Overclock.net › Forums › Graphics Cards › NVIDIA › Tutorial: Power Target limit hardware mod (Shunt mod) for Titan X and many other Nvidia GPUs
New Posts  All Forums:Forum Nav:

Tutorial: Power Target limit hardware mod (Shunt mod) for Titan X and many other Nvidia GPUs - Page 21

post #201 of 299
For those of you that are shunting your 10x series cards, can you weigh in this?

The KingPin power shunting guide at Xdevs- Extreme OC modifications for EVGA GeForce GTX 1080Ti FE specifically calls out to avoid shunting the RS1, RS2, RS3 5MO sites, qualifying it with a lack of stability, "Don’t do anything with these shunts, just understand what they are for. Common “modifications” like applying liquid metal paste or using pencil are bad idea, as they do not provide reliable and static resistance change."

Instead they recommend "Adding 10Ω 0805 size resistor on top of every larger ceramic capacitor next to INA3221 U26 will reduce power levels about x3 times." which is three IC Power sensors.

Their recommendation is a hard-mod though, having to solder the 0805 caps in parallel as tops to the existing IC chips. Questions I have are:
  • Are most people doing the LM shunting of the 5MO sites because it is mostly reversible?
  • Has anyone taken this route to go mount 0805 caps on the IC power sensors instead of the 5M0 sites?
  • I did notice that in kingpincooling's older article they actually recommend using 5ohm 0603 on the ICs (from a 680 article). I read somewhere (can't find it) the 10x cards are better power rated, so I am assuming that's why they recommend the 10ohm 0805 chips now.

Weighing my options while I wait for my water block to arrive. I'm pretty sure I'm hitting volt limit just as much as power limit so either route may be moot. But, once I get temps under control if it does turn out to be more power limit I may want to try this option.

Anyone do this yet?
post #202 of 299
Quote:
Originally Posted by PeterOC View Post

Shunt mod lets card pull more, but the question is how safe and to what extent it is safe to pull more than 300 W from an 8+6 pin config. My understanding is that recommended power draw limits are 75W for PCIE, 75W for 6pin and 150W for 8pin.

Watch gamer nexus shunt mod analysis. The guy is a electrical engineer iirc and he goes over all the power components and what they're rated at.

For example, the vrm are capable of going up to 40 watts and 150C but we're running them at 16W, not one to their maximum capacity.

How safe? Well we're still doing something to our card that's not meant to be done, but the components are equal and similar to partner cards that are running at 375-400 watts.

If your concern is about the pins themselves, they can take more than 150/75 watts.

https://xdevs.com/guide/pascal_m_oc/

"STEP 3 – Power limit overrides
Often gamers and users are mistakenly referring to 6-pin or 8-pin MiniFit.JR connectors as 75W or 150W capable inputs. Nothing can be further from truth. These power levels are nothing but just way for NV determine how capable is used board hardware in terms of power delivery system. It’s imaginary target number and have nothing to do with actual power taken from connector nor power input capability. Software and NV BIOS will handle GPU clocks and reduce voltages if measured power hitting programmed BIOS limit (which can and usuall is different value than 75/150W per connector!).

If you intend to do serious overclocking and benchmarking, it may be required to trick power monitoring circuitry to report lower power reading, so you not run into power throttle. Also to make sure we are not at any physical limit of power connector itself, check Molex 26-01-3116 specifications, which have specifications both 13A per contact (16AWG wire in small connector) to 8.5A/contact (18AWG wire).

This means that using common 18AWG cable, 6-pin connector specified for 17A of current (3 contacts for +12V power, 2 contacts for GND return, one contact for detect). 8-pin have 25.5A current specification (3 contacts for +12V power, 3 contacts for GND return and 2 contacts for detection). This is 204W at +12.0V level or 306W for 8-pin accordingly.

Now if somebody tells you that 6-pin can’t provide more than 75W, you know they don’t understand the topic very well. It’s not the connector itself or cable limit the power, but active regulation of GPU/BIOS/Driver according to detection of used cables and preprogrammed limits. So now we getting to know how actual power measured?"
Edited by SlimJ87D - 4/16/17 at 1:55pm
post #203 of 299
Quote:
Originally Posted by SlimJ87D View Post

There's an nvidia command you type but I'm not sure it'll work because I think it relies on your cards power percent reading.

