Can GPUs degrade/get hotter with age

[UltimaBoB]

I've had the following computer for 16 months:

Antec P180 Case
Enermax 550w PSU
Intel E6700 CPU
2 x HIS ATI 1950XTX GPU in Corssfire
Soundblaster X-Fi Extreme Music
ASUS P5W DH Deluxe

For 13 months or so it worked fine. But now it seems to have gone to hell. The temps started rising and the CPU seemed to slow down. When I installed games, old games, the games would tell me my CPU did not meet the min requirements. Every program would take about 5 to 6 times its normal CPU load.

The tech I had come in told me that the GPUs were too hot and that I could not run both of them. He took one out and everything is fixed.

But how can there have been no problem for 13 months and then a problem?

BTW i didn't do anything to help with heat. I installed no extra fans, no thermal paste, no 3rd party fan speed programs etc... All my coolers are stock. The tech says that two 1950XTXs were too hot to have run without taking extra heat precautions - that I had no margin for error - when the system was first built. So it ran okay for awhile, but over time stuff degrades and now it is just too hot without adding cooling etc...

Does this make any sense? Should I have been adding cooling from the beginning?

I want to put my second card back in. But I don't really want to add fans or thermal paste or do anything complex for heat management. Are 1950XTXs hot cards? Should I just get a single new card then and that should solve heat issues and provide comparable power - say a 8800 GTS G92?

 

[tvdang7]

dust?

 

[Fallen Kell]

Yes, the cooling power can degrade. You have a moving part in the cooling, the fan. You also have to deal with that moving part sucking in dust, dirt, hair, etc., which wraps around the motor and causes it to need to exert more force to maintain the speed of the fan, but in reality, the fan's motor and controller don't know this, since it is probably not RPM and power monitoring the motor to maintain a constant RPM, but just has certain voltage set-points for the thermals (i.e. from 0-30c voltage=x, from 30-35c voltage=y, etc., and not 0-30c RPM=x, 30-35c RPM=y, etc.).

So, yes cooling devices can start to degrade over time. This is true for any moving parts. It is also why most people recommend taking apart the computer every few months and dusting it out and check the fans that they are still working/moving freely.

As for your issues, well, I don't think you have a choice. If the thermals were that close to the limit that a few months of use and a little degradation of system/component fans would cause the entire internals of the computer to hit their thermal limits, then you NEED more cooling. You don't want to be pushing that limit, it shortens the life of the components.

Now a graphics card upgrade might be something you want to consider. The 1950XTX's are starting to show their age. They also produce as much heat as a 8800GTX . The G92 GT or GTS will both produce less heat in a single card than the 1950XTX does in a single card, and will perform about as fast (or faster in some cases) then the dual crossfire setup.

 

[UltimaBoB]

Everything is super clean now - no dust. A tech thoroughly cleaned everything and took it apart. But he didn't check fans or anything beyond cleaning.;

But still the problems persist. I'm a computer newb working in Korea dealing with Korean speaking techs. My "translator" knows zero about computers. So the communication gap is what is killing me. It is hard to know if the tech I am using is an idiot or not.

Why would my set-up work fine for 12 months and suddenly not work?

Could fans and barings have degraded? Thermal compound degraded?

So you are saying that the original tech was right that while I did have enough cooling with all stock components to get by - there was no marin for error? As soon as stuff degraded and cooling dropped i hit the red zone and my CPU throttled?

Trying to communicate adding fans and complex cooling solutions might be hard. I have the money, what about ditching the two 1950s for a single nvdia? Would that cool things down enough you think seeing as a single 1950 seems to be running fine?

Or given that I have run the comp at two high a temp fopr awhile, are there some things that simply have to be done? A friend at home said communication gap or not, after running at overheat for so long I simply have to check the heatsink, thermal paste (and reapply new stuff), and fan barings for my CPU fan. He said stuff could be degrading and even if the single GPU seems to be working okay now, heat could still rise if mainteance isnt done and throttle me with only a single GPU.

Does that make sense? Do I need someone who knows what he is doing to look at all my fans and heatsinks and thrmal compunds even if I don't add more? I hate complicated crap lol

 

[CP5670]

I think you may just need to clean out the dust in the heatsinks on those cards. Those X1950 coolers get filled up with it pretty quickly.

[edit] never mind, didn't see your last post before I submitted this.

 

[v8envy]

X1800s and X1900s also tended to die an early age from constantly running at > 85C (standard operating temperature if you didn't crank the fan). I know my card started artifacting at 'normal operating temperature' after 11 months, and I wasn't alone.

But in your case I'm guessing your PSU isn't long for the world, and was undervolting everything causing the extra heat/performance weirdness. Removing 10 amps from the 12v rail(s) and the additional heat from inside the case gave it a new lease on life.

Like all other components PSUs wear out and become less efficient. As they get less efficient they generate more heat. Which causes them to be even less efficient and to wear out faster... Unfortunately I'm not familiar enough with the 550w Enermax to know whether it's a quality unit or a rebadged 300 watter.

