How To Overclock A Video Card: NVIDIA & AMD Graphics Card (GPU) Overclocking Guide

[waggythegeek]

i think it's normal for your temperatures to increase like that especially since your running a stress test that forces your gpu to go under 100% load. the r9 280 gigabyte windforce edition is most likely voltage locked. your overclock will be limited as you won't be able to increase the core voltage

 

apparently, the r9 200 series cards run very hot by nature. they can supposedly run up to a safe max operating temperature of 95 degrees according to the article below:

 

http://www.pcgamer.com/amd-radeon-r9-290x-review/

 

as long as your temps stay below 95 degrees or even 90 degrees, i would continue pushing an overclock and trying to find the optimal settings for your card

I know, I just don't think the temperature readings are accurate. If I leave it at stock, it hovers in the high seventies under load. however, if I UNDERCLOCK by anything, the temperature spikes as described above.

[BigDay]

I know, I just don't think the temperature readings are accurate. If I leave it at stock, it hovers in the high seventies under load. however, if I UNDERCLOCK by anything, the temperature spikes as described above.

 

it doesn't matter if you underclock it. you're stress testing. temps go up.

[thekeemo]

Shouldn't the warning be at the start?

[BigDay]

Shouldn't the warning be at the start?

 

excellent post. 5 stars. love it. post again plz.

[waggythegeek]

it doesn't matter if you underclock it. you're stress testing. temps go up.

I know, but stock+stress = high 70's after a couple of minutes. Any under or overclock+stress= 85 within seconds. Shouldn't underclocking make the temps rise slower?

EDIT: issue clarification: Underclocking and stress gets a significant amount hotter than stock and stress.

[xMishax]

is shader complexity 7 important for a stress test of amd card? because I have some problems using it. When I use shader complexity between 4 and 7 my gpu fails to keep at max clocks(1050mhz). It jumps from 1050 to 900 to 990 and rince and repeat, but when I set shader complexity to 3 everything seems to be okay and core clocks are always at max values. Should I be worried?

update: ok it got really weird. There is some connection between msi afterburner and occt. at first I thought it was because of shaders. I had to lower down shader complexity to have stable core clocks under 100 gpu load in occt. Now when I reached 1075mhz mark on core and try to go higher e.g 1085mhz gpu cant keep stable core clocks even with 1 shader complexity. WTH is happening? Note: in benchmarks such as heaven, kombuster,furmark, valley clocks seem to be stable all the time even above 1075mhz with 100proc gpu usage

[BigDay]

is shader complexity 7 important for a stress test of amd card? because I have some problems using it. When I use shader complexity between 4 and 7 my gpu fails to keep at max clocks(1050mhz). It jumps from 1050 to 900 to 990 and rince and repeat, but when I set shader complexity to 3 everything seems to be okay and core clocks are always at max values. Should I be worried?

update: ok it got really weird. There is some connection between msi afterburner and occt. at first I thought it was because of shaders. I had to lower down shader complexity to have stable core clocks under 100 gpu load in occt. Now when I reached 1075mhz mark on core and try to go higher e.g 1085mhz gpu cant keep stable core clocks even with 1 shader complexity. WTH is happening? Note: in benchmarks such as heaven, kombuster,furmark, valley clocks seem to be stable all the time even above 1075mhz with 100proc gpu usage

 

have you turned off all of the power saving features native to your amd card? i give instructions on how to do this in the guide. it could be a power saving feature issue in the sense that you still have something enabled

 

what kind of card do you have? which resolution are you running occt at?

 

i have a 7950 hd and i have no problems running with shader complexity at 7 with a resolution of 1080p

[xgn]

My gpu wont overclock past 70mhz gtx 780 100mhz clock is unstable no matter what voltage i use no voltage or 10 voltage eventually driver crash.

Do i need to overclock the pci lane or something else which can fix it? the temps stay below 40c trough valley benchmark.

[xMishax]

have you turned off all of the power saving features native to your amd card? i give instructions on how to do this in the guide. it could be a power saving feature issue in the sense that you still have something enabled

what kind of card do you have? which resolution are you running occt at?

i have a 7950 hd and i have no problems running with shader complexity at 7 with a resolution of 1080p

If I turn off all the power saving settings youve mentioned on amd cards here what happens - pc crashes after I press apply oc that is higher than 1030mhz, black screen must restart.if I uncheck option about display reset after applying unofficial oc, everything seems to be fine, I can apply oc, but with oc higher than 1030mhz fans wont spin even when set to 100proc and when I try to run occt it BSOD's. You know anything about that?

Ps. Card is r9 380 from msi

One more thing Ive noticed - if i disable all power savings and and put clocks below 1030mhz (lets say 1025)clocks wont apply. Voltages go higher, power limit goes higher but in monitoring its shows thats clocks are topped at 980mhz(my cards stock clock out of the box)

[BigDay]

My gpu wont overclock past 70mhz gtx 780 100mhz clock is unstable no matter what voltage i use no voltage or 10 voltage eventually driver crash.

Do i need to overclock the pci lane or something else which can fix it? the temps stay below 40c trough valley benchmark.

