Gigabyte G1 Gaming 970, only +25 Core Voltage

Status
Not open for further replies.

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640
A couple days ago I received my GIGABYTE GV-N970G1 GAMING-4GD GeForce GTX 970. I have been screwing around with overclocking it quite a bit and have noticed two things that seem off:

1. Factory Default Voltage for the card is 1218 mv, but when I use MSI Afterburner to raise the voltage by +87 mv, the actual voltage rise from sensors has never gone past 1243 mv. This is only a +25 mv increase :??:

2. Even under extreme overclocking and benchmarks, I have not seen the card TDP go above 89%, This doesn't seem right considering many reviews for GTX 970's were claiming the overclocking limitation would be TDP.

From my experience, this GPU is just as stable running 1500 Mhz core clock and 8000 Mhz memory on stock voltage and power limit than not. Thats right, raising the voltage and power limit to maximum does not help whatsoever on overclocking. I have pushed this card right up to where it starts to artifact on stock voltage and power limit, then raised the voltage and power limit to maximum (+87 mv and 112%) and the card gets no more stable! :pfff:

Normally when overclocking you are able to go a fair amount farther when you raise the voltage and power limit...

Anybody else with experience know whats going on here?
 
Solution
G
Silicon lottery I guess.

It does seem awfully odd that you're not reaching the maximum TDP potential of your card. That would vary by how lucky you get but with that high of a clock it's strange things aren't going higher. Mine topped out at 108% under Heaven.
Have you tried using Furmark and seeing what it gets up to in that? It's not a good representation, but it tends to push the TDP wall as high as it will go.
You're also absolutely certain things aren't throttling down to that level?

It's difficult to answer your questions when there isn't much of a logical answer to begin with. 1500MHz on the core and 8000MHz on the effective memory frequency is, as I said, about as high as I've seen anyone go to date; give or take a few...
G

Guest

Guest
Also experienced this on mine. I haven't pushed it quite as high as yours clock wise, but it's just as stable at its stock mV than with a slight increase.

Honestly I've only increased it by about 25mV for peace of mind.

Absolutely no throttling either.


Also I'm a little bit concerned. What voltage levels does everyone have their 970's out of the box? Most of the people I've seen are just a notch over 1200mV.
The default for this one was 1175mV.
Currently running 1200mV. Never seen it go near that with the various monitoring tools.
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Ive been all over the internet looking for information on the limitations, but even on overclocking forums there isnt much right now about these cards because they are so new.

Does anybody have any good sources that would confirm that the +87 mv allowed in afterburner doesnt actually occur on the GPU itself? I have never overclocked any other cards, was this the case for other gpus in the past (say 700 series?). It wouldnt make much sense if the overvolting allowed in afterburner never actually occurred on the gpu...
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Accidently chose this as the best answer... :( If someone could unselect it that would be great.

Thanks for the input though, my GPU came at a stock of 1.218 mv which I thought was a little high, I wanted to push it anyways.

I understand the reasons for being careful, but I know a couple experienced overclockers who have run their 700 series cards at max overvolt for years with no problems so I thought I would try it. Thats when I found out that +87 mv is actually +25...
 


Unselected as requested.
 
G

Guest

Guest


Dang, and there was me feeling all giddy.

I wouldn't mind seeing if someone can set up a few thermal cameras and/or temp probes on these things. My tiny brain can't comprehend that they're handling this much without effort.

What programs besides afterburner have you tried to monitor the levels?
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Ive been monitoring using quite a few programs actually, Afterburner, HWiNFO64, and GPU-Z. All show the same Voltage...

Anyways, even though I would love to continue troubeshooting this problem, its 6 in the morning where I live and I need to get some sleep. Il be back on this thread tomorrow :)
 
G

Guest

Guest


That works. Mainly interested in overclocked VRM and memory temps but I suppose with the added voltage of the Gigabyte OC edition it would be about the same.

I'm mainly concerned about the memory and the single VRM phase that delivers it. Slapping 800-1000MHz on top of default frequencies (Effective rates... duh) has got to be cooking something right? =/ Surely.

Anyway this is getting off-topic I guess. My bet would be just plain ol' lack of support in its current form. I mean, GPU-Z still lists it as having 32 colour rops. Either that or it is genuinely the limit for the OP's card, those clocks are about as high as I've seen anyone go so far so it makes sense that going further would be difficult.
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Makes sense to me that the temp of the VRM's could be limiting my overclocking potential, but im still quite confused on why I am only able to achieve a +25 mv increase. I was actually hoping someone with a lot of overclocking experience would drop in and tell me if this is normal or not.

My other question still stands too, what's the point of a 112% power limit if the card never even gets above 90%? I was reading reviews of the Gigabyte 970 where people were claiming that their cards were throttling under load because they reached max TDP. Why isn't that the case for me?

I guess at the end of the day I should just be happy with my overclock, which is pretty darned good already. Im just a little baffled about my questions.

