GTX 970 Owners Club

[willegard]

TRIPLE POST INCOMING;

 

I guess my NON K cpu is bottlenecking me.. Right?
https://i.imgur.com/1nEDaxF.jpg

[tomaatvk]

After doing some more small bios tinkering I figured out a few things about my inno3d ichill 970:

1. it'll stall out at around 260 watts no matter the bios settings (it will go lower if I tell it to).

2. if I use constant voltage and set the powermode in the nvidia config screen to optimal prestations my max vcore is 1.1v and if I set the powermode to adaptive it is 1.212v at it's max.

3. I can overclock the GPU to 1562MHz and the memory to 3580MHz but if the fan speed goes over 25% it'll throttle as low as ~1250 at 100% fan speed.

[willegard]

After doing some more small bios tinkering I figured out a few things about my inno3d ichill 970:

1. it'll stall out at around 260 watts no matter the bios settings (it will go lower if I tell it to).

2. if I use constant voltage and set the powermode in the nvidia config screen to optimal prestations my max vcore is 1.1v and if I set the powermode to adaptive it is 1.212v at it's max.

3. I can overclock the GPU to 1562MHz and the memory to 3580MHz but if the fan speed goes over 25% it'll throttle as low as ~1250 at 100% fan speed.

 

Sounds like your GPU would benefit from water A BUNCH. Look into getting the Kraken G10 and a water cooler (Corsair H55 for example). You'd hit alot better numbers with it.

Cheers

[tomaatvk]

Sounds like your GPU would benefit from water A BUNCH. Look into getting the Kraken G10 and a water cooler (Corsair H55 for example). You'd hit alot better numbers with it.

Cheers

 

May I ask why it would benefit from watercooling? Thermals aren't the problem at all, I'm going to run the GPU fan power cables from my PSU directly and use the GPU's PWM and RMP cables so my GPU can cool itself without sacrificing it's own power and I'm going to do the resistor thing soon so I can maybe get some extra wattage into it.

[Amazingasian]

May I ask why it would benefit from watercooling? Thermals aren't the problem at all, I'm going to run the GPU fan power cables from my PSU directly and use the GPU's PWM and RMP cables so my GPU can cool itself without sacrificing it's own power and I'm going to do the resistor thing soon so I can maybe get some extra wattage into it.

Your GPU is showing signs that it can easily go higher on the clock and memory.

[tomaatvk]

Your GPU is showing signs that it can easily go higher on the clock and memory.

I'm not that great in overclocking, please explain how I could make it go higher. The voltage is already maxed (can't go above it even with bios tweaks) and I will have to hard mod to increase the powerlimit.

[Vellinious]

After doing some more small bios tinkering I figured out a few things about my inno3d ichill 970:

1. it'll stall out at around 260 watts no matter the bios settings (it will go lower if I tell it to).

2. if I use constant voltage and set the powermode in the nvidia config screen to optimal prestations my max vcore is 1.1v and if I set the powermode to adaptive it is 1.212v at it's max.

3. I can overclock the GPU to 1562MHz and the memory to 3580MHz but if the fan speed goes over 25% it'll throttle as low as ~1250 at 100% fan speed.

Have you had GPUz sensors tab open while you're running your tests?  Why is it throttling with the fans at 100%?  PWR?  Temps?  Watch the perf cap reason line.

[Amazingasian]

I'm not that great in overclocking, please explain how I could make it go higher. The voltage is already maxed (can't go above it even with bios tweaks) and I will have to hard mod to increase the powerlimit.

It's throttling the clock speeds when you get to a certain temperature.  Either flash a custom BIOS and remove turbo boost or get a Kraken G10 bracket with an AIO watercooler and you'll see your temps fall.  That's good because under load, your temps won't go too high and your clock will stay put.

[tomaatvk]

Have you had GPUz sensors tab open while you're running your tests?  Why is it throttling with the fans at 100%?  PWR?  Temps?  Watch the perf cap reason line.

It's throttling the clock speeds when you get to a certain temperature.  Either flash a custom BIOS and remove turbo boost or get a Kraken G10 bracket with an AIO watercooler and you'll see your temps fall.  That's good because under load, your temps won't go too high and your clock will stay put.

 

The throttling happends because the GPU gets too little power. I looked at every sensor on it. If the fans spin up they use up a part of the total power the GPU can draw so the chip itself can't use that power anymore. Not because of temperatures or whatever, just because of the maximum power that can be used.

