Increasing GPU Voltage has no effect on overclock? [GTX 780]

[THEINTUNITY]

I've been running an Asus GTX 780 DC2 videocard for 2 months now whilst running on a 24/7 overclock of 1175 Mhz Clock and a 6750 Mhz on the memory clock. I haven't increased the voltage on the videocard, but the weird thing is, is that adding voltage does not let me overclock any higher without seeing artifacts. At 1200 Mhz it does not matter whether I have +38 Mv on the GPU or am using stock voltage because in the end they both show artifacts. It also has no effect on memory overclock. What could be the cause of this?

The software I'm using, if it matters, is Asus Gpu Tweak. If someone comes here to say that my PSU could be the limiting factor, than no, my PSU is fine. (EVGA SuperNova 850 Watt G2, gold rated)

Thanks in advance!

 

[NightAntilli]

Maybe the voltage isn't actually changing. It could be locked, or it could be bugged.

 

[THEINTUNITY]

Maybe the voltage isn't actually changing. It could be locked, or it could be bugged.

No, I actually did monitor the GPU voltage, Power Target, Temperature and clock speeds using GPU-Z while playing games.

 

[JackNaylorPE]

Your PSU is fine .... I have two water cooled Asus 780 DCII's and your experiences mirror my own.... I have done better with the newer MSI 780s air cooled on other builds. One question though, when you say 1175, I assume you are talking boost clock, not base core clock as 1175 would be a way extreme 36% OC. That would be one helluva accomplishment on an air or even water cooled build, without modding the voltage control more than is allowed by the DCII series.

Here's what I see when running Unigine Valley ....

Unigine Valley shows 1410 MHz
GPU Shark Shows 1251 Mhz
GPU_Z shows 1231 MHz

But those are all boost clocks. The base core clock setting I am using now for stability in all games is 1071 because, while all the benches and other games are fine up well past 1100, BF4 crashes anywhere from 15 minutes to 2 hours in at higher settings. Same with memory, at 7200-7440 was stable in benchmarks but BF4 gets cranky at over 6800. I don't even play the danged game, my son does . In the benchmarks like Valley, my temps never break 39C .... Furmark can make them hit 44C at max 850 rpm on the fans (39C at 1200 rpm).

The thing is this:

1. nVidia has placed heavier and heavier legal and physical design restrictions on the cards the with each generation making it harder and harder for individuals to distinguish themselves. 38mV is really a very small increment

2. As such many vendors have oft (i.e 780 Ti) not even opted to use custom PCBs and their offerings have equaled and even in some cases (MSI 780 Ti) outperformed competing vendor's cards.

3. Even water cooling a card has had ever diminishing performance gains with each generation. As I said above, air cooled builds I have been involved with since building by SLI water coooled 780 build have equaled or slightly exceeded my results.

4. Both CPUs and GPUs (well nVidia ones anyway) are getting more efficient such that cooling is not the "wall" so much anymore..... it is more of a voltage thing and as discussed abve, nVidia has put the clamps on here. I see a max VDDC of 1.086 while running Furmark. That +0.38v is merely 3.5% of that.

5. Different utilities as shown above report different results, even versions of GPU_z report different results at the same settings .... while GPU shark was reporting 1.086, when using GPU_Z, I never saw above 1.0

6. You can download custom BIOSs for the DCII's which will give you much more voltage boost .... however, while users have reported that they get much higher OC's, what I have not seen is a corresponding increase in actual benchmark or gaming performance.

 

[NightAntilli]

Ok. Well, the software used doesn't really matter either so... It might be a limit that has been set on the current regulator on the graphics card, which means increasing the voltage only allows a lower amount of amperage, starving the GPU/memory, creating artifacts.

 

[JackNaylorPE]

Voltage / Amperage really doesn't matter from throttling standpoint as the throttling is controlled by either temperature or power (V X A) limits. Power in a circuit is generally limited by amps.... increasing the voltage while maintaining amperage limits allows more power to reach the CPU / GPU.

 

[THEINTUNITY]

Thanks guys for the helpful answers and informative information. @jackNaylorPE , yes I'm talking about the boost clock. And also, you said that 38 MV is a very small amount, I can get to a maximum of 63 MV, could this voltage do any harm to the card for a 24/7 overclock? Also, no, I might be an overclocking enthousiast, but I'm not planning on using custom bioses since I cannot take the risk to brake the card, it has to do at least one year until the next 16 nm Nvidia GPU's come out next year. And last thing...about the software, if different software read different readings, which software is the most accurate would you say? GPU-z?

Thanks again guys

 

[JackNaylorPE]

I have no way of knowing which is more accurate w/o setting up the card in an open test bench and taking physical measurements. I dont remember GPU Tweak, haven't launched it in a while but the max slider I see is +38 mv

http://www.guru3d.com/articles_pages/asus_geforce_gtx_780_directcu_ii_review,26.html

Original This sample Overclocked
Core Clock: 863MHz Core Clock: 889MHz Core Clock: 1034MHz
Boost Clock: 906MHz Boost Clock: 941MHz Boost Clock: 1215MHz
Memory Clock: 6008 MHz Memory Clock: 6008 MHz Memory Clock: 6732MHz


To get the tastebuds going here's an example of what you can achive with a tweak or two. We applied:

Power Target 110%
Priority at Temperature target
Temp Target 95 Degrees C
CPU clock +145 MHz
Mem clock +360 MHz
Voltage +37 Mv