GTX 570 | i5 2500k Small Overclock

[Vario]

100% agree with vayra, couldn't have said it better.

edit: there are airflow tweaks that you can do that will help with cooling, might help stabilize overclocks, and will definitely extend the lifespan of your components. For example, removing PCI slot covers. Considering placement of HDD. Tucking wiring.

 

[INSTG8R]

So simply, the offset value just gives the CPU the ''go ahead'' to add more volts that it wouldnt otherwise add? I would be really suprised if I needed even 1.35 to get the 4.2 stable. But its nice to know if I ever want to pull more out of it.

Well when I was OCing mine was using 1.35@4.6 but I ran the smallest negative offset available (like 0.01) to compensate for the MoBo increasing the voltage with the clocks(so 1.40v if left on its own) I think you’re not even close to pushing it 1.30v is great IMO. I will also echo don’t touch the Bus that could have easily caused your instability.

 

[las]

My experience is similar: CPU overclocks are incredibly reliable and GPU overclocks are generally just that final push that an old GPU needs to get pushed over the edge and die fast.

There are noticeable differences between a CPU and GPU.

- GPUs have VRM and VRAM chips rather close together. Many VRAM chips are extremely temperature intolerant; and their max rated temps are always lower than the VRM specs. EVGA's 1080 FTWs started dying because of exactly that: insufficient VRM cooling > heat radiated onto VRAM > dead cards
- GPUs are clocked much closer to their limits out of the box AND get binned more tightly than CPUs + used for different card tiers
- Older GPUs have reduced cooling performance in general. Only those meticulously cleaned and perhaps even repasted can get some of the old glory back, but never everything. A CPU heatsink is much easier to keep at 'level' performance over time.
- GPUs are generally run at max performance and utilization most of the time, and for older GPUs that is even more likely than it is for new ones, further increasing heat/wear/tear. CPUs however, even under gaming loads, are mostly *idle*. Even a game that hammers one thread will get moved to different cores all the time. The GPU uses all resources in parallel.

Casual OC or not, all these products do stop working at some point and for GPUs the life expectancy is much lower than it is for a CPU. If you haven't seen this yet, you either haven't owned and kept using older GPUs for a long time or you've not experienced stock or AIB air cooling and its behaviour over time. Or just incredibly lucky / a combination of these options.

Another dead giveaway is the number of threads about dead or dying GPUs compared to those about dead or dying boards/CPUs. That alone is 'nuff said' and no it ain't mining that causes this, those threads existed long before BItcoin.

But more important than all of this: the performance gain on a GPU OC of that calibre is non-existant. So why risk it.

Performance gain from OC on 570 is greater than 1070. I remember my 580 gained tons of performance from stock to max OC (1 GHz).
It makes much more sense OC'ing older GPU's, because they are already slow and performance is bad. 15-20% higher fps is not what I'd call "non-existant" when fps is already low.

I have OC'ed GPU's for more than 20 years and never seen a GPU dying from casual OC. Only from custom firmware or voltmods over time (still took YEARS).
My 980 Ti has been running 1580 MHz since day one, 2015, with custom firmware (max voltage). Zero issues. Rock solid. No signs of degradation whatsoever. Performance is aprox. 40-45% above 980 Ti reference/stock speed. Matching 1080 perf in pretty much all games.

If GPU overclocking was dangerous you would not see custom cards ramp up the clocks like crazy. Even reference cards can handle much more than stock speeds no problem. They will just be more noisy.
Not exactly sure why people are afraid of casual OC - Zero risk involved. If it's stable and temps are decent, it will all be fine.

 

[cameronh779]

I will get pictures of my case tomorrow and we can all talk about airflow and other tweaks. I'm now kinda confused based on @INSTG8R post. You are saying that you decreased the offset value because the motherboard would use too high voltage for the clock that you had correct? Also you said its good im not going past 1.3.. why is that? Isn't 1.28 pretty close to that?

