Turning up the heat on GPUs
Posted on 5th Aug 2009 at 11:18 by Mark Mackay with 47 comments
The first graphics card I ever bought as a stand-alone purchase was an ATI Radeon 9800 Pro with an Arctic Cooling VGA Silencer bolted to it. The game that I was oh-so-thoroughly addicted to at the time was one of my first PC games, and my first MMO, Asheron’s Call 2.
The effects and textures of AC2 were exceptionally advanced, especially for an MMO and especially for one released that many years ago. When you cranked up the detail settings and AA, it was very demanding on the hardware of the day.
The jump in visual splendiferousness from the crappy low-end Nvidia GPU that I'd been ripped off for by my local PC store was immense. It looked like a completely different game. However after about 10 minutes of gaming, my PC would BSOD and restart itself, telling me that it had recovered from a serious error.
After some browsing forums and experimenting with different drivers I decided that the card must be broken. So when RMAing it I thought: ‘Screw it, I’m gonna spend even more money’ and managed to obtain the illusive Radeon X800 XT Platinum Edition. So, I had my shiny new £260 graphics card all set up. I fired up AC2 and low and behold, the same thing happened.
I got the notion that maybe it was overheating, as turning down the settings a little solved the problem. So I bought myself a 12in desk fan and pointed directly into my case with the side panel off. Problem solved. It never shut down and restarted when settings were maxed out again.
Over the years I’ve owned a Radeon X1900XT and a Radeon HD 4870 and both have, at times, shut down on me when I've been playing games with absolutely maximum settings.
It's happened even when the cards have had aftermarket coolers and not been overclocked, which makes it all the more difficult to work out why I've had such a bad run. Have other people had similar problems with overheating or have I just been unlucky with my graphics cards?
The effects and textures of AC2 were exceptionally advanced, especially for an MMO and especially for one released that many years ago. When you cranked up the detail settings and AA, it was very demanding on the hardware of the day.
The jump in visual splendiferousness from the crappy low-end Nvidia GPU that I'd been ripped off for by my local PC store was immense. It looked like a completely different game. However after about 10 minutes of gaming, my PC would BSOD and restart itself, telling me that it had recovered from a serious error.
After some browsing forums and experimenting with different drivers I decided that the card must be broken. So when RMAing it I thought: ‘Screw it, I’m gonna spend even more money’ and managed to obtain the illusive Radeon X800 XT Platinum Edition. So, I had my shiny new £260 graphics card all set up. I fired up AC2 and low and behold, the same thing happened.
I got the notion that maybe it was overheating, as turning down the settings a little solved the problem. So I bought myself a 12in desk fan and pointed directly into my case with the side panel off. Problem solved. It never shut down and restarted when settings were maxed out again.
Over the years I’ve owned a Radeon X1900XT and a Radeon HD 4870 and both have, at times, shut down on me when I've been playing games with absolutely maximum settings.
It's happened even when the cards have had aftermarket coolers and not been overclocked, which makes it all the more difficult to work out why I've had such a bad run. Have other people had similar problems with overheating or have I just been unlucky with my graphics cards?
47 Comments
Discuss in the forums ReplyAnd props to you ever getting an X800XT PE, those things were amazing.
I've never really had a problem with GPUs overheating personally, well mainly because most of my cases has good ventilation and I only did a mild overclock. To this day I've ran most games on max settings and the GPU's been fine, of course this was post 2006 and it was when I finally got a gamer card of my own(X1950XT), it worked beautifully.
8800GTX feels close to overheating and with vents INSIDE the case, it cooked my USB/Firewire backplate. gtx260 had been fantastic, zero issues.
with my experience with 8800GTX, i wrote a guide on how to adjust fan speeds. it uses Rivatuner so should also work on ATI cards
http://www.overclock.net/nvidia/273072-how-set-auto-fan-profile-most.html#post3140251
I do kinda regret getting my HD4850 though, the 9800GTX+ was only marginally more expensive at the time..
And the HD4870, a tad more...
That's interesting. I cant imaging the through-flow was particularly great in the Dell case either.
