Overclock.net banner

181 - 200 of 592 Posts

·
Registered
Joined
·
195 Posts
Please post a screenshot of your wattage rails in HWinfo64, doing a full load test when you are hitting the power limit hard.
Then we will know what shunts you should use.
It depends on the ratio of slot power to single 8 pin power.

It seems like Asus and Founder's Edition cards use a smaller ratio. FE seems to be around Slot=40% of 8 pin, so 166 watts on 8 pin #1 = 66 watts on PCIE. (e.g. 166 + 166 + 66 = 400 watts).

So if 40% is your ratio, then 15 mOhm shunts will keep you at 87.78 mhz PCIE slot with a 1.33x multiplier for 15 mOhms
If we assume your max power limit is 400 watts:

(66 * 1.33 + 166 * 1.33 + 166 * 1.33) = 529 watts instead of 400 watts.
With 87 watts for your PCIE slot.

@bmgjet
Can you kindly update your shunt calculator to allow people to enter a multiplier for the PCI express slot?
It seems different boards use a different multiplier, which throws off your calculator.
The calculator currently assumes a 50% multiplier for PCIE -->8 pin#1.

The FE cards and I think Asus cards use a lower multiplier if I'm not mistaken, from looking at @olrdtg 's posts of his PCIE to slot ratio earlier in this thread.
(this can be determined by anyone by dividing the PCIE watt usage into the 8 pin #1 watt usage when the card is at max power limit).

I think someone already determined eVGA cards use a higher ratio==the 50% ratio (which is why the 500W Bios caps at 450W on their card but caps at 500W on an Asus or MSI three-pin board).
2465242


Here you go... so looks like my ratio is 20% for PCIE slot and 40% for each of the 8 pin pcie power connectors. So with 8ohm shunt on PCIE, I would be pulling max around 120W from the PCIE slot. I am leaning towards going this route. Do you guys think a board like the Z490 Unify will be OK with pulling 120W on the pcie slot? I would rather have all 8mohm shunts, than have 5 5mohms and 1 15mohm on the PCIE shunt. The more power I can get "safely" the better... LOL. The above screenshot is a PNY reference board on gigabyte 390W OC bios, voltage and power limit maxed out in afterburner. But after Shunt, I plan to use the 480W strix bios, Thanh Nguyen has the exact same card, with 5mohm shunts on all the resistors, and he is using the 3X8pin 480W strix bios with no issues at all. With the Gigabyte bios, he was still hitting power limit for some reason, and his card couldn't hold high clocks...
 

·
Facepalm
Joined
·
9,199 Posts
View attachment 2465242

Here you go... so looks like my ratio is 20% for PCIE slot and 40% for each of the 8 pin pcie power connectors. So with 8ohm shunt on PCIE, I would be pulling max around 120W from the PCIE slot. I am leaning towards going this route. Do you guys think a board like the Z490 Unify will be OK with pulling 120W on the pcie slot? I would rather have all 8mohm shunts, than have 5 8mohms and 1 15mohm on the PCIE shunt. The more power I can get "safely" the better... LOL. The above screenshot is a PNY reference board on gigabyte 390W OC bios, voltage and power limit maxed out in afterburner. But after Shunt, I plan to use the 480W strix bios, Thanh Nguyen has the exact same card, with 5mohm shunts on all the resistors, and he is using the 3X8pin 480W strix bios with no issues at all. With the Gigabyte bios, he was still hitting power limit for some reason, and his card couldn't hold high clocks...
Why not just 10 mOhm? 70 * 1.5 = 105W, that keeps you nicely under 120W and still allows close to 600W of power draw (depending on how much comes from the two plugs)
Are you trying to exceed 600W? I don't think that's safe on a 2 power pin board. The 12 pin FE plug (and the 12 pin Seasonic cable) is rated up to 600W and it's still going to get mighty warm there.
 

·
Registered
Joined
·
195 Posts
Why not just 10 mOhm? 70 * 1.5 = 105W, that keeps you nicely under 120W and still allows close to 600W of power draw (depending on how much comes from the two plugs)
Are you trying to exceed 600W? I don't think that's safe on a 2 power pin board. The 12 pin FE plug (and the 12 pin Seasonic cable) is rated up to 600W and it's still going to get mighty warm there.
Haha, Great! I did order 10mohm resistors from digikey last night... so the plan is to shunt(stack) the 5 resistors with 8mohm and the pcie resistor with 10mohm, and I should be golden? Can't wait for my order to arrive, want to really see what my card can do with the power limit unlocked!

