Overclocking with newbie (me) 970G1 Gigabyte

Toggle sidebar Toggle sidebar
P

Pablo1990

Reputable
Oct 23, 2015
18
0
4,510
Hello
I was thinking about overclocking my GPU so i read plenty of topics, watched hours of youtube videos.

I have Gigabyte Geforce GTX970 G1 Gaming GPU and Be Quiet Pure Power 630 Wat PSU with +80 Bronze.
It gives a lot of power , and should be sufficient so I turned up the GPU and CPU future.

So i started from raising up power limit to 112%.... and i get core clock +190 stable with +25 mV more Core Voltage.

So i started raising memory. From GPU-Z i read that i have Samsung memory so i thinked that +500 will be reachable.
I was wrong. With +450 i get crash when runing Furmark. So i get +400 stable.

I wanna ask, did my memory isn't that good, or i need to raise voltage or sth? Also i see in furmark that core clock isnt 1519 as GPU-Z shows, but i have 1355? Why?

It's my first overclock so if you can please explain me this as simple as you can :)

Best Regards
Pablo
 
Solution
Vellinious
May want to try lowering the memory just a tad. Try 460 or 470 on the memory offset. The core clock should be ok. I've noticed with the cards on air, that 2000 is about the limit it likes to run at in Valley.
Sort by date Sort by votes
1. Recognize the difference between core clock and boost clock.

2. Just slide Power and voltage sliders to max , you won't hurt the card.

3. Once you determine max stable settings, you can try lowering voltage.

http://www.guru3d.com/articles_pages/gigabyte_geforce_gtx_970_g1_gaming_review,26.html

With AfterBurner we applied:

Temp Target 80 Degrees C
GPU clock +150 MHz
Power limiter 112%
Mem clock +500 MHz
Volatge + 87Mv
FAN RPM default
Click to expand...

They got:

Core Clock 1328 MHz
Boost Clock: ~1516 MHz
Memory Clock: 8002 MHz
Click to expand...
 
Upvote 0 Downvote
And did i need to change settings in furmark? What resolution i need to pick? Antialiasing? Other settings?

When i set +150 on core, and +500 on memory i get 1328MHz normal clock, 1479MHz boost clock, and memory 2003 Mhz. Is that good? Can i set more Mhz on core clock?
 
Upvote 0 Downvote
First....stop using furmark. It's a heat machine, and NVIDIA has a built in throttle to keep the GPU from damaging itself. It's not good for much of anything any more. If you own Firestrike, use Firestrike Ultra graphics test 1 and 2 for stability testing. Have it loop for an hour. If there's any instability in your clock, it'll find it. If you don't own Firestike, you can use Unigine Valley for stability testing. Run it 3 times in a row on the Extreme HD preset for stability tests.

Also....please don't use offsets. Read what GPUz tells you the clocks are, and tell us what those are. Use the GPUz sensors tab. It'll show you everything you need to know to overclock your card. Pay particular attention to the "PERF CAP REASON" line. Have it open during your stress tests.

Ok. There is absolutely NOTHING in the stock bios of that card that's going to hurt it. So, crank up the voltage as high as it'll go, which will likely only go to 1.256v, and maybe 1.262v and try your clocks again. Keep the power limits up as far as it'll go and try your overclock again. And yes, with the Samsung memory those memory clocks should be easily reachable.

It could be that you've started to hit a PWR perf cap reason...that's green...if you hit that, it'll throttle the card. The variation in the clock speeds can cause instability and make it crash. Try to get the clocks set so that you're not hitting the power limit perf cap.

IF you're not satisfied with the clocks you get, without hitting the PWR perf cap, or the Vop, voltage restriction, then you'll need to run a modified bios to any higher.

As with any overclocking, you'll need to make sure your temps stay as cool as possible, so try setting up a custom fan curve to keep the GPU as cool as possible. Maxwell LOVES running cool, and the warmer they get, the more unstable they get.

There it is.
 
Upvote 0 Downvote
As a general rule of thumb, most Maxwells can reach 1500+ on the core clock, and between 2000-2100 on the memory. Depending on how cool you keep it.

That's a good initial target to shoot for. Obviously, some can go much higher than that, but only a very small portion of cards aren't able to reach those clocks.
 
