Gigabyte GTX 1070 G1 Gaming - 24/7 Overclocking Guide (Basic method)

[RodroG]

HOW TO GET A 24/7 OVERCLOCK FOR GIGABYTE GTX 1070 G1 GAMING

***Remember that any overclocking process is risky, so be aware that you try this on your own risk***

A) Setout

What is a 24/7 Overclock (OC)?
In easy words, it is the overclock profile for your graphic card that can be used 24 hours a day, seven days a week and 365 days a year, as it is stable under full load conditions and does not reduce the lifetime of your device.

The process to determine this special overclocking profile is not fast. It requires patience and a lot of time to properly complete all its stages.

How can we trust that we will not reduce the lifetime of our graphic card?
Through two essential preventions:

1. Keep a good GPU temperature, applying a proper fan mode or a custom fan curve that will always keep it between 60 and 70 Celsius degrees.
   
2. Do NOT modify the stock GPU voltage.

How do we measure graphic performance improvement?
Through the following benchmarking and graphic stress programs:

• 3DMark[www.futuremark.com]
  
• Heaven 4.0[unigine.com]
  
• Valley 1.0[unigine.com]
  
• FurMark[www.geeks3d.com]
  
• OCCT[www.ocbase.com]

All of them must be installed and updated before starting the overclocking process.

In addition you can also use these ones:

• EVGA OC Scanner X[www.evga.com] (you must be logged in to download)
  
• MSI Kombustor[www.geeks3d.com] (you must download the 32 bits or 64 bits version depending on your system)

How to get this bench/stress softwares?
It is simple, you can get them by clicking on the names of the list above or searching by their names through your internet browser.

How to check the stability of your OC profile?
Through these same programs using them as graphic stress tests. That is to say:

• 3DMark: Running its Fire Strike Stress Test, consisting of 20 uninterrupted loops of the Fire Strike benchmark Graphic Test 1.
  
• Heaven: Running it in loop mode for at least 30 minutes.
  
• Valley: Running it in loop mode for at least 30 minutes.
  
• FurMark: Running its GPU Stress Test at least 1 hour.
  
• OCCT: Running its GPU stress test at least 30 minutes (tick Error Checking box only when you are testing Memory OC).

Other programs needed
Our main overclocking utility, one GPU monitoring software and one spreadsheet tool.

Specifically, these are the ones I used:

1. Gigabyte - Xtreme Gaming Engine (XGE)[www.gigabyte.com.es]: To configure and apply overclocking profiles.
   
2. TechPowerUp GPU-Z[www.techpowerup.com]: To know key information about our graphic card and its maximum clocks, temperature, TDP and cap or throttling reasons.
   
3. Microsoft Excel[www.microsoftstore.com] or a free equivalent[es.libreoffice.org]).

Instead of the Gigabyte utility you can use other overclocking tools like MSI Afterburner[es.msi.com] or EVGA Precission X[www.evga.com]. Just use the OC program you prefer.

GPU-Z must be open for registering all key data related with the applied OC profile while Heaven, Valley, FurMark and OCCT are running as stress tests.

With Excel we will create a worksheet where we will record all the relevant data for this overclocking process and with which we will identify the most stable OC level. Here[1drv.ms] is a empty but formatted spreadsheet sample you can use.

Process to identify our 24/7 Overclock
The full process consists of 3 main stages:

1. GPU Overclock (GPU OC)
   
2. Memory Overclock (Memory OC)
   
3. Games & 24/7 Overclock (Final)

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

B) First Stage - GPU Overclock

At this stage we will check the stability / performance of the different GPU OC profiles.

Please, follow the procedure below step by step:

1. Checking the factory profiles stability and its performance scores. This profiles are called Gaming Mode (stock) and OC Mode by Gigabyte. Ensure you through XGE that the default values of these two profiles are applied (by clicking Default or Apply button).
   
2. Set through XGE these common values for each of the following OC variables:
   
   - GPU Voltage: +0%
   - Power Limit: +11% (max)
   - Temperature Limit: -3 ºC (target: 80 ºC)
   - Memory Clock: +0% (stock)
   - FAN: Choose a suitable fan mode or set a custom fan curve. The fan mode or curve we set must keep GPU temperature below 70 Celsius degrees in any stressful graphics conditions.
   
   Record all of these values in the spreadsheet for each OC profile.
   
3. Increase the stock GPU clock by 5% through XGE and then continue with 10 MHz increments per step. Note each increment in our spreadsheet.
   