Yes, I had thought about this and tried cooling the Northbridge independently to find out. I tried bolting whiny little fans to it and also aftermarket heatsinks. I'm confident that it has been the GPUs at fault.
Of course, ever since I built my Athlon XP rig, I've been running some fairly high-airflow cases. A Thermaltake Xaser (that thing has so many fans in it! Pity they're all 80mm...) and then the Antec P180.
(NOT IN USE) Geforce 2 mx 64mb
(NOT IN USE) Geforce 4 mx 64mb (overheated only once)
(NOT IN USE) Geforce 5700 LE 256mb (ran hot, so changed out the stock cooler for a BFG cooler)
(DIED due to OC) Geforce 6800GT 256mb, (ripped off the stock cooling with in a week as the pissy fan annoyed me to hell, arctic cooling nv silencer 5)
(NOT IN USE) Geforce 7600GS 256mb, stock cooling replaced for same reason as 6800GT
(SOLD) Geforce 8800GTS 320mb, kept stock cooling, ran fine but it killed a DFI NF4 motherboard due to overheating the northbridge.
(IN USE) Geforce GTX280 runs cool, but the fan is starting to get to me...
(IN USE)Quadro FX 3450, stock fan ripped off, waterblock fitted for silence
(IN USE)Radeon X800XT PE, still got stock cooling for now.
So one hell of a list, and never really had cooling issues however only a few keep the stock cooling solution..
For you it's ATi cards... for me it's nVidia - I tried an 8800GTX and two 8800GT's before the third 8800GT worked. The other three were DOA. :(
What a piece of trash that was! But back then i didn't know any better. Also, that card works to this very day while all my ATi cards died way before i decided to upgrade them. So yeah, i'm pretty much sticking with Nvidia. Gainward at that.
GeForce2 GTS 64MB, GeForce FX5900 256MB, Radeon 9800XT 512MB, Radeon X800 PRO 512MB, 8800GT 640MB, 8800GT 512MB, 9600GT 512MB, GTX 275 896MB. All of these with stock coolers and worked hard.
My one bone of contention was a Sapphire motherboard with an overheating northbridge. I tried desk fans into the case, a new case, and several other methods before finally bodging a 40mm fan right on top of the block. Needless to say, it worked and the instability went away. Then I ditched the board...
I have a P182 case, so there isn't much of an airflow probem, but I recently had to drop back from a 4850 to a 9600GT because I just couldn't get the ATI card stable.
However, about 10 months later or so the card died literally from one moment to the next. I left the room for 5 minutes and when I came back the screen was dark. Turns out that at one point the fan connector got loose so the card went on to only be passively cooled. I didn't have a clue anything was amiss, since thermal probes on ATi cards were first introduced as of the 9800*XT*.
That thing was above 80C and the friggin card was a low end one.
I never had one die on me altogether before, though sometimes I wish it would so I could use it as an excuse to get a new one =)
the only one of them I've had issues with has been the x1650pro. from day one I had trouble getting catalyst installed and the card would overheat sometimes (it had a stupidly small heatsink with a very thin 40mm fan on it). the fan eventually ceased right up and now takes an indecent amount of force to rotate, but a 92mm fan sitting over the heatsink fixed that.
my x1950 has been perfect though, with a zalman vf1000 cooler on it with the fan speed set to minimum (at max speed, it sounds like a hair dryer and I've yet to see the gpu break 45°C)
I got a X1300 Pro like that but passively cool it with a arctic cooling accelero S1 rev 2 in a HAF 932 and you'll get less than 40C constantly :D
Come to think, I might consider using that X1300 to power a second monitor. Any ideas on how to do so?
I think it's just his case is providing a dead spot in airflow where the graphics cards are.
For AAA gaming, I always recommend cases that are built for the purpose of having high airflow. Whether it is a watercooled system or not.
I am a hardcore gamer as well. However, the only time I have ever experienced a blue screen due to overheating, is when my 9800 GTX fan failed while gaming. Straight away I rebooted and checked temperatures for the GPU which was around 90 degrees even after it had time to cool down.It must have reached well over 100 degrees before the crash.
