During the course of playing FarCry, I noticed several times that random 
polygon edges ended up in the wrong place sometimes. (By "noticed" I 
mean "now and then I couldn't see half of the room because one of the 
wall polygons was stretched across the entire width of it".)

I had assumed this was a software glitch - FarCry is rather old now 
after all. I had assumed something in the game's coding was interacting 
badly with my video card.

However, I've just been playing TF2, and the exact same thing happened. 
This has *never* happened before in TF2 with this video card. So now I'm 
thinking there is actually something wrong somehow.

I see three possibilities:

1. The GPU is overheating.

2. There's a bug in the version of the video driver I have installed.

3. Valve have made an update to TF2 which introduces some kind of 
rendering bug.

Apparently the nVidia control panel no longer tells you your GPU 
temperature; you have to download and install an additional tool. 
Anyway, having just stopped using the GPU, the temperature was reading 
55 °C. Over the course of the last 5 minutes, that's fallen to 53°C.

To me, that sounds rather damned hot. Is it usual for a GPU to get this 
warm?

OTOH, the fact that it isn't cooling down especially fast seems to 
indicate that this isn't an issue. (The nVidia tool allows you to set a 
temperature warning, and the default is 127 °C.)

At this point, I'm not sure if my GPU is failing, or whether I just need 
to update my drivers...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v8 wrote:
> Apparently the nVidia control panel no longer tells you your GPU 
> temperature; you have to download and install an additional tool. 
> Anyway, having just stopped using the GPU, the temperature was reading 
> 55 °C. Over the course of the last 5 minutes, that's fallen to 53°C.
> 
> To me, that sounds rather damned hot. Is it usual for a GPU to get this 
> warm?

Very hot to touch, not so hot to run.  IIRC, most chips are fine up to 
65-70C, and several can go higher w/o a problem.

Myself, I try to keep them around 45C.  If they go higher than that, 
then I look at additional cooling.

...Chambers

Post a reply to this message

Shutdown. Let it cool for a few minutes. Open your tower, clean dust out. 
And I mean *clean* dust out. Open your power unit, (yes, open your power 
unit) and clean the dust out of that too. Open your fan covers, and clean 
the grime off the blades of the fan/s. Pull your graphics card out, and with 
a small screwdriver, (or similar), get as much grime off the fan blades as 
possible. (Hard to get in there if it's a sealed unit, but you've got to do 
it - be careful).

  I did exactly this the other day. You should have seen the crap in mine! 
Get a hoover in there to mop up. Be careful of the wires.  ;)

   If you don't do what I said, bye-bye graphics card. Hello new one. 
Seriously. I learnt this a long time ago.

    Edit: Oh, you're using a laptop aren't you? That must suck with dust in 
one of those. :o/

   ~Steve~

Post a reply to this message

St. <dot### [at] dotcom> wrote:
> Shutdown. Let it cool for a few minutes. Open your tower, clean dust out. 
> And I mean *clean* dust out.

  Get one of these too:
http://www.extendyourlashes.com/images/air-blower.jpg

  They are really handy when cleaning your computer of dust. You might also
want to use a vacuum cleaner at the same time if the amount of dust is large.

-- 
                                                          - Warp

Post a reply to this message

Warp wrote:
> St. <dot### [at] dotcom> wrote:
>> Shutdown. Let it cool for a few minutes. Open your tower, clean dust out. 
>> And I mean *clean* dust out.
> 
>   Get one of these too:
> http://www.extendyourlashes.com/images/air-blower.jpg

There's also "canned air", which is just what it sounds like. A bit more 
vigorous than a bulb for camera lens cleaning.

>   They are really handy when cleaning your computer of dust. You might also
> want to use a vacuum cleaner at the same time if the amount of dust is large.

I've seen vacuum cleaners advertised as static-free. I don't know if 
that makes any difference, but you might want to consider it if you're 
going to vacuum directly inside the case.

-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message

Darren New <dne### [at] sanrrcom> wrote:
> There's also "canned air", which is just what it sounds like. A bit more 
> vigorous than a bulb for camera lens cleaning.

  But limited contents.

-- 
                                                          - Warp

Post a reply to this message

On Sun, 17 Aug 2008 14:17:21 -0400, Warp wrote:

> Darren New <dne### [at] sanrrcom> wrote:
>> There's also "canned air", which is just what it sounds like. A bit
>> more vigorous than a bulb for camera lens cleaning.
> 
>   But limited contents.

