Tips for overclocking GTX 970
you should join our GTX 970 owners club, you will get lots more help as thats all we do is overclock, and now we have new bios for most 970 gpu brands. hope to see you there. :) [url]http://www.overclock.net/t/1514085/official-nvidia-gtx-970-owners-club[/url]
you should join our GTX 970 owners club, you will get lots more help as thats all we do is overclock, and now we have new bios for most 970 gpu brands. hope to see you there. :)
http://www.overclock.net/t/1514085/official-nvidia-gtx-970-owners-club

#1
Posted 10/30/2014 03:27 AM   
Nvidia caps the amount you can adjust the voltage by a lot, so with a modern Nvidia it's not possible to kill it with adjusting voltage; even with 400 series cards (not sure how far back though). AMD afaik gives you full control there, so its possible with AMD cards. Increase it as need be with no worry. [color="gray"][i][size="L"]~ In a nutshell ~[/size][/i][/color] [list] [.][color="orange"][b][u]Core Clock:[/u][/b][/color]For the first time, increase in increments of +20 MHz or so. Eventually, you will start to get display crashes, system crashes, throttling, artifacting. When it happens, keep nudging off until you appear stable, so keep decreasing it by -5 or so each go. Set it back to stock to start testing memory.[/.] [.][color="orange"][u][b]Memory Clock:[/b][/u][/color] Same as above. Keep in mind memory can be clocked considerably higher than the core each and every time. So for this card here, you can probably start with +200 MHz and then keep increasing it marginally.[/.] [.][color="orange"][b][u]Testing All Together:[/u][/b][/color] Apply both the highest core and memory clocks you found that were stable, and do a few runs with multiple programs. Chances are with both clocks applied, stuff will happen, so you'll have to nudge off -5 MHz on each clock until stable.[/.][/list] Save what you find/are happy with so you always have it. [color="gray"][i][size="L"]~ When You Want More ~[/size][/i][/color] [olist] [.]In Afterburner, or other software, you need to unlock the voltage control as it's disabled by default. In Afterburner > Settings > Tick "Unlock Voltage Control" and "Unlock Voltage Monitoring" and let the program restart.[/.] [.]For Afterburner, make sure you're not using an old custom skin that predates GPU Boost, or use one of the stock skins. If you're using an old skin, new parameters like "Power Limiter", "Temperature", and "Link..." will not be available.[/.] [/olist] [size="S"][u][i]Random Tip[/i][/u]: Press the Arrow next to Power Limiter and it will expand to show a few new things, including Link, Temperature, and a new arrow. [.][color="green"]Link[/color] keeps the Power Limiter and Temperature proportionally together (untick it unless you desperately want this).[/.] [.][color="green"]Power Limiter[/color] is basically how much headroom the GPU has to increase it's own boost clock, something you have no control over and is dynamic (feel free to set it to max if you want). When this target is reached, the boost clock will stop increasing and adjust itself as need be. Do not confuse this with GPU utilization.[/.][.][color="green"]Temperature[/color] is sort of like power limiter. When the designated GPU temperature is reached, GPU Boost will dynamically adjust itself so it doesn't surpass the temperature and goes under it.[/.][.][color="green"]Prioritize Arrow[/color] will tell the GPU which setting to hold itself to when it comes to dynamically overclocking itself, its boost clock. GPU Boost 1, which was used with the 600 series, was usually limited in capability due to Power Limiter being the forced prioritized target. With GPU Boost 2, we can designate if we want Power Limiter to still be the main target, or temperature. With Temperature as the target, our capabilities are increased. Unless you have some specific reason, set the arrow so its pointed down at temperature. Set temperature to whatever you like, such as 80, 90, 76, etc. Setting it too low will impact your own overclocking.[/.][/size] [b][u][color="orange"]Core Voltage:[/color][/u][/b] When you hit the point you want higher clocks, you increase this until problems subdue (just like before, core only, then memory only, then both). It's trial and error. Maybe you want +110 core/+250 mem, but artifacting is occurring/crashing. So increase the voltage by a few mV and see if the problem goes away. Keep nudging it down until problems starts up again and give it a bit more for some headroom if needed. When you combine both clocks, you will likely need to increase the voltage. Don't worry about applying too much, you can't do any damage at these levels. [size="L"][color="gray"][i]~ Useful Programs ~ (clicky)[/i][/color][/size] [url=http://www.geeks3d.com/20140908/furmark-1-14-1-gpu-burn-in-utility-stress-test-download/]FurMark[/url] [size="S"]Use 0xMSAA when testing core, 8xMSAA when memory and both together. Disable PostFX. Enable Xtreme BurnIn and Dynamic Background and FullScreen. Use Burn In, not Benchmark. I think F3 or F7 will show an error counter (unreliable, one of the F# keys). If crash, black screen, then overvolt/reduce clock. PRO TIP: Stare at center of screen and if you see what seems like weird tearing with colors and stuff, increase voltage or cut back on clock. If you see white pixel snow, do same.[/size] [url=http://www.ocbase.com/]OCCT[/url] [url=https://unigine.com/products/valley/]Unigine Valley[/url] [url=https://www.evga.com/ocscanner/]OC Scanner[/url] [size="S"]Tick artifact scanner. Use Run, not Benchmark and in Fullscreen. Same AA rules as Furmark. Look for artifacts, not sure how reliable the counter is.[/size] When that's all done and you have your max total overclock, and multiple profiles if you want, run some games, preferably heavy hitters like Metro 2033, Crysis 1, Crysis 3, and other modern games. If you ever see those artifacts, [b]EXTREMELY[/b] bad tearing, white pixelated snow, weird flashing and stuff that didn't happen before, reduce your clocks and or increase voltage. Memory clocks that are too high will act like core clocks too high, and more often crash if anything. I personally use mostly Furmark, memory with another program, then games. Keep in mind your max may cause problems in some games and not others. I.e. Bioshock Infinite doesn't like mine, but has no problems with other games so far. In those cases, modify them as need be and reduce. [i]Lastly, feel free to assign hotkeys to your profiles in Afterburner. I use Ctrl+Alt+1-5 for my profiles to change them on the fly. If you have questions, no matter how dumb you may feel it is, ask away. It's late right now as I write this, so I think I covered pretty much everything. You can also use figures of others for a general ballpark figure, but don't expect to get that high, and head towards it, not starting at it.[/i]
Nvidia caps the amount you can adjust the voltage by a lot, so with a modern Nvidia it's not possible to kill it with adjusting voltage; even with 400 series cards (not sure how far back though). AMD afaik gives you full control there, so its possible with AMD cards. Increase it as need be with no worry.