Edit: Have been doing some research, someone on Overclock.net claims his Gigabyte 970 hits max 1.26 volts...
http://www.overclock.net/t/1517316/extract-and-flash-gtx-970-and-980-firmware-gpuz-test-build-and-new-nvflash-5-190/40
 
G

Guest

Guest
Silicon lottery I guess.

It does seem awfully odd that you're not reaching the maximum TDP potential of your card. That would vary by how lucky you get but with that high of a clock it's strange things aren't going higher. Mine topped out at 108% under Heaven.
Have you tried using Furmark and seeing what it gets up to in that? It's not a good representation, but it tends to push the TDP wall as high as it will go.
You're also absolutely certain things aren't throttling down to that level?

It's difficult to answer your questions when there isn't much of a logical answer to begin with. 1500MHz on the core and 8000MHz on the effective memory frequency is, as I said, about as high as I've seen anyone go to date; give or take a few MHz. I would put it down to a software bug more than anything else, personally.

Have you also tried returning to default clocks and running the same tests? Does the power usage push past 89% with that?

Finally, correct me if I'm wrong (I used ASUS's GPU-Tweak utility to overclock mine as opposed to Afterburner), is that slider not suggesting the maximum allowed voltage? That would explain why it isn't using that amount under load.
The only other idea I have on that front is GPU Vdrop/Vdroop.
 
Solution

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


I used to use heaven for benchmarks until it started to cause Nvidia Control Panel to crash :(

Anyways, I took your advice and just took a 15 minute burn test with Furmark, it is indeed giving me different results than the tomb raider benchmark I was using! I am now reaching a power limit of up to 109%, however the core voltage is staying quite low, not even to 1.2 volts (is this because it is throttling?).

The following is a bunch of pictures of the tests and results for comparison.

To start off, here are the settings my GPU is running at:

Afterburner:
Overclock_Msi_Afterburner.png


GPU-Z:
Overclock_GPU_Z.png


And here is my Furmark score (quite good I think :D )
Furmark_Score.png


I also used HWiNFO64 to take data during the Furmark benchmark and tomb raider benchmark (both at same GPU settings) but I cant attach the logs to this post. Here is a link to both of them, they took data every second and recorded gpu temp, core clock, memory clock, power usage, fan speed, Vram used etc. if anyone wants to look at them:

Overclock HWiNFO64 Logs.zip

So the results of this are:
1. Power usage seems normal, differs depending on benchmark type.
2. Voltage Limit is still unanswered (I get even lower voltage in Furmark than tomb raider?)
3. Core clock seems to differ depending on benchmark as well. I get 1500 constant in tomb raider but only ~1418 constant in Furmark (again, is this throttling?)

Note: I saw no artifacts during either benchmark.

 
G

Guest

Guest
I wouldn't worry about Furmark. The whole point to that program is to go balls-to-the-walls. To the point where both AMD and Nvidia auto-detect the executable and throttle accordingly (Which probably explains why engine clock drops significantly while running it). Or that's what I'm told anyway.
I only ran the 1min bench a few times with it but saw mine go as low as high 900MHz. It kinda decreases the more you run it, almost as if the cards know what they're in for.

Heaven maxed out is about as stressful as things are going to get for now. Even trumps maxed Crysis 3 as far as gpu load goes, for me anyway. Valley is just under it, despite being more recent.

Here's something interesting - Your card boosting up past its set boost rate, heck, past 1500MHz (Which is very good). That'll be a testament to the Gigabyte version. I have mine listed at 1295MHz in GPU-Z and the highest I've seen it boost is 1447. Fair play to that card.
These kinds of clocks are knocking on the 980's door really.

In short. No crashing or no artifacts = No worries. I would say watch the temps but..... Duuude, 145w TDP card and a Windforce cooler, they are so not going to be an issue.

Edit: Try 3DMark11 too.
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Alright this makes me feel a little better :)

So to be clear, the voltage dropping in Furmark is expected?
 
G

Guest

Guest


If you watch the reader in the Furmark bench itself you'll notice that it will frequently trip over the maximum power target you set. Usually by about 1-4%. This will most likely cause it to throttle back, couple that with deliberate throttling and it's very explainable. Though I can't speak for everyone else.

Put it this way, when I was experimenting with my OC, I stuck with the same clocks and changed nothing but the power target limit - 120 on one run and 110 on another. It would always trip over the line, it's what the program is designed to do.
I say a real world benchmark is the best way to test. Furmark is only useful for a quick double check that it doesn't crash with your OC.

If you go through a 3DMark11 run drop the final score here :F It would be interesting to compare when we both essentially have the same CPU, what with the 4.5GHz overclock n' stuff.
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


So because my power limit is set at 112%, during Furmark it is tripping over to about 113-116% and then throttling back?

And as far as 3dmark goes, I just downloaded it. I will run a benchmark in a couple hours and post back here :)

 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640
Just ran 3dmark, kinda scared me a little because of the frame rate at the end for the combined test :(

Here are the scores:
Firestrike.png


Im a little worried about this score considering the normal score of a stock clocked Gigabyte 970:
Gigabyte GeForce GTX 970 G1 Gaming review - DX11: Futuremark 3DMark 2013

Im not sure, but my guess is my score is lower because of my CPU. Intel CPU's win out over AMD's when it comes to pure performance from what ive seen. For the Firestrike score above the test system had a Core i7 3960 Extreme (Sandy Bridge-E) @ 4.6 GHz on all six cores, pretty crazy CPU.