I'm hearing some pretty good reasoning from both sides of this debate about OCing the GPU. Thanks for debating in a civil manner everyone. I'm leaning towards leaving it alone simply out of fear from differing opinions. But im still listening and open to hearing more.

 

[Vario]

1.2-1.35 voltage is kind of the safe zone, 1.3 puts you solidly in it. In truth you can go even higher, but your motherboard is older. You are getting conservative advice.

He is saying that if he didn't run negative offset and just ran voltage on the auto setting it would be too high so he runs a tiny bit of negative offset and it put his load vcore where he wanted it, being 1.35V.

 

[Vayra86]

Performance gain from OC on 570 is greater than 1070. I remember my 580 gained tons of performance from stock to max OC (1 GHz).
It makes much more sense OC'ing older GPU's, because they are already slow and performance is bad. 15-20% higher fps is not what I'd call "non-existant" when fps is already low.

I have OC'ed GPU's for more than 20 years and never seen a GPU dying from casual OC. Only from custom firmware or voltmods over time (still took YEARS).
My 980 Ti has been running 1580 MHz since day one, 2015, with custom firmware (max voltage). Zero issues. Rock solid. No signs of degradation whatsoever. Performance is aprox. 40-45% above 980 Ti reference/stock speed. Matching 1080 perf in pretty much all games.

If GPU overclocking was dangerous you would not see custom cards ramp up the clocks like crazy. Even reference cards can handle much more than stock speeds no problem. They will just be more noisy.
Not exactly sure why people are afraid of casual OC - Zero risk involved. If it's stable and temps are decent, it will all be fine.

If a game runs 20 fps now and you get 20% that makes 24 and doesn't change all that much about the experience... but in reality its not even that, its less because Fermi is ill equipped for newer engines. Back then, it was definitely worth doing. Today? Nah. The card is just too old.

But anyway, OP has the opinions from both ends, choice and power to him whichever choice it is

 

[BiggieShady]

Be gentle to your fermi, my 560 Ti kicked the bucket after 5 years of light oc ... iirc one vram ic died

 

[John Naylor]

The GTX 570 was the 1st rush of toasted VRMs we saw on nVidia cards whereby some AIB manufacturers allowed for significant increases in power / voltages. Fried VRMs were very common on reference boards because the VRMs were not robust enough for overclocking. As the EVGA SC series has just about always been a stock nVidia PCB with a nice cooler. Since then we have not recommended the SC series to our users. And while nVidia has made the reference VRMs a bit more robust in recent generations, when ya paying the same price, and Asus, MSI, Gigabyte, etc almost all have beefed the power delivery up a bit for there gaming cards, I like the beefier option. But yes, it is still overclockable, but don't go web searching and copy voltage / power settings from cards with more robust power delivery.

http://www.overclock.net/forum/69-nvidia/912139-fried-my-evga-gtx580-superclocked.html
http://www.overclockers.com/forums/showthread.php/695321-My-gtx570-caught-on-fire-for-real

 

[INSTG8R]

1.2-1.35 voltage is kind of the safe zone, 1.3 puts you solidly in it. In truth you can go even higher, but your motherboard is older. You are getting conservative advice.

He is saying that if he didn't run negative offset and just ran voltage on the auto setting it would be too high so he runs a tiny bit of negative offset and it put his load vcore where he wanted it, being 1.35V.

Exactly what I was doing. ASUS is a little ham fisted with the voltage so putting a small negative offset reeled it in to reasonable levels.

 

[cameronh779]

Okay so my computer ran P95 for a while and I watched movies on it all night no crashes. I'm still @ 4.2Ghz with 1.28 volts. Should I drop the volts a bit now? By what increment? I plan on cleaning my case and working on airflow this week and we will follow up on that. I think for now, im going to leave the GPU alone. I think considering its old and that it was given to me and I dont know much about its history I dont want to push it, also considering that im only playing Overwatch and its running well enough for me. I do want to get a good fan profile and some help with that here soon in this thread that we can talk about. On an auto fan profile it reaches 85-87C and I know its ''safe'' but I still am a bit uncomfortable with it and want to work on fan profile that keeps it under a certain temp while being as quiet as possible.

