Post Your GTX 680M Overclocks and Temperatures

I would like to gather all the data in one thread so others who would like to see how much their new GTX 680M can handle and give us a some idea of the statistics of the GTX 680M overclocking ability. Also please mention if you have a HM or EM and if you applied the custom vbios.

Thanks

Also I know that I don't have GTX 680M at the time of this post but I am hoping to get one soon .

Also mention if you are trying to OC your card to the max or just a mild OC.

 

Just an idea, but wouldnt be better call this thread in the same way as meaker did for 7970m?

Instead of: 7970M benchmark results (ONLY) it could be
680M benchmark results (ONLY)

 

I felt like this would include benchmarks and gaming results as well but the focus I want to put on is the speed the card is running at since benchmarks can vary +/- 100 to 200 depending on system specs. But I really appreciate the advice

 

Sig, 89c and 195w peak consumption in games.

 

Thats nearly the same that i get on my HD 6970m when OCed but i think i will buy the 680m over the HD 7970m just because its easier to get it in Norway.
If i go with the AMD offering i got to disable Enduro myself or get it abroad meaning outrageous shipping prices / taxes.

 

Should have bought mine.

 

For some reason my OC is 89 degrees never going up more then that at +135, +750.

Stock clocks I get around 70~75. I'm also using Arctic MX-4 btw, on stock thermal paste I was always hovering 89 degrees.

 

It throttles any higher, I have taken it to that point and calibrated it on demanding games.

 

Odd thing is that I could run +700 with the stock BIOS, but I think there was a lot of error correction going on, primarily trying to make everything stable. With the revised BIOS's I think it's more realistic and just fails if it fails, so you're likely to get much lower RAM clocks than with the stock. In any case, the best OC I can do is ~ 1050 GPU and 2150 vRAM for ~ 2050 3DMark 11. I'm going to stick with the modded BIOS at stock voltage.

 

I could bench higher for some reason with the modded bios lol.

Went from 2575 to 2600 and even 2625.

Have you tried jumping up to 2500mhz as that's the rated speed for the memory chips.

 

Weird. I ran at 2400MHz and it ran fine. Much more than 2400MHz and I get graphical anomalies. Guess I just got some bum RAM. But at ~ 2150-2300 it crashes all the time. Weird.

edit: Seems the biggest culprit is my GPU core. I still can't exceed +130 (850) even with more voltage. Strangest thing is that it fails the physics test but not the other things, and thought the physics test was supposed to tax the CPU only not the GPU at all. But sure enough if I bump up the core past +130 it crashes every single time.

edit edit: I don't get it. Physics score is higher when using 680m vs HD 4000 but I thought it was supposed to be exclusively a CPU test. It seems it uses GPU vRAM though. I can't pass Physics with core past +130 or vRAM past +340. Everything else I've run works fine. Guess I just don't get to run 3DMark11. Thing is I can't really exchange the GPU because it works great at stock speeds. Guess I just don't have an overclocker.

 

Are you running stock vbios?
Are you running Saltius Vbios?
Are you running SVL Vbios with higher voltage than Saltius?
Are you trying a different voltage than the three mentioned?

Stop confusing us with all the different testing.

As far as I know, if you are using Saltius Vbios, you are the only one with stability problems, even with +135/+200 additionally

Meaker also noted that more than around 1275ish for the GDDR5 was too much. The stock clock with Saltius Vbios (902MHz) is super stable for everyone, even at 1037MHz. If you go much higher than that, I think you need to feed the GPU more voltage than what the Saltius VBIOS give you, like Meaker is doing with his GPU now. Plus you need his mod to unlock higher overclock since +135 is max for Saltius Vbios. But the memory doesn`t go much higher than 1275ish anyways

 

If you turn off the 680M then the integrated is working harder and taking some more of that precious TDP of the CPU.

 

I have run SVL BIOS both stock voltage and +200 with overvolt.

The issue is primarily with 3DMark 11 Physics test. It is not stable at +200 with higher voltage. It is only stable at ~ +130-135 (~850) at stock or higher voltage. I get random crashes anything more than that. Same with vRAM. Anything more than +300MHz (~2100) I get random crashes.