~ In a nutshell ~
  • Core Clock:For the first time, increase in increments of +20 MHz or so. Eventually, you will start to get display crashes, system crashes, throttling, artifacting. When it happens, keep nudging off until you appear stable, so keep decreasing it by -5 or so each go. Set it back to stock to start testing memory.


  • Memory Clock: Same as above. Keep in mind memory can be clocked considerably higher than the core each and every time. So for this card here, you can probably start with +200 MHz and then keep increasing it marginally.


  • Testing All Together: Apply both the highest core and memory clocks you found that were stable, and do a few runs with multiple programs. Chances are with both clocks applied, stuff will happen, so you'll have to nudge off -5 MHz on each clock until stable.


Save what you find/are happy with so you always have it.


~ When You Want More ~
  1. In Afterburner, or other software, you need to unlock the voltage control as it's disabled by default. In Afterburner > Settings > Tick "Unlock Voltage Control" and "Unlock Voltage Monitoring" and let the program restart.
  2. For Afterburner, make sure you're not using an old custom skin that predates GPU Boost, or use one of the stock skins. If you're using an old skin, new parameters like "Power Limiter", "Temperature", and "Link..." will not be available.


Random Tip: Press the Arrow next to Power Limiter and it will expand to show a few new things, including Link, Temperature, and a new arrow.
  • Link keeps the Power Limiter and Temperature proportionally together (untick it unless you desperately want this).
  • Power Limiter is basically how much headroom the GPU has to increase it's own boost clock, something you have no control over and is dynamic (feel free to set it to max if you want). When this target is reached, the boost clock will stop increasing and adjust itself as need be. Do not confuse this with GPU utilization.
  • Temperature is sort of like power limiter. When the designated GPU temperature is reached, GPU Boost will dynamically adjust itself so it doesn't surpass the temperature and goes under it.
  • Prioritize Arrow will tell the GPU which setting to hold itself to when it comes to dynamically overclocking itself, its boost clock. GPU Boost 1, which was used with the 600 series, was usually limited in capability due to Power Limiter being the forced prioritized target. With GPU Boost 2, we can designate if we want Power Limiter to still be the main target, or temperature. With Temperature as the target, our capabilities are increased. Unless you have some specific reason, set the arrow so its pointed down at temperature. Set temperature to whatever you like, such as 80, 90, 76, etc. Setting it too low will impact your own overclocking.