I'd like to thank everyone for all the help. I wouldn't go anywhere else but here for the opinions and knowledge.

 

[Vya Domus]

Fermi can just die on higher than stock voltages. But if you just mess with clocks I can't see why you wouldn't do it. To be honest the card is so old it might just as well kick the bucket at any moment anyway , I wouldn't worry over a small OC.

You may also want to change the thermal paste.

 

[Vario]

Okay so my computer ran P95 for a while and I watched movies on it all night no crashes. I'm still @ 4.2Ghz with 1.28 volts. Should I drop the volts a bit now? By what increment? I plan on cleaning my case and working on airflow this week and we will follow up on that. I think for now, im going to leave the GPU alone. I think considering its old and that it was given to me and I dont know much about its history I dont want to push it, also considering that im only playing Overwatch and its running well enough for me. I do want to get a good fan profile and some help with that here soon in this thread that we can talk about. On an auto fan profile it reaches 85-87C and I know its ''safe'' but I still am a bit uncomfortable with it and want to work on fan profile that keeps it under a certain temp while being as quiet as possible.

I'd like to thank everyone for all the help. I wouldn't go anywhere else but here for the opinions and knowledge.

Just drop the CPU vcore an increment, you could set the offset to an additional -.01, then observe in hwmonitor and see what the load voltage is. Ideally it will be dropped to ~1.27 or ~1.26 or similar, just drop it a bit and see if it is stable. Eventually it will become unstable, then you raise it back up two increments to be sure. Always check the voltage with hwmonitor after you make the change to make sure you aren't drastically overvolting it or something.

If it runs well enough leave the graphics card alone. What can drop the graphics card temperatures and cost you nothing is to remove the PCI slot covers, then the graphics card can pull in air through them.

 

[cameronh779]

Just drop the CPU vcore an increment, you could set the offset to an additional -.01, then observe in hwmonitor and see what the load voltage is. Ideally it will be dropped to ~1.27 or ~1.26 or similar, just drop it a bit and see if it is stable. Eventually it will become unstable, then you raise it back up two increments to be sure. Always check the voltage with hwmonitor after you make the change to make sure you aren't drastically overvolting it or something.

If it runs well enough leave the graphics card alone. What can drop the graphics card temperatures and cost you nothing is to remove the PCI slot covers, then the graphics card can pull in air through them.

So would this be setting my voltage to auto then adjusting another setting being the offset voltage?

 

[Vario]

So would this be setting my voltage to auto then adjusting another setting being the offset voltage?

You'd have to see what your voltage is on auto first and then dropping offset until its in the ballpark 1.25-1.28 range and then stability test. How are you doing vcore adjustments right now? Does the voltage drop when you are idle the way you are doing it now?

 

[cameronh779]

You'd have to see what your voltage is on auto first and then dropping offset until its in the ballpark 1.25-1.28 range and then stability test. How are you doing vcore adjustments right now? Does the voltage drop when you are idle the way you are doing it now?

I have it manually set to 1.28. During P95 it will bounce around like 1.272 or 1.288 then 1.28. Every few seconds or so. Also, I just noticed its not lowering @ idle like it was. It would go down to like 1.4Ghz and such when I used OC tuner. But since ive gone in and changed manually my multi, bus, and voltage its not clockinjg down like was.

When I was using the auto OC and even before overclock it would always downclock when idle. My volts would go down to around 0.9 or so and maybe 1.4GHz. I actually want that back.

 

[WhiteNoise]

That 2500K is a beast cpu. I still own and operate one and it has been running at 4.6GHz since the CPU released and it is running 24/7. Great CPU. I wouldn't bother OCing the video card myself.

 

[cameronh779]

I just read that it may be the windows power settings. Its always been on high performance and downclocked volts and speed. I changed it to balanced and power saver trying both. It clocks it down now to 1.6GHz but still isnt bringing down the volts which still even at idle are bouncing around from 1.272, 1.28, 1.288, and even up to 1.296.

