How to make graphics cards run cooler and quieter. Part 1

[JezEngland]

Part 2 of this project is in here: http://forum.overclock3d.net/index.php?/topic/34209-more-cooler-and-quieter-reference-cooled-cards/

Sorry I deleted some PHOTOBUCKET pictures and most here are gone, sorry I'll add stuff later. Still look and read. Post 12 shows comparisson results and mods pics then Youtube vids & watever

may the future prove why this happen'd

Hi,

Amateur thoughts are in my posts here and the idea is effective though not for every scenario. Thanks

Reference cooler can keep the GPU cooler and the theory in this thread is more effective with the 480 and hopefuly helps the 69XX's as these cards are running hotter than the rest.

I use 2 GTX 480's, they've run very hot, not anymore

The first card is an EVGA SC+ edition. It's factory overclocked and I reduced its frequency's to NVIDIA's refernce stock frequencys. The second is a stock 'Point of View' card. Both cards have same back-plates. For my tests here the 480's are at NVIDIA's refernce stock frequencys set with MSI's Afterburner or EVGA Precisionwich also are used for the monitoring.

All my results in this thread are with game settings maxed and in NVIDIA Control Pannel 3D settings all apart from Anti-Aliasing (Application controlled) and Anti Aliasing Transparancy (set to 2 x Supersample) are at max settings. I played the games for around 30 minutes - 1 hour or more before taking result pictures. The temeratures became stable within 30ish minutes.

There is a comparison post for GTX 480 results in Bad Company 2 on the 12th post in this thread showing cooler and quieter results

Default fan settings:

My 480's get hot, it's expected. The fan noise can be controlled. The default fan setting makes the fan spin loud/fast when the GPU is early 90 degrees C. In many games the GPU gets to 90's and the fans stay loud for ages, alternating a little, dropping speed alot when a level is loading, then when it's back to real-time rendering the fans are back to high speed and loud. Both cards ran at similar temperatures when they were +92 degrees C. The temps can be lowered and that means adding to the blazing noise with increased intake fan speed. The setup has a BeQuiet case fan blowing from the side of the cards to take heat away from the 480 heatsinks out the back of the case. Tweeking and monitoring I've found that the referance settings and custom settings control the fans using intelligence because when the temperature drops from a high temperature the fan stays at the higher temperature speed until the temperature has dropped enough.

My custom fan settings, results are from playing Bad Company 2 maxed + NVIDIA 3D setting enhancements, V Sync'd, 1920 x 1080p:

My custom fan setting tweekings led to geting a noise level thats tolerable with high fan speed. 70% fan speed suits me, my case has sound proofing. Now the fans run at 70% from a low temperature up to a v high temperature. The GPU rarely hits 90 Degrees. At first it's very slow to heat up. When the cards are loaded for ages the 2 480's alternate between 72-89 degrees mostlty (89 is from a different game), 75-82 more, and if the GPU spikes to the 90's it's heat is quickly dispersed.

Not allowing the heatsink to hold lots of heat in the begining is what helps

The EVGA ran at 78 degrees and the POV (stock card) ran at 72 degrees. On the HARVEST DAY map, we had 2 tanks working together with engineers firing at us etc, the cards max temp were 82 & 76 degrees C. The difference between the cards temps can be caused by the 1st card warmed by the 2nd card wich is underneath. After hours of gaming the difference dropped to 3-4 Degrees C. The GTX 480 has exposed heatsink wich emits heat inside the case.

Here is a pic of my MSI's Afterburner custom fan settings:

For custom settings with other GPU chips first know the GPU's thermal limit

For long time benchmarking etc this idea is helpful though these settings aren't suitable because being more cautiouse is necessary

I played LOST PLANET 2 last week with the default fan setting. Episode 5, chapter 1 and the temps were around 89-93 degrees for both cards.

Now I'm usin the custom profile, I've just done Episode 5, chapter 1 and am now on chapter 2. The temps 70-74

Max temp i saw on GPU 1: 86

Max temp I saw on GPU 2: 79

Screen shots:

At the bottom of each picture is MSI's afterburner monitoring readings.