For you, it might say 150 watts because you're at 50%.

http://www.overclock.net/t/1624521/nvidia-gtx-1080-ti-owners-club/4680#post_26003001

Please test this while running a benchmark in window mode on the side. We'll get to see how the shunt mod works.
sweet thx will turn the power up and see whats good under the hood
Aurora B
(18 photos)
  
Reply
Aurora B
(18 photos)
  
Reply
post #204 of 299
Quote:
Originally Posted by Tinman12 View Post

For those of you that are shunting your 10x series cards, can you weigh in this?

The KingPin power shunting guide at Xdevs- Extreme OC modifications for EVGA GeForce GTX 1080Ti FE specifically calls out to avoid shunting the RS1, RS2, RS3 5MO sites, qualifying it with a lack of stability, "Don’t do anything with these shunts, just understand what they are for. Common “modifications” like applying liquid metal paste or using pencil are bad idea, as they do not provide reliable and static resistance change."

Instead they recommend "Adding 10Ω 0805 size resistor on top of every larger ceramic capacitor next to INA3221 U26 will reduce power levels about x3 times." which is three IC Power sensors.

Their recommendation is a hard-mod though, having to solder the 0805 caps in parallel as tops to the existing IC chips. Questions I have are:
  • Are most people doing the LM shunting of the 5MO sites because it is mostly reversible?
  • Has anyone taken this route to go mount 0805 caps on the IC power sensors instead of the 5M0 sites?
  • I did notice that in kingpincooling's older article they actually recommend using 5ohm 0603 on the ICs (from a 680 article). I read somewhere (can't find it) the 10x cards are better power rated, so I am assuming that's why they recommend the 10ohm 0805 chips now.

Weighing my options while I wait for my water block to arrive. I'm pretty sure I'm hitting volt limit just as much as power limit so either route may be moot. But, once I get temps under control if it does turn out to be more power limit I may want to try this option.

Anyone do this yet?

Another guy did this in the 1080 Ti thread and F his card up. But he blamed his workmanship, not the mod.
post #205 of 299
Quote:
Originally Posted by Jbravo33 View Post

sweet thx will turn the power up and see whats good under the hood

So did you do the command and was it showing 200 watts and reading it wrong or did it show 330+?
post #206 of 299
Quote:
Originally Posted by Tinman12 View Post

For those of you that are shunting your 10x series cards, can you weigh in this?

The KingPin power shunting guide at Xdevs- Extreme OC modifications for EVGA GeForce GTX 1080Ti FE specifically calls out to avoid shunting the RS1, RS2, RS3 5MO sites, qualifying it with a lack of stability, "Don’t do anything with these shunts, just understand what they are for. Common “modifications” like applying liquid metal paste or using pencil are bad idea, as they do not provide reliable and static resistance change."

Instead they recommend "Adding 10Ω 0805 size resistor on top of every larger ceramic capacitor next to INA3221 U26 will reduce power levels about x3 times." which is three IC Power sensors.

Their recommendation is a hard-mod though, having to solder the 0805 caps in parallel as tops to the existing IC chips. Questions I have are:
  • Are most people doing the LM shunting of the 5MO sites because it is mostly reversible?
  • Has anyone taken this route to go mount 0805 caps on the IC power sensors instead of the 5M0 sites?
  • I did notice that in kingpincooling's older article they actually recommend using 5ohm 0603 on the ICs (from a 680 article). I read somewhere (can't find it) the 10x cards are better power rated, so I am assuming that's why they recommend the 10ohm 0805 chips now.

Weighing my options while I wait for my water block to arrive. I'm pretty sure I'm hitting volt limit just as much as power limit so either route may be moot. But, once I get temps under control if it does turn out to be more power limit I may want to try this option.

Anyone do this yet?

