GTX 970 overclock results!

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

jty0yt

Reputable
Sep 3, 2014
363
0
4,960
So on my GTX 970 I have upped the core my 120 MHZ observing a boost clock of 1479mhz. MY fans are around 50 percent at 100 percent usage and the temps are 52 degrees! I have only upped the voltage by 15 MV and the power limit to 105%. I love this card, give me your results below!
 
Got my gigabyte g1 970 to 1550 mhz core + 7,36ghz mem +25mv 112% power, stable. the mem is a bit wierd on my card if i even go as far as 7,45 games will crash after about 5 mins, and i found out it would even crash on the 7,36 so i just put it back to 7.0 . what im running on my 24/7 oc is 1500 core and 7 mem, with +0mv and 100% power
 
2 x Asus Strix +125mhz GPU clock and +100mhz memory.

PC spec in sig.

screenshot_363.jpg
 
Gigabyte G1 Gaming GTX 970 Rev. 1.1
ASIC: 58.6%

MAX Bench stable OC: 1504/3802 +.025mv
MAX Long Term Game STable OC: 1452/3758 stock voltage
Zero (0) coil whine.

I noticed with low ASIC quality on this particular sample the stock boost voltage was already high @ 1.225v core. It was not happy running at any speed with any voltage bump for more than 30-45 mins.

Added thermal pads between backside ram and backplate. This seemed to help with stability over a 2+ hour full load stress test. SO I just went for the max I could get on stock voltage. The results are exactly where I hoped to end up. Would have liked a little more overhead, but happy with these results.

What a beast of a card. :)
 



On the Gigabyte G1 Gaming, 1/2 the ram is on the backside of the card and not touching any kind of heatsink or the backplate. I added some 3M high quality thermal pads and this seemed to help with stability. So my guess is the backside ram is getting head soaked after long term use.


The gap is about 1.5mm, if you are considering adding pads. You will need 4 total. I would suggest maybe getting a few extra of different sizes just to be safe. Maybe 8 each of 1mm, 1.5mm and 2mm.

*** BE CAREFUL with the thin pads on the front side *** These can be re-used but are very thin and soft.


I cleaned and re-mounted the heatsink using Arctic Silver 5. Load temps on the core were the same as with stock TIM.

*** BE CAREFUL to make sure the mounting studs position in the center of the holes *** If they are misaligned even a little, you will see the shoulder getting stuck and not completely tightening down. Looking between the PCB and heatrsink you should see only the straight side of the stud for the screws. If youy see any one of the 4 studs showing the L-shape of the shoulder, you are misaligned.


For those wondering: "Dude, why did you mod a berand new GPU? Won't that void your warranty?" The mod is reversible and was done by a professional technician. So it will be fine. Besides, they don't call me 'mad' for nothing. }:)
 
I've had my MSI GTX 970 Gaming 4 since September now and I've had plenty of time testing it and overclocking as well as flashing my own custom BIOS's. I have to say i'm very dubious about people that claim 1550mhz 8000mhz is truely stable with no voltage or TDP limit bump. I found that running benchmarks and gaming for short periods or on certain games this was achieveable by simply increasing tdp to 110% but under real stress it would always crash or throttle back the GPU.

In the end I modified the entire voltage scale on the cards BIOS, topping out at 2.81v @ 1550mhz / 8000mhz with a TDP max limit of 300w & i have seen it actually reach 260w several times. This allows the card to run rock solid stable for hours on end in anything I've thrown at it and it never scales back the gpu clock. FC4, BF4, valley bench and heaven all run perfectly with no artifacts, hangs or crashes.

Here are my scores to prove results:
Capture.JPG

Capturee.JPG
 
My SLI MSI gtx 970's run heaven 4.0 fine at 1541mhz. Been gaming at that frequency too single card, no artifacts or crashes. Ran fine in watch dogs, metro last light redux, the evil within, and alien isolation so far running for hours. If I see an artifact anywhere my next drop is to 1531mhz. Memory is at +500mhz which is 8 ghz I believe? Says 4ghz on heaven, not sure why I see 8ghz and 7 stock but whatever.

I have GPU-Z sensors and evga precision running and notice no throttling.

Maddofargo's gigabyte being at 1450 seems a bit low. 1500mhz is a guarenteed safe stable clock with these cards so that would be my minimum overclock. I prefer to take advantage of all the power I can get out of these 970's.
 