MSI's moniotring readings are in this order: GPU temperature, GPU usage, cooler fan speed, GPU core clock frequency

The game is more demanding in the movie clips and this shot is from a boss ending movie.

A few runs round Unigine Heaven 2 at extreme, 8xAA settings made GPU 1 around 90 Degrees and GPU 2 around 81 Degrees and both were very settled at those temperatures and both cards had around 97% usage and 70% fan speed

These temperature readings are from a fully cased PC

All this cooling with a lower speed intake fan

Conclusion:

Marketing? NVIDIA and AMD and custom card/cooler manufacturers making naughty deals users don't know about. Many people have owned a GTX 480 and got the GTX 580 with high hopes about using the quiet PC solution. AMD and NVIDIA control the market, they will work together to keep alternative solutions for media purposes thinking like more is more. The card will break quicker running at a hotter temperature. NVIDIA made retrictions for the 580 in Furmark benchmark etc, they should make a cautiouse fan setting too, for all cards (maby they have?). A big mistake is there isn't a vapour cooler system attatched to the mighty 480 heatsink.

I've informed NVIDIA and EVGA about this and I suggested to integrate custom fan profile advice and options to their cards software. NVIDIA and AMD Control Pannel's able to do it all is what we want.

A friend has used the custom fan setting advice for his 8800Ultra and reports 20% lower temps

Comment your views please

Thanks

 

[tinytomlogan]

Nice info but tbh the 480's are too loud no matter what you do, 70% fans must drown out your hoover! 500 series ftw, always said the 400 cards were for fan boys and Nvidia would spend the money the fan boys paid fixing the 400's properly..... few months later and boom 500 series and its all fixed. Case closed.

 

[JezEngland]

Blimey, OK thanks

The 480 is a problem, the 480 has to be tolerated.

The Antec P183 sound proof'd case keeps the noise in, I play with low volume speakers without bother

I'll install Metro 2033 soon, give that a whirl and I'm gona complete the ranger mission hopefully

The theory is take a fan speed the card uses when loaded gaming for ages, contemplate a lower speed and use that speed from a lower card temp thus restricting the build up of heat. I want results for other cards. 480 at 70% is quite humble, screams a little, the faster the fan spins the noise diefference accelerates, going past 70% the noise screams and screams but not as loud as that

.

NVIDIA's reference setting doesn't prevent the heatsink getting fully loaded with hot heat and raises the fan speed too late thus cooking the hardware, AMD's too. Gaming with the low fan noise is the factor with that though once it's cooking it's a long, hard and loud rescue attempt wich sometimes isn't effective

The reference setting raises the fan speed to around 90% and temps don't lower to what I've got with the custom fan setting, they stay put at high 80's low 90's, the cooler doesn't cope well, depending on the application behaviour.

A good part of my gaming friends comes with the 480, I don't see me waiting for the 580 in the next life

I could have replaced it though that idea is limited with expense.

My findings gaming the Fermi chip is a pleasure.

NVIDIA could have matched a 5870 market on Windows 7 release, NVIDIA needs a team to compete like that and could have set similar and better prices as the technology would be less high-tech as Fermi

 

[JezEngland]

I have more results made today. Both cards are running at similar temperatures. I ran Battlefield Bad Company 2 for about an hour. The temperatures got to the max of 80 and alternated below that. Here's a picture from that session:

 

[hmmblah]

Nice info but tbh the 480's are too loud no matter what you do, 70% fans must drown out your hoover! 500 series ftw, always said the 400 cards were for people who fold and Nvidia would spend the money the people who fold paid fixing the 400's properly..... few months later and boom 500 series and its all fixed. Case closed.

There, fixed it for you

. I do agree though. If I didn't fold I would have looked elsewhere or waited for the 500 series.

 

[JezEngland]

 

[JezEngland]

More information. EVGA staff are assisting with this.

