Overclock.net banner

241 - 260 of 382 Posts

·
Facepalm
Joined
·
8,623 Posts
I soldered and thoroughly cleaned up with isopropyl and scraped off the edges of each shunt

I'll have another look on the pcie shunt and probably apply more solder on the edges
Yeah check what's going on there.
You can see that your TDP Normalized is much higher than your TDP. I believe TDP Normalized is based on some factor with a rail being close to an individual power limit. If normalized is much higher than TDP, you have an issue going on. For example if your TDP is 103.8% and normalized is 121%, then you KNOW one of the power rails is totally out of balance.
 
  • Rep+
Reactions: dante`afk

·
Registered
Joined
·
718 Posts
thanks.

how sure are you it's the pcie shunt?
Is there any tool I could buy to test each shunt seperately?
 

·
Facepalm
Joined
·
8,623 Posts
thanks.

how sure are you it's the pcie shunt?
Is there any tool I could buy to test each shunt seperately?
The board power draw is a combination of the 8 pins and PCIE slot.
One person didn't do the PCIE shunt and his MVDCC shot up to 200W.
But when he did do the shunt, MVDDC dropped.

So it's definitely one of the three shunts on the backplate itself.

I'm not sure what's going on but you may need to look at his two posts.

 
  • Rep+
Reactions: dante`afk

·
Registered
Joined
·
18 Posts
Is it bad for the card to not shunt the pcie? Because I didn't shunt mine lol. Only the top two power connector shunts. I am still waiting on a bykski block to come from china. I might shunt it when I install it. This is a gigabyte card btw. As you can see the normalized tdp is very high.



2466285
 

·
Facepalm
Joined
·
8,623 Posts
Is it bad for the card to not shunt the pcie? Because I didn't shunt mine lol. Only the top two power connector shunts. I am still waiting on a bykski block to come from china. I might shunt it when I install it. This is a gigabyte card btw. As you can see the normalized tdp is very high.



View attachment 2466285
You're being power throttle limited by the PCIE slot.
When normalized TDP is much higher than TDP, it means one of the wattage rails is out of spec. Normalized TDP is not well defined, but Martin of Hwinfo said it's something in the Nvidia API with how it's used.

Your bios thinks it's drawing only 148W and 139W through the two 8 pins, even though it's clearly pulling more, but it's being limited from going higher by the PCIE slot triggering power limit throttling.

My question for you is, what shunt did you use to shunt the two 8 pins?
Your vbios thinks they're at ~145W or so each. However your port royal score is acting like you're pulling around 420 watts total. At 420 watts, I would expect the 8 pin "uncorrected" values to be about 120 watts or so from the 8 pins. But that also depends on what mOhms you used to shunt. 10 mOhm shunts on the two 8 pins should definitely be reporting less power draw than that (10 mOhms is a 1.5x multiplier).
 
  • Rep+
Reactions: Keninishna

·
Registered
Joined
·
18 Posts
You're being power throttle limited by the PCIE slot.
When normalized TDP is much higher than TDP, it means one of the wattage rails is out of spec. Normalized TDP is not well defined, but Martin of Hwinfo said it's something in the Nvidia API with how it's used.

Your bios thinks it's drawing only 148W and 139W through the two 8 pins, even though it's clearly pulling more, but it's being limited from going higher by the PCIE slot triggering power limit throttling.

My question for you is, what shunt did you use to shunt the two 8 pins?
Your vbios thinks they're at ~145W or so each. However your port royal score is acting like you're pulling around 420 watts total. At 420 watts, I would expect the 8 pin "uncorrected" values to be about 120 watts or so from the 8 pins. But that also depends on what mOhms you used to shunt. 10 mOhm shunts on the two 8 pins should definitely be reporting less power draw than that (10 mOhms is a 1.5x multiplier).
I soldered 8 mohm shunts on them. Hmm, if I put a 10 mohm on the pcie slot do I have to shunt the other 3 shunts as well?
 

·
Facepalm
Joined
·
8,623 Posts
I soldered 8 ohm shunts on them. Hmm, if I put a 10 ohm on the pcie slot do I have to shunt the other 3 shunts as well?
Put the 10 mOhm shunt on the PCIE slot first.
Remember to scrape the edges of the original shunt so it shows silver to ensure proper contact.
You don't need the same shunts on everything. You should see my board.

My eight pins are acting like it's a 11 mOhm shunt stacked on it. And I just have a lot of MG 842AR paint over them (no shunts). I tried shunts but they didn't work properly due to the edges of the original shunts being lower than the middle. I had them stacked with 5 mOhms before and 8 pin #1 was reporting 175W (throttle point) while 8 pin #2 was reporting 135W, so I removed the shunts and applied new 842AR paint over the dried paint that was on the edges of the original shunts (barely any was on the shunts i pried off) and boom: it worked!

