Overclock.net › Forums › Graphics Cards › NVIDIA › ~~The GTX 670 Overclocking Master-Guide~~
New Posts  All Forums:Forum Nav:

~~The GTX 670 Overclocking Master-Guide~~ - Page 16

post #151 of 3357
Quote:
Originally Posted by brettjv View Post

Anything anyone says about 'what happens' from OC'ing core or memory ... it's just a rough guideline of what to expect. In reality, an unstable OC of either Core or Memory could cause anything from artifacting to driver crashes to bsod to system freezes. It's not written in stone exactly what's going to happen from either core or mem being too high.
And even just your 'stock' clock of 1241 (including the KB) is pretty decent clock on a 670 thumb.gif

ah, gotya. I'll have to mess around with it a bit more and see if I can eek out anything extra from it.

and agreed! May not get to that 1300 number it seems everyone wants to hit, but considering that makes like a 5 fps difference, by the time that performance becomes relevant in gaming, it'll be time for a new card upgrade anyways biggrin.gif

Thanks for your responses man, appreciate the knowledge!
post #152 of 3357
Does the long version guide work for GTX 670s SLI?
post #153 of 3357
Because, when I turn up my GPU clock offset, for both 670s on Precision X, only the first GPU Boost Clock goes up in GPU-Z..
post #154 of 3357
I'm loving the wealth of information provided by OP teaching.gif +rep, I've got a lot of reading to do.
post #155 of 3357
If my voltage is left at auto, it starts at 1.175 but drops to 1.162 and then to 1.150 as the temperature rises. I'm curious as to why nvidia implemented this since the reduction in voltage seems to have more potential to cause instability than lower temps.

(Running Unigine 3.0)
Voltage set to 1.175 (steady at 1.175): max temp 80
Voltage on Auto (1.175 dropping to 1.150): max temp 80

With auto fan control, fan speed maxes out at 57% in both cases.

One question: Does enabling software fan control have any performance impact? The tooltip in the settings makes it sound like it could possibly cause stuttering. Also, does Precision X need to be running for the fan control to work?

One more question: Where does the <85 degrees recommendation for graphics cards come from? Nvidia's listed max temperature for the 670 is 97C. It doesn't seem likely to me that they would post a max of 97C if 97C actually would damage the card. They would probably leave some headroom right? They would get absolutely no benefit from setting too high a temp ceiling and then seeing lots of cards fail. If 97C would really screw up the card, why wouldn't they set the temp ceiling at 92C? If, say, 86C would damage the card, I would see them setting the ceiling at 81C.

Of course I'm just blindly speculating. Great guide btw, will be posting my results in a bit.
Edited by dizzyorange - 6/11/12 at 1:38pm
post #156 of 3357
Thread Starter 
Sorry guys, I've been busy with Guild Wars 2 beta the last couple days. I just went through the last 10 pages and I'll try to respond to everyone. I'm not a huge fan of multi-qoute as it forces people to read my entire post instead of the information aimed directly at them, but I'm not going to post 10 different posts in a row tongue.gif. To make things easier, i will bold the name of each poster I'm responding to so you can quickly scan through this post and find where i respond directly to your post.
Quote:
Originally Posted by Italianguy View Post

I've been following the guide and when I increased the gpu clock a second time I got the below crash.
Is that the error that shows up when you've overclocked too high? Is there anything that I can tweak to make it possibly work? Voltage and power are already maxed out. Or at this point is the only option bringing down the clock? Thanks.
What hurts is the monitor graphs look so good right up until it crashed. frown.gif
Is "D3D10 App window::swap_window():Device removed." the error that you guys got when overclocking too high? And if not, what does that error mean? Warning: Spoiler! (Click to show)
7170449327_20ae4ea6de_b.jpg

Yes, that error means your overclock is unstable. If you already have the voltage slider maxed out, there's nothing you can do except reduce the offset until it's stable. The locked-voltage just isn't enough to stabilize your GPU at that frequency.

Quote:
Originally Posted by Seltox View Post

So, i've got a Asus GTX670 DirectCU II.. Not TOP, just stock speeds.
Heaven is able to pass a run fine at and a little over +200MHz offset (and I have ~130MHz Kepler Boost), but during a second run through directly after it would crash. Fair enough, wasn't totally stable. So I went down to 195 offset, and it passed two runs. I thought "sweet", and started raising it by 1MHz testing two runs each time. Long story short, I got to 199MHz before it crashed. then on 198 I tried for a third consecutive run (by 'consecutive run' I just mean starting another benchmark as soon as one finishes), and it crashed. Even longer story short, i'm now on 185 offset (1165 boost + ~130 Kepler Boost, it sits stable at 1296MHz according to the chart), and let Heaven run continuously for about 20-30 minutes (not benchmarking, just running through) and it was fine. Sat at about 62 degrees the whole time.
Anyway, what I eventually wanted to get to, was to suggest that you recommend people let Heaven run two or 3 times through once they think they've found their maximum stable clock - just to be sure.

