I've just downloaded RivaTuner. (rather than nTune since nTune has a pisstakingly huge file size and I have a completely crap 56k internet connection)

I took a look at the readme file, and was immediately feeling lost and confused. It doesn't make much sense, and seems to be written for only experts to understand. It also tells people to look at the HELP before doing things. It doesn't tell you that you can only do that by right-clicking on various things. I only just realised that.

So... basically I was trying to improve my GPU's temperature. I have an 8800GTX. It has the standard fan and heatsink unit that ASUS provides with their GTX cards.

I get about 80-85 centigrade under load. 70 idle.

What I want to know is basically... I found the utility in RivaTuner to change my GPU fan speed. But I'm not sure how safe it is to use. I told it to detect my default fan settings, or whatever RivaTuner calls that. Now it gives me a slider which lets me change the fan speed. But rather than RPM it seems to be a percentage.

Is 100 percent the maximum SAFE speed for the fan to run at? Or if I set it to higher than normal, does it cause huge wear and tear and/or burn out the fan?

Is it some kind of stupid bug that makes the card never seem to run its fan faster even when it's over 80 degrees centigrade? I know the card can take much higher temperatures, but you'd think the speed would at least increase slightly.

Anyway, could someone tell me if it's perfectly safe to make my fan run faster? I noticed the temperature dropping like a stone when I put it on 100 percent speed for about 30 seconds, so it would be nice if it's ok to keep it running extra fast.

Also, has anyone overclocked using Rivatuner? I'm not sure my GPU needs overclocking, but I'd like to know how safe the utility is for that purpose too. My understanding is that it's meant to somehow test automatically how far you can go safely. Is that right?

RivaTuner doesn't seem to be up to date for the latest Nvidia drivers, so I get an error message about that. It does seem to run fine though, so I'm not sure what to make of the complaint. Could it cause a problem?

EDIT: Also, my new PC case has a fan speed controller on the front panel. I didn't have the right connector to make it control the GPU, which is why I'm using RivaTuner. But I can control the case fans and the CPU fan. But what speeds should I be using? Can a fan speed controller set the fans to run TOO fast? For example, my CPU fan is an 120mm fan. What kind of RPM should that be using? The speed controller seems to say it's normally at 1600RPM. Just ONE touch of the "increase speed" button causes it to increase to 2300RPM. Is that speed safe? it seems a bit high. ;;
It certainly improves the system temperatures though... What setting should I be using? Or how do I find out?

(Weird... it seems that after it went up to 2300 it then slowly decreased back to the original 1600RPM. Maybe there's meant to be something you press to lock it at the higher speed...)

Sorry for being a complete noob at this stuff. I never had a speed controller before. Can anyone help?

ARGH! It's suddenly on 2300RPM again. I don't understand this! The bloody thing didn't come with proper instructions!

Ah, thank goodness for that. I was a bit worried about using 2300rpm rather than 1600rpm. Seems like a big increase. But I just noticed that the sound it makes at 2300rpm is a lot like it sounded in my old case. I wonder if it used to run at 2300, and the fan controller slowed it down.

One thing I do wonder... Why does the speed controller only give me 1200rpm max for my case fans, but 2300rpm for the CPU fan? Surely it can't magically detect that they're different? I don't see how... That said, they're all 120mm fans, so I wonder what the difference IS.


I think I worked out what's going on with the speeds. When you set it, it seems to go up to the maximum rated speed, and then decrease until it's at the speed you wanted. Then it will stay there. Though it does vary slightly, and I'm not sure why that is. Seems to work ok overall, but the speed wobbles between the two nearest 100rpm figures. Say if you set it to 1700rpm, it might keep swapping between 1700 and 1800. I think it's just rounding up or down respectively due to small speed variations.

This makes me curious though. If a fan speed controller can't make the fan go over it's limit, then how WOULD you do that? Not that I want to. Noooo sir.

Speedfan never controlled my fans at all, which is interesting. In fact, speedfan only showed the speed for one of the fans. Maybe the others weren't linked to the motherboard.

I don't KNOW of any fan control in my BIOS. I'm not sure about that. I have been in there and looked, but I may have missed something. Might take another look next time I'm booting up.