The result I got this morning was with a very low intake fan speed with results of equal temperature readings on both cards. This evening I've raised the intake fan speed to a guess of 30ish%. With my case it's quiet. The front door blocks alot of sound. I'm using EVGA's monitoring software for these results called EVGA Precision.

Here is a pic of my EVGA Precision custom fan settings and they're almost identicle to the previouse custom settings:

These results a with a warmer room temperature than this mornings tests.

Then I put the intake fan speed to maximum, and the PC isn't quiet though I didn't have to raise the low speaker volume much and it's quieter than some sessions with NVIDIA fans blazing and runs much cooler, like 25 degrees cooler. Metro is gonna be tested, i got over 100 degrees records in Metro 2033:

I opened the case door and fan filter door to improve intake flow and the minimum temerature is 64 degrees C:

I ran Unigine Heaven 2 with 100% GPU and intake fan speed and open door and side off case. Running the application I let the cards warm to a stable temperature then ran the benchmark test. It's not quieter than reference fan settings though reference settings will run at 92% max, 100% isn't much different and we're getting 60's degrees C and every time I checked the GPU's usage was over 90%. Consider music isn't the point of this benchmarking and this custom setting or other high constant fan speeds is a good solution. The temperatures reached around 60 before I ran the test and alternated around 59-60 continuously. It's a shame and unfair that necessary precautions for benchmarking and other scenarios isn't catered and AMD'S 69XX are hitting that wave of problems. Put a user guide in the BOX please manufacturer. It takes engineering knowledge to sus and fix these details, the market is appealing to unaware people.

Teach or learn the hard way and thats a long time

 

[JezEngland]

More info. They're cooler

read read

These results are with game settings maxed and in NVIDIA Control Pannel 3D settings all apart from Anti-Aliasing (Application controlled) and Anti Aliasing Transparancy (set to 2 x Supersample) are at max settings.

These results are done in a warm room.

Today I did a good deed. I took the dog for a walk. Later I recieved thoughs to mod my rig with cardboard

Look:

That stops the heat emitted from the 480's exposed heatsink flowing into the fan. When the case is closed, inside I have a fan at the side of the cards blowing the heated air away from the exposed heatsinks and out the back

The temps:

Then I swapped the 480's round to see. The POV card was first used April 2010. The EVGA card was first used in November 2010 and it's now in the 2nd slot. Look:

Suggestions please, what can I get or engineer to replace the cardboard?

Transformers: War for Cybertron is made with Epic's Unreal 3 engine. It's and old and well optimised engine and powers the Bulletstorm game to be released soon.

The results I got for Transformers: War for Cybertron show GPU 2 working lightly. The cards are SLI'd and the 2nd GPU probably is doing the PhysX.

For this test I played the campaign for around 30 minutes and the temperatures were the around the same all the way.

The results:

 

[Rastalovich]

Nice info but tbh the 480's are too loud no matter what you do, 70% fans must drown out your hoover! 500 series ftw, always said the 400 cards were for fan boys and Nvidia would spend the money the fan boys paid fixing the 400's properly..... few months later and boom 500 series and its all fixed. Case closed.

Hoover is probably an understatement.

The fan boy argument is really for fan boys to throw back'n'fore at each other. Many of the enthusiasts that got the 480 and tamed them simply wanted the best single gpu out there and had the 'enthusiasm' to do more with their kit than just slot it in and install drivers. Cos it's what they do.

The 570 is a 480, in real terms.

Nice read, it's good to see people running their own tests and finding out stuff for themselves or how stuff works for them. Hayley will notice the 7-10 degrees that the thread was asking about yesterday. Pics included - everyone's a winner.

 

[JezEngland]

Ye it's an interesting pleasure.

With my case about 1m away with the fans at 60% it's as loud to my ears as me breathing blowing out

70% breathing blowing out harder and my speakers are quite well heard at low volumes. My case has sound proofing and front door.