Lots of people have done the shunt mod and it has worked fine for them, I see no reason not to do it. xdevs.com may claim that it reduces stability but they have yet to provide evidence.
post #207 of 299
Thread Starter 
Quote:
Originally Posted by PeterOC View Post

Wouldn't 400 W cause an issue with 6+8pin FE? Its safe power pull limit is 300W =75+75+150

from what I recall, an 8-pin PCI-E connector can do about 275w within spec of the connector and wires from the PSU, and a 6-pin can do about 200w. The only reason people say 75w and 150w is because SIIG says that is spec for PCI-Express. But the cables and connectors are all rated far higher than the bus spec.





Quote:
Originally Posted by Tinman12 View Post

For those of you that are shunting your 10x series cards, can you weigh in this?

The KingPin power shunting guide at Xdevs- Extreme OC modifications for EVGA GeForce GTX 1080Ti FE specifically calls out to avoid shunting the RS1, RS2, RS3 5MO sites, qualifying it with a lack of stability, "Don’t do anything with these shunts, just understand what they are for. Common “modifications” like applying liquid metal paste or using pencil are bad idea, as they do not provide reliable and static resistance change."

Instead they recommend "Adding 10Ω 0805 size resistor on top of every larger ceramic capacitor next to INA3221 U26 will reduce power levels about x3 times." which is three IC Power sensors.

Their recommendation is a hard-mod though, having to solder the 0805 caps in parallel as tops to the existing IC chips. Questions I have are:
  • Are most people doing the LM shunting of the 5MO sites because it is mostly reversible?
  • Has anyone taken this route to go mount 0805 caps on the IC power sensors instead of the 5M0 sites?
  • I did notice that in kingpincooling's older article they actually recommend using 5ohm 0603 on the ICs (from a 680 article). I read somewhere (can't find it) the 10x cards are better power rated, so I am assuming that's why they recommend the 10ohm 0805 chips now.

Weighing my options while I wait for my water block to arrive. I'm pretty sure I'm hitting volt limit just as much as power limit so either route may be moot. But, once I get temps under control if it does turn out to be more power limit I may want to try this option.

Anyone do this yet?
  • yes, easily reversible is a big plus. It is also far easier as most people dont have good solder skills and feel safer doing it this way.
  • One person here mentioned it, but IDK if they ever actually did it.


Kingpin does liquid nitrogen benching for world records. Those temps will cause a liquid metal mod to completely fail and not be reliable in any way, so of course he does not recommend using it. His application of the mods are drastically different than anyone here. And while what he says is true, that LM will not provide uniform decrease because the drop depends somewhat on how much LM you apply, it is still plenty reliable for any of our purposes.
Gaming
(17 items)
 
Gaming PC
(20 items)
 
 
CPUMotherboardGraphicsRAM
7700K AS Rock Z170 OC Formula Titan X Pascal 2050MHz 64GB DDR4-3200 14-14-14-34-1T 
Hard DriveHard DriveHard DriveCooling
950 EVO m.2 OS drive 850 EVO 1TB games drive Intel 730 series 500GB games drive Custom water cooling 
OSMonitorKeyboardPower
Win 10 Pro x64 AMH A399U E-Element mechanical, black switches, Vortex b... EVGA G3 1kw 
CaseMouseAudioAudio
Lian-Li PC-V1000L Redragon M901 LH Labs Pulse X Infinity DAC Custom built balanced tube amp with SS diamond ... 
Audio
MrSpeakers Alpha Prime 
  hide details  
Reply
Gaming
(17 items)
 
Gaming PC
(20 items)
 
 
CPUMotherboardGraphicsRAM
7700K AS Rock Z170 OC Formula Titan X Pascal 2050MHz 64GB DDR4-3200 14-14-14-34-1T 
Hard DriveHard DriveHard DriveCooling
950 EVO m.2 OS drive 850 EVO 1TB games drive Intel 730 series 500GB games drive Custom water cooling 
OSMonitorKeyboardPower
Win 10 Pro x64 AMH A399U E-Element mechanical, black switches, Vortex b... EVGA G3 1kw 
CaseMouseAudioAudio
Lian-Li PC-V1000L Redragon M901 LH Labs Pulse X Infinity DAC Custom built balanced tube amp with SS diamond ... 
Audio
MrSpeakers Alpha Prime 
  hide details  
Reply
post #208 of 299
Unlocking the FE's true potential

After hitting power limits, I could not take it anymore. I knew that it was time for be to do the shunt mod. My card could bench at 2114 Mhz in heaven, and I believe it could go higher. After all, I've seen it boost to 2138 Mhz before without crashing.