EDIT: Went in and changed voltage back to auto so that it clocks down correctly, running prime95 now to check load voltages. There are about four different voltages ranging from 1.296 to 1.312. I'm going to leave it for now until @Vario lets me know what I should now change the offset to. FYI, the idle voltages bounce around from 0.888 to 1.0 with lots of different numbers inbetween.

 

[Vayra86]

Bouncy voltages are good, it means Speedstep and power saving states are functional

 

[Vario]

I just read that it may be the windows power settings. Its always been on high performance and downclocked volts and speed. I changed it to balanced and power saver trying both. It clocks it down now to 1.6GHz but still isnt bringing down the volts which still even at idle are bouncing around from 1.272, 1.28, 1.288, and even up to 1.296.

EDIT: Went in and changed voltage back to auto so that it clocks down correctly, running prime95 now to check load voltages. There are about four different voltages ranging from 1.296 to 1.312. I'm going to leave it for now until @Vario lets me know what I should now change the offset to. FYI, the idle voltages bounce around from 0.888 to 1.0 with lots of different numbers inbetween.

Try running -.01 offset and see if it drops to 1.29. If so, then try -.02 and try for 1.27 and so on. Since you tested 1.28 and it worked aim for that or a little less.

 

[mouacyk]

The GTX 570 was the 1st rush of toasted VRMs we saw on nVidia cards whereby some AIB manufacturers allowed for significant increases in power / voltages. Fried VRMs were very common on reference boards because the VRMs were not robust enough for overclocking.

This was exactly how I lost mine, which was an SC edition and it even had the beefy aftermarket ZALMAN VF3000F cooler on it. The GPU temp was cool (<70C), but the VRM blew with a moderately overvolted OC. I baked it once and it came back to life for about 3 hours in Witcher 2 and then died for good. Only afterwards did I google, and found testimonies that the reference 570 had a notoriously weak VRM with plenty of other similar experiences.

 

[cameronh779]

@Vario

Changed voltage to offset mode and -.01. Ran P95 for about ten minutes. Volts are topping @ 1.28. Going to keep my computer going all night and will make another 0.01 change in the morning if recommended.

 

[Vario]

Nice. Looking good. Make that reduction and see if it is stable, eventually you will find the sweet spot. Reducing the voltage should reduce your temperature and that will help with component life span.

 

[cameronh779]

Nice. Looking good. Make that reduction and see if it is stable, eventually you will find the sweet spot. Reducing the voltage should reduce your temperature and that will help with component life span.

What temps should I be aiming for considering the situation? I may have to go down on the clock until I can get better fans. My 120mm in the back of my case that was pulling the hot air out conked out. I still have one above it that pulls air out, as well as one in front that pulls air in. My case definitely needs to be cleaned. Very dusty. My cooler seems to be seated perfectly but im not sure if temps should be so high with this 212 on it. What do you think? I don't have any thermal paste except for some white unbranded stuff in a small syringe. If these temps make ''sense'' ill assume its seated right and leave it be.

As far as voltage.. Should I just go down 0.01 on the offset until it doesnt boot.. then go in and add a point.. then wait 24 hours and so on? Or do the way im doing now where I see if it lasts the 24 hours and then lowering it?

 

[Vario]

I would slowly lower the offset until it doesn't boot and then add a point and run Prime for an hour and see if stable, then try adding another point if not, once you think its stable then I'd play games on it or use it as you would normally and if it blue screens, add another point. Long P95 sessions will likely just wear your motherboard without much gain. Good temps would be something like 30*C idle and 65*C load. Maybe 60*C in games.

 

[BraveSoul]

ran 2500k at 4.5ghz for years with beefy air cooler(noctuaD14 & 1.39V in bios , amazing CPU would still have it if not for recent video editing hobby

GTX 570 big overclocks will likely bring more risk than gains, especially in modern games, I would keep it closer to stock if longevity is a priority