When I overclock the cards enough for me noise will be a problem issue and the headset will get a dusting

More LOST PLANET 2 results. The EVGA card is in the 2nd slot. I played Episode 6, Chapter 2 and then Chapter 3 where I took the screen shot. I think the results are showing about max temps I'll get with this setup as the real-time rendered movies befor the chapter are very intense and this chapter vary good, loads of animated monsters, explosions etc. It's excellent.

The monitoring results are in this order: Temperature, GPU usage, fan speed, core clock, shader clock, memory clock, memory usage

The first result is biased as the bottom cardboard wall fell over. The results:

I checked the cardboard walls and the bottom one fell over, I put it straight and the result:

Maby the 1st GPU is rendering the Anti Aliasing and other stuff as there is an option for SLI Anti Aliasing wich I think means making both cards do it. I think that will give the 1st GPU a hotter run

Some thermal compounds require hundreds of hours use to perform properly. POV card has been used alot. I got it last April. The EVGA card is about 2 months old so the thermal compound is less used.

EDIT: 5/2/2011, 22:51: I've just played the same levels and completed the game. Max temperatures were 84 and 78. Room temperature affect results, it's not cold in here

 

[JezEngland]

More results. Here I re-ran the some Episode 5 levels and compare the previouse results on this thread with these new results with the cardboard mod.

The older results in this post were taken at night, the central heating was off for a while. The new results a very fresh, the central heating is on, candel is burning and mum cooking next door.

These results are with game settings maxed and in NVIDIA Control Pannel 3D settings all apart from Anti-Aliasing (Application controlled) and Anti Aliasing Transparancy (set to 2 x Supersample) are at max settings.

You've seen this picture wich was taken without the cardboard walls mod in my first post on this thread:

SORRY

These results show that when a game is very demanding the chip gets hot, that's expected, and these mods and results are great. When the game is moderatley demanding the difference in temperature is substantialy more.

Here are more pics from these 2 chapters in LOST PLANET 2:

 

[JezEngland]

___________COMPARISON__________

mods are 2 pieces of cardboard and 1 92mm BeQuiet case fan blowing from the side of the cards and the custom fan setting.

Here's what it looks like:

The monitoring results are in this order: Temperature, GPU usage, fan speed, core clock, shader clock, memory clock, memory usage

These results are with game settings maxed and in NVIDIA Control Pannel 3D settings all apart from Anti-Aliasing (Application controlled) and Anti Aliasing Transparancy (set to 2 x Supersample) are at max settings.

This result is using the mods:

In the same round of Battlefield Bad Company 2 i removed the cardboard and fan mods and set the fans to default NVIDIA settings. The result:[/size]

[/size]

Both cards have an EVGA Backplate that has a thermal pad for the GPU area. 21 degrees hotter and rising and a higher fan speed. There was maby 10 minutes between the 2 pictures. I didn't use tanks in this round and using tanks can raise the temperatures. If I kept that standard gaming for hours the whole rig gets hot thus heating the cards more and that makes the rig hotter and at about 95 degrees the fans go like turbo chargers at max possible speed wich is 92% and that rarely solves the problem, depends on the application behaviour, leaving users rigs in a very loud muddle

These cards are for gaming and we can switch off AFTERBURNER when we stop gaming.

Motherboards that don't have a slot or more space between 480's will suffer alot more.

I'm not testing LOST PLANET 2 with the default un-modded setup. When I had the BeQuiet fan and NVIDIA default fan setting the cards were around 95 on episode 6 chapters where the screen is chaos

. Removing the fan the temperatures will rise. These tests are done with an acceptable intake fan noise for normal/quiet speaker volume in an Antec P183 case

 

[JezEngland]

Metro 2033 Maxed with Advanced PhysX & NVIDIA Control Pannel 3D settings all apart from Anti-Aliasing (Application controlled) and Anti Aliasing Transparancy (set to 2 x Supersample) are at max settings.

Again I've edited the custom fan settings:

Both cards are overclocked.