I used to see cans of compressed air that came with a hand pump so they 
could be refilled.  I never managed to get my hands on one to see if they 
worked well, and haven't seen them for a while (the store that sold them 
has since gone out of business).

But I imagine they still exist *somewhere*.

Jim

Post a reply to this message

> I see three possibilities:
>
> 1. The GPU is overheating.
>
> 2. There's a bug in the version of the video driver I have installed.
>
> 3. Valve have made an update to TF2 which introduces some kind of 
> rendering bug.

4. Hardware fault on the graphics card (eg bad RAM)

> Apparently the nVidia control panel no longer tells you your GPU 
> temperature; you have to download and install an additional tool. Anyway, 
> having just stopped using the GPU, the temperature was reading 55 °C. Over 
> the course of the last 5 minutes, that's fallen to 53°C.

Try showing the temperature in parallel with running the GPU (eg 3D game in 
a window), IIRC when I did those sorts of tests before, the GPU temperature 
dropped rapidly very quickly (eg 10-15 degrees in just a second after 
exiting a game).

> To me, that sounds rather damned hot. Is it usual for a GPU to get this 
> warm?

Actual chip core temperature of 55 degrees seems fine to me, if it got to 80 
or 90 I would start to get worried.

> At this point, I'm not sure if my GPU is failing, or whether I just need 
> to update my drivers...

One very easy and cheap way to check...

Post a reply to this message

>> I see three possibilities:
>>
>> 1. The GPU is overheating.
>>
>> 2. There's a bug in the version of the video driver I have installed.
>>
>> 3. Valve have made an update to TF2 which introduces some kind of 
>> rendering bug.
> 
> 4. Hardware fault on the graphics card (eg bad RAM)

Surely that's a rather rare failure mode on a graphics card? That seems 
to me more like the kind of thing where you get a brand new card and 
it's DOA. This card has been installed and running without issue for years.

>> Apparently the nVidia control panel no longer tells you your GPU 
>> temperature; you have to download and install an additional tool. 
>> Anyway, having just stopped using the GPU, the temperature was reading 
>> 55 °C. Over the course of the last 5 minutes, that's fallen to 53°C.
> 
> Try showing the temperature in parallel with running the GPU (eg 3D game 
> in a window), IIRC when I did those sorts of tests before, the GPU 
> temperature dropped rapidly very quickly (eg 10-15 degrees in just a 
> second after exiting a game).

While idling, it got down to about 53°C. I ran nVidia's "GPU stress 
test" and it climbed as high as 63°C. Didn't seem to go any higher.

>> To me, that sounds rather damned hot. Is it usual for a GPU to get 
>> this warm?
> 
> Actual chip core temperature of 55 degrees seems fine to me, if it got 
> to 80 or 90 I would start to get worried.

Oh, OK. Generally my CPU doesn't get anywhere near that hot - but then, 
I guess my CPU doesn't develop several hundred GFLOPS...

>> At this point, I'm not sure if my GPU is failing, or whether I just 
>> need to update my drivers...
> 
> One very easy and cheap way to check...

Well, I did it anyway, just for the hell of it. Next time I play, I'll 
see if it makes any difference.

(The wonderful thing about intermittent faults is that you can never 
truly know if they are fixed.)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

>>> Shutdown. Let it cool for a few minutes. Open your tower, clean dust 
>>> out. And I mean *clean* dust out.
>>
>>   Get one of these too:
>> http://www.extendyourlashes.com/images/air-blower.jpg
> 
> There's also "canned air", which is just what it sounds like. A bit more 
> vigorous than a bulb for camera lens cleaning.

I own this. I'll give it a try later. If it *outside* of my PC is 
anything to go by, it's probably about time... o_O

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

> Surely that's a rather rare failure mode on a graphics card?

About the same as getting some bad RAM bits in your normal PC RAM probably.

> That seems to me more like the kind of thing where you get a brand new 
> card and it's DOA. This card has been installed and running without issue 
> for years.

I had the RAM in my desktop machine at work partially fail after some time. 
It started when dealing with big zip files, it would complain that the CRC 
check didn't match up, even though I had just downloaded it.  Then it would 
blue-screen randomly, and gradually got more frequent.  I did the memcheck86 
or whatever it was and there was an area of RAM that was totally corrupt, 
it's a wonder Windows even booted, let alone worked normally most of the 
time.  Replaced the RAM and very reliable now.