I found it as a OC member, it was irresponsible for me to have so much potential thrown away. My theory was that I had a pretty nice golden card. That theory was proved to me today.

What I needed to prepare for

The shunt mod, if done right without too much CLU, will drop your 100% to 75-80%. This means that at 75%-80% you are drawing 300-310 Watts.This can be dangerous because at 120%, that is 450 Watts!

Because I set out to perform this shunt mod, I need to upgrade my cooling. My x41 was nearly at its limits. The liquid temperature was at 40C and my GPU temperatures stock was at 48C. If efficiency was 100%, then liquid and GPU temperatures would be equal according to newton's energy equation. Thus, I knew there was things I could do.

1. I could improve the TIM in between my shim and heat pump. Originally, I was planning to CLU the die and other side of the shim, but after looking online, CLU is really bad for dies from a RMA stand point, it leaves stains. So I decided to just use Gelid Extreme on the die and CLU on the shim and pump. Eureka, this got liquid temperatures from 40C to 44C, this means that more heat was being transferred to the radiator.

2. The midplate is a heatsink itself. It is connected to VRMs, VRAMs and other components. If I was going to introduce more watts to this card, I need to keep them cool. Originally I had placed 8 heatsinks on my midplate. This dropped my core temperature from 51C to 48C. Taking thermal measurements, the shims were at 36C. I knew that if I added more heatsinks, specially closer to the VRMs, it would help with the extra 50 to 75 Watts.



3. Finally, the backplate was not absorbing any heat whatsoever. The aluminum plate is somewhat thin, my theory is that the plate was too thin to absorb heat from the GPU, it would cause warping and flexing. But using my engineering knowledge and deduction, I knew that my AIO was keeping my GPU at 48C, so 45-60C wouldn't harm the back plate at all. So I added a thermal pad in between the back of the GPU and the backplate.




The shunt mod

For my shunt mod, I used CLU and liquid electrical tape.





My results

My previous calculations were correct. From the shunt mod, running a light heaven, I was drawing 300-310 Watts at 75-80% power draw. That means at 100%, I would be drawing 400 Watts.

At max, my card was drawing 95-100%, which mean I was drawing 400 Watts. This is confirmed below!



Here are my thermal results

Stock at 300 Watts, 48C.

Shunt modded at 400 Watts, 50C

All my modifications have successfully fought off the incoming 100 Watts.

After some light benchmarks, I can easily hit 2152 Mhz @ 1.093v with no limiting.



I have decided to call it a day and spend time with my wife. I have not pushed this card to is maximum core clock yet.

With great power... comes great responsibility... Conclusion

This card is now drawing 100 extra watts, which is very dangerous. Because I do not want a $700 paper weight, my next step is to find a balance between my maximum core clock and voltage and try to limit my power draw to 350-375 Watts.

Do I recommend this mod?
To people on water, yes!

To people on AIO, maybe. You must take the extra steps I did to ward off that 100 Watts of extra heat that will be coming to your card.

To people on AIR, No. I do not recommend this mod to people on air. It can potentially be dangerous.

To people with partner cards. Maybe... you have to think about it, you're adding 33% more power and that can increase your temperatures by 33% as a rough estimate. It can be counter productive due to thermal throttling. And if you don't take extra pre-cautions to cool your card, it can blow up!
post #209 of 299
nice work! thumb.gif
Aurora B
(18 photos)
  
Reply
Aurora B
(18 photos)
  
Reply
post #210 of 299
Quote:
Originally Posted by SlimJ87D View Post

Unlocking the FE's true potential


What did you use to measure power draw?
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: NVIDIA
Overclock.net › Forums › Graphics Cards › NVIDIA › Tutorial: Power Target limit hardware mod (Shunt mod) for Titan X and many other Nvidia GPUs