The monitoring results are in this order: Temperature, GPU usage, fan speed, core clock, shader clock, memory clock, memory usage

The results:

SORRY

The game and NVIDIA 3D settings aren't suitable for all-round good gameplay performance. I have a 9800GT to do PhysX that I may install for this game though that makes the 2nd GTX 480 hotter as it's slot is directly under that 480. It's easier and safer to lower settings. The Metro 2033 Advanced Depth of Field (DOF) setting is a killer and some Advanced PhysX effects especialy when the smoke particles are lit-up with flood light

Thanks OC3D for leeting this project be stored here

Now I'll add results as and when

 

[JezEngland]

These results are for Metro 2033 at descent playable setting. In the game settings Advanced Depth of Field and Advanced PhysX are off and all other settings are at max & NVIDIA Control Pannel 3D settings all apart from Anti-Aliasing (Application controlled) and Anti Aliasing Transparancy (set to 2 x Supersample) are at max settings.

I overclocked the GPU cores to 735 Mhz

Monitoring results are in this order: GPU temperature, fan speed usage, core clock frequency

SORRY

This picture shows up to 10 degrees cooler GPU compared to the same scenario picture earlier in the thread where the game isn't running at descent frame rates and with lower clocks on the GPU's

The Advanced Depth of Field isn't missed, the game is great. In the Stone Giant benchmark we can toggle on/off Depth of Field and it almost halves the frame rate and does almost nothing to the looks and it's almost useless I mean it looks better without it

I have records of over 105 Degrees C with standard NVIDIA driver fan settings massive noise is included

 

[TMaksimov]

No matter what you do sooner or later the back of your case will look like this with two clocked reference 480s

 

[GBcf]

Haha! That would be awsome!

Jerome, why not try something like this... It generates more pressure to the intakes of the 480's...

Keep you're testing going! I like it

Cheers!

 

[JezEngland]

Thanks GBcf I've got some spare fans I'll try it tonight

Todd

 

[JezEngland]

Success

That's a noisy setup. The white fan hum's loudly and the red fan screems

I ran Unigine Heaven 2 maxed with NVIDIA Control Pannel 3D settings all apart from Anti-Aliasing (Application controlled), Anti Aliasing Transparancy (set to 2 x Supersample), V Sync (OFF) and Tripple Buffering (OFF) are at max settings.

The monitoring results are in this order: Temperature, GPU usage, fan speed, core clock, shader clock, memory clock, memory usage

The results:

Then i removed the white fan and the red fan leaving this:

I ran Metro 2033 and the temp are better

. I didn't run the game for an hour before this test though with this custom fan setting the rig doesn't get hot and the cards have stayed at low temps recently, reaching a normal average temp within a few minutes of running the application. I did leave the resource heavy Metro 2033 main menu running befor this test. The temps did spike higher than what we see on these readings. The spike heat was quickly dispersed.

The monitoring results are in this order: Temperature, GPU usage, fan speed, core clock, shader clock, memory clock, memory usageThe results:

This next pic is taken less than 10 seconds later and the temps are lower:

These tests were done with the hardware in a complete case

I've informed NVIDIA through Youtube. I've informed others. I want NVIDIA to fix the issue and I want you users to try the custom fan setting idea with your cards. Post your results here please. It's evident the drivers aren't qualified for the hardware

I've played this game with 107 Degree C on both GPU's and posted evidence here: http://forum.overclock3d.net/index.php?/topic/31250-gtx-480-high-temperatures/

Thanks

 

[JezEngland]

More

The setup is this:

1 fan instead of 2 and is biased to supplying the 480 fans

Next we make a 1 piece cardboard sheild with minding to not creat an electric circuit sticking foil to the cardboard

Results:

I'll add results as and when

 

[JezEngland]

Earlier I had early 80's in a strong stressing area of GTA 4. This guy uses GTX 580, gets up to 88 in this vid in an average GPU usage area with the case open and the fan usage is quite high.

http://www.youtube.com/watch?v=HAAa1cWFG3Q

An idea is to put a foam strip between the cards to stop warm air from the back of the second card flowing into the 1st card's fan like whats shown here on the 480:

That particular chip/card build could be wrong