2465245
 

·
Facepalm
Joined
·
9,199 Posts
Haha, Great! I did order 10mohm resistors from digikey last night... so the plan is to shunt(stack) the 5 resistors with 8mohm and the pcie resistor with 10mohm, and I should be golden? Can't wait for my order to arrive, want to really see what my card can do with the power limit unlocked!

View attachment 2465245
That should work.
 

·
Registered
Joined
·
846 Posts
what do you think of of the hot glue method framechasers is using? He's using hot glue gun to apply the shunts.

does that even give proper conductivity?
 

·
Facepalm
Joined
·
9,199 Posts
what do you think of of the hot glue method framechasers is using? He's using hot glue gun to apply the shunts.

does that even give proper conductivity?
MG Chemicals Conductive Paint 842AR just works. No need to stack shunts when the paint acts like a 15 mOhm shunt, and 15 mOhm is a very conservative mod for a daily (just remember to scrape the edges of the original shunts to remove the conformal coating first).
 

·
Registered
Joined
·
103 Posts
Discussion Starter #187
what do you think of of the hot glue method framechasers is using? He's using hot glue gun to apply the shunts.

does that even give proper conductivity?
Yeah I would definitely not recommend using hot glue. Seems like it'd be very hit or miss on contact. It's important the new resistors make contact to the originals when stacking, so like Falkentyne said above, definitely go with some conductive paint + conformal coating if you want to stack without soldering. Also, these cards get VERY hot, so depending on the glue used, low temp for example might melt under load. Hot glue apparently melts anywhere between 80 C to 200 C. It's unlikely the glue would melt from the heat of the PCB, but if it happens to be brushing up against the block or something it might. Just go with conductive paint + conformal coating or soldering. Stack soldering is not as hard as it looks as long as you take it slow. Just apply some no-clean flux to each of the resistors on the board, dab some solder on each end of every resistor, then use tweezers to gently hold a resistor ontop of the old and melt the solder on both ends until its nice and stuck. :)

edit: framechasers also mentioned shunt modding the small shunt resistors -- this is absolutely not necessary. I tested this myself the other day. Though two of the tiny shunts on the FE board are directly connected to the larger shunt resistors, modding them makes zero difference. If you go through my post history you can see my test results. Removing the tiny resistors was a ***** and a half too, I actually broke one of them, so now im stuck with one permanently modded tiny resistor and a blue-tac + gorilla taped full sized resistor next to it with some ugly wires running to where the tiny one used to be (at least it works lol), at least until I can source a replacement 3mOhm or 5mOhm tiny shunt resistor.
 

·
Registered
Joined
·
846 Posts
thanks

is isopropyl enough to clean up the conformal coating from the shunt contacts or do i need to scrap them off?
Holding the soldering iron into it and scraping through it?
 

·
Registered
Joined
·
103 Posts
Discussion Starter #189

·
Registered
Joined
·
846 Posts

·
High Clocker
Joined
·
3,631 Posts
Really all you need to do is, Use some isopropyl alcohol which you should already have for cleaning the die off.
That softens it enough so it basically just rubs off.
 
  • Rep+
Reactions: dante`afk

·
Registered
Joined
·
103 Posts
Discussion Starter #193
thanks, perfect.


edit: btw, am I risking melting anything off if I stack 5ohms on all 6 shunts?
Well, if you stack 5 ohms on any of them I think you are going to have a bad time :p What you want is to stack 5 mOhms (milli-ohms)

It really depends on the quality of your motherboard with it's PCIe slots and how much power you are already drawing from the 24-pin. So if you have any other PCIe cards or m.2 drives and a butt load of LEDs, I would probably use an 8 or 10 mOhm on the PCIe slot shunt resistor. to try and keep it under 100W on that.
 
  • Rep+
Reactions: dante`afk

·
Registered
Joined
·
846 Posts
thanks. This I can do :)

2x m2 cards, 4 pumps, 4 fans. no other pcie cards
motherboard is a x570 asus CH hero VIII

correct mOhms :D
 

·
Facepalm
Joined
·
9,199 Posts
You don't need any of that.
Just small flat blade screwdriver and alcohol.
You're overthinking it :)
The conformal coating isn't a resin covering the entire shunt.
No need to spend more money than you need.
 
