TR Forums

Hey gerbils,

My gaming rig isn't exactly, you know, top of the range these days (In sig) - but it still can hold it's own at 1200p! It's latest upgrade was to a GTX 750Ti, but performance is a bit iffy compared to my Radeon HD 5770 that was in there beforehand.

My issue revolves around the clock speed of the core, and bare with me on this. My card is a KFA2/Galaxy GTX 750Ti, it's a small form factor card but is overclocked out the box to 1072Mhz on the GPU core. I've noticed that (for whatever reason) the card never runs at this speed - as you can see from the screenshot below it is at 1215Mhz but fluctuates (the Unigine tool says 1346Mhz - Erm...) - Is that normal? I wanted to OC the core a little bit (I've already got the memory stable at 6000Mhz effective) but the core temp is already to high - once it hits 80C (and it does this without any overclocking, only as a result of the speed it's decided to run at) it starts clock throttling and my FPS plummets.



So, can someone help explain why my cards core speed is higher than what I've set, and is there anyway I can stop it clock throttling so early on (like increase it's thermal cut-off threshold perhaps?) or stop the core speed fluctuation? I'm determined at get another solid year of gaming out of this system before it's shuffled off to server duties or something - even if it kills me (or the card!).

Nvidia cards monitor the GPU temperature and set the overclock based on that. My GTX 660 will is supposed to run at 1085 overclocked but will run run 1200 on its own. In short, the better the cards cooling, the faster it will run. The light throttling you see is from the GPU overclocking a bit high and throttling due to the temp.

Yeah, looks normal to me.
GPU-Z is a better tool for monitoring dynamic clocks in realtime.

Thanks both - if this behaviour is normal then that's ok. It just seemed a little weird compared to my old Radeon card that was also OC. It didn't care how hot it was, as you just set a speed and it run to it. What you both say makes sense though now thinking about how the dynamic clock speed seems to work.

Once it hits 80C though, there is a hard and fast sudden down clock. I've read pages online that show you can increase the temperature threshold from the 80C it's set to right up to 90-95C or so. I've not found out how to do this myself, and is that even a good idea (those higher temps don't sound like a good idea)? Should I look at getting a better cooler or seeing if I can change it's fan cooling profile or something so it's more aggressive instead?

Edit: My GPU cooler has a single, weedy 40mm fan. Epic design win, not...

I wouldn't go over the set limits by the manufacture... That extra 10 degrees isn't likely going to yield much and could harm the card or shorten it's life. One of the things I learned early on is to get a really good cooler. That being said, if the fan is not that great, I would run a custom profile. I always had my GTX260ti and my GTX560ti fans running about 70% when gaming. Both were the factory overclocked cards and the 260 in particular ran very hot. Used to heat up my spare bedroom quite nicely! I set the ramp profile pretty aggressive in Afterburner. I haven't done this with the GTX660 as it runs very fast and very cool. It is a very good card and I haven't felt the need to push it hard has it will run BF4 with a good mix of medium, high and ultra settings in excess of 60 fps. The game looks and plays great on my computer so why push it.

For more info on how the boost works, read the old Titan reviews and search for GPU Boost 2.0. TR covered it very briefly, Anand has a lot more info on that particular feature.

If it's a non-reference half-height card you're going to struggle to cool it any better that that. I assume you need a half-height one?

If not, it's probably easier to ebay it and buy a full-size model like the reference boards or one with an aftermarket cooler. Buying a 3rd-party cooler and fitting it yourself will be as expensive and additionally void your warranty.

Pville_Piper wrote:

I wouldn't go over the set limits by the manufacture... That extra 10 degrees isn't likely going to yield much and could harm the card or shorten it's life. One of the things I learned early on is to get a really good cooler. That being said, if the fan is not that great, I would run a custom profile.

Yeah, I was worried too - I found a website via Google that used EVGA's Precision X to modify the thermal limit - apparently it will allow you to set the temp to 95C before throttling starts. Yikes. I can't easily change the cooler because the card doesn't use the reference design board, and it's very small. Water-cooling is an option, but I have nowhere for the radiator as the CPU's cooler / fan & radiator is in the way.

I made a custom fan profile in Afterburner, where the fan kicks in at up to 70% speed with temps below 60C - Then ramps up 7.5% per 5C thereafter. It works as the card stays cooler by 3-5C than before. There's a strange bug though with it, because if the custom fan profile is on, after 1-2 minutes of rendering time the GPU clock falls to 1072 Mhz (stock) and never recovers. It's temp will be between 60-68C. Weird.

Chrispy_ wrote:

For more info on how the boost works, read the old Titan reviews and search for GPU Boost 2.0. TR covered it very briefly, Anand has a lot more info on that particular feature.

If it's a non-reference half-height card you're going to struggle to cool it any better that that. I assume you need a half-height one?

If not, it's probably easier to ebay it and buy a full-size model like the reference boards or one with an aftermarket cooler. Buying a 3rd-party cooler and fitting it yourself will be as expensive and additionally void your warranty.

Fantastic read about how GPU Boost works, thanks a lot. I'm no longer worried about clock fluctuations as it's clearly by design, just when it throttles back to silly speeds and FPS plummets. As for the card/case situation, I do need a half-height card to fit inside the case. The whole cases dimensions are about that of an Xbox One. It's a PITA to keep it cool, hence my CPU cooler and the hack job I took to the side panel to mount the rad and fan!

My next build will be a case that can take full width cards I think, but there's still plenty of life in this one yet. I could possibly try adding some venting area to the side panel where the card is...?

geekl33tgamer wrote:

My next build will be a case that can take full width cards I think, but there's still plenty of life in this one yet. I could possibly try adding some venting area to the side panel where the card is...?

I did that on a previous build and mounter a 120 mm fan pointing right at the card. Really helped with the temps. I drew up the hole pattern in Cad and the machine shop kindly cut it out with a water jet. Came out sweet!

My case is very narrow - it's not much larger than the Xbox One overall. I've template a pattern over the side and used a riveter to punch tiny holes in it right above the GPU area. It's worked a treat, seeing as under the side casing, the GPU is all but a few mm away (there would be no room to mount a fan).

Pics to follow.