4. At each OC level start by running the 3DMark Fire Strike Stress Test. This point is passed only if this stress test ends with a result of at least 97% in its Frame Stability Rate (FSR) AND without noticing artifacts during the test.
   Recommendation: during this test look at the screen almost all the time to detect possible artifacts.
   Save the result / report that 3DMark gives you, take note of the maximum temperature shown in the chart and click the Compare Result Online button to get an extra online report. In the online report find the values for FSR, Core Clock and Memory Bus Clock.
   In our worksheet we will write Yes / No in the column called Passed? and all previous data.
   Follow this writing pattern:
   
   - In Passed? write the affirmative results in bold and the negatives in red and bold.
   - Write in bold the clocks that have increased in each level.
   
   
5. If the OC profile seems stable in 3DMark continue by running Heaven for 30 minutes in loop mode with GPU-Z been opened.
   Take a look at the temperature and the maximum theoretical clocks shown by this program during the test to also confirm in this way that the OC profile has been applied.
   Recommendation: Observe the screen repeatedly but only occasionally and for a few minutes to detect possible graphics errors.
   We will close the program if we observe artifacts or if we exceed 30 minutes without seeing errors.
   Whenever we exit to desktop due artifacts or program / driver crashes or when loop time end successfully, we will go to GPU-Z Sensors tab and we will register Max Core, Max Mem Clock and Max Temp values.
   In our worksheet write Yes / No in the column called Stable? and all previous GPU-Z data.
   
6. If it seems stable in Heaven continue by running Valley for 30 minutes and do the same.
   
7. If it seems stable in Valley run FurMark for 1 hour and do the same.
   
8. If it seems stable in FurMark run OCCT for 30 minutes with error checking off and do the same.

Common GPU OC Recommendation: Once you detect instability in any test do not continue testing with the remaining programs.

Common GPU OC Guidelines: Once instability is detected we must record it in the spreadsheet and we will not continue with the other programs, after that we will apply the immediately preceding stable GPU OC profile and we will re-check its stability. If it was stable we will stop and we will stay at this step. If it was not stable we will roll back a step further and we will check its stability again. We will proceed thus until we find a GPU OC level that is stable for sure.

First Stage Final Recommendation: For less risk you should reduce 4-5 MHz extra the maximum stable clock reached and then apply it.

Congratulations! At this point you will have identified the maximum stable GPU OC of your GTX 1070 G1 in stability tests.

NOTE: The 24/7 GPU OC does not necessarily match with the maximum stable GPU OC in stability tests, since it will depend on the third stage of the process that involves intense and long gaming sessions (at least 2 hours) with your most demanding games.

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

C) Second Stage - Memory Overclock

Set our new maximum stable GPU OC, we will check the stability / performance of different Memory OC profiles.

Procedure:

1. Set through XGE these common values for each of the following OC variables:
   
   - GPU Clock: final GPU OC clock identified in Stage 1.
   - GPU Voltage: +0%
   - Power Limit: +11%
   - Temp Limit: -3 ºC
   - FAN: your suitable fan mode or fan custom curve.
   
   Note all this values in your spreadsheet.
   
2. Increase the Memory Clock by 24-48 MHz per step using XGE and then write it down in your Excel spreadsheet.
   
3. Run and observe the Fire Strike Stress Test running and register your data for each Memory OC profile.
   
4. If it was stable, continue with Heaven for 30 minutes with GPU-Z in the background and write all data down.
   
5. If it was stable, continue with Valley ...
   
6. If it was stable, continue in the same way with the rest of programs.
   Common Memory OC Guidelines: Once instability is detected we must record it in the spreadsheet and we will not continue with the other programs, then we will apply the immediately preceding stable Memory OC profile, we will re-check its stability and if it was stable we will stop and we will stay at this step. If it was not stable, we will roll back a step further and we will check its stability again. We proceed thus until we find a Memory OC level that is stable for sure.
   Writing Guidelines: Same as in the previous stage.
   
7. Run 3DMark Fire Strike Benchmark for each stable Memory OC level.
   If the benchmark ends correctly and without seeing artifacts or unexpected crashes, note the overall score and the average frame rate in Graphic Test 1 and 2 and Physics and Combined tests.
   Writing Note: Write in red the overall score when you observe any signs of instability.
   
8. Write in bold the overall scores which are better than those of the first stable Memory OC level and which are also better and at the same time in FPS 1, FPS 2 and FPS Combined partials scores (corresponding to those of the Fire Strike benchmarking tests that are only sensitive to graphic overclocking).
   Note: Sometimes we will check that although we raise the memory clock we get lower scores than a previous stable step.
   Guideline: identify stable Memory OC levels with which the performance improves and between these stay with the one that give you the highest score.
   