 

What is your ASIC score? Use GPU-Z to find it. You could just have a bad video card.

 

Are you using OCCT to overclock your card? Are you using MSI AB? Make sure you have the latest drivers for your gpu.

[xgn]

What is your ASIC score? Use GPU-Z to find it. You could just have a bad video card.

 

Are you using OCCT to overclock your card? Are you using MSI AB? Make sure you have the latest drivers for your gpu.

It says 59,9% Not sure what it means the pci lane on mobo should handle 4 times as fast  gpu but maybe its worth trying to tweak it a little anyway.

Its a zotac card.

[BigDay]

If I turn off all the power saving settings youve mentioned on amd cards here what happens - pc crashes after I press apply oc that is higher than 1030mhz, black screen must restart.if I uncheck option about display reset after applying unofficial oc, everything seems to be fine, I can apply oc, but with oc higher than 1030mhz fans wont spin even when set to 100proc and when I try to run occt it BSOD's. You know anything about that?

Ps. Card is r9 380 from msi

One more thing Ive noticed - if i disable all power savings and and put clocks below 1030mhz (lets say 1025)clocks wont apply. Voltages go higher, power limit goes higher but in monitoring its shows thats clocks are topped at 980mhz(my cards stock clock out of the box)

 

Unfortunately, I've never had to deal with your particular situation before. Try playing around with MSI AB. If you find that OCCT is not working well for you, try using the EVGA tool that I mentioned in the guide.

 

These guys managed to get the r9 380 from msi up to 1200 on the core:

 

http://www.guru3d.com/articles_pages/msi_radeon_r9_380_gaming_2g_review,21.html

 

Try resetting all of the MSI AB settings to default and try overclocking the card with your power saving features enabled. What happens then? Are you still able to achieve consistent core and memory clocks during stress testing?

[BigDay]

It says 59,9% Not sure what it means the pci lane on mobo should handle 4 times as fast  gpu but maybe its worth trying to tweak it a little anyway.

Its a zotac card.

 

some say that the asic quality is a measure of how good your particular gpu is. 59.9% is not that high so this may be a good indication that your card may not overclock that well.

 

which stress testing tool have you been using? when you say that it won't overclock past 70mhz on the core, what do you mean? are you getting errors in occt?

[xgn]

some say that the asic quality is a measure of how good your particular gpu is. 59.9% is not that high so this may be a good indication that your card may not overclock that well.

 

which stress testing tool have you been using? when you say that it won't overclock past 70mhz on the core, what do you mean? are you getting errors in occt?

If i go 90-120 it will work a while then driver crash. Once i got it to run on +500 which is 1780 and it passed whole valley benchmark.

When i tried playing a game instantly crash and screen black out had to restart. so for now the highest stable overclock is +70 i dont get it.

This was with +800 vram boost +500 clock and +22v and it never worked again weird i know.

I tried with msi ab and zotac firestorm seem to be same result the auto voltage function in msi ab dosen't help either.

I should mention i use +70 clock and +30 vram but the vram dosent make any difference judging by msi ab sensors .

[BigDay]

If i go 90-120 it will work a while then driver crash. Once i got it to run on +500 which is 1780 and it passed whole valley benchmark.

When i tried playing a game instantly crash and screen black out had to restart. so for now the highest stable overclock is +70 i dont get it.

This was with +800 vram boost +500 clock and +22v and it never worked again weird i know.

I tried with msi ab and zotac firestorm seem to be same result the auto voltage function in msi ab dosen't help either.

I should mention i use +70 clock and +30 vram but the vram dosent make any difference judging by msi ab sensors .

 

have you followed my guide exactly as it's been laid out?

[xgn]

have you followed my guide exactly as it's been laid out?

Nope. I will try it now thanks.

[BigDay]

Nope. I will try it now thanks.

 

i highly recommend you try using my guide and following everything exactly as it's been laid out. once you've done this, let me know how it goes.

 

running valley is not a good measure of stability

[xgn]

i highly recommend you try using my guide and following everything exactly as it's been laid out. once you've done this, let me know how it goes.

 

running valley is not a good measure of stability

I see i will look at it now.

Also the best benchmark i did so far is this http://pastebin.com/iHyGF32N but not stable so i guess it dosen't matter.

[xgn]

msi ab wont save profiles for some reason.

OZ scanner artifact scanner sometimes work sometimes not it says disabled when i start the benchmark.

[xMishax]

Unfortunately, I've never had to deal with your particular situation before. Try playing around with MSI AB. If you find that OCCT is not working well for you, try using the EVGA tool that I mentioned in the guide.

 

These guys managed to get the r9 380 from msi up to 1200 on the core:

 

http://www.guru3d.com/articles_pages/msi_radeon_r9_380_gaming_2g_review,21.html

 

Try resetting all of the MSI AB settings to default and try overclocking the card with your power saving features enabled. What happens then? Are you still able to achieve consistent core and memory clocks during stress testing?