Upvote 0 Downvote


No...Green is bad. Green means you're hitting the power limit performance cap. Clocking any higher when you start hitting the green, will be useless, because the card will just throttle even more to keep it under the Power limit perf cap.
 
Upvote 0 Downvote
This is bad. See the green there? See how the clock in the top graph moved down? That's power limit throttling to get the card back under the power limit.

8x4sZTX.jpg

 
Upvote 0 Downvote
Ok so i get Valley Bench tool, i make +190 core (GPUZ shows 1519 boost) +500 on memory (GPUZ shows 2008) get voltage and power limit to max (+87mV and 112%). Let's see what card i got 😉

So i get artifacts and perfcap reason is half blue and half orange ( VREL and VOP ) i need to lower my core clock right?
 
Upvote 0 Downvote
May want to try lowering the memory just a tad. Try 460 or 470 on the memory offset. The core clock should be ok. I've noticed with the cards on air, that 2000 is about the limit it likes to run at in Valley.
 
Upvote 1 Downvote
Solution


Vrel isn't any big deal. Now raise your core clock some more. Try to push it to 1550 and see if that works.
 
Upvote 0 Downvote
Also...make sure you do a reboot after a driver crash. The soft reset the driver does after a crash isn't very good, and will cause problems. So, during your testing, you'll invariably see a driver crash once in a while. If it does, reboot your pc and start again. It's best to keep track of what clocks / memory clocks are working, what doesn't work etc, on a sheet of paper, along with what FPS they were able to achieve during those runs. It'll help in the end find your GPUs "happy place". Every card has one...just have to keep playing with the settings until you find yours.
 
Upvote 0 Downvote
Ok, stable +220 (1549 in GPUZ), no artifacts. Perfcap reason blue, but i saw in gpuz sensors tab, that my core clock under load (in bechmark) is 1408 only??
Any suggestions for that?

And i got 56-57 degrees after 1 benchmark loop. It's that temp ok?
 
Upvote 0 Downvote
Valley doesn't read the clocks right, unless you're running a custom bios. That's normal. Only pay attention to what GPUz tells you.

I think blue is Vrel...cursor over it and it'll tell you, but......sounds like you have more room to get more core clock out of it. Just play with the settings and keep going with it. Your card is officially overclocked. Congrats!
 
Upvote 0 Downvote
See where it's showing the number in GPUz? Click on it. Little green letters will come up and show "min", "max" and "avg". Set it to "max", and tell me what it says. OR, the high points in the red line, where the GPU is under stress from Valley, mouse over those and it'll tell you what core clock it was running in the area where the cursor is over.
 
Upvote 0 Downvote
Also...go into your NVIDIA control panel. In the adjust 3d settings tab, for power management control, turn it to "performance" instead of "adaptive" and make sure that VSYNC is off.

Temp is good, btw.....try to keep it under 75c for good stability.
 
Upvote 0 Downvote
Ok, i select show highest reading in GPUZ, set up vsync of, power management to performance, also set up core clock to +240 😀 I run valley one loop, GPU was stable, GPUZ shows highest core clock reading about 1600?? i get min 32,7 FPS, max 115,4 and score 2627. When i set higher clock i get dxdiag errors. So i finally found G spot for my GPU 😀
 
Upvote 0 Downvote


That clock likely won't be fully stable, depending on the game you're playing...some may like less clock, some won't mind the overclock at all. It's just a matter of figuring out which games you want to see a little more frames for, and finding settings that will work in each one.

There's always tweaks that can be done, but....sounds like you've got it dialed in pretty close. There ya go, congrats. = )

2600 is decent....some tweaking in the NVIDIA control panel, and on the overclocks should yield a little bit better score.
 
Upvote 0 Downvote
After reboot i get more MHz in GPUz: 1669,7!?! and in Valley i get 10 points more. Thank you very much for help.
I also need to start looking about right clock's for my game's. But i learned many things which help's me a loot.

Thank you for help :)
 
Upvote 0 Downvote


I would say that certain 970s can go to 1500+ or that most cards in certain model lines can go to 1500+. It depends just how good the non-reference model line in question is. From site to site and test to test, certain model lines fare much better than others ... consistently.