  • Core Voltage: When you hit the point you want higher clocks, you increase this until problems subdue (just like before, core only, then memory only, then both). It's trial and error. Maybe you want +110 core/+250 mem, but artifacting is occurring/crashing. So increase the voltage by a few mV and see if the problem goes away. Keep nudging it down until problems starts up again and give it a bit more for some headroom if needed. When you combine both clocks, you will likely need to increase the voltage. Don't worry about applying too much, you can't do any damage at these levels.


    ~ Useful Programs ~ (clicky)
    FurMark Use 0xMSAA when testing core, 8xMSAA when memory and both together. Disable PostFX. Enable Xtreme BurnIn and Dynamic Background and FullScreen. Use Burn In, not Benchmark. I think F3 or F7 will show an error counter (unreliable, one of the F# keys). If crash, black screen, then overvolt/reduce clock. PRO TIP: Stare at center of screen and if you see what seems like weird tearing with colors and stuff, increase voltage or cut back on clock. If you see white pixel snow, do same.
    OCCT
    Unigine Valley
    OC Scanner Tick artifact scanner. Use Run, not Benchmark and in Fullscreen. Same AA rules as Furmark. Look for artifacts, not sure how reliable the counter is.

    When that's all done and you have your max total overclock, and multiple profiles if you want, run some games, preferably heavy hitters like Metro 2033, Crysis 1, Crysis 3, and other modern games. If you ever see those artifacts, EXTREMELY bad tearing, white pixelated snow, weird flashing and stuff that didn't happen before, reduce your clocks and or increase voltage.

    Memory clocks that are too high will act like core clocks too high, and more often crash if anything. I personally use mostly Furmark, memory with another program, then games. Keep in mind your max may cause problems in some games and not others. I.e. Bioshock Infinite doesn't like mine, but has no problems with other games so far. In those cases, modify them as need be and reduce.



    Lastly, feel free to assign hotkeys to your profiles in Afterburner. I use Ctrl+Alt+1-5 for my profiles to change them on the fly. If you have questions, no matter how dumb you may feel it is, ask away. It's late right now as I write this, so I think I covered pretty much everything. You can also use figures of others for a general ballpark figure, but don't expect to get that high, and head towards it, not starting at it.

    Motherboard ~ Gigabyte GA-Z77X-UD5H Rev1.1 (BIOS F16mod11)
    Graphics Card ~ Gigabyte Aorus GTX 1080 Ti (2053 MHz Core | 12120 MHz)
    Processor ~ Intel i7-3770K @ 4.5GHz (1.225 V)
    Memory ~ G.Skill RipjawZ 16GB 2400MHz CAS10
    Sound Card ~ HT Omega Claro Halo
    Power Supply ~ Corsair AX760i
    Monitor ~ ASUS PG279Q - GSync
    Operating System ~ Win10 Enterprise x64

    ~ Rob ~

    #2
    Posted 10/30/2014 03:44 AM   
    Just a general question opinion about how good a GPU is. I have an EVGA GTX 970 SC ACX 2.0. I seem to be able to overclock it to +125 core, +250 memory & 110% (the max) Power Target. At this OC level, I can run Titan Fall, BF4 & Ghost without crashing. I can also run above clocks using Unigine Valley OpenGL benchmark test with settings of Ultra Detail, 8X AA, Fullscreen and 1080p. But, even a mild OC can cause TDR (I think), Events Viewer Warning Message regarding driver not responding when running DX11 test of Unigine with the same settings above except for AA, which I put to None. The same goes true to when playing Starcraft 2: Wings of Liberty. Any OC, mild or miniscule can inconsistently cause driver not responding then responding again. I think this is TDR. Not sure. So, what is your take on my card? It OC's fine or exceptional, but not in certain specific situations. Thanks. PS--I basically want a reason to RMA my card. LOL. even, though, I am not sure if it's even bad or worth it since I don't get coil whine and have read that it is prevalent with these cards.
    Just a general question opinion about how good a GPU is. I have an EVGA GTX 970 SC ACX 2.0. I seem to be able to overclock it to +125 core, +250 memory & 110% (the max) Power Target. At this OC level, I can run Titan Fall, BF4 & Ghost without crashing.