Well, if the fan speeds are all fine running at maximum... this is very nice. The GPU core temperature is 58 degrees Centigrade right now.

The CPU temperature isn't so impressive. That's 44 for both cores, idle. Still, that's not bad. I just hoped it would be even cooler with the big fan. And the weather is pretty hot here right now. Room temperature is 27 Centigrade at the moment, though I'm not sure how much of a difference that makes.

I actually found a guide of how to do that... Did seem a little complicated, but it seems you have to set up limits on the temperature recording graphs.

I could just do it myself. Turn up the fan speed before I start a graphics intensive game. I hear the automatic monitoring method takes up quite a lot of system resources. Did you find that was a problem?

Holy crap, 58C with only 25 percent fan speed? Did you have a custom cooling unit on your card or something? Could be the GTX is just hotter... Mine is about 70C idle with the fans at about 50 percent (or whatever the default is). And that's with the standard ASUS 8800GTX cooling unit. Which is a fan with a heatpipe leading to a great big huge heatsink. The whole thing's enclosed in plastic casing. I had it open yesterday and removed some fluff, so it should really be running as well as it ever could right now.

As for the case fans, I know bugger all about what bearings they use. The companies who send out the cases never tell you any useful information like that. In fact, I didn't even know the case was going to show up with a built in 120mm case fan and the fan controller unit. For the price, I'd only expected the case and the huge 400mm fan.

The fans are just 120mm basic fans. One of them came with the new case, so it certainly shouldn't be worn out. The other, I took off my old case. It was the original 120mm case outtake fan. I put that one on the spare vent holes at the front of my case.

I wonder if this is anything to do with it... The two case fans had to plug into two of those white connector socket/plug things. But both of them were on the same cable. I'm not an expert with the wiring, so I'm not sure if those are for data or for power...
Surely it wouldn't be sharing the power between the two fans and thus having the speed shared between the two also?

I can't take the fans out without a massive amount of rearranging crap in the case. One of them's in behind my hard drives. The other is at the back of the case and it's not exactly simple to get to. After putting everything together I'd prefer not to be dismantling things again.

You'd think it would be easier to look the parts up. Does it help if I say that one of them is a black 120mm fan with the word Aerocool across it? Aerocool made the case. You'd think there would be information on their website about what fan comes with the case. But I couldn't find anything. Actually, there was sod all useful information about the fan speed controller either.

Oh well, never mind then. So long as the fans are running, it's fine.

Hm. I seem to have lost the information I found about setting up the auto fan speed changing in RivaTuner.

Could you tell me how you did it?

Oh, so that's why...

I was trying to do it through "low level" control. Didn't like the thought of screwing with the drivers. Is it completely safe? Assuming I don't do something unbelievably dumb like telling it to use 25 percent fans at 90C.
Anyway, thanks for telling me how.

Oh, and I found another problem. I have certain games that I have to play on Vista 32bit instead of XP. Crysis for example. So I just booted up Vista, with the new Nvidia drivers installed. I tried to use RivaTuner, in Vista, from the folder I installed it on XP in. I get a message on the main screen saying that there are "no supported drivers detected for this display adaptor".

Does that just mean some idiot didn't make that version of RivaTuner compatible with the latest Vista drivers? Or is something else wrong?
Do I need a fresh seperate install of RivaTuner or something?

As a result of it not finding drivers, I can't use the driver level controls in Vista at all.

Also, there's something weird going on. When I opened up the hardware monitoring window, I found that in Vista, RivaTuner is reading my GPU core temperature to be 58C idle. That's with the default fan settings! In XP it's more like 70C under the same conditions.

Any idea why that is? Is it just reading it wrong because it doesn't understand my driver? Surely it can't possibly really BE that much cooler. There's no reason for it to be different that I know of.
Also, in Vista it's reading the "ambient" temperature to be over 50C when I know it's more like 30C. The hell?

That's the thing. I already DO have the latest drivers for Vista. That's why I can't understand what the problem is.

Well, I found another program (eVGAPrecision) that lets you do the fan control on Vista with no problems. It seems to read the temperatures right, unlike RivaTuner.

