Looking to cool down a HD4870

the_syco

Hava a "1GB HD4870 [Factory Overclocked]" by Club3d (I think), and it keeps heating up. After a certain amount of time, my games becomes unstable, and crash. At first, I thought it was just drivers, but as similair crashing has happened on other games (such as Crimecraft) as well as Bad Company 2, and some games just randomly crash (Company of Heroes), I'm begining to think it's the GFX card. Apart from the crashes, I'm happy with the card, so I'm thinking of getting a new fan or cooler of some type.

Would anyone have any recommendations?

Also, the room I'm in is a converted garage, so poor ventelation. Because of this, I'm thinking of going the water cooling route, and wondering how expensive this would be? If I do go this route, I'll probably water cool the CPU as well (ah, sure why not).

cherryghost

have you adjusted the GPU fan speed?

the_syco

cherryghost wrote: »

have you adjusted the GPU fan speed?

Can't seem to It's a "factory overclocked" card, and thus any attempt to change the fan setting has had no real luck - option is grayed out in the CCC, and 3rd party tools (ATT) did have some success in XP (duel boot), but no such luck in Windows7

It seems ATI Tray Tools doesn't work in Windows 7, as the ATI Tray tools cannot connect to a "low level driver". Have read the ATT unofficial forum a while back with no luck, and skimming over it now to see has anyone been able to get it working in Windows 7, but no luck so far.

Aith ATT in XP, i was able to put the fan at 100%. Loud (could hear it hum away), but meh, the games ran okay. Need to get this sorted sooner rather than later, as i don't want random crashes during Midlans.

/edit
I can disable driver signing via F8 upon boot for a session, but doing it everytime I boot is bloody annoying.

/edit2
Info for myself - must check out "test mode" later - http://forums.guru3d.com/showpost.php?p=3520626&postcount=6 and http://forums.guru3d.com/showthread.php?t=313629

PogMoThoin

Here's a driver signature overrider, sign the driver Youyrself

Try Afterburner if that don't work

the_syco

PogMoThoin wrote: »

Here's a driver signature overrider, sign the driver Youyrself

Try Afterburner if that don't work

Didn't want to go down the route of running Windows in Test Mode, so I tried Afterburner. Seemed to allow me pop the fan up to 100, from 71. Will try this for a while.

cherryghost

What temperatures are you getting by the way? Before and after adjustments.

It just occured that if it was only Bad Company 2 that it's happening, the game has been broken since the last patch. My BC2 constantly crashes after about 20mins/half an hour

the_syco

cherryghost wrote: »

What temperatures are you getting by the way? Before and after adjustments.

It just occured that if it was only Bad Company 2 that it's happening, the game has been broken since the last patch. My BC2 constantly crashes after about 20mins/half an hour

Temps? Well, just played BC2 just there, and when the game crashed (after Fraps told me there was no more space to record games), it was 67 to 70 Celsius with the fan speed at 69%.

Random crashes also happened in CoH, and weird objects would appear in Crimecraft, after it heated up.

marco_polo

Hmm not sure heat is your problem, 70 degrees isn't excessively hot for a HD4870 and well short of what should be causing overheating crashes, fairly normal temperatures for a period under load in fact.

PogMoThoin

Try replacing thermal compound?

the_syco

marco_polo wrote: »

Hmm not sure heat is your problem, 70 degrees isn't excessively hot for a HD4870 and well short of what should be causing overheating crashes, fairly normal temperatures for a period under load in fact.

70 degrees when I have the fan on high. Also, today, the room was cool. Other days, the room gets a lot hotter, and that's when the crashes will usually happen more frequent.

PogMoThoin wrote: »

Try replacing thermal compound?

Of the GPU? Hrm. Have to say I haven't done that yet. Will get around to it shortly.

marco_polo

Ah sorry the fan speed didn't register when I was reading it .