  • Rep+
Reactions: dante`afk

·
Facepalm
Joined
·
9,199 Posts
thanks. This I can do :)

2x m2 cards, 4 pumps, 4 fans. no other pcie cards
motherboard is a x570 asus CH hero VIII

correct mOhms :D
You're using water cooling?
5 mOhms stacked on 5 mOhms will double the TDP, e.g. 400W-->800W, or 350W-->700W. Keep that in mind.
Of course at 1080p, the card will probably not draw more than 500W
But at 4k, with full RTX and geneworks (??), it probably would try to. Watch the cooling if you go higher than 600W !!!!

You can use the MSI Afterburner slider to lower the power limit.
 
  • Rep+
Reactions: dante`afk

·
Registered
Joined
·
12 Posts
Hi Colleauges,

Owner of 3080 Ventus there...

I posted my problem in that thread:

Unfortunately, still didnt find any solution. I tried almost everything:
1. shunted 5 resistors on the top + PCIe, 10 mOhm all of them: result, my card throttles at ca. 80% of TDP, 260...270W readings in GPU-Z, it means arround 380...400W in real, confirmed by energy meter out of the wall

2. shunted 5 resistors on the top + PCIe + 3 tiny resistors on the back, 10 mOhms all of them: result as above

3. point no 1, but 5 mOhm all of resistors: result, card throttles at 60% of TDP, 190...200W readings in GPU-Z, it means arround 380...400 in real (again), confirmed by energy meter out of the wall

4. point no 2, but 5 mOhm all of resistors including tiny ones: result, card throttles at 60% of TDP, 190...200W readings in GPU-Z, it means arround 380...400 in real, confirmed by energy meter out of the wall

5. crossflash of various bioses - nothing happens, still 190...200W reading in GPU-Z.

6. desoldering all: 300...320W reading in GPU-Z, so inline with specs of 3080 Ventus, stock value

7. crossflash of various bioses without shunt mod - doesnt help, max board power draw reading in GPU-Z is 300...320W. I.e. I flashed TUF BIOS, 370W. And the new power limit emerges, 86% perf.cap wall. 86% x 370 = 320W. Evidently something outside the BIOS controll the power draw.

Some other findings:
  • the power draw for stock card is like that: 290W from 2x8 pins, 30...35W from PCI-e slot
  • even with shunt I cant go higher from PCI-e slot then 40..45W
  • crossflasing doesnt help. The new limit is higher, but perf.cap kicks earlier, below 100%

  • that is quite interesting: the power limit which kicks in point 1,2,3 and 4 is concreted on the same absolute value of the real power draw, look on that:

case A: 10 mOhms resistors
TDP 320W stock, , theoretical new TDP is 1.50 x 320 = 480W, but new power limit kicks at 80%. So 480W x 0,8 = 384W of the real power draw

case B: 5 mOhms resistors
TDP 320W stock, , theoretical new TDP is 2.00 x 320 = 640W, but new power limit kicks at 60%. So 640W x 0,6 = 384W of the real power draw... again...

it couldnt be a coincidence... No matter which resistance of shunts you choose, the absolute value of the real power draw is constant. 384W rock-stable, sure it's 64W more then on stock, but... my idea was to go higher, 450...480W.

hypothetical case C: 3 mOhms resistors (not tested yet...)
TDP 320W stock, , theoretical new TDP is 3.00 x 320 = 960W, but I can bet - the new power limit will kick at 384/960 so 40%. So 960W x 0,4 = 384W... it could be tested, but I have no 3 mOhms resistors at home and they are hardly available.

Literally speaking - I have no f....ing idea, what could be done more. It seems, something is throttling that GPU. The card is watercooled, temps below 50 centigrades Celsius, I use it for VR and the higher power limit is pretty welcome by me. Will appreciate any good idea from your side. Thanks in advance.
 

·
Registered
Joined
·
195 Posts
Did the shunt mod with the silver paint. Did not work at first, but I forgot to shake and mix the paint. After reapplying, my 3090 reference pcb seems to be pulling 530-550w approx from the wall based on readings from my kilowatt meter. So very important step: DO NOT FORGET TO SHAKE AND MIX THE PAINT BEFORE USE.
 

·
Registered
Joined
·
128 Posts
Do you need to remove the conformal coating when soldering? And also doesn't the solder material affect the resistance? And why is it safe to stack a 5mohm, wouldnt the 2x power draw blow the 20amp fuses?
 

·
Facepalm
Joined
·
9,199 Posts
FE does not have fuses.
 
  • Rep+
Reactions: olrdtg
181 - 200 of 592 Posts
Top