[Vellinious]

The throttling happends because the GPU gets too little power. I looked at every sensor on it. If the fans spin up they use up a part of the total power the GPU can draw so the chip itself can't use that power anymore. Not because of temperatures or whatever, just because of the maximum power that can be used.

So, you're seeing green in the "perf cap reason" line.  Indicating a power limit throttle?  Increase the power limit more in the custom bios...

[tomaatvk]

So, you're seeing green in the "perf cap reason" line.  Indicating a power limit throttle?  Increase the power limit more in the custom bios...

 

Yup but you think I didn't already? I think I may have hit a maximum of what the card can actually draw for whatever reason, might play around a bit more with the numbers, see if I can get some result but I don't think so.

[Vellinious]

Yup but you think I didn't already? I think I may have hit a maximum of what the card can actually draw for whatever reason, might play around a bit more with the numbers, see if I can get some result but I don't think so.

Try this.  No idea why the voltage is so high.....but, I didn't mess with it.  Just upped the power limits and adjusted the 6 / 8 / pcie slot power.  I probably should have raised the TDP too, but......if it needs it, change it to 300 to match the power limit.  May even need more.  Experiment with it and see how it goes.

https://drive.google.com/file/d/0B6zqzZ0qTCB5Y2QwcE9Cd1RlVDg/view?usp=sharing

[tomaatvk]

Try this.  No idea why the voltage is so high.....but, I didn't mess with it.  Just upped the power limits and adjusted the 6 / 8 / pcie slot power.  I probably should have raised the TDP too, but......if it needs it, change it to 300 to match the power limit.  May even need more.  Experiment with it and see how it goes.

https://drive.google.com/file/d/0B6zqzZ0qTCB5Y2QwcE9Cd1RlVDg/view?usp=sharing

 

The voltage is so high because Gofspar did that for me (without any real effect sadly), I'm currently already running a TDP of 320, a power limit of 290 and a PCI-e 6 pin power of 115, 8 pin of 175 and pcie of 75 and for whatever reason it keeps saying power limit reached at only 260 watts.

[Gofspar]

The voltage is so high because Gofspar did that for me (without any real effect sadly), I'm currently already running a TDP of 320, a power limit of 290 and a PCI-e 6 pin power of 115, 8 pin of 175 and pcie of 75 and for whatever reason it keeps saying power limit reached at only 260 watts.

If you're card scales to 1.3v you have a really good card, most cards scale to like 1.243 or 1.262v

[tomaatvk]

If you're card scales to 1.3v you have a really good card, most cards scale to like 1.243 or 1.262v

 

Well, mine does to 1.212 and it already did before.

[tomaatvk]

I thought I'd explain my situation well for once:

 

I have my power limit set to 290 watts with the TDP at 320 watts. Through the PCI-e, PCI-e 6 pin and 8 pin connector my GPU could draw upto ~360 watts.

 

When my GPU is cold the fans are turned off and they turn on before it gets close to the thermal throttling point so it doesn't throttle because of the temperature, never.

 

Loading up a stresstest with my GPU starting at a low temperature it will use boost up quite high using around 260 watts with the occasional spikes to 290 but also drops to 250 (anywhere above the 260 it'll say that it reached it's max power limit eventhough it's set to 290 watts in the bios and msi afterburner).

 

After a tad bit of time the GPU will have warmed up to the point that it turns the fans on. Because of some overshooting I put in the fan curve they'll spin up to 100% for 3 seconds and after that they'll spin down to around 60% (with the temperatures still within limits, no thermal throttling).

 

If the fans are set to 100% the GPU uses around 220 watts with spikes upto 240 and drops to 200 (anywhere above the 220 it'll say that it reached it's max power limit eventhough it's set to 290 watts in the bios and msi afterburner) and if the fans are set to 60% the GPU uses around 250 watts with spikes upto 270 and drops to 230 (anywhere above the 250 it'll say that it reached it's max power limit eventhough it's set to 290 watts in the bios and msi afterburner).

If I were to set the fan speed to 0 RPM after this with the test still running it'll use it's "full" 260 watts again.

 

Using Corsairlink the ammount of power pulled by my system stays between 10 watts no matter the fan speed and watts pulled by my gpu (during the testing).