I usually recommend spacious Lian Li or Coolermaster aluminium cases with large intake air grills and more than one intake fan at the front intaking chilled air into the case. Aftermarket coolers on the CPU and GPU. I dislike cases with front doors and cases that only allow a single intake fan at the front.
I lol'ed =)
I guess it could have been, but the symptoms have been identical across so many boards and the first couple weren't overclocked. Pretty sure the GPUs were to blame.
If the problem has spanned across many different motherboards, then I would definitely have another look at things like the type of case designs you have been using, and the general airfow.
Gone with water now, no longer any problems with high temps and noise. I've also made sure there's a nice breeze through the case to avoid hotspots.
I have found all the ATI cards I've ever had run too hot and easily overheat. Users turning up their fans is a poor solution. There's also the issue of the power circuitry on stock 4870s not allowing the GPU to hit it's max TDP. With PhysX games (and a small selection of other games) I've noticed the power consumption of my GTX280 has rocketed up - what if ATI release a similiar feature, will the drivers have to downclock the GPU for those games?
My GTX280 automatically turns itself off if any part overheats (VRMs/GPU), and that's been tested more than once by the pump being too low and turning off. I wouldn't like to try that with an ATI card. It's like the P3 vs. Athlon. The P3 turns itself off, but the Athlon just burns out.
Hate to appear like I'm bashing ATI cards, since I think they're pretty good. I especially like how ATI don't boast about the GFLOPS like Sony would. But ATI really need to get a lid on the heat, their heatsink solutions, and emergency procedures should the worst happen.
But my current ATi 4870 will happily run at 80*C at load, with no signs of instability.
One time, after I found the manual fan controls, it hit 91*C at medium load and still had no troubles, but I made sure that didn't happen again.
People say the HSF with the card is loud, but it can certainly keep it cool, at 60%, only loud music/gameplay will cover it up but it can keep the card @ 770/1100 below 60*C.
Yes, it's been with various rigs over the years. The first ones I ever had admittedly lacked proper airflow but even as I got more into hardware and had better cases it was still an issue, albeit a less frequent one.
How is the house wiring, or the electrical current. Back when we were still using Via chipsets and Win98, people always claimed crashes were caused by Windows. The day I put in a quality UPS with a line conditioner, things changed. A short time later I went with a higher end board and psu as well. While most were crashing Windows I was running overclocked and as stable as Windows 2000 was for most.
Studies have shown that not only are systems more stable, but the parts last longer when there is clean, consistent power. Contrary to common thinking, power spikes, while bad, are not as common and damaging as people think and that power dips are a large cause of problems with hardware failing. Also keep in mind that if you have an ups, and it is underpowered, it can create issues for it is teh same as a power dip.
Interesting. I've never used a UPS before. Maybe I should get one.
Yup, it's perfectly possible. :) Should just be a case of popping both cards in, hooking up monitors and hitting the power button. They're both ATi cards (although since the X1300 is now considered 'legacy'... you'll be stuck running Catalyst 9.3 if you're playing with Vista...)
I'll have to remember that, thanks.
I recommend APC. They usually have better line conditioning and (much) longer lasting batteries than others. I have had some that barely lasted a year while my APC's lasted several. Also while some will claim lightening protection, APC has a much more standard/higher lightening rating.
A cheap ups is exactly that, cheap.
Thanks! Just making sure +rep.
u basically have two choices, either running the fan at speeds that makes ur case move slowly forward, or water cooling.
im still at figuring out the third solution, getting efficient quiet airflow going, but just using and aftermarket aircooler so far has been a huge mistake, letting my vrm temps rise above 140 !!! degrees...
Had a MX4000 fail on me but I had a huge overclock on it so that was more my fault...
Yeah strangely all the X1k.05 cards ran a bit hotter.
The X1550(from the X1300), the X1650, and the X1950s. Good times though, I had an X1550 almost explode on me, it literally was hot enough to burn your hands.
I can't speak for any of the 1x50 cards other than the X1950XTX... but that ran a lot cooler (and a lot quieter!) than my X1900XT. It idled around 5*C cooler than the 1900XT, and ran about 12*C cooler loaded... but it was their top-end card, and the first of ATi's favoured cooler style that didn't sound like a leafblower even at idle...