Good recommendation. I was honestly trying to simplify some steps in my guide because running heaven 2-3 times after each offset change will bring the total time from 3 hours to about 6 hours. I will mention in the fine-tuning section of the guide (where you are increasing the offset by 1MHz per step) to run Heaven 2-3 times after each increase as an optional (but recommended) step. thumb.gif

Quote:
Originally Posted by InfoWarrior View Post

I wish these cards wouldnt throttle at only 70 degrees. Thats too soon and limits the OC. AFAIK 70 degrees is perfectly safe. I put in hundreds of hours on my GTX 460's at 80 degrees without a problem.
Yeah, i don't get why they decided to have the card throttle based on temperature at all. I can see the importance of throttling at 85C+ but not at 70C and 80C. I would also like to point out that i think the Asus TOP 670 doesn't throttle until 79C. I made a lot of posts about this in the Asus TOP thread. My Asus TOP specifically doesn't throttle until 79C and then again at 80C.

Quote:
Originally Posted by bern43 View Post

So is everyone using precision x for these cards? I like the ui much more in afterburner. Is pinning the voltage that critical to getting a good overclock?
Sent from my iPad using Tapatalk HD
Yes, it's important. You can get about 13-26MHz higher overclock on some cards that struggle with stability when the voltage drops to 1.162. The Precision-X UI is actually quite nice once you get used to it. I actually prefer it over the Afterburner UI now.

Quote:
Originally Posted by Italianguy View Post

Also no matter what I do I can't seem to get any of the Precision X OSD monitoring numbers to show up when running the Heaven Benchmark (its running in full screen mode). Anyone know what could be wrong? Thanks.
Not sure if the other suggestion of installing the windows update fixed this for you, but you could try setting up a keybind in Precision-X to toggle the OSD on and off and then when heaven is running you can press that key and see if it turns it on.

Quote:
Originally Posted by bern43 View Post

Quick question. The guide mentions that you should back off your gpu clock if you have oscillation as little as 13mhz that doesn't correspond to a gpu usage drop. I'm on a Asus GTX 670 DC2 with a Kepler boost of +144. At +115 to the boost clock (1239) I'm getting 2 or 3 oscillations of 9mhz over the course of one Heaven run. Should I be shooting for no oscillation or is there an acceptable number??? Any help would be greatly appreciated.
Hard to say what exactly will give the best performance. If the oscillations only last a split second each (not a constant down-clock) then you might get better performance with a higher max boost with some down-clocks. But from what i've seen, when the card is forced to throttle itself, the throttle is pretty much constant. For example, your Max Boost might be at 1239 but if the card runs at an average of 1233 over the entire course of the benchmark then that will give you less performance than if you dropped the offset down a bit and could run at a constant max boost of 1236. You could try setting GPU-Z's core-clock setting to 'average' right before each run and see which gives you the highest average max boost. You should also look at the Heaven score at the end and decide based on which ones gives a better score.

Quote:
Originally Posted by Omni322 View Post

Asus GTX 670 TOP Warning: Spoiler! (Click to show)
Stock Boost Clock 1137
Stock Max Boost Clock: 1241
Memory Clock: 3004
Max power and voltage
First off I'd like to say I am very much a newbie when it comes to all things about overclocking. I've read the guide and have a somewhat increased understanding of whats going on. From my numbers I gather my Kepler boost is 104, which i guess is on the lower side. I am actually running Heaven on loop now while I'm at work at stock settings and hoping that I'll find it on when I get home later today. About a 11 hour loop.
One question I have is when I run Heaven at above stock settings, say +20 to the core clock, sometimes it will run and others it will fail(making me think it is unstable at that frequency). What I don't understand is what's causing the failure. I've never seen temps go above 70C, and my power percentage usually stays right around 100%, even though I have it set at the max of 117% and voltage at 1.1175. Shouldn't the card be able to draw a higher power percentage? The EVGA Precision X tool shows that its running at a fairly consistent 1.1175 voltage, so I think that's fine, just not sure why setting the power target higher isnt resulting in the card drawing more power.

Brett already explained this, but in short, the locked-volage of only 1.175V is not enough to stabilize your GPU at that higher frequency. There's nothing you can do to fix it other than reduce your overclock.

Quote:
Originally Posted by djmikea View Post

Thanks for this guide, really useful thumb.gif
I think I got a bit unlucky with my card as I only get 65 MHz Kepler boost, which looks to be low for a Gigabyte card. Still, managed to get up to 1227 Mhz core (+104) and 7108 MHz memory (+550), which I'm happy enough with.
One question, do people actually run at their max overclock 24/7 or is it more sensible to clock it a bit below that?? i.e. 25Mhz less than max, seeing as the fps difference would be negligible but then you're not constantly pushing the card to it's limit?
A 25MHz downclock is probably around 1FPS decrease, it's not completely negligible tongue.gif. But i wouldn't recommend downclocking. As long as your GPU isn't running at 80C+ (which would degrade the life of your GPU, but not enough to actually be meaningful before you replace the gpu in 3-4 years), then there's no reason not to run at your max overclock. When the card is at idle it will automatically downclock itself to 324MHz and .987V no matter what your overclock is at, so you aren't increasing the stress on the card at idle by overclocking.