I have a 5 mOhm stacked shunt on GPU core. That seems to be working fine.
I have 10 mOhm shunts on PCI Express (this started triggering power at 68W, but slowly lowered to 64W I guess as the board heated up and the shunt/paint worked itself in(?), and 10 mOhm on Current Sense and MSVDD (attached by MG 842AR paint).

This is how your system should look.

I used a "1.47x" multiplier for everything in Hwinfo64, after doing some calculations. Although I'm not sure if the PCIE factor is 1.47x. I think it's probably more like 1.40x, because the PCIE I think signals throttle at 68W, while the 8 pins signal throttle at 170W....but I'm not sure. So I used 1.47x after doing Time Spy testing pre-mod and after mod. According to the TDP and normalized values, I think I'm pretty close.

2466289
 

·
Registered
Joined
·
18 Posts
Yeah, the main thing I worry about is with the unbalanced power draw if its stressing out the power controller or the vrms or caps if I don't do all the shunts. Your mvddc is super low compared to mine, well 0w in that screenshot lol. I do get thermal throttling if I push the card harder but its not from the GPU chip itself because the GPU temp is below thermal threshold so its from some where else on the board.
 

·
Registered
Joined
·
103 Posts
Discussion Starter #250
which one did you get?
From Bitspower I got this one: BP-VG3090FE

I'm extremely happy with it so far. The fins are not very dense, though my card never goes over 50 C with all my rad fans on super quiet mode. If I set the fans to a performance curve, it usually never peaks 47 C with a 114% power limit and +100mv core voltage.

I am planning on ordering the EK 3090 FE block as soon as it goes up for pre-order though. The EK block looks far nicer IMO
 
  • Rep+
Reactions: dante`afk

·
Facepalm
Joined
·
8,623 Posts
also do you need to short all of the shunts?
Yes you do need to repaste. If the pads are of high enough quality and undamaged, they can be re-used, although that depends on the quality of the stock pads.
All shunts must be modded.
 

·
Registered
Joined
·
93 Posts
anyone know where the SRC resistor is located on the reference PCB, I think I am still getting power limited due to SRC power.
2466352
 

·
Facepalm
Joined
·
8,623 Posts
anyone know where the SRC resistor is located on the reference PCB, I think I am still getting power limited due to SRC power. View attachment 2466352
It's either resistor 1 or 2 on the first page of this thread.
I know #2 is linked to 'current sense' so I'm not sure if that's the same as pwr_src or not, but the OP has a line connected to it.
#1 is by nvvdd_pwm but i don't know if that's MVDDC power or not....


What does your hwinfo "normalized" TDP look like in comparison to the other TDP value?
 

·
Registered
Joined
·
93 Posts
It's either resistor 1 or 2 on the first page of this thread.
I know #2 is linked to 'current sense' so I'm not sure if that's the same as pwr_src or not, but the OP has a line connected to it.
#1 is by nvvdd_pwm but i don't know if that's MVDDC power or not....


What does your hwinfo "normalized" TDP look like in comparison to the other TDP value?
Here is a screenshot of HWinfo for my card. PCIE slot power, PCIE #1 and #2 are corrected with multipliers. Trying to figure out why the card is still power limited. 10mohm shunt on PCIE express slot, 5mohm shunt on the rest.

2466355
 

·
Facepalm
Joined
·
8,623 Posts
Here is a screenshot of HWinfo for my card. PCIE slot power, PCIE #1 and #2 are corrected with multipliers. Trying to figure out why the card is still power limited. 10mohm shunt on PCIE express slot, 5mohm shunt on the rest.

View attachment 2466355
Your core is throttling. I am sure it's the SRC. It's throwing off the other NVVDD readings and causing you to throttle, probably not on SRC itself but on the rails related to GPU Core Power.

Your GPU Core output power (the very last one) is higher than GPU Core (NVVDD) Input power (sum). I believe GPU Core NVVDD Input power (sum) is the same value as "GPU Core Power" in GPU-Z. You of course want this way below board power--for example this usually throttles at 300W in GPU-Z. So if GPU-Z were reporting 350W (out of 400W maximum) for board power and 220W (out of 300W maximum) for GPU Core power, you did good with the core power shunt, but the SRC is messing stuff up.

Someone correct me if I'm wrong!!!!!
But I think ,

You want your GPU Core NVVDD Output value (that last one before Sram output power) to be lower than your GPU Core Power (NVVDD Input Power (sum), after you 'correct' all the wattage values. Let's assume you 5 mOhmd the PCIE slot so all power inputs have 2.0x multiplier. Now you can double all the wattage values. But if you double those two just like you doubled the 8 pin and slot, and thus you double board power draw, you get 300W for Core NVVDD Input power (sum) and 486W for GPU Core NVVDD Output Power, which is not what you want.