I had the same with my nVidia Ti4200 graphics card, actually identical 
symptoms to you, random triangles flickering across the screen.  Gradually 
got worse over time, and also while playing a game until it crashed.  Tried 
it in my sisters PC and the same symptoms.  I didn't bother trying to figure 
out the cause, but just assumed it was bad RAM.

Post a reply to this message

>> Surely that's a rather rare failure mode on a graphics card?
> 
> About the same as getting some bad RAM bits in your normal PC RAM probably.

Well, you'd think. ;-)

I've bought brand new RAM sticks to find that they're defective, and 
I've occasionally damaged one inserting or removing it, but never seen 
one just fail during normal use. They seem to be quite reliable.

> I had the RAM in my desktop machine at work partially fail after some 
> time. It started when dealing with big zip files, it would complain that 
> the CRC check didn't match up, even though I had just downloaded it.  
> Then it would blue-screen randomly, and gradually got more frequent.  I 
> did the memcheck86 or whatever it was and there was an area of RAM that 
> was totally corrupt, it's a wonder Windows even booted, let alone worked 
> normally most of the time.  Replaced the RAM and very reliable now.

Yeah, RAM failures cause weird crap to happen. Depending on where in RAM 
the problem is, it often only causes problems when you try to do certain 
things. (But what with virtual memory, it's quite hard to predict when 
it will or won't show up a problem.) Diagnosing RAM faults without a 
real RAM tester is just Not Fun...

> I had the same with my nVidia Ti4200 graphics card, actually identical 
> symptoms to you, random triangles flickering across the screen.  
> Gradually got worse over time, and also while playing a game until it 
> crashed.  Tried it in my sisters PC and the same symptoms.  I didn't 
> bother trying to figure out the cause, but just assumed it was bad RAM.

Well, since the RAM isn't replaceable on a graphics card, I guess it 
really just boils down to "does this graphics card work properly or not?"

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

On Mon, 18 Aug 2008 15:43:22 +0200, Invisible <voi### [at] devnull> wrote:
>>  4. Hardware fault on the graphics card (eg bad RAM)
>
> Surely that's a rather rare failure mode on a graphics card?

Not if it has been overheating.


> While idling, it got down to about 53°C. I ran nVidia's "GPU stress  
> test" and it climbed as high as 63°C. Didn't seem to go any higher.

Have you checked the temperature of the video RAM?



-- 
FE

Post a reply to this message

>>>  4. Hardware fault on the graphics card (eg bad RAM)
>>
>> Surely that's a rather rare failure mode on a graphics card?
> 
> Not if it has been overheating.

True, but since the GPU generates far more heat than a block of RAM does...

>> While idling, it got down to about 53°C. I ran nVidia's "GPU stress 
>> test" and it climbed as high as 63°C. Didn't seem to go any higher.
> 
> Have you checked the temperature of the video RAM?

I don't have a way to do that.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

On Mon, 18 Aug 2008 17:25:30 +0200, Invisible <voi### [at] devnull> wrote:
>>>>  4. Hardware fault on the graphics card (eg bad RAM)
>>>
>>> Surely that's a rather rare failure mode on a graphics card?
>>  Not if it has been overheating.
>
> True, but since the GPU generates far more heat than a block of RAM  
> does...

The GPU also has far better cooling. Modern high-performance RAM can get  
rather hot, particularly video RAM.



>>> While idling, it got down to about 53°C. I ran nVidia's "GPU stress  
>>> test" and it climbed as high as 63°C. Didn't seem to go any higher.
>>  Have you checked the temperature of the video RAM?
>
> I don't have a way to do that.

Try touching a VRAM chip (very briefly) while the game is running.



-- 
FE

Post a reply to this message

>> True, but since the GPU generates far more heat than a block of RAM 
>> does...
> 
> The GPU also has far better cooling. Modern high-performance RAM can get 
> rather hot, particularly video RAM.

Well, if they don't bother fitting cooling to a component, that 
presumably means they don't think it's necessary. ;-)

> Try touching a VRAM chip (very briefly) while the game is running.

Wouldn't simply having the case open radically alter the operating 
temperature of the components inside? ;-)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

scott wrote:
> Actual chip core temperature of 55 degrees seems fine to me,

Compare it to your CPU temperature. I dunno about my GPU, but my CPU is 
close to 55 just sitting idle.