9. Final Memory OC setting: Set through XGE the intermediate clock between the best and maximun stable Memory OC profile and the immediately preceding stable profile and test its performance with Fire Strike Benchmark.
   Guidelines: If the benchmark finish successfully note the new profile and its scores, then check if the scores are better than the ones of the preceding best and maximum stable Memory OC profile. If there was an improvement stop the memory overclocking in that new profile and otherwise roll back to the best and maximum stable Memory OC profile.

Congratulations again! You have identified the maximum and best stable Memory OC profile for your GTX 1070 G1 in stability / performance tests.

NOTE: The 24/7 Memory OC does not necessarily match with the maximum and best stable Memory OC in stability / performance tests, since it will depend on the third stage of the process that involves intense and long gaming sessions (at least 2 hours) with your most demanding games.

Personal Note: In my case the maximum and best stable OC profile in stability tests also results stable with Games, so if you have properly followed the steps of this two stages the same thing may happen to you.

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

D) Third Stage - Games & 24/7 Overclock

In this final stage we will check the stability of our maximum and best stable OC profile in stability tests through intense and long gaming sessions with demanding video games.

Setout: Install at least 4-5 demanding video games which together make it possible to check the stability under DX11, DX12, OpenGL and Vulkan API.

Prcedure:

1. Apply through XGE your max stable GPU OC in stability tests and check clock frequencies with GPU-Z.
   
2. Test its stability in each game by 2 hours of uninterrupted play.
   Guidelines: Run the game at least on Ultra and look for artifacts or unexpected crashes. If there are not you can consider the profile stable and you can write it down in your spreadsheet but if there are errors you must reduce -5 MHz the GPU Clock and re-check stability. We will continue this way until we find an stable level.
   For each game and each GPU OC profile add all the OC Settings in your spreadsheet and write down Yes / No in Stable? column.
   The maximum stable GPU OC in Games will be our 24/7 GPU OC.
   
3. With our 24/7 GPU OC been set apply via XGE the max and best stable Memory OC in stability tests and check clocks with GPU-Z.
   
4. Test its stability in each video game by 2 hours of uninterrupted play and note the results.
   Guidelines: Same as previous but if you notice errors lower -12 / -24 MHz the memory clock and re-test stability. We will continue this way until we reach stability.
   For each game and each Memory OC profile add all the OC settings in your spreadsheet and write down Yes / No in Stable? column.
   The maximum stable Memory OC in Games will be our 24/7 Memory OC.

We did it! The union of the maximum stable GPU OC and the maximum stable Memory OC in Games will be the desired 24/7 Overclock for our Gigabyte GeForce GTX 1070 G1 Gaming. Enjoy it!

NOTE: Here is my 24/7 OC profile:

- GPU Clock: +115 MHz
- Memory Clock: +476 MHz
- GPU Voltage: +0%
- Power Limit: +11%
- Temperature Limit: -3 (target: 80 ºC)

Please use it ONLY as a reference. You should find your own 24/7 OC profile! Remember that all Systems and Graphics Cards (also those of the same model) are physically different. These differences make the OC profiles not directly extrapolable. In fact, what was valid for me, does not have to be valid for you.

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

I hope you find this guide useful and do not hesitate to comment it!

[[MVM]HunterKiller]

Yeah I under volt it, it is at 60% power and try to get a nice OC out of it to be able to mine as fast as it can with the least amount of power draw and because I profit switch each algorithm is different and uses the GPUs resources differently so this is a sort of all rounder OC.

[Monk]

Lowering the power limit isn't the same as undervolting, you'll need to unlock voltage control for that.

[[MVM]HunterKiller]

Oh sorry my bad, I mis-read that, yeah im only power limiting, so still going for maximum OC for least amount of power draw.

[RodroG]

Monk の投稿を引用：

Sorry but on modern cards, you are totally safe to unlock voltage control and max it out along with max temp and power sliders to get the best OC.
You should also start the overclocking process from stock, go up in 10MHz steps for the core and 50MHz steps for the memory, pull back once there's an issue.
The cards cannot pull enough voltage even unlocked to damage themselves, hell myself and other enthusiasts looking g for the ultimate performance even shunt/power mod our cards to get round the very limited voltage controls.

I will say getting an extra 1.3GHz out of your memory is very impressive though.
Though I must say as detailed as your methods are (and everyone has their own system), I feel there is alot of wasted time there.

Also for mining you tend to want the cards undervolted for maximum efficiency not over clocked, unless you don't pay your power bill that is.