Tails142

I have a HD4870 and it often runs at over 90 degrees C with no issues. I try to keep it lower generally but the fans on the standard model are so loud at full blast =P

the_syco

Tails142 wrote: »

I have a HD4870 and it often runs at over 90 degrees C with no issues. I try to keep it lower generally but the fans on the standard model are so loud at full blast =P

Using a factory overclocked version. Cos it's a bit warm outside, the temp is a steady 55 Celsius, with the fan on 71% - just browsing the net.

marco_polo

the_syco wrote: »

Using a factory overclocked version. Cos it's a bit warm outside, the temp is a steady 55 Celsius, with the fan on 71% - just browsing the net.

Sh*t that is not so normal alright :eek:, My Saphire 4870 holds that temperature with the fan @ about 25%-30% (it has a better fan than the reference version) which does seems to run a bit hotter (Unlike Tails142 I never see tempratures north of 80-81 degrees).

A FO version like yours with prpbably an even better fan again should certainly be in the same fan speed / temperature ballpark.

As PogMoThoin suggests I would definately pop that sucker off and reapply it.

Cuddlesworth

Have you cleaned the dust out of the heatsink?

PogMoThoin

the_syco wrote: »

Didn't want to go down the route of running Windows in Test Mode, so I tried Afterburner. Seemed to allow me pop the fan up to 100, from 71. Will try this for a while.

You didn't do it right With this app You digitally sign the ATT driver so You don't need to enable test mode

the_syco

PogMoThoin wrote: »

You didn't do it right With this app You digitally sign the ATT driver so You don't need to enable test mode

Probably not, but no time to f**k around with it atm.

Cuddlesworth wrote: »

Have you cleaned the dust out of the heatsink?

I did so a while back. Will probably do it soonish, and see how much dust is in it.

Cuddlesworth

the_syco wrote: »

I did so a while back. Will probably do it soonish, and see how much dust is in it.

I'd do that first. Blow down through the exhaust towards the fan. You might just have a matted lump of dust blocking the heat sinks channels at the start.

the_syco

Cuddlesworth wrote: »

I'd do that first. Blow down through the exhaust towards the fan. You might just have a matted lump of dust blocking the heat sinks channels at the start.

Last time I cleaned it, I had to take it apart to clean it (hover + blow air through the heatsinks).

Cuddlesworth

the_syco wrote: »

Last time I cleaned it, I had to take it apart to clean it (hover + blow air through the heatsinks).

good set of lungs will do it. To get it perfect you need compressed air or the above, but a quick blow down the exhaust will tell you how bad it is.

the_syco

Cuddlesworth wrote: »

good set of lungs will do it. To get it perfect you need compressed air or the above, but a quick blow down the exhaust will tell you how bad it is.

Aye. Compressed air is what I used.

Overheal

How dusty is it? I've since learned that my first PC caked dust because I had the fans set up wrong so it was negative-pressure: Dust particles came in, they never went out.

Is there any way this is a CPU or RAM issue?

the_syco

Overheal wrote: »

How dusty is it? I've since learned that my first PC caked dust because I had the fans set up wrong so it was negative-pressure: Dust particles came in, they never went out.

Is there any way this is a CPU or RAM issue?

Still failing. Fails every 20 seconds if playing the Crysis2 demo :eek:

Going to put it into another machine over the weekend, and if I still get the same problem, I get put through to ATI's 2nd level people ^.^

Cuddlesworth

I have the same card I think and it started crapping out on me too recently. It came pre-overclocked so I put the clocks down to normal and it seems fine now.

Overheal

now i have to ask who the hell are Club3D

the_syco

Cuddlesworth wrote: »

I have the same card I think and it started crapping out on me too recently. It came pre-overclocked so I put the clocks down to normal and it seems fine now.

What program did you use, or did you do it through BIOS somehow, or what?

the_syco

Overheal wrote: »

now i have to ask who the hell are Club3D

This is my card: http://www.club-3d.com/index.php/products/reader.en/product/hd4870-overclocked-edition.html. Seems it needs "600Watt High Quality Power Supply"... and I think I have an 800Watt PSU...

Overheal

not exactly a big player in the GPU industry. /shrug

http://www.club-3d.com/index.php/support/warranty.en.html

my next step of advice before you go voiding the warranty or anything.