-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message

On Mon, 18 Aug 2008 17:39:57 +0200, Invisible <voi### [at] devnull> wrote:
>>> True, but since the GPU generates far more heat than a block of RAM  
>>> does...
>>  The GPU also has far better cooling. Modern high-performance RAM can  
>> get rather hot, particularly video RAM.
>
> Well, if they don't bother fitting cooling to a component, that  
> presumably means they don't think it's necessary. ;-)

It can also mean that they chose not to include VRAM coolers because it is  
cheaper that way. Whether they are necessary or not depends on many  
factors outside the control of video card designers. Also, simply having a  
cooler is not the same as having an effective cooler.



>> Try touching a VRAM chip (very briefly) while the game is running.
>
> Wouldn't simply having the case open radically alter the operating  
> temperature of the components inside? ;-)

It can change the temperature, but not radically, unless you are talking  
about the video card casing. In any event I doubt you could get an  
accurate temperature reading using only your fingertip under any  
circumstances. The point of the exercise is simply to get a feel for how  
hot those things get.


-- 
FE

Post a reply to this message

>> Actual chip core temperature of 55 degrees seems fine to me,
> 
> Compare it to your CPU temperature. I dunno about my GPU, but my CPU is 
> close to 55 just sitting idle.

Really? Mine is usually about 30°C or so...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

>>> Try touching a VRAM chip (very briefly) while the game is running.
>>
>> Wouldn't simply having the case open radically alter the operating 
>> temperature of the components inside? ;-)
> 
> It can change the temperature, but not radically, unless you are talking 
> about the video card casing. In any event I doubt you could get an 
> accurate temperature reading using only your fingertip under any 
> circumstances. The point of the exercise is simply to get a feel for how 
> hot those things get.

Ooo... Maybe I should use my IR thermometer? :-D

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:

>> Try touching a VRAM chip (very briefly) while the game is running.
> 
> Wouldn't simply having the case open radically alter the operating 
> temperature of the components inside? ;-)

Also, I've read somewhere that that is a very bad thing to do. (Touch an 
operating chip to check its temperature) it has something to do with 
heat saturation at the point of contact and can cause a chip that was 
operating within its parameters to actually overheat.

Of course, that's just hearsay...

Post a reply to this message

Invisible wrote:

> 
> Ooo... Maybe I should use my IR thermometer? :-D
> 

Now you're thinking :)

Post a reply to this message

Chambers wrote:
> Very hot to touch, not so hot to run.  IIRC, most chips are fine up to
> 65-70C, and several can go higher w/o a problem.

My Pentium IV used to be at 72C constantly (full load; I never let a CPU
idle :P). According to specs, 75C is the limit.

Post a reply to this message

>> Compare it to your CPU temperature. I dunno about my GPU, but my CPU is 
>> close to 55 just sitting idle.
>
> Really? Mine is usually about 30°C or so...

Where is that temperature measured though?

Post a reply to this message

>> Really? Mine is usually about 30°C or so...
> 
> Where is that temperature measured though?

I don't know. I assume it's an on-die heat sensor. AFAIK, most modern 
CPUs have them, so that they can turn up the fan speed or go into lower 
power states if the die temperature gets too high...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

>>> Apparently the nVidia control panel no longer tells you your GPU 
>>> temperature; you have to download and install an additional tool. 
>>> Anyway, having just stopped using the GPU, the temperature was 
>>> reading 55 °C. Over the course of the last 5 minutes, that's fallen 
>>> to 53°C.
>>
>> Try showing the temperature in parallel with running the GPU (eg 3D 
>> game in a window), IIRC when I did those sorts of tests before, the 
>> GPU temperature dropped rapidly very quickly (eg 10-15 degrees in just 
>> a second after exiting a game).
> 
> While idling, it got down to about 53°C. I ran nVidia's "GPU stress 
> test" and it climbed as high as 63°C. Didn't seem to go any higher.

OK, I was just playing TF2 and noticed a few stray polygons. (I.e., the 
driver update hasn't fixed this.) When I quit the game, I opened up the 
temperature monitor. It was reading 70°C (and dropping quickly).

Next time I play TF2, I must remember to open the monitor *first*. ;-)

I wonder why it gets so much hotter running a real game (which is 
optimised to run smoothly) than running a stress test (which is designed 
to use up as much GPU resources as possible)...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

> Next time I play TF2, I must remember to open the monitor *first*. ;-)

Can't you play TF2 in windowed mode?  Or download some game demo that can be 
played in windowed mode.