I've discovered an unpleasant suprise. Playing a game such as Crysis... Actually, it's ONLY Crysis so far... Makes the GPU temperature go up to about 82C. Ok, so that's not high enough to make it explode, but with 100 percent fan speed, that's... well, worrying. I know Crysis has a reputation for being the game that no system can actually play on full settings yet, but with the fans in my case you'd think it would be a little cooler...

EDIT: Hm. I got a program that lets you read your GPU bios.
Using it I found out that apparantly my GPU can get to 110C before it reaches the critical threshold temperature and mostly shuts itself down. Seems 80C isn't anything to worry about.

What's weird is, I looked up the fan settings with the same utility, and according to the readout, the FAN on the GPU thinks the critical temperature at which it should turn itself on at 100 percent speed is 220 degrees!?
It doesn't give a unit, but I shouldn't think it meant Farhenheit.

What the hell is that all about? The fan only goes into panic mode long after the GPU core itself has! In fact, the threshold for the fan seems to be about 110 degrees after the GPU has either shut down or caught fire!

This message was edited by DG on Jul 05 2009.

Not very good? You can say that again. It seems to only have a critical threshold, no milder thresholds...

Then again maybe the card uses a seperate fan speed controller unit instead of letting the card's drivers/bios do it. I'm pretty sure the fans do get faster, though the bios says they don't. Weird, really.

Oh well, the main thing is how hot it gets, maximum. Crysis appears to be the most straining game in existance for graphics cards. It's not going past 85C so I think it's probably fine.

I'm... not sure I understand why you're telling me to convert into Fahrenheit.

In the UK we always use Centigrade. Well, I do anyway. Did you say that because of what I said back in my other post about the fan speed threshold in the GPU bios? It didn't actually say anything about a unit at all. I just noticed that if it was in Centigrade, then that means the fan only starts to go really fast much later than the GPU shutting itself down. Which is... pretty dumb.

What exactly do you mean by GPU cooler? The 8800GTX comes with a great big cooling unit by default. It's a heatsink on the processor, linked to a really BIG heatsink by a heatpipe. There's a fan that blows air over the heatsinks.

Did you have something more than that in mind? I know you can get some truly huge cooling units, but to be honest I have bugger all space in my case already. The GPU is nearly touching my Hard drive cage thing. (If I ever need a bigger GPU I'm going to have to move my hard drives into Caddies and somehow remove the cage with a power tool...)

Also I'm not sure how much extra cooling one of those things would really get me. What would you suggest I did get to mount on an 8800GTX, and what kind of improvement would I get?

Oh, and dangerman200k... Try running Crysis in that resolution with AA off and all settings on VERY HIGH. That's what I played (Past tense because I finally completed it last night. Time for Warhead.) it on. I'm curious as to how your system handles it.

Meanwhile mine is at 80... Did you replace your stock cooler with something else then? If so, which?

Oh, just wondering by the way, but why do you prefer to play Crysis on High? Very High surely looks better?
Just turn down the resolution a bit and it won't lag.

800x600!? Good grief, I've not played anything in THAT resolution since Diablo 2!
1024x768 is the minimum I use. I go for the highest resolution I can without getting lag. Crysis looked like CRAP when I tried it in 1024X768 on this screen.
(22 inch HDTV. Max resolution is 1680x1050)
On a high resolution, Crysis looks BEAUTIFUL. No doubt that's why it nearly makes graphics cards explode in flames.

Anyway, am I being an idiot? I'm looking at that newegg coolers page. I can't see how you'd actually fit any of those products TO the graphics card. Some seem to just be fans. The card already has a fan built in. Removing that would mean removing the whole stock cooling unit, surely? But that would mean there was no heatsink. There doesn't seem to be a giant heatsink with the fans on the newegg page.

And does a GPU cooler need to be specially made for that specific card, or not?

Hmm. Well, I found a couple on there for the 9800 and 8800 series cards.

There's one huge massive drawback with said cooling units. They both take up an enormous amount of space. Apparantly they make one card take up the space of 3 PCI slots.

