Overclock.net banner

421 - 440 of 637 Posts

·
Registered
Joined
·
861 Posts
this was superposition 4k with extreme settings.

oh damn, so the 8mOhm is not really acting like 8mOhm right, but more like 1 mOhm?

2467900
 

·
Registered
Joined
·
313 Posts
So conclusion time folks...at least for reference PCB, not sure about FE PCB or other Custom PCB's, but just my experience with my PNY reference card....

It does not matter what resistors you use to shunt... this PCB will still draw about 515W-535W max on the stuck bios. History of my shunting shenanigans...

1. First shunt - used conductive paint on all six resistors --- max power draw (verified with GPU-Z and kill-a-watt meter - 520w-535w (depending on benchmark)
2. Second shunt - used 5mohms on all resistors except on PCIE resistor (10 mohms) --- Max power draw - 520W-535W (depending on benchmark)
3. Third shunt attempt - Used 5mohms on all resistors including PCIE ---- max power draw 520W-535W (depending on benchmark)

Now the interesting part lol....

1. first shunt with conductive paint on PCIE slot - 85W max power draw on PCIE SLOT ( max power draw on GPU-Z on average 64W)
2. second shunt with 10mohms on PCIE slot - 85W max power draw PCIE SLOT ( max power draw on GPU-Z on average 55W)
3. Third shunt with 5mohm on PCIE slot - 85W max power draw from PCIE slot... * ( max power draw on GPU-Z on average 42.5W)

You see the pattern here folks???? You cannot outsmart NVIDIA engineers, LOL. There are other power limits on these cards that shunting cannot defeat. Only a proper bios can disable these hidden limits.

These stats above are all results using the stock PNY reference bios. THE ONLY WAY I can get this thing to draw more power (600W) is to flash either the EVGA 500W bios or the kingpin 520W bios.

So conclusion --- IT DOES NOT MATTER what resistor you use to shunt the PCIE Slot or even the entire card for that matter, your card will draw a certain amount of power, and changing shunts will just result in the card adjusting values based on your shunt resistor values. The only way to get more power is to flash a bios that allows more power draw.

have a good weekend folks!!!!!

LOL.
 

·
Registered
Joined
·
2,783 Posts
So conclusion time folks...at least for reference PCB, not sure about FE PCB or other Custom PCB's, but just my experience with my PNY reference card....

It does not matter what resistors you use to shunt... this PCB will still draw about 515W-535W max on the stuck bios. History of my shunting shenanigans...

1. First shunt - used conductive paint on all six resistors --- max power draw (verified with GPU-Z and kill-a-watt meter - 520w-535w (depending on benchmark)
2. Second shunt - used 5mohms on all resistors except on PCIE resistor (10 mohms) --- Max power draw - 520W-535W (depending on benchmark)
3. Third shunt attempt - Used 5mohms on all resistors including PCIE ---- max power draw 520W-535W (depending on benchmark)

Now the interesting part lol....

1. first shunt with conductive paint on PCIE slot - 85W max power draw on PCIE SLOT ( max power draw on GPU-Z on average 64W)
2. second shunt with 10mohms on PCIE slot - 85W max power draw PCIE SLOT ( max power draw on GPU-Z on average 55W)
3. Third shunt with 5mohm on PCIE slot - 85W max power draw from PCIE slot... * ( max power draw on GPU-Z on average 42.5W)

You see the pattern here folks???? You cannot outsmart NVIDIA engineers, LOL. There are other power limits on these cards that shunting cannot defeat. Only a proper bios can disable these hidden limits.

These stats above are all results using the stock PNY reference bios. THE ONLY WAY I can get this thing to draw more power (600W) is to flash either the EVGA 500W bios or the kingpin 520W bios.

So conclusion --- IT DOES NOT MATTER what resistor you use to shunt the PCIE Slot or even the entire card for that matter, your card will draw a certain amount of power, and changing shunts will just result in the card adjusting values based on your shunt resistor values. The only way to get more power is to flash a bios that allows more power draw.

have a good weekend folks!!!!!

LOL.
I mean, thats proven not to be universally true for all card. Point being the OP and his 600+ watts on the fe.

Edit: I did not read you first paragraph correctly. Ignore the above.

Are you actually measuring power from the wall or just trying to use hwinfo and modifying multiplier values?

It is very possible on a reference, or even your specific card, there are other shunts you need add aside from the obvious 6 already in play.
 

·
Registered
Joined
·
2,783 Posts
this was superposition 4k with extreme settings.

oh damn, so the 8mOhm is not really acting like 8mOhm right, but more like 1 mOhm?

View attachment 2467900
No, it probably isn't, nor is the 5 ohm you are shunting it with. Current will take the path of least resistance and it will for sure be that solder bridge lol. (To be clear, yes, there is current still going through the 5 and 8mohm resistors, but it will pale in comparison to the solder bridge.)
 