Solitaire

They're not huge but still a popular vendor for nVidia cards in Europe. A bit like how a lot of people over here won't even have heard of Rosewill

Overheal

Doesnt mean I buy Rosewill either :pac:

Hellrazer

Syco-check yer pms.Have a spare card here you can borrow if youre stuck.Or even if you just want to use it to test the rest of the system.

the_syco

Hellrazer wrote: »

Syco-check yer pms.Have a spare card here you can borrow if youre stuck.Or even if you just want to use it to test the rest of the system.

May do so. I think I need a PSU with more wattage... Will give you a shout later.

PogMoThoin

Overheal wrote: »

now i have to ask who the hell are Club3D

They're all reference cards with stock Ati coolers, Club3d just put their sticker on it. I've had one of their cards before, when shopping I usually opt for the cheapest card and later add an aftermarket cooler if the stock one is noisy, usually works out cheaper than buying one with a Vapour X etc, then I do my own overclock

Overheal

Vapor-X for me worked out cheaper as a refurbished item I reckon the previous owner chucked it back when he found out the card doesnt like to overclock beyond the default OC very much.

Cuddlesworth

the_syco wrote: »

What program did you use, or did you do it through BIOS somehow, or what?

I'd forgotten about this thread, this is the program I used

http://event.msi.com/vga/afterburner/ (Works with all cards, not just MSI ones)

Default cpu and memory clocks for the 1gig model should be
Core Clock: 750MHz
Memory Clock: 3,600MHz (effective).

the_syco

Cuddlesworth wrote: »

I'd forgotten about this thread, this is the program I used

http://event.msi.com/vga/afterburner/ (Works with all cards, not just MSI ones)

Aye, I've used it in the past, to ensure the fan goes at full whack. I barely hear the fan, so I don't notice it at all when gaming, or even when watching a movie.

Cheers for the default clock speeds, etc. May readjust it if all else fails.

=-=

Have just got a 2TB drive. My setup is now 3 WD HDD's: 2TB, 1.5TB, and a 1TB. All "Cavier Green". Thinking that maybe 5 HDD's, a DVD drive, and an overclocked card on a 800Watt PSU may have been too much, so it'll now be only 3 HDD's, a DVD drive, and the overclocked card on the 800Watt PSU.

Installing Windows now, and will then copy the stuff from my old drives. Added benefit is that one of the drives houses a lot of my photos, so now I have a "backup" photo drive

Should have it all copied over by tomorrow, and then I'll have a go at the Crysis2 demo (was freezing & crashing all over the place on it before), and get back to ye.

the_syco

Update:

Got one mini crash, and a few slight freezes last night when playing BLOPS. Had fan on 100%. Usually it'd freeze, and then black screen, but this time, just a few freezes.

the_syco

Hellrazer wrote: »

Syco-check yer pms.Have a spare card here you can borrow if youre stuck.Or even if you just want to use it to test the rest of the system.

Stuck it in, had a game of BLOPS, had another black-out-crash. Very odd. As this doesn't need any extra power, I don't think it's a lack of power that is causing it.

Overheal

yeah but something something voltage rail something - hell, I don't ****ing know, Solitaire does

the_syco

Overheal wrote: »

yeah but something something voltage rail something - hell, I don't ****ing know, Solitaire does

Sent Solitaire a message, so hopefully

=-=

The crashes affect the 4670 which needs no extra power, as well as the 4870, which had two 75Watt 6-pin PCI Express power cables plugged into it.

/edit
MW2 lagging/freezing/BSOD 14/01/2010
Computer won't start 18/02/2010
Looking to cool down a HD4870 13/07/2010

This system is giving me no end of troubles, it would seem

I do wonder what the problem could be, though?

Solitaire

Did it affect the HD4670 too? Could be a bad mobo with dodgy PCIe :eek:

Did you have an equally huge array of drives back when you were using the HD4670? And what model PSU is that anyway? Its not in your sig...

the_syco

Solitaire wrote: »

Did it affect the HD4670 too? Could be a bad mobo with dodgy PCIe :eek:

Did you have an equally huge array of drives back when you were using the HD4670? And what model PSU is that anyway? Its not in your sig...

Actually, I went from the 4870 to the 4670, to see if the problem occurred with a card that took less power. Had 3 HDD's when I went from the 4870 to the 4670. So far, I have had one freeze & black screen with Black Ops when running the 4670.