They are two different benchmarks - Heaven & Valley. Unfortunately I dont have a before overclock result saved and it would mean reflashing stock bios if i was to get it which i dont want to do really. From memory though my original score (before overclock) on Heaven bench under the extreme preset (as shown on my overclocked pic) was around Score: 1,500, Ave: 60 (Min: 19, Max: cant remember) fps so it is a massive increase.

As far as flashing custom BIOS's goes, its easy but just make sure you do a lot of reading up first. All you need is the 'Maxwell Bios Editor V1.36' and NVFLASH (latest version). You can extract your BIOS from your card using GPUZ and modify it in the maxwell bios editor, then smply flash it back onto the card using NVFLASH. There's quite a lot of info on the GTX970 thread on the overclockers forums. These helped me a lot:

http://www.overclock.net/t/1517316/extract-and-flash-gtx-970-and-980-firmware-nolimits-and-gamestable/750
http://cryptomining-blog.com/3757-how-to-raise-the-power-target-limit-on-geforce-gtx-970-and-gtx-980/
 


Not sure if i was unlucky but even at 1450mhz gpu i started getting crashing and driver errors if i didnt manually increase the tdp limit and voltage significantly. My card kept hitting the ceiling of the TDP as well even at 110% (dont you notice this too in GPUZ?). I can't speak for others but in my case the only way i could get 1550 / 8000 stable was with 1.81v and a much higher TDP max limit. I kept increasing incrementally until it stopped crashing and even at 1.75v BF4 would crash occasionally at this overclock. All i can say is consider yourself lucky hah! :)
 


My power limiter is at +110% and during bench i'm seeing like 96 watts max. In game I think my max was like 106%. My voltage is at +87 since it's locked on this card at stock 1225mv or 1250mv whether it's at +20 or +87. So I keep it running at 1250mv.

The thing is the voltage adds to the clock speed so without voltage it's running at 1528mhz, adding voltage raises it to 1541mhz. Anything past 1546 I believe produces the smallest artifact. My max OC is sitting at +200core/+500mem/+87 voltage. I got good cards. The voltage increase gives me 1-2 fps more but it's stable so whatever :)
 


Don't know to be honest. I have the MSI versions and running fine at 4005 MHZ (+500mem) from 3505MHZ. Gives me a nice performance boost along with +200 core (1541mhz). Rivals stock 980 speed easy. Running it with a 4790k at 4.5ghz and good air flow case. What reference company is your card from? MSI and gigabyte versions are the best overclockers it's really surprising if you can't atleast get 1500mhz on the core.
 
What is your ASIC quality? my 2nd Card is 56.6% however the first card is 74% - because im running in SLI this may be the reason why the memory won't o/c as high as other cards. I agree the MSI and Gigabyte cards are the best performers in overclocking terms, the gigabyte G1 is the best.
 


It would stay bench stable at over 1500/3900, for 30-45 minutes. The limiting factor on my particular card seems to be the RAM. Half is on the backside and isn't cooled very well. It would never go to 8000, even for short runs. 7800 was the highest I could get it stable for benchmarking. 7200-7400 long term for gaming. Plus my ASIC was 58.6%, which is a bit low. This is reflected in the high stock boost voltage of 1.225v gpu core.

I'm wondering if Gigabyte changed their binning between the 1.0 and 1.1 cards. Chances are they had to go to a higher cost VRM to fix the coil whine issue with 1.0, and this raises costs. A way to offset this is put less expensive chips on the 1.1 revision.

I'm still very happy as 1450 is the most I would want for long term use.

The MAIN reason I got this particular model is the cooler. I wanted the same or similar performance to my HD 7950 crossfire, but in a single card, which ran cooler and was quieter. It nails all those points, and overclocks to 1450. I can't complain.
 
I just got my msi gtx 970 and I got it to be stable at +250 core and, +400 memory. which gave me a core clock of 1623 and a memory clock of 7800mhz. Which isn't bad consider is liquid cooled with a universal cooler. Voltage are only +30mv. I want to try to see the max I can get it to but I might wait till I get a full cover block for it.
 


I don't believe you lol liquid cooled or not. The power limit is the limiting factor. Post a screenshot of heaven 4.0 results with it at that clock and evga precision next to it. Please blow me away. Also the MSI versions voltage is a fixed value between 1225mv and 1250mv whether it's at +20mv or +75mv the result is always the same.
 
Well I think I have found out why my card is bench stable at 1500 or so but not game stable. The cause is unexpected, but makes sense.