    I can also run above clocks using Unigine Valley OpenGL benchmark test with settings of Ultra Detail, 8X AA, Fullscreen and 1080p.

    But, even a mild OC can cause TDR (I think), Events Viewer Warning Message regarding driver not responding when running DX11 test of Unigine with the same settings above except for AA, which I put to None.

    The same goes true to when playing Starcraft 2: Wings of Liberty. Any OC, mild or miniscule can inconsistently cause driver not responding then responding again. I think this is TDR. Not sure.

    So, what is your take on my card? It OC's fine or exceptional, but not in certain specific situations.

    Thanks.

    PS--I basically want a reason to RMA my card. LOL. even, though, I am not sure if it's even bad or worth it since I don't get coil whine and have read that it is prevalent with these cards.

    #3
    Posted 10/30/2014 05:10 AM   
    [quote="portofdismay"]Just a general question opinion about how good a GPU is. I have an EVGA GTX 970 SC ACX 2.0. I seem to be able to overclock it to +125 core, +250 memory & 110% (the max) Power Target. At this OC level, I can run Titan Fall, BF4 & Ghost without crashing. I can also run above clocks using Unigine Valley OpenGL benchmark test with settings of Ultra Detail, 8X AA, Fullscreen and 1080p. But, even a mild OC can cause TDR (I think), Events Viewer Warning Message regarding driver not responding when running DX11 test of Unigine with the same settings above except for AA, which I put to None. The same goes true to when playing Starcraft 2: Wings of Liberty. Any OC, mild or miniscule can inconsistently cause driver not responding then responding again. I think this is TDR. Not sure. So, what is your take on my card? It OC's fine or exceptional, but not in certain specific situations. Thanks. PS--I basically want a reason to RMA my card. LOL. even, though, I am not sure if it's even bad or worth it since I don't get coil whine and have read that it is prevalent with these cards.[/quote] it is known this time around that EVGA did not do very good with the 970, I have a Gigabyte G1 Gaming GTX 970 that overclocks like a beast. it came with a factory overclock that hits 1380mhz boost clock. I have reached 1587mhz with no tdr's in any program game or benchmark I run. my settings:core voltage +60 power limit +112 core clock +150 memory clock +250, max temp 66c now if your not really into overclocking your gpu and it runs fine for you then keep it, but if you want a good overclocker thats also stable then the top 3 are Gigabyte, MSI, and ASUS. you can read reviews about all 3 on guruof3d website, you can also go there and read the poor review on the EVGA also about it's heat sink they cut corners on to save themselves money. EVGA has admitted to the heatsink issue and is offering free backplates for their card, so make sure you contact them for your free backplate if your keeping the card.
    portofdismay said:Just a general question opinion about how good a GPU is. I have an EVGA GTX 970 SC ACX 2.0. I seem to be able to overclock it to +125 core, +250 memory & 110% (the max) Power Target. At this OC level, I can run Titan Fall, BF4 & Ghost without crashing.

    I can also run above clocks using Unigine Valley OpenGL benchmark test with settings of Ultra Detail, 8X AA, Fullscreen and 1080p.

    But, even a mild OC can cause TDR (I think), Events Viewer Warning Message regarding driver not responding when running DX11 test of Unigine with the same settings above except for AA, which I put to None.

    The same goes true to when playing Starcraft 2: Wings of Liberty. Any OC, mild or miniscule can inconsistently cause driver not responding then responding again. I think this is TDR. Not sure.

    So, what is your take on my card? It OC's fine or exceptional, but not in certain specific situations.

    Thanks.