Of course, it's safe to unlock voltage control and max it out in a short term but it isn't in a mid-long term if you're in a real 24/7 OC scenario, which is something different to a punctual overclock profile to get your highest bechmark score. In short, you can max the GPU voltage out safely when performing a punctual OC, but not when you're keeping a 24/7 OC over time.

[Monk]

no its safe to do long term, i have done for every gpu i have ever owned, right now my 1080ti's are in a very heavily overclocked state and power modded, realistically the shortening of its life span is only at the stage where the GPU is well beyond its expected useful lifespan anyway, the card wont allow you to give it to much power to the point it would cause any harm, even when ytou unlock voltage control it is only a minor increase and still WAYYY below dangerous levels.

only when you begin power modding or running custom bios is it ever going to be an issue, abd even then, there is no evidence that it will significantly shorten its lifespan.

[RodroG]

Monk の投稿を引用：

no its safe to do long term, i have done for every gpu i have ever owned, right now my 1080ti's are in a very heavily overclocked state and power modded, realistically the shortening of its life span is only at the stage where the GPU is well beyond its expected useful lifespan anyway, the card wont allow you to give it to much power to the point it would cause any harm, even when ytou unlock voltage control it is only a minor increase and still WAYYY below dangerous levels.

only when you begin power modding or running custom bios is it ever going to be an issue, abd even then, there is no evidence that it will significantly shorten its lifespan.

No prob dude, anyway that's a debatable issue... Regards!

[Irohen]

Hi man, thanks for the tutorial, do you think it would work for another GTX 1070 model or is it just for the G1?

[RodroG]

Irohen の投稿を引用：

Hi man, thanks for the tutorial, do you think it would work for another GTX 1070 model or is it just for the G1?

Hi! This method would work for all GTX 1070 models and Pascal models too.

[[MVM]HunterKiller]

Irohen の投稿を引用：

Hi man, thanks for the tutorial, do you think it would work for another GTX 1070 model or is it just for the G1?

It will although other models and brands may not be stable at those clocks but the principal is the same, just gradually go up in OC profiles until you find the right spot where it doesnt artifact or crash, if it does then drop back until its stable.

[LuckyBuster]

- GPU Clock: +115 MHz
- Memory Clock: +476 MHz
- GPU Voltage: +0%
- Power Limit: +11%
- Temperature Limit: -3 (target: 80 ºC)

I have tried to overclock to 125mhz on the base clock but this is extremely subjected to screen tearing and crashes and I have found 123mhz compared to 125mhz runs better on the average benchmark but anything over 115 is extremely subjected to game crashes and even full pc crashes but perhaps overclocking the voltage may fix this I will test.

-For me even after increasing the volts and following your recommendations many of my games crash if the gpu is overclocked over 105-110 mhz on the base clock

What works best for my gpu is
+100mhz on gpu base clock
+500mhz on mem
power and temperature target to default.

-PS I haven't tried to play with the memory clock to much if you want you could possibly increase this but for me this is the most stable and 8gb vram is already plenty.

[Al_Chafai]

i have the same graphics card but honestly sometimes i find OCing to be useless,when OC this card will only give u from 5 tp 10 fps and for someone like me whos playing @1080p this card is more than enough,it eats every game i throw at it,with no issues at all.
so i put a modest OC of +60 /+200 with power and temp limit to the max,runs great and cool.

[RodroG]

6xdarksiderx6 の投稿を引用：

i have the same graphics card but honestly sometimes i find OCing to be useless,when OC this card will only give u from 5 tp 10 fps and for someone like me whos playing @1080p this card is more than enough,it eats every game i throw at it,with no issues at all.
so i put a modest OC of +60 /+200 with power and temp limit to the max,runs great and cool.

Agreed.

[Volesprit]

Hi,

First thank a lot for this topic.

I'm totally newbie and are on step two at +140 Mhz for the moment.

Please can you explain to me why we don't touch to GPU Voltage? Indeed I've see in anothger Video that with recent Geforce we can set it to maximum because protection are set by Nvdia to ensure that no issue with voltage are possible (Not for AMD warning!!).
Can you confirm and explain to me why we don't touch the voltage to permet maybe more stable performance?
Please read agfain: I'm newbie it's my first overclock so just explain me the reason i'm not saying it's better :p.

Thanks'

[Monk]

He said that as he doesn't actually know how to oc very well, your safe to max out those sliders (power limit / voltage / temp) just remember to go up in small steps, all gpu's are different so will take different frequencies so, test slowly and good luck.

[Volesprit]

Oh ok thanks for this quick response Monk. GPO Voltage +100% also now.