The PSU is a 850Watt Coolermaster, model RS-850-ESBA.

Solitaire

Try the cards on a different mobo or a completely clean new install, that's all I could suggest. The HD4670 doesn't run too hot or draw much power and the smallest of the 12V rails on the M850 can supply at least 216W so if you can vouch for the CPU and RAM you have a dodgy mobo chipset (or possibly PSU) on your hands

Last-ditch attempt: Try cranking the chipset voltages up a few notches and see if boosting the volts to a marginal (as opposed to outright defective) PCIe controller can keep it stable under high stress...

the_syco

Solitaire wrote: »

Try the cards on a different mobo or a completely clean new install, that's all I could suggest.

Recently just reinstalled Windows 7. Yup, gotta try the card on another machine. May see if Hellrazer (whom I got the 4670 off for a loan to test teh system) could try it.

Solitaire wrote: »

The HD4670 doesn't run too hot or draw much power and the smallest of the 12V rails on the M850 can supply at least 216W so if you can vouch for the CPU and RAM you have a dodgy mobo chipset (or possibly PSU) on your hands

New mobo would be about €133, new PSU around €300-€500, so hope if it's anything, it's the mobo :S

Solitaire wrote: »

Last-ditch attempt: Try cranking the chipset voltages up a few notches and see if boosting the volts to a marginal (as opposed to outright defective) PCIe controller can keep it stable under high stress...

I take it that that is a voltage increase on the PCI-E is "PCI-E OverVoltage Control" on page 52, which I can increase 0.05V to 0.75V at 0.05V increment. How much should I increase it?

Solitaire

No idea! :pac: Try 0.15V to start with, and if you can keep an eye on southbridge temps Its really a last-ditch effort

And why on earth would a new PSU cost €300?! :eek::p

the_syco

Solitaire wrote: »

And why on earth would a new PSU cost €300?! :eek::p

Looking at prices. I think I was looking at the wrong site Oh, and I think I was looking at the 1200Watt PSUs, thinking I needed moar poweh

Solitaire

Not unless you want to replace the HD4870 with three GTX480s You'd probably need €90-150 worth of PSU including delivery, max

the_syco

Solitaire wrote: »

No idea! :pac: Try 0.15V to start with, and if you can keep an eye on southbridge temps Its really a last-ditch effort

Increased by 0.15V, temp went up 1 or 2 degrees, I think, and still suffered a freeze/black screen when playing BLOPS.

Is there any way that I could rule the PSU out as the problem? Digital Voltmeter, or something?

As the current rig cost me about €800 back December 2009, I don't want to do another cpu/mobo/ram/psu/etc upgrade anytime soon.

As I have a full case, I may as well go for the full ATX motherboards. Will keep the socket 775. The PC2-8500 DDR2 RAM, though, may be replaced, as can't seem to find many LGS775 mobos that take PC2-8500 DDR2 RAM.

Can't seem to find any X48 boards, but I'm finding G31, G41 and P965 boards on hardwareversand.de and G41 boards on elara.ie Would probably prefer elara.ie or some Irish site, as HWV charges €30 for delivery. And no, I'm not going to touch Pixmania with a f**king bargepole.

I think I'll replace the mobo first, and then the PSU, as the mobo is the "cheap" part of the required parts replacement.

bonkey

Have you tried running something like Prime95 in Torture Mode to make sure that its not something else failing under load?

It'd be a real bummer to go through all of this and discovered that the problem was in yoru CPU or memory, but only showing up when games put it under sufficient load....

the_syco

bonkey wrote: »

Have you tried running something like Prime95 in Torture Mode to make sure that its not something else failing under load?

It'd be a real bummer to go through all of this and discovered that the problem was in yoru CPU or memory, but only showing up when games put it under sufficient load....