AFAIK EVGA wont work with AMD cards  With all the power saving features and without constant voltage Im able to get stable 1020hz (+0mV) and 1075mhz(+50mV), but when testing with 1075mhz Im using 1 shader complexity as anything higher than that makes my card jump core clocks like mad (300-1075mhz) and thus it cannot show errors. Also going more than 1075mhz with shader complexity one does same - core clocks jump.

I found on the internet how to extend power limit on amd cards from 20proc to 50proc. I tried it. When I go e.g. +80mV and 1100Mhz with 7 shader complexity everything seems to be fine at first. But as far as card reaches somewhere 70-71C it starts throttling for unknown reason and clocks go down as far as 300mhz... 

UPDATE: yap its gpu throttling. No matter what benchmarks I use on what settings, whenever it reaches 70-71C it will throttle down to 2D clocks.

[xgn]

Ok so OZ scanner artifact check only work on normal mode not benchmark. Im able to go +220 mhz without changing the voltage which is a HUGE improvement.

I think turning off power saving made all the difference. So far temps are on 43c.

The pink number in top left screen of OC scanner is that the power usage? is there magic number for it? should i try to stay under 100?

so when ti goes to 103-105 i turn up the voltage? Also why do you have so long testing times in you're tutorial? im doing 1-2 min tests now with small increments.

Obviously i will be testing longer once i get where i want to be.

Also should i edit tdrdelay in regedit to solve some crashing issues or is that considered not stable? because im coming to a point where it just crash over and over

when i pass 250mhz i try different voltage and i get crashes.

Ok my conclusion is that it just caps out somewhere between +90-110 for me. Either the card sucks or its because im on pci2

I overclocked the pci lane a bit but only to 118 it wont let me boot if i chose 119. Overall though im pretty happy about it stock settings are good.

[BigDay]

AFAIK EVGA wont work with AMD cards  With all the power saving features and without constant voltage Im able to get stable 1020hz (+0mV) and 1075mhz(+50mV), but when testing with 1075mhz Im using 1 shader complexity as anything higher than that makes my card jump core clocks like mad (300-1075mhz) and thus it cannot show errors. Also going more than 1075mhz with shader complexity one does same - core clocks jump.

I found on the internet how to extend power limit on amd cards from 20proc to 50proc. I tried it. When I go e.g. +80mV and 1100Mhz with 7 shader complexity everything seems to be fine at first. But as far as card reaches somewhere 70-71C it starts throttling for unknown reason and clocks go down as far as 300mhz... 

UPDATE: yap its gpu throttling. No matter what benchmarks I use on what settings, whenever it reaches 70-71C it will throttle down to 2D clocks.

 

i'm sorry, i wish i could help, but i've never run into this problem before

[BigDay]

Ok so OZ scanner artifact check only work on normal mode not benchmark. Im able to go +220 mhz without changing the voltage which is a HUGE improvement.

I think turning off power saving made all the difference. So far temps are on 43c.

The pink number in top left screen of OC scanner is that the power usage? is there magic number for it? should i try to stay under 100?

so when ti goes to 103-105 i turn up the voltage? Also why do you have so long testing times in you're tutorial? im doing 1-2 min tests now with small increments.

Obviously i will be testing longer once i get where i want to be.

Also should i edit tdrdelay in regedit to solve some crashing issues or is that considered not stable? because im coming to a point where it just crash over and over

when i pass 250mhz i try different voltage and i get crashes.

Ok my conclusion is that it just caps out somewhere between +90-110 for me. Either the card sucks or its because im on pci2

I overclocked the pci lane a bit but only to 118 it wont let me boot if i chose 119. Overall though im pretty happy about it stock settings are good.

 

sorry, i don't know what the pink number is that you're referring to. stress testing is designed to emulate short and intense gaming sessions. 30-45 minutes is advised. i wouldn't edit anything in your registry.

[Albatross]

Alrighty, so I'm trying to OC so that the Star Citizen Alpha 2.1 gets me over 28-30fps (before I was 40-45) so I thought "time to overclock". So I stumbled across this guide (very well laid out by the way). So onward, I followed the settings listed in the OP (can't get passed "2. Start & Progression" of the guide) to the dot and I can't seem to get it to allow me to change core voltage in Afterburner or TRIXX. I've edited my CFG, tried restarting the PC, reinstalling etc etc but it just won't budge. I've got a Sapphire 280 with the latest drivers from AMD.

Here's some screenies:

 

(I cycled through themes hoping that would help lol)

 

 

Am I missing something?

[BigDay]

Alrighty, so I'm trying to OC so that the Star Citizen Alpha 2.1 gets me over 28-30fps (before I was 40-45) so I thought "time to overclock". So I stumbled across this guide (very well laid out by the way). So onward, I followed the settings listed in the OP (can't get passed "2. Start & Progression" of the guide) to the dot and I can't seem to get it to allow me to change core voltage in Afterburner or TRIXX. I've edited my CFG, tried restarting the PC, reinstalling etc etc but it just won't budge. I've got a Sapphire 280 with the latest drivers from AMD.

Here's some screenies:

 

(I cycled through themes hoping that would help lol)

 

 

Am I missing something?

 

what is the model number of your video card? your video card may be voltage locked. have you tried researching this possibility online?