> I wonder why it gets so much hotter running a real game (which is 
> optimised to run smoothly) than running a stress test (which is designed 
> to use up as much GPU resources as possible)...

I guess the stress test wasn't written very well, or without the knowledge 
of your particular card.  Lots of things in the GPU can be the bottleneck, a 
well-written stress test should work this out and adjust the test details so 
that every part is almost at maximum throughput.  If it's not very well 
written, then - for example - the vertex shader might be loaded at 100% 
while the pixel shader and texture lookup units are sitting idle for 90% of 
the time.

BTW what is the CPU load during the stress test and TF2?  Maybe that is 
affecting the GPU temperatures?

Post a reply to this message

scott wrote:
>> Next time I play TF2, I must remember to open the monitor *first*. ;-)
> 
> Can't you play TF2 in windowed mode?  Or download some game demo that 
> can be played in windowed mode.

I guess I could - but it would be very hard to play it and enjoy it.

FWIW, last time I played, I did remember to start up the monitor. I've 
done this a few times, and the temperature never seems to go abve 70°C. 
Which sounds pretty damned hot to me, but hey...

>> I wonder why it gets so much hotter running a real game (which is 
>> optimised to run smoothly) than running a stress test (which is 
>> designed to use up as much GPU resources as possible)...
> 
> I guess the stress test wasn't written very well, or without the 
> knowledge of your particular card.

Given that it's an nVidia GPU running a stress test written by nVidia, 
you'd think so. ;-)

> BTW what is the CPU load during the stress test and TF2?  Maybe that is 
> affecting the GPU temperatures?

AMD Athlon64 X2 4200+ socket 939.

(When I built the system, everybody said that socket 939 was going to be 
the future. Now it appears that actually it's AM2... oh well!)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Orchid XP v8 wrote:
>> BTW what is the CPU load during the stress test and TF2?  Maybe that is
>> affecting the GPU temperatures?
> 
> AMD Athlon64 X2 4200+ socket 939.
> 
> (When I built the system, everybody said that socket 939 was going to be
> the future. Now it appears that actually it's AM2... oh well!)

That's the CPU model; and the load while playing the game? :)

Post a reply to this message

>> Can't you play TF2 in windowed mode?  Or download some game demo that 
>> can be played in windowed mode.
> 
> I guess I could - but it would be very hard to play it and enjoy it.

I meant just so you could see the temperatures while playing.

Post a reply to this message

Orchid XP v8 wrote:
> (When I built the system, everybody said that socket 939 was going to be 
> the future. Now it appears that actually it's AM2... oh well!)

Yes, and a lot of people were p***ed off about that...

...Chambers

Post a reply to this message

Chambers wrote:
> Orchid XP v8 wrote:
>> (When I built the system, everybody said that socket 939 was going to 
>> be the future. Now it appears that actually it's AM2... oh well!)
> 
> Yes, and a lot of people were p***ed off about that...

AMD claim to have learned their lesson there. Certainly the new AM2+ 
chips all work in socket AM2 motherboards (although obviously not as 
well as they could do).

But then, I heard people saying things like "this could be the end for 
AMD" (referring to the way Phenom pales in comparison to Core 2).

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:
> But then, I heard people saying things like "this could be the end for 
> AMD" (referring to the way Phenom pales in comparison to Core 2).

I doubt it.  A company the size of AMD doesn't just disappear overnight, 
no matter how hard times are.

They've got time to get their act together.

...Chambers

Post a reply to this message

>> But then, I heard people saying things like "this could be the end for 
>> AMD" (referring to the way Phenom pales in comparison to Core 2).
> 
> I doubt it.  A company the size of AMD doesn't just disappear overnight, 
> no matter how hard times are.
> 
> They've got time to get their act together.

Heh. I guess "oh noes, this is teh endzors!" just makes better headlines 
or something. ;-)

They just bought ATi, so they can't be doing that badly. (Or, 
alternatively, ATi were doing *really* badly!)

That put me in a bit of a spot. I always buy AMD CPUs, and I always guy 
nVidia GPUs. So AMD buying ATi was a problem. Of course, currently AMD 
is nowhere in terms of CPU performance, so if I were to upgrade sometime 
soon, I'd have to go with Intel...

(Now and then I get these crazy ideas that I might buy a Mac or 
something... This usually wears off when I look at the price list.)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message