Garbbed the 64bit version. Gonna leave it run whilst I go drinking

/edit

[Fri Apr 01 20:02:41 2011]
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

[Apr 1 20:02] Worker starting
[Apr 1 20:02] Setting affinity to run worker on logical CPU #0
[Apr 1 20:02] Beginning a continuous self-test to check your computer.
[Apr 1 20:02] Please read stress.txt. Choose Test/Stop to end this test.
[Apr 1 20:02] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.
[Apr 1 20:03] FATAL ERROR: Rounding was 0.5, expected less than 0.4
[Apr 1 20:03] Hardware failure detected, consult stress.txt file.
[Apr 1 20:03] Torture Test completed 0 tests in 0 minutes - 1 errors, 0 warnings.
[Apr 1 20:03] Worker stopped.

[Apr 1 20:02] Worker starting
[Apr 1 20:02] Setting affinity to run worker on logical CPU #1
[Apr 1 20:02] Beginning a continuous self-test to check your computer.
[Apr 1 20:02] Please read stress.txt. Choose Test/Stop to end this test.
[Apr 1 20:02] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.
[Apr 1 20:02] FATAL ERROR: Rounding was 0.5, expected less than 0.4
[Apr 1 20:02] Hardware failure detected, consult stress.txt file.
[Apr 1 20:02] Torture Test completed 0 tests in 0 minutes - 1 errors, 0 warnings.
[Apr 1 20:02] Worker stopped.

The other two "workers" were error free.

[Apr 1 20:02] Worker starting
[Apr 1 20:02] Setting affinity to run worker on logical CPU #2
[Apr 1 20:02] Beginning a continuous self-test to check your computer.
[Apr 1 20:02] Please read stress.txt. Choose Test/Stop to end this test.
[Apr 1 20:02] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.
[Apr 1 20:04] Test 2, 4000 Lucas-Lehmer iterations of M19922943 using FFT length 1024K.
[Apr 1 20:05] Test 3, 4000 Lucas-Lehmer iterations of M19374367 using FFT length 1024K.
[Apr 1 20:07] Test 4, 4000 Lucas-Lehmer iterations of M19174369 using FFT length 1024K.
[Apr 1 20:08] Test 5, 4000 Lucas-Lehmer iterations of M18874369 using FFT length 1024K.
[Apr 1 20:10] Test 6, 4000 Lucas-Lehmer iterations of M18874367 using FFT length 1024K.
[Apr 1 20:11] Test 7, 4000 Lucas-Lehmer iterations of M18474367 using FFT length 1024K.
[Apr 1 20:13] Test 8, 4000 Lucas-Lehmer iterations of M18274367 using FFT length 1024K.
[Apr 1 20:14] Test 9, 4000 Lucas-Lehmer iterations of M18274369 using FFT length 1024K.
[Apr 1 20:14] Torture Test completed 8 tests in 11 minutes - 0 errors, 0 warnings.
[Apr 1 20:14] Worker stopped.

[Apr 1 20:02] Worker starting
[Apr 1 20:02] Setting affinity to run worker on logical CPU #3
[Apr 1 20:02] Beginning a continuous self-test to check your computer.
[Apr 1 20:02] Please read stress.txt. Choose Test/Stop to end this test.
[Apr 1 20:02] Test 1, 4000 Lucas-Lehmer iterations of M19922945 using FFT length 1024K.
[Apr 1 20:04] Test 2, 4000 Lucas-Lehmer iterations of M19922943 using FFT length 1024K.
[Apr 1 20:05] Test 3, 4000 Lucas-Lehmer iterations of M19374367 using FFT length 1024K.
[Apr 1 20:07] Test 4, 4000 Lucas-Lehmer iterations of M19174369 using FFT length 1024K.
[Apr 1 20:08] Test 5, 4000 Lucas-Lehmer iterations of M18874369 using FFT length 1024K.
[Apr 1 20:10] Test 6, 4000 Lucas-Lehmer iterations of M18874367 using FFT length 1024K.
[Apr 1 20:11] Test 7, 4000 Lucas-Lehmer iterations of M18474367 using FFT length 1024K.
[Apr 1 20:13] Test 8, 4000 Lucas-Lehmer iterations of M18274367 using FFT length 1024K.
[Apr 1 20:14] Torture Test completed 7 tests in 11 minutes - 0 errors, 0 warnings.
[Apr 1 20:14] Worker stopped.

the_syco

Going to pop the 4870 back into it, and test it with 3DMARK Vantage, which seems to be designed for DX10 cards.