So how does this compare to your system with the FX-8350 @ 4.5 Ghz?

Edit: Forgot to add that when examining the log taken during the benchmark I never saw the Power usage go over 90%. Core clock boosted right up to 1500 though.
 
G

Guest

Guest


Ah crap, I haven't ran Firestrike yet :F I just ran the general 11 one.

I'll drop a result tomorrow. 2am here and I require lots of beauty sleep to maintain my dashing good looks.
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Sorry haha, I went to download 3DMark and steam told me there was a newer one available.

Here are my scores for 3DMark11:
3_DMark11.png


Examination of sensor logs shows that the gpu ran at a constant 1500 on the core, and hit a maximum of about 99% TDP. Again, the physics test wrecked my CPU, causing it to hit almost 100% on all 8 cores.

 
G

Guest

Guest
Right then.
First thing's first - Firestrike is bleedin' demanding.

Anyway, here's how the scores played out.

Firestrike:
http://www.3dmark.com/3dm/4439138

3DMark11:
http://www.3dmark.com/3dm11/8845775

I gave it about 8 runs in total. Something interesting is that GPU-Tweak would dial back the voltage to 1213mV if I attempted to go over that, upon clicking Apply. That fits in line with what you were saying in the OP and what the first poster said about Nvidia voltage limits. Some 980 and previous 700 series cards have/had voltage points on the card itself to which you could use to get around these limits.

I adjusted a few settings from my original overclock but eventually came back to 1205mV and 110% on the power limit. Firestrike did use up to 118% of the TDP with the power limit raised, but there was no observable difference in stability. Considering everything else but Furmark used up to 108% I dialled it back.

Notice the obvious, with your higher clocks the graphics score on both tests is higher for you despite our similar processors.

Both scores can change up or down each time you run them I noticed.



I wouldn't worry too much about Piledriver's lack of oomph. The review that you linked to gave nearly a 1000 increase evidently down to the CPU alone, as you said it's an extreme edition processor. The results page on the 3DMark website has a handy comparison tool, according to it, both your score and my score beats a machine with an i7-4770k and GTX Titan. Now theoretically this should be obvious, with these clocks the 970 is coming awfully close to a reference 980, the Titan was only slightly inbetween the 780 and 780TI - the 980 is about 10% above the 780TI.
It's not ideal, it's a bit sucky to have a card slightly limited by the CPU, but It's really not that bad.

Either way it appears people with Piledriver chips and 970's are getting very similar scores going by the same comparison method. I can sleep easy knowing that.

Finally, I have to ask because it bugged me - During one of the graphics tests in Firestrike, the one where the female is walking on what I can only describe as a street, there's a bit at which a robotic bug creature appears behind her very near to the end of that test - Does your screen flash black once or twice during this part? It's very very fast, probably less than a second, but it has me wondering.
I'm not particularly ready to jump the gun and lower the OC because of it, no crashes or artifacting in every benchmark and real-world game, I'm not about to drop it for a few flashes, I just find it interesting.
 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Thanks for reminding me about the comparison tool, I just looked into it and its interesting to try different combinations to see the scores.
Something I thought was particularly interesting is that a stock clocked FX-8320 scores between 1000 to 2000 less than me with my 1 Ghz overclock :D

As far as the flashing you are talking about: I did not see anything like that when I ran the benchmark but I did see a couple of small red flashes later on in the test during a part before you enter the temple. It doesn't worry me because I am yet to see any problems in real game performance. If I do see problems I will lower my overclock, but for now I am happy.

Edit: My questions pertaining to this thread have basically been answered, so I selected one of your posts as the answer. Thanks for all the help! :)
 

munchy22

Reputable
Oct 31, 2014
27
0
4,540
According to my over clock and using cpuz, gpus, 0 evga precision 16 my gtx970 g1 is running at 1556mhz and about 4gig mem and I see voltage at 1.2 all the time with the occasional peak at 1.43. I also noticed in heaven and valley bench mark it shows gpu core at 1673 but im guessing that's something to do with cpu, I don't really know cause im a noob at this, I just love my overclock. All air no special cooling never higher that 72 and I think a score of 1600 in heaven is respectable situation lol. Hope that helps


 

Quentin Chalmers

Reputable
May 11, 2014
91
0
4,640


Dang, sounds to me like you won the silicon lottery...

But on the other hand the fps difference between a couple hundred Mhz really isn't that much anyways.
 

reg66

Honorable
Nov 8, 2012
49
0
10,530
thinking about overclocking my G1 970. (currently reading up on it, complete noob). haven't tried anything yet. but, prior to oc'ing...

Q. my GPU stock voltage is 862mv - tested with precisionX and GPUZ, is this potentially a good or bad thing? i see others are getting much higher @ 1200mv or so (at default/stock)
 
Status
Not open for further replies.