GPU Core NVVDD2 Input power (sum) and GPU Core NVVDD Output power (sum) should be relatively close to GPU Power (corrected). But you're getting 192W for the first and 324W for the second. If you close to double them...that's 384W and 648W...

And you GPU Core NVVDD1 input power (sum) is reporting an amazingly low value. So yeah you have work to do :)
 
  • Rep+
Reactions: motivman

·
Registered
Joined
·
93 Posts
Your core is throttling. I am sure it's the SRC. It's throwing off the other NVVDD readings and causing you to throttle, probably not on SRC itself but on the rails related to GPU Core Power.

Your GPU Core output power (the very last one) is higher than GPU Core (NVVDD) Input power (sum). I believe GPU Core NVVDD Input power (sum) is the same value as "GPU Core Power" in GPU-Z. You of course want this way below board power--for example this usually throttles at 300W in GPU-Z. So if GPU-Z were reporting 350W (out of 400W maximum) for board power and 220W (out of 300W maximum) for GPU Core power, you did good with the core power shunt, but the SRC is messing stuff up.

Someone correct me if I'm wrong!!!!!
But I think ,

You want your GPU Core NVVDD Output value (that last one before Sram output power) to be lower than your GPU Core Power (NVVDD Input Power (sum), after you 'correct' all the wattage values. Let's assume you 5 mOhmd the PCIE slot so all power inputs have 2.0x multiplier. Now you can double all the wattage values. But if you double those two just like you doubled the 8 pin and slot, and thus you double board power draw, you get 300W for Core NVVDD Input power (sum) and 486W for GPU Core NVVDD Output Power, which is not what you want.

GPU Core NVVDD2 Input power (sum) and GPU Core NVVDD Output power (sum) should be relatively close to GPU Power (corrected). But you're getting 192W for the first and 324W for the second. If you close to double them...that's 384W and 648W...

And you GPU Core NVVDD1 input power (sum) is reporting an amazingly low value. So yeah you have work to do :)
yeah, this is totally confusing, lol. I am getting 550W currently with this as it is ( PNY bios), close to 600W with the EVGA 500W bios, card is scoring high on the leaderboards, Not looking to tear down my loop honestly, too much work... As long as my card doesn't die, I am good, lol. Maybe I need to remove the 10mhom shunt on PCIE and replace with 5mohm? I dunno.
 

·
Registered
Joined
·
93 Posts
Your core is throttling. I am sure it's the SRC. It's throwing off the other NVVDD readings and causing you to throttle, probably not on SRC itself but on the rails related to GPU Core Power.

Your GPU Core output power (the very last one) is higher than GPU Core (NVVDD) Input power (sum). I believe GPU Core NVVDD Input power (sum) is the same value as "GPU Core Power" in GPU-Z. You of course want this way below board power--for example this usually throttles at 300W in GPU-Z. So if GPU-Z were reporting 350W (out of 400W maximum) for board power and 220W (out of 300W maximum) for GPU Core power, you did good with the core power shunt, but the SRC is messing stuff up.

Someone correct me if I'm wrong!!!!!
But I think ,

You want your GPU Core NVVDD Output value (that last one before Sram output power) to be lower than your GPU Core Power (NVVDD Input Power (sum), after you 'correct' all the wattage values. Let's assume you 5 mOhmd the PCIE slot so all power inputs have 2.0x multiplier. Now you can double all the wattage values. But if you double those two just like you doubled the 8 pin and slot, and thus you double board power draw, you get 300W for Core NVVDD Input power (sum) and 486W for GPU Core NVVDD Output Power, which is not what you want.

GPU Core NVVDD2 Input power (sum) and GPU Core NVVDD Output power (sum) should be relatively close to GPU Power (corrected). But you're getting 192W for the first and 324W for the second. If you close to double them...that's 384W and 648W...

And you GPU Core NVVDD1 input power (sum) is reporting an amazingly low value. So yeah you have work to do :)
here is a screenshot of another members reference card after shunt mod. He used 5mohm on all resistors... compared his values to mine, they look very similar, so its either the bios or something else going on that is limiting power....
 

Attachments

·
Registered
Joined
·
93 Posts
compared the values again with the other member, and noticed he does not have a performance limit for thermal, and I do... his max temp is 51C, mine is 55C ( I am running my fans very low (800rpm) and my ambient temp is 29C) ... from what I know, he is on alphacool block and I am on EK... maybe I need to replace my thermal pads?
 
241 - 260 of 382 Posts
Top