Read all 5 pages here, paying particular attention to post #20 and also a later confirmation by another user that it works:

https://forums.geforce.com/default/topic/784294/boost-feature-causes-gtx-970-980-instability-in-low-utilisation-situations/


Appears the voltage changes between boost states are not so smooth and that is what causes instability in some games. This is a non-issue when benchmarking because you're always operating at max/near max voltage. But the midrange throttling may be a bit too low for overcloocked cards, especially when you factor in how leakage is engineered into overclocking enthusiast cards.

Basically, you can use either software or a BIOS tweak to bump the minimum voltage up. This will help improve stability across the midrange loads you will sometimes see when v-sync is on for gaming.

Since the G1 Gaming has a failsafe dual bios, I'm going to try the BIOS tweak and see if it works. I may try bumping the minimum up, or I may try editing the voltage tables to give the midrange values a small bump, maybe .0125-.025v per level. We'll see. I'll report back with my findings. :)
 


That would make a lot of sense actually. Would this not be something for Nvidia to patch?
 
I got some pretty sweet results with my Gigabyte G1 970 after a day's work,
I used a bios reflash to bump up my TDP to effective 125% and a slight bump in voltage maximum 1.275 over 1.262 stock.
I was able to do some kombuster tests at 1601mhz core and 8400 mem so I transcribed my overclock to heaven 4.0.

Unfortunately the high memory clock resulted in artifacting and eventual crash, I found 8090 to be the sweet spot where I finished this heaven run. The core clock might have a little headroom, but my perfcap was bouncing off of TDP limit and voltage limit the entire run anyways. Still... needs.... more... powwweerr!!

L7N8qmG.jpg


edit: I reflashed the BIOS for 1606 core and it turns out my TDP limit was not properly raised for that previous run, for 1606/8090 I made 85.8 fps, score 2162. It was pinned to max voltage the entire time.
 


lol, gotta say nice edit. almost believable XD now if only you edited the bigger score to the bottom pic with the higher avg FPS. i mean really? how does 65FPS avg score higher then 82FPS avg? XD nice try trying to troll.
 


At minimum, his bottom score is legitimate with his numbers and score being very close to mine. But I don't see an Extreme HD preset anywhere.

For AAx8 1080p, Ultra and Extreme Tessellation, I get:
TiQ0D7K.jpg



 
i

I have recently bought an MSI GTX 970 4 GB Gaming and I have started overclocking it, and I would like to hear other people'e experiences clocking this card.

I have gotten these results in MSI Afterburner

Clocks:

GPU Core Clock: +200 MHz
Memory Clock: +550 MHz

I have not unlocked core voltage, but I use these power/temp settings:

Power Limit: 110%
Temp Limit: 80 C

(Power Limit and Temp Limit are unlinked)

If I clock either GPU or Mem further by just 50 MHz the system crashes or becomes unstable, but at these settings the system is stable.

In MSI Kombustor benchmark I get these results:

GPU Core Burner v2 (Furry PQTorus)
Score [Factory]: 4936 (82 FPS)
Score [Overclocked]: 5572 (92 FPS)

(Factory is Core and Mem clock set to +0)

So roughly 13% increase in this test

GPU Memory Burner (3072 MB)
Score [Factory]: 3061 (51 FPS)
Score [Overclocked]: 3727 (62 FPS)

So almost 22% increase in this test.

3DMark - Firestrike
[Factory] Score: 9415
[Overclocked]: Score: 10404

11% increase

I would like to hear from other people with this same card. It does not have to be an MSI card, just a GTX 970, though I have read that the MSI card is the most willing to be overclocked, perhaps due to a more effective cooling system?

Stupid question: The reason why MSI Afterburner shows Memory Clock to be half of what it is in other Tweaking apps in this post is because it shows the clock rate of the memory bus with the 2x multiplier of a DDR bus (Dounbe Data Rate)?

I cannot get Core Clock above +200 MHz, even at +210 MHz the drivers crashes as soon as GPU Usage reaches 100%, but at +200 MHz the system is very stable.

What benefit is there to gain from unlocking core voltage? And do the two power connectors on the card need to be feed from separate cords, each connected directly to the PSU? When I first installed the card I could not find two separate power cords so I connected both power inlets to the same cord and yet the card ran flawlessly. I have found the other cord, so now the card IS connected with separate cords. Does this have any influence on performance and/or overclocking potential? (Corsair 750w PSU)