 

[UltimaBoB]

The tech said that he tested every part separately and they all work. They just won't work together anymore - must be heat.

I'm thinking that complex or not, it is probably beter to add cooling and get my two 1950s in there and working rather than going to a G92 since won't that be a lot of money for something that benchmarks in the same area as crossfired 1950s? And extra cooling never hurts - so maybe best to just add the cooling and use what I have and upgrade later when a really sweet new card comes out?

What about getting a bigger full-size tower? I have the room? Would that help? Is the P-180 too small? Am I going to have trouble getting new cards onto the mobo in a P180 later on?

 

[chizow]

Temp and seasonal changes can certainly impact OC'ed rigs. I tend to notice hotter ambient temps in the summer and dead of winter when the heat is blasting. Have you noticed any changes in ambient temps? Can try to increase cooling by adding a side panel fan or something, or running the side panel off and using a house fan to simulate better cooling to see if that's the problem.

I'd also take a look at your PSU. 550W seems to be cutting it kinda close with that rig and 2 1950XTs. If your PSU is going flaky and not providing enough power under load that could certainly impact performance and cause some throttling. Pulling one card and seeing the slow-down problems go away point more in this direction.

 

[UltimaBoB]

PSU could have been an issue too?

The sound card also seemed to be an issue. The original builder had it sandwiched between trhe 1950s for some reason. And it kept having issues.

Could the dirty truth be that if you play with high level stuff, you simply have to go beyond the stock cooling offerings?? Do you always need to get better fans and coolers, and extra ones, and use fan monitoring programs etc... if you want to use the high level stuff?

Is it possible that lots of my parts have been damaged because I did not invest the time to learn about heat management after I bough what was at the time a top top end machine? I had all my gear in Oct 1 of 2006.

Complex or not, I'm starting to feel that I need to learn about fan speeds, monitoring heat, cooling, etc... if I want high end rigs - even if I don;t overclock. And I'm starting to fear that all of my computer parts may be damaged to some degree with no 100% for sure way to find out. This kinda sucks.

 

[UltimaBoB]

BTW nothing was ever overclocked- my GPUS were a bit factory overclocked.

Oh yeah and my PSU was in fact a 620 watt Enermax Liberty PSU - considered good at the time.

 

[v8envy]

Thermal management and reliable power are the first lessons a computer enthusiast learns, sooner or later. That way lie many an obsessive-compulsive disorder, as well.

The sound card issues may have been IRQ conflicts. I know many AGP motherboards used to share the AGP slot's interrupt line with the PCI slot immediately below (read: they shared a physical IRQ, and no amount of bios twiddling would change that) and that may still be the case with PCIe. I could be full of crap here though since I know approximately 0 about PCIe (never had to troubleshoot a PCIe equipeed board). Creative Labs cards never liked to share, so that may explain why the sound card flaked.

 

[Arcanedeath]

The P180 has horrible GPU cooling in the upper cage you need to put a 120mm intake fan infront of the HDD' in the top bay (a 1200+ rpm one) as the grill is very restrictive w/out this I very much doubt you'll be able to do x1900 crossfire w/ reasonable temps a 1600rpm + fan in that front slot would be even better but may be to noisy, that should solve your overheating issue w/ the P180 (I had to do this w/ a single 8800GTS 640mb or it got to insane temps) P180 is great for CPU cooling but subpar for GPU cooling.

 

[UltimaBoB]

If I were to get a new case to make things easier, what would you recommend?

The thing is, I just want high end performance - I never wanted to be an "enthusiast" and know what went on inside the box. This may not be an option however lol

 

[Fallen Kell]

I think most of it was probably heat related. I forget the tool to use (Throttle Watch?), but you can check to see if the Intel CPU is throttling its power due to heat. That would prove to you that heat is your problem in the system (at least for CPU slowness from thermal issues).

I am surprised that in the P180 case that you have this issue. I would really think that one or more fan has possibly died or was never connected. Do you have the 80mm fan connected that goes over the expansion slots and pulls in cool air over the video cards? Look halfway down this link if you don't know what I am talking about: http://www.silentpcreview.com/article249-page6.html

 

[UltimaBoB]

At the corner between the top of the case and the back there are two fans. One at the back of the top and the other at the top of the back.

There is a fan on the CPU and then there is another small fan, very small, coming out of the mother board at a 90 degree anle right behind the CPU. Weid, no idea what it is.

Then there is a PSU fan.

There are no fans in the front of the case.

Dust is gone and everything has been cleaned - or so I have been told. But I don't know how good a job he did or didn't do.

The wires and cords seem to be managed from the initial build - but from an amateur's pov don't seem to be managed ideally and may be messy and interfere with airflow.

There are NO fans whatsoever in the front of the case. No drive fans were added despite most of the drive slots being empty and covered up (only single DVD burner/player and a 150 GB Raptor HDD).