Quote:
Originally Posted by Lutsk92 View Post

Because, when I turn up my GPU clock offset, for both 670s on Precision X, only the first GPU Boost Clock goes up in GPU-Z..
I'm not really familiar with GPU-Z in relation to SLI, maybe that's a normal bug with GPU-Z? Anyways, just uncheck the 'sync' button in Precision-X and overclock each card independently. Then physically move your best clocking card to the first GPU slot in your SLI array (since GPU1 is ran the hardest).

Quote:
Originally Posted by dizzyorange View Post

If my voltage is left at auto, it starts at 1.175 but drops to 1.162 and then to 1.150 as the temperature rises. I'm curious as to why nvidia implemented this since the reduction in voltage seems to have more potential to cause instability than lower temps.
(Running Unigine 3.0)
Voltage set to 1.175 (steady at 1.175): max temp 80
Voltage on Auto (1.175 dropping to 1.150): max temp 80
With auto fan control, fan speed maxes out at 57% in both cases.
One question: Does enabling software fan control have any performance impact? The tooltip in the settings makes it sound like it could possibly cause stuttering. Also, does Precision X need to be running for the fan control to work?
One more question: Where does the <85 degrees recommendation for graphics cards come from? Nvidia's listed max temperature for the 670 is 97C. It doesn't seem likely to me that they would post a max of 97C if 97C actually would damage the card. They would probably leave some headroom right? They would get absolutely no benefit from setting too high a temp ceiling and then seeing lots of cards fail. If 97C would really screw up the card, why wouldn't they set the temp ceiling at 92C? If, say, 86C would damage the card, I would see them setting the ceiling at 81C.
Of course I'm just blindly speculating. Great guide btw, will be posting my results in a bit.
It could also be reducing the voltage slightly to save power and not just reduce temperatures. Hard to say without graphs. As for your questions though: I don't think the fan control will have any meaningful impact on performance. And yes, you do need to have Precision-X open for the custom fan control to work. The 85C recommendation is just an arbitrary number i chose. Technically speaking, any increase in temperature will degrade component lifespan even if it's well below the safe maximum. For example, if the safe max is 97C then you could run the card at 97C but the card will probably die after 1-2 years of use. If you were to run it at 85C it would probably die after 2-4 years. If you were to run the card at 75C it would probably die after 4-6 years. Each 10C increase in temperature usually leads to a 50% decrease in lifespan.
Edited by SeanPoe - 6/11/12 at 2:14pm
post #157 of 3357
I'm still very new to overclocking and I was wondering, do I just keep my voltage and power target at max for my overclock? Is this not hurting my card?
post #158 of 3357
I'm having some problems with the guide although i'm following everything to the letter. I set the fan curve as shown in the image, maxed out voltage and power. Then I set my offset to +41 for a 1100 boost clock and ran Heaven benchmark with everything maxed out ofc. It gave me a score of 1179 (46.8 FPS). Then I increased my offset by +20 but after running another test, it gave me a slight lower score of 1176 (46.7 FPS). I set my offset back to +41 but my score and FPS were only decreasing with each test. After, I tried this with an offset of +82 and without changing it in between tests, I had the same problem too, score kept decreasing with each test after I started off with a score of 1291 (51.2 FPS). I also noticed that after restarting my PC, running the test with the last offset tested would bring back the score at its highest peak (or somewhere around) but then it would still keep decreasing. What's going on? Is this happening because i'm running 670s SLI with sync turned on in X-precision? If not, please help me out because I don't want to end with a brain tumor.
post #159 of 3357
Lutsk, are you shutting down Heaven and retrying w/higher clocks, or alt-tabbing out, changing clock, and then going back into that same running session? If it's the latter, do the former. Using Alt-Tab even once messes up the way Heaven runs for me. Other than that, I'd watch the graphs for each run, try to reckon what's changing from one to the next. There's likely some kind of a hint there that would point you in the right direction thumb.gif
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
xeon X5675 6-core @ 4.1ghz (1.29v, 20x205 +ht ) rampage iii extreme msi rx470 gaming X (the $159 budget king) 3 x 2gb corsair xms3 pc12800 (9-9-9-24-1T@1600MHz) 
Hard DriveOptical DriveCoolingOS
hynix 250gb ssd (boot), 2tb deskstar (apps),1tb... plextor px-712sa - still the best optical drive... corsair h8o v2 aio W10 home 
MonitorPowerCaseAudio
asus vw266h 25.5" (1920x1200) abs sl (enermax revolution) * single 70A rail 850w silverstone rv-03 XFi Titanium 
  hide details  
Reply
post #160 of 3357
I'm doing none of what you just mentioned.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: NVIDIA
Overclock.net › Forums › Graphics Cards › NVIDIA › ~~The GTX 670 Overclocking Master-Guide~~