 

TL:DR

max GPU power set = 290 watts

max GPU power used @ 0% fan speed = 260 watt

max GPU power used @ 60% fan speed = 250 watt

max GPU power used @ 100% fan speed = 220 watt

No limits besides power limits are reached. Thermals and voltages are OK.

[Vellinious]

Do you have a screenshot of the GPUz sensors tab during one of your test runs?

[tomaatvk]

Do you have a screenshot of the GPUz sensors tab during one of your test runs?

 

Yup, at the start the fans are manually set to 100% and after that they turn off for some time and go up to 60%, 100% TDP=200 watts FYI

[Vellinious]

Have you tried to raise the TDP and power limit to like...350 or so?

[tomaatvk]

Have you tried to raise the TDP and power limit to like...350 or so?

 

Actually planned on going something like 400 tomorrow to make sure I'd see a difference.

[tomaatvk]

So I did the 400 watts GPU bios mod, nothing changed compared to before except that I can set 400 watts in MSI afterburner (with no effect).

 

This is the bios I'm currently using with the 400 watts (eventhough I only gain 260).

http://www.mediafire.com/download/1tnqdb8zqw3i61a/gtx9702.rom

 

Also, my PSU is an RM750i and I run my GPU off of 1 split PCI-e power cable with 2 ends, 2 6+2 pin connectors, I think I'm gonna try running more power cables to it so that resistance is dropped a bit (if it would make any difference).

 

EDIT: I ran more cables to my GPU, nothing changed, again. I'll need to short some resistors in the GPU to make it have more power.

[SkilledRebuilds]

So this is nice and all... using the 1.275v or 1.3v bios settings (Can't remember)

But does that TDP value mean I can go further, or maximize it further in any way.

Thats around the highest TDP goes... averaging 60-70% TDP at maximum.

 

/Memclocks are low cos card does not like it (compared to every1 elses) Anything over +100 memclock or so, and in-game it eventually artifacts.

The Core last time I tested, went up to 1565Mhz and It crashed Battlefront (could be isolated to SWBF tho) but as that was the only game installed at the time... I just backed off to 1550Mhz and called it a day.

 

So back to the question...Does that TDP value mean I can go further, or maximize it further in any way.

No perf-cap signs are happening in GPU-Z

[christianled59]

I get very bad artifacting with any of the settings you posted while benchmarking. Also, during gameplay, the screen just goes blank and I have to do a hard reset.

[Monarch]

So this is nice and all... using the 1.275v or 1.3v bios settings (Can't remember)

But does that TDP value mean I can go further, or maximize it further in any way.

Thats around the highest TDP goes... averaging 60-70% TDP at maximum.

 

/Memclocks are low cos card does not like it (compared to every1 elses) Anything over +100 memclock or so, and in-game it eventually artifacts.

The Core last time I tested, went up to 1565Mhz and It crashed Battlefront (could be isolated to SWBF tho) but as that was the only game installed at the time... I just backed off to 1550Mhz and called it a day.

 

So back to the question...Does that TDP value mean I can go further, or maximize it further in any way.

No perf-cap signs are happening in GPU-Z

 

Your TDP is around 70% in Valley, but that doesn't mean it won't go higher in everything else. I get much higher TDP in Firestrike 1.1 than in Valley. It depends on what the GPU is rendering. In BF4 for example, TDP is low, but once explosions start to happen on screen it jumps up to 100%, which is 310W in my case, at 1.262. 

[Monarch]

So I did the 400 watts GPU bios mod, nothing changed compared to before except that I can set 400 watts in MSI afterburner (with no effect).

 

This is the bios I'm currently using with the 400 watts (eventhough I only gain 260).

http://www.mediafire.com/download/1tnqdb8zqw3i61a/gtx9702.rom

 

Also, my PSU is an RM750i and I run my GPU off of 1 split PCI-e power cable with 2 ends, 2 6+2 pin connectors, I think I'm gonna try running more power cables to it so that resistance is dropped a bit (if it would make any difference).

 

EDIT: I ran more cables to my GPU, nothing changed, again. I'll need to short some resistors in the GPU to make it have more power.

 

Try to set 400W as default, rather than setting it as max value and having to use Afterburner to increase the power limit. When I modded my BIOS I think I had issues with the power limit as well, I'm not sure, but I think I fixed it by setting the max power limit I wanted as default value. Keep in mind that max power limit has to be higher than default, otherwise you'll get a BSOD, so set it to 410W or something.