There's another interesting one there though, the Aerocool FANLESS cooling unit, on page 3 of the newegg devices. I can't really understand how it could work all that well. It seems to just be a really big heatsink on its own.
I wonder, though, is there a way to use it AS WELL as the normal heatsink? That could work...

Well, it depends where the space is being taken up. I currently have a 56k modem under my graphics card so there's no space for a large cooler UNDER it. Some of the passive coolers appear to have the giant heatsink part OVER the card.

For example: http://www.aerocool.us/cooler/vm-101/vm-101-feature.htm

Holy crap that's big.

Oddly, the other one that I found on newegg seems to be no longer on the Aerocool website.

What I really can't understand is how anyone SLI links graphics cards these days. The cards get ridiculously hot with the stock coolers, and the larger coolers take up so much space that you wouldn't be able to fit in the second GPU.
Surely replacing the FAN on your card can't make that much difference?

You mean hot melt glue from a glue gun? Hmm. Would I need to be super careful about where the glue itself went? I assume it's an insulator, but I wouldn't want to mess up the card somehow.

Any idea roughly how many degrees Centigrade of improvement I'd get by doing that? Sounds like it's worth thinking about, at least. Though it's a pity you can't just BUY a fan that comes with clips to attach to a card like this. You'd think it would be a pretty obvious product to sell.

Hmm. I think I DO have a spare 80mm fan. The one from my front panel in my old case.

http://www.neoseeker.com/forums/17/t1321391-easy-graphics-card-temperature-drop-modification/

Is that the guy? From his post it looks like he had the fan SUCKING air from the back of the GPU rather than by blowing onto it. Any advice on which way around would actually be best? For that matter, directly onto the card, or blowing across it?

Well, I found the thread you meant. At least, A thread where someone did the same thing. On neoseeker, and you posted in it, so I assume it's the one.

I sent the thread creator a PM to see if has any advice on the matter.

My main concerns are how much space there is above the card, though honestly there should be craploads. Then how to connect the power to the fan. I assume I just plug it directly into one of the MANY cables hanging off my PSU.

What I actually need advice on is whether it would be best to have it blowing onto the card or sucking. The other guy had it sucking, which is why he didn't need an air space.

If I have it blowing I guess I'll have to rig up some kind of stilts made of plastic. Or... just a big pile of glue? What do you think?

What I'm not super happy about is having a really hot glue gun on my card.

And for that matter... What stops the card from MELTING the glue. I don't know how hot the glue's melting point is.

Well now that's just weird. I wonder why you get cooler temperatures than I do, if we have the same cooler, and I've got more fans. (Or do you have lots of case fans near the GPU or something?)

Meanwhile I'm currently using...

A front 120mm fan. A rear 120mm fan. A side 400mm fan.

Apparantly your card is just awesome.

Oops. I forgot you had the GT. Yeah, I think that's probably a bit cooler. Still, quite a difference.

I've actually managed to overclock the GTX now. Not too far, not far enough to heat it up more noticably. I find myself wondering why anyone bothers. It doesnt seem to make much difference.

WHOA. My friend also has a GT. His is currently running at 75C IDLE! Should be scared crapless? I suspect his fan might have fluff in it.

This message was edited by DG on Jul 07 2009.

Well, I left the shader clock linked to the core clock.

My original clock speeds were:

Core: 576Mhz
Ram: 900Mhz.

I overclocked to
Core: 620Mhz
Ram: 1000Mhz

In comparison to the overclock I've seen which had a core speed of over 750Mhz, that's nothing.

I just had Crysis Warhead playing on 1680 by 1050 Resolution, on full settings for everything apart from Anti-aliasing, which the game doesn't even need. NO LAG. I got a bit of lag just now on a mission where I'm riding a speeding train at insane velocities while blasting everything in sight, but I think that was pushing it.

That was using the overclocked settings. I'd not tried Warhead without overclocking. Maybe I should test it.

Hard to say, really, whether it made a difference. I think I'll see how Warhead runs without the overclock.

Unfortunately I uninstalled the original Crysis, so I can't test that WITH it.

Hmm... I wonder if there's a "show FPS" setting.

Now that you mention it, I've got an outdated version of Futuremark around here somewhere... Must find it...