However, I can run Kombustor easily at 1050MHz GPU and 2450MHz vRAM stable for 15 minutes+. Crysis 2 bench runs stable too. It's something with 3DMark physics test. I even uninstalled, reinstalled the 304.79 beta drivers, uninstalled, reinstalled 3dmark 11. Same result. I swapped system RAM and it seemed to improve to +450MHz vRAM, but it ended up being a fluke because then it started crashing again.

I may use throttlestop to actually limit the CPU speed to 2.9GHz or something and see if that's the culprit. But still not clear why GPU or vRAM would affect Physics test since it should be all about CPU.

If it seems to be CPU related, I may swap my CPU from my NP6110 (it's out of commision at the moment anyhow) and see if that fixes it. If it does then I will have to see about RMA'ing the CPU hopefully without my whole system.

 

You should stick with the one cloud has and try running the stock clocks (902/2200) that come with vbios

 

I'm just going to make sure it isn't my CPU first. Signs are starting to point that way.

 

What happened to your W110 HTW ? I'm going to test SVL7 +200mhz vbios tonight but will probably keep Saltius vbios (1.037v) for 24/7 use.

 

Yeah, the VBIOS I have (Saltius) seems spot on in terms of performance/reliability and with the right amount of voltage. I have shipped it to HTWingNut to see if he gets better results with it.

 

Could I also gain access to said Bios xD. I'd like to see how far I could oc mine as well. Also the bios mod doesn't raise the voltage does it or have any potential harm?

 

I think I need to link to a post and put it in my sig, lol. I've been asked that probably a dozen times already. But bottom line is I think it was my fault, I dragged some pliers across my mobo routing a wire and think it shorted something out. Power was unplugged, but battery still attached. New motherboard is incoming. Hopefully here by end of week.

 

It raise the voltage from 0.987V to 1.037v and have a stock clock of 902MHz (instead of 719MHz) and 1200MHz for the memory (instead of 900MHz).

We have 5-6 people using it in our forum right now without hassle so no bricking so far. But I take no responsibility for any problems you may encounter.

MSG me your email adress and I`ll send it to you. Well if you have the 4GB version that is. Won`t work on 2GB

 

TBH I have no idea if I do have the 4GB version, it says on CPU-Z that its 2gb but on nvidia inspector it says 4gb. I'll send you a pm either ways though.

 

I thought the bios you have had a voltage of 1.012 , do you have two different ones or was the 1.012v abandoned since it prove to be unstable on a few systems?

 

You should have the 4GB version.. XoticPC only sells 4GB versions, unless I am utterly wrong and you bought 3rd party

 

Oh sorry. The VBIOS isn`t at 1.037V, its at 1.012V like you say. Sorry, my mistake. Too much to do at work today

 

Hi all, I have a np9150 that I ordered from Sager with the 680m and ic diamond compund...when i first got the comp, the temps were hovering around 85 celsius in demanding games (bf3, jc2, etc.). Even games with vsync turned off would push it sometimes to 90 celsius. I knew something was up, because at least 2 other people i found had sub 80 celsius temps OC'd! So, I repasted with at least 3 diffrent types of thermal compound in many different application methods. Imust have tried like 12 times. The closest I have come to good temps is a max of 81 celsius in furmark at stock clocks with shin etsu x23 in the X-application form. I even taped the heatsinks with aluminum tape and tried to apply mini ram heatsinks. NOTHING IS WORKING!!! The cpu temps are the same. I have the 3820qm but temps shouldnt be that different from the 3720qm or 3610qm. Is there some secret I dont know about????

 

Well tug my nipple and call me a cow. The Saltius vBIOS that I got from Cloudfire stabilized my machine. Strange thing is I did get one crash in physics but only one, ran through 3dm11 a few times with no crashing. Woot! Thanks Cloudfire!

 

did you try overclocking or did you keep it at 902/2200?

 

Well dammit I spoke too soon. Now it's crashing again. I kept it at 902/2200. No rhyme or reason. Was working perfectly fine now crashing again... This is getting quite frustrating. Although it really is making zero sense that physics would crash due to GPU clocks at all. It's supposed to be completely CPU based other than the GPU, and in this case it should be the HD 4000, displaying the frames.

 

Did you swap the CPUs?