    PS--I basically want a reason to RMA my card. LOL. even, though, I am not sure if it's even bad or worth it since I don't get coil whine and have read that it is prevalent with these cards.



    it is known this time around that EVGA did not do very good with the 970, I have a Gigabyte G1 Gaming GTX 970 that overclocks like a beast. it came with a factory overclock that hits 1380mhz boost clock. I have reached 1587mhz with no tdr's in any program game or benchmark I run. my settings:core voltage +60 power limit +112 core clock +150 memory clock +250, max temp 66c
    now if your not really into overclocking your gpu and it runs fine for you then keep it, but if you want a good overclocker thats also stable then the top 3 are Gigabyte, MSI, and ASUS. you can read reviews about all 3 on guruof3d website, you can also go there and read the poor review on the EVGA also about it's heat sink they cut corners on to save themselves money. EVGA has admitted to the heatsink issue and is offering free backplates for their card, so make sure you contact them for your free backplate if your keeping the card.

    #4
    Posted 10/30/2014 01:57 PM   
    The Palit Jetstream is also very good in my case! without overclocking it hits 1380mhz boost clock and reached only 57°C.+ it is very silent(I added a selfmade fancurve with afterburner) + it is one of the cheapest gtx970. I`m very happy with it!
    The Palit Jetstream is also very good in my case! without overclocking it hits 1380mhz boost clock and reached only 57°C.+ it is very silent(I added a selfmade fancurve with afterburner) + it is one of the cheapest gtx970. I`m very happy with it!

    #5
    Posted 11/10/2014 07:47 PM   
    Hello mate, I bought a zotac gtx 970 a standard one 2 weeks back and it was fine until I overclocked it with a moded bios bypassing the stock voltage and it was running fine for a couple of days but now suddenly it had crashes so I reverted it back to its original bios and now at stock clocks while running valley benchmarks I'm experiencing artifacts...did I fried my card or shud give for RMA
    Hello mate,
    I bought a zotac gtx 970 a standard one 2 weeks back and it was fine until I overclocked it with a moded bios bypassing the stock voltage and it was running fine for a couple of days but now suddenly it had crashes so I reverted it back to its original bios and now at stock clocks while running valley benchmarks I'm experiencing artifacts...did I fried my card or shud give for RMA

    #6
    Posted 12/18/2014 08:54 PM   
    [quote] it is known this time around that EVGA did not do very good with the 970, I have a Gigabyte G1 Gaming GTX 970 that overclocks like a beast. it came with a factory overclock that hits 1380mhz boost clock. I have reached 1587mhz with no tdr's in any program game or benchmark I run. my settings:core voltage +60 power limit +112 core clock +150 memory clock +250, max temp 66c now if your not really into overclocking your gpu and it runs fine for you then keep it, but if you want a good overclocker thats also stable then the top 3 are Gigabyte, MSI, and ASUS. you can read reviews about all 3 on guruof3d website, you can also go there and read the poor review on the EVGA also about it's heat sink they cut corners on to save themselves money. EVGA has admitted to the heatsink issue and is offering free backplates for their card, so make sure you contact them for your free backplate if your keeping the card.[/quote] I'm having a different experience with my EVGA ACX 2.0 card. I currently run it at 110% power, +250mhz GPU, and +250mhz mem for a total of 1528 mhz core clock and 1879 memory clock. My max temp is jumping between 59 and 60 deg. This is all done without touching the voltage. I'm pretty happy with it.

    it is known this time around that EVGA did not do very good with the 970, I have a Gigabyte G1 Gaming GTX 970 that overclocks like a beast. it came with a factory overclock that hits 1380mhz boost clock. I have reached 1587mhz with no tdr's in any program game or benchmark I run. my settings:core voltage +60 power limit +112 core clock +150 memory clock +250, max temp 66c
    now if your not really into overclocking your gpu and it runs fine for you then keep it, but if you want a good overclocker thats also stable then the top 3 are Gigabyte, MSI, and ASUS. you can read reviews about all 3 on guruof3d website, you can also go there and read the poor review on the EVGA also about it's heat sink they cut corners on to save themselves money. EVGA has admitted to the heatsink issue and is offering free backplates for their card, so make sure you contact them for your free backplate if your keeping the card.


    I'm having a different experience with my EVGA ACX 2.0 card. I currently run it at 110% power, +250mhz GPU, and +250mhz mem for a total of 1528 mhz core clock and 1879 memory clock. My max temp is jumping between 59 and 60 deg. This is all done without touching the voltage. I'm pretty happy with it.

    #7
    Posted 05/07/2015 06:36 PM   
    Scroll To Top