WTF. It was on the pc already! Stupid Windows Vista just refused to find it when I searched.

.... And Vista refuses to run it, giving a BS excuse about a missing dll file. I'll have to try in XP.

The futuremark/3Dmark 2006 benchmark software.

Ok, I got some 3dmark results. It seems to get a few more FPS but I'm not sure how much better it really is. I could try clocking it higher, but I can't find a program to test stability... ATI tool is meant to do it, but the bloody thing crashes even WITHOUT overclocking.

Before overclocking, I got:

3DMark Score 9593 3DMarks
SM2.0 Score 4582
HDR/SM3.0 Score 4063
CPU Score 2346
Game Score N/A
GT1 - Return To Proxycon 36.947 FPS
GT2 - Firefly Forest 39.412 FPS
CPU1 - Red Valley 0.740 FPS
CPU2 - Red Valley 1.191 FPS
HDR1 - Canyon Flight 41.237 FPS
HDR2 - Deep Freeze 40.014 FPS
Fill Rate - Single-Texturing 3455.196 MTexels/s
Fill Rate - Multi-Texturing 17199.814 MTexels/s
Pixel Shader 336.899 FPS
Vertex Shader - Simple 206.663 MVertices/s
Vertex Shader - Complex 139.102 MVertices/s
Shader Particles (SM3.0) 132.345 FPS
Perlin Noise (SM3.0) 115.477 FPS
8 Triangles 22.218 MTriangles/s
32 Triangles 89.085 MTriangles/s
128 Triangles 254.302 MTriangles/s
512 Triangles 256.606 MTriangles/s
2048 Triangles 251.306 MTriangles/s
32768 Triangles 250.494 MTriangles/s


AFTER overclocking to 620Mhz core, I got:
3DMark Score 10107 3DMarks
SM2.0 Score 4859
HDR/SM3.0 Score 4410
CPU Score 2345
Game Score N/A N/A
GT1 - Return To Proxycon 39.542 FPS
GT2 - Firefly Forest 41.442 FPS
CPU1 - Red Valley 0.738 FPS
CPU2 - Red Valley 1.193 FPS
HDR1 - Canyon Flight 44.855 FPS
HDR2 - Deep Freeze 43.343 FPS
Fill Rate - Single-Texturing 3724.880 MTexels/s
Fill Rate - Multi-Texturing 18567.752 MTexels/s
Pixel Shader 365.190 FPS
Vertex Shader - Simple 221.905 MVertices/s
Vertex Shader - Complex 150.439 MVertices/s
Shader Particles (SM3.0) 145.858 FPS
Perlin Noise (SM3.0) 124.840 FPS
8 Triangles 22.156 MTriangles/s
32 Triangles 87.956 MTriangles/s
128 Triangles 273.728 MTriangles/s
512 Triangles 276.980 MTriangles/s
2048 Triangles 271.634 MTriangles/s
32768 Triangles 270.850 MTriangles/s

The heat gets up to about 83C during the artifact scanning in ATI tool, but I think it does that anyway even without overclocking it. That scanning of the spinning cube heats the card up insanely for some reason.

Hmm. Strange. I tried taking the clock higher. I got the Core to 650Mhz with no problems. But I can't use the core at 650Mhz AND the memory at 1000Mhz. It starts freezing up the system during the ATI tool test.

I don't really understand that... Every guide to overclocking I've read says that you're meant to test out the core and the memory independantly, but that doesn't make sense. If I do that, I can get both to insanely high levels with no problem. Then when I put those values together, I get big problems.

Any idea why?

Wait, maybe I was wrong. It freezes up after a while with 640Mhz, no matter what the memory is set to, even though the heat isn't any worse. I think it's something to with the shader clock being too high since it's connected to the core by default. Can change that, but not in ATI tool.

OK...

I set the clock in RivaTuner and TESTED it in ATI tool this time.

Core clock 640Mhz.
Shader clock 1450Mhz.
Memory 1000Mhz.

Oddly, it found no artifacts, didn't get insanely hot, and yet it froze up for a few seconds.

Somehow I think I'll be sticking with my tried and tested overclock.

This message was edited by DG on Jul 09 2009.