 

No, but I limited the speed of the CPU and it still crashed. However I did find this: Process exited unexpectedly on Physics Test GTX 680 SLI ONLY

It does seem to be CPU related. I guess I'll try to swap CPU's now. Perhaps its the mainboard power issue to the CPU. I guess I'll find out.

 

I am such a FREAKING MORON!!! I'm sure many of you could have told me that already. Let me tell you what the issue was. I didn't even heed my own advice. KISS method... Keep it simple stupid... you know?

After all that, it WAS THE DRIVERS! I reinstalled the beta 304.79 once. This time I reverted back to the stock ones from Sager 302.71. It works stable. Completely stable. GAAAA!

I lose some performance, but it's stable.

 

Hey I should mention, I was also having those crashes and I decided to look it up and apparently it was a software issue where during the physics test it would crash for no reason. I did a software update and it fixed mine.

 

I`m glad it worked out for you HTWingNut
Please post your 3DMark11/Vantage/whatever results so that other can see what this VBIOS can do. The more the better.

You too City when you have flashed my VBIOS

 

Well I know the 304.79 offers a decent performance boost over the 302.71, but if it's not stable, not gonna bother. I'll push it a little and see what's best performance I can get with this BIOS. It seems it caps at about ~ 950MHz and my vRAM at about 2450MHz so likely not too many gains of the stock 900/2200 of this BIOS.

 

Even though the graphic DRAM is rated at 1250 (2500MHz), but that's for 1.5V, there's a big chance it's running at 1.35V on the Clevo GTX680M which is rated at 900 (1800MHz).

Clevo GTX680M is using K4G20325FD-FC04. But this one has 2 voltages, 1.5V for 1250 and 1.35V for 900: Samsung Semiconductor Product Selection Tool

 

Well stock frequency of this vBIOS gives plenty of extra performance too. I know a few in this forum who have settled there

 

The good news is that the oc of the memory could not damage them since there are rated at higher speed.

 

I'm sure I will settle there too for regular use. Played BF3 last night and GPU didn't exceed 80C without Fn+1.

 

HTWingNut: I`d really like to see you put in FPS numbers in the games you tested in your review, but with the FPS you got with Saltius vBIOS. That would be so awesome since you have stock bios with stock clocks and you have stock bios with overclocks. All thats missing now is Saltius vBIOS with stock clocks, and perhaps Meaker could fill in the rest with his extra overclock or something. Your review would be the ultimate guide to 680M for future owners

I know you are probably fed up with the testing because you worked so many hours on your review, but please, do it for the kids

---------------------------------------------------------------------------------

I`m also curious to what temperatures one would get with undervolting this GPU based on stock vbios...

 

I will go for testing with the Saltius vBIOS at stock clocks (900/2200) and update my benchmarks.

 

impatient to read, your review of the P150EM is just perfect

 

Thats super HTWingNut. Your review covers everything

 

I sense something bad happening.. a machine running several hours a day with such a vBIOS... can't be normal

 

It is normal if Nvidia didn`t want to give us the best performance they can offer.
Which I think is the case considering the stability and the low temperature

 

Slight voltage bumps and overclocks to stability limits is nothing new. I wouldn't be too concerned. It's a 2.5% voltage bump. That's well within margin of tolerances I'm sure. Temperatures are also well under control. I have run many desktop systems much harder than that, significantly harder for extended periods. If there's any concerns then don't do it, but I don't see an issue. I personally wouldn't go much more than this, not to mention the temps probably skyrocket if you start bumping up voltage.

 

on 580m 0.87 -> 0.92 caused a ton of dead cards.. just warning.

 

Yet the 570M to 0.9V caused none that I knew of.

 

It may have been temperature related. The 580M reached 96c in Furmark even undervolted to 0.80v...

 

I bet Nvidia forced the most voltage they could to the 580M for it to operate, since it was the most they could squeeze out of Fermi. No wonder why people killed it when they added more voltage than the limit. GTX 570M like Meaker notes above, was not the highest performing card of Fermi and it took the extra voltage just fine.

GTX 680M on the other hand is the first ever high end mobile Kepler card, shows lots of overclocking potential and even runs cool with an overclock up to 902/2200.
I bet we will see the same thin line when we have our GTX 780M too with very little room for more voltage when that time comes



My 2 cents. I may or may not be wrong.