The Gigabyte G1 and MSI Gaming series breaks 1500 in almost every review I have seen ... the EVGA and Asus, not so mush. If ya wanna see why these two can manage what most of the competition most oft can not read the bottom third of pages 2 thru 4 here:

http://www.bit-tech.net/hardware/graphics/2014/09/19/nvidia-geforce-gtx-970-review/14

As for cooling, that is one reason why I go to Furmark first ... I wanna see if heat is a factor and it has not been on any of our 9xx builds (using Gigabyte / MSI cards). I want to know just how hot things can get and Furmark is the fastest way to do that.

Now it might seem contradictory as I won't use P95 for that reason (I do use P95 to set TIM) .... it generates heat unlike anything your CPU will ever see with real applications. But there is an extreme difference between P95 and CPU based application benchmarks whereas the difference between furmark and gaming is not all that much. MSI's 970 for example goes from 192 watts in Metro to 213 watts in Furmark (+21 watts). The MSI 980 Ti actually drew less power 281 watts in gaming, 279 in Furmark. OTOH, a Gigabyte 980 Ti can pull almost 70 extra watts. Gigabyte however, on their 980 ti goes from 293 to 359

http://www.techpowerup.com/reviews/MSI/GTX_970_Gaming/25.html
http://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Gaming/28.html
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_G1_Gaming/28.html

It's useful to get the memory / GPU hotter than normal usage as this gives an opportunity to help set the TIM and quickly improve thermal transfer for subsequent testing. It's also useful for determining what, if any, case air flow modifications are necessary.

The 970 had been artificially nerfed by nVidia at release because it was just too danged fast compared to their flagship 980. The 980 Ti has a max temp of 92C and starts throttling at 85C .... leaving 7C of headroom. The 980 gets 13C of headroom (98C / 85C). So why does the 970 with a max temp of 98C start throttling at just 80C ? .... cause otherwise, if they used just the 13C or even the 7C, there'd be no reason to cough up the extra $$ for a 980.

As for relying on Valley, Heaven, Firestrike, while they are good for bragging rights perhaps, as a stability tool, I find them quite unreliable. Don't get me wrong, I do use them to tune OCs, but if you walk way after showing stability in all of them thinking you will be stable in every game you play, you are going to be disappointed. There's a reason why Afterburner has all those profile buttons.... they come in handy as your Firestrike, Heaven, Valley stable OC settings will not be stable in many games.

I don't play a lot of games (3 - 4 a year perhaps) but do troubleshoot a lot of builds and also set up a lot of OCs for user builds .

Profile 1 = Max benchmark stable GFX OCs (4+ hours heaven / valley etc)
Profile 2 = 3/4 of max core and memory Ocs
Profile 3 = Max benchmark stable Core OCs
Profile 4 = Max benchmark stable Memory OCs
Profile 5 = Stock Settings

If there's a problem with 1, then go to 2.....if still a problem, will use 3 and 4 to see whether core or memory causing the issue

Some games are particularly problematic, some examples:

1. BF4 - worst of all time; many users I talk to just go to "Profile 5" which is "stock settings"

2. Witcher 3 - I had no problem with W3 up thru version 1.06 on profile 1 .... 1.08 was introduced some problems .... I'd crash in 1.08 after about 2 hours but only on some days ... other days not at all. Profiles 3 and 4 were not immune ... not often but enough to be annoying .... I was so close to finishing that I just finished game at stock settings (was still getting 80 - 100 fps).

3. Metro 2033 - Would not run at profile 1 ... couldn't get past opening credits.... trimming memory just a notch was good all thru game till one point where you come out from underground. I could change settings to stock, get past this point, save and exit then reset back to OC and all would be fine. Weird cause all that was happening on screen was changing your view from ladder and sky to landscape.

In summary, your max OC will change from tool to tool and from game to game and even from patch to patch. You're not going to spend a day or a weekends fine tuning your settings with benchmarks and then not have to tweak them when you dive into your game collection.

Best that you can do is:

a) Find and read tear down articles to find out which non-reference boards provide the better componentry and how well or even if they take any extra steps to cool that componentry.
b) See if they
 
Upvote 0 Downvote
NVIDIA has throttles build in for furmark. Doesn't do any good to even use it.

I don't need articles and reviews, man. I've been creating bios files for them, and helping people get their cards tuned in for quite a while now. Firestrike Ultra looped for an hour will find any instability in the overclocks that any game will throw at them.

As for the rest? TLDR.....you started off with "reviews", and lost my interest right away. /wink
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom