what makes you think they'd read log files?

As you yourself note, "if they would work on fixing what generates most of the errors ... SL would be more stable." Do they seem to be doing so to you? I contend they have an endemic cultural problem believing that SL is buggy and when you are infused with that sensibility, nothing positve happens. This is my attempt to "be the change". I'm sorry that it offends you, but nothing else I've done seems to work either.

I would like to start by saying that this post is not intended to rip or knock down anyone else. It just bothers me to see someone else get torn into while trying to provide technical support which by nature is a difficult task. Computers are a different animal and do not follow any of the rules per say as most other things go when it comes time to diagnose problems. Those books say one thing but in reality it is a totally different issue.
I must say kudos to the Linden tech support team in all their approaches to tech issues. Since I work in the computer repair business as an on-site/in-house technician I totally understand how difficult it is to diagnose any issues related to poor PC performance without actually seeing the computer firsthand. Wile I can agree that a lot of it has to do with client software having issues and core bugs, but those bugs can have a greater effect on a system that is already buggy as opposed to a system that is more rock solid as far as hardware/software configs go and can compensate by refreshing ram faster, cycling out CPU and GPU cache faster and fetching data from the hard drive faster. The majority of the fixes posted here are 100% valid and in my 10+ years I have seen them all and had diagnosed/fixed them all( including the cooling issues). I would like to stress the fact again that it is very difficult to diagnose any system bug be it client or hardware/drivers/other apps when 1. LL did not build your computer system, let alone recommend a specific brand/config to purchase. 2. LL tech support staff do not have physical access to your computer to see that your power lead extensions are in fact cut and spliced with electrical tape or see the fact that there is a space heater/mini-fridge that is tied into the same circuit as your computer is. 3. LL is only handing out the minimum hardware specs for running the SL client, it is up to the end user to check their mainboard manuafacturer's Hardware Compatability list to see if that $1,000 awesome totally roxorrzz AGP 55X 1GB nVidia 89000XMPLOPM video card with sub-space dynamic quantum cooling fan is going to work with their system's BIOS/Core Chipset, just because the specs are cool, that does not mean it will play nice with your mainboard.

I agree that the SL client needs cleaned up and spit-shined but this is the result of a progressive, moving technology. You cannot have innovation or progress without the associated potholes.

I would also like to add if it is ok to (not in violation of the TOS), if there are any tech questions that I could be in any way help with, please feel free to IM me in-world I would be happy to help in any way that I can or refer you to a website/other source of info that would be better help.

Dear Goodness!

Andrew Linden posts a useful tip for people using their graphics cards for intensive graphics operations (Hands up everyone who uses SL) and he gets attacked?

Let the man speak, and if you have to reply, make it a "Thanks".

Thanks Mike, well said.

Linden Labs is a fairly small company trying to make a pioneering application work on thousands of different system configurations. Nobody said it would be done today.

I try to be helpful, not clever. Being clever in a forum environment just gives me trouble; it makes my responses less productive, and a little too self-serving for my own tastes.

Since statistics have already been run on everything I've said on the forums, I'm curious about the process. I see that not every post was examined, only a sample (however represenative it might be; I don't know how many made a sample). Were terms such as "firewall" and "driver" simply searched for? Does this include posts where I made a suggestion that had not yet been made? Or, is it the percentage of posts where I stated that a condition outside of Linden's control was certainly the only source of the problem? I enjoy statistics, but I'm always curious when I see qualifiers that can be interpreted broadly. As such, I'm intrigued by the math behind a statement along the lines of "30% of Lee's posts place blame on things beyond Linden's control".

I say things like "this does happen" and "have you checked this" quite a bit. By the same token, I can understand how (in the proper perspective) "this has been a problem for many people" can be read as "this is what your problem is."

Statistical curiosity aside, it looks like most of the rest of this (the assertation that Second Life should be more stable than Mozilla because its filesize is smaller, for example, or suggesting that my response was not appropriate for public discourse) is better left alone. As far as I'm concerned, Lindens are too busy trying to make Second Life better, facing (and trying to conquer) the things we're not doing right day after day, to be absorbed in some theoretical wave of arrogance that we can do no wrong. I'm sorry you disagree, but that's where I'm going to leave this thread.

I would also like to add this:


The moment I stop seeing bugs in SL is the day I start worrying about the future of SL. As long as bugs continue to exist then get fixed and new ones emerge to follow the same pattern, I am happy with the knowledge that LL is still pushing the envelope in the technology, I am happy knowing that the software engineers and technicians at LL are doing their jobs and ensuring that the monthly fees are worth paying.

I consider myself dismissed, Lee; Thanks. But your refutation was more handwaving about how you are making things better whist they manfestly get worse. Congratulations on form over substance. You win, but not in any sense that matters.

I have to say I agree with Andrew there. I don't have any problem with the current version of SL, or indeed any other application, unless there's a hardware failure.
I spent the whole summer logged on to SL for hours at a time and although it may have crashed like once or twice, I don't really recall any instability.
And yeah, I'm sure that anyone with half a brain realizes that 90% of support issues are user error. It is a fact. Just about every email I get about Stagecoach Island is a result of people misunderstanding how the system works, or some weird latency with the email/rpc backend.
It probably doesn't keep them from thinking I'm a crap coder

Not that LL needs defending, but I have a lot of tech support experience and the trouble shooting method described here is indeed the only logical approach one can take. The idea is to eliminate possibilities one by one starting at the bottom of the tree.

To approach it any other way is to approach it ass backwards and to produce the opposite of the intended result.

I can't even imagine what you read between the lines of the innocuous, extremely helpfull and even cheerful post that started this thread that caused you to think these things. All I can get from your posting is that you seem to think the Lindens were conspiring to make up excuses so they did not have to do their job? Do you realise how nutty that sounds?

I commend the obvious restraint used by the Lindens in the first reply.

Where I work, your file would be dropped into the "high-maintenance" file for further dust collection whereas the Lindens seem to be much more charitable.