·
Facepalm
Joined
·
9,518 Posts
So conclusion time folks...at least for reference PCB, not sure about FE PCB...

It does not matter what resistors you use to shunt... this PCB will still draw about 515W-535W max on the stuck bios. History of my shunting shenanigans...

1. First shunt - used conductive paint on all six resistors --- max power draw (verified with GPU-Z and kill-a-watt meter - 520w-535w (depending on benchmark)
2. Second shunt - used 5mohms on all resistors except on PCIE resistor (10 mohms) --- Max power draw - 520W-535W (depending on benchmark)
3. Third shunt attempt - Used 5mohms on all resistors including PCIE ---- max power draw 520W-535W (depending on benchmark)

Now the interesting part lol....

1. first shunt with conductive paint on PCIE slot - 85W max power draw on PCIE SLOT ( max power draw on GPU-Z on average 64W)
2. second shunt with 10mohms on PCIE slot - 85W max power draw PCIE SLOT ( max power draw on GPU-Z on average 55W)
3. Third shunt with 5mohm on PCIE slot - 85W max power draw from PCIE slot... * ( max power draw on GPU-Z on average 42.5W)

You see the pattern here folks???? You cannot outsmart NVIDIA engineers, LOL. There are other power limits on these cards that shunting cannot defeat. Only a proper bios can disable these hidden limits.

These stats above are all results using the stock PNY reference bios. THE ONLY WAY I can get this thing to draw more power (600W) is to flash either the EVGA 500W bios or the kingpin 520W bios.

So conclusion --- IT DOES NOT MATTER what resistor you use to shunt the PCIE Slot or even the entire card for that matter, you card will draw a certain amount of power, and changing shunts will just result in the card adjusting values based on your shunt. The only way to get more power is to flash a bios that allows more power draw.

have a good weekend folks!!!!!

LOL.
Very interesting find.

Yeah I noticed something strange going on. Notice Time Spy Test #2? Even though no rail hit the power limit, there was still a power flag being triggered from some strange place that happened twice--both times only around the beginning section of the test! But the end of the test--not the beginning--drew more power!

My Kill-a-watt actually registered 815W from the wall (!!) close to the end of Timespy graphics test #2, when I already know that 590-600W on the meter is 400W.
So that means the video card was pulling 600W. You can see what I'm talking about here on this run I did this morning. The "High load" section you see on the middle graphs is test #2. The idle part on the very left edge of the graph is the loading screen to test #2.

You can see the power draw increase slightly by the end of the test, where the KAW was picking up 800W-815W from the wall, which means the card was reaching 600W. Also notice my normalized is still lower than 114%, so none of the main rails were at their limit. So something else triggered the PWR alert (this only caused a small drop in voltage and -15 mhz clocks) at the beginning. The "idle" part in the middle of the test was the voltage not using the highest tier voltage V/F tier anymore (the highest tier used = vOP+vREL simultaneously).

2467909


Was just messing around in Overwatch at 4k (1080p + 200% render scale) and when my GPU hit 73C, the GPU was pulling about 575W. A power limit "warning" was triggered around 550W (power limit bar in GPU-Z only being a small part of the way up from the bottom of the graph area), with no throttling. Normalized TDP was around 106%. I can probably hit 600W before it actually starts tryharding the throttling.

The overwatch test is even more interesting.
Some throttling happens right after joining the server and the watts hits 540W. Then as the load increases, it actually tapers off and turns into a "warning" (no throttling), which you can tell just by looking at the slope of the graphs and seeing the watts slowly rise. Normalized 110.7% TDP, main TDP 105.2%, so still no main rail is at its power limit yet. I ended the test because the temps are getting unmanageable. Stock cooler can't handle 580W, even with good thermal pads and Kryonaut Extreme....

The fact that both the Timespy graphic test #2 early throttling and the Overwatch initial throttling happens early on (while the temps are still pretty low) then tapers off (in OW, it seems to pick up again once I get to 590W) is interesting. Makes you wonder if it's actually a voltage rail that's throttling ...but I had VREL+VOP through the entire Overwatch test along with the "throttle warning" with the small throttle mini bar...

2467910


I mean, thats proven not to be universally true for all card. Point being the OP and his 600+ watts on the fe.

Edit: I did not read you first paragraph correctly. Ignore the above.

Are you actually measuring power from the wall or just trying to use hwinfo and modifying multiplier values?

It is very possible on a reference, or even your specific card, there are other shunts you need add aside from the obvious 6 already in play.
This "issue" definitely happens on the FE also, see the timespy test #2 "early" power limit throttle, which suddenly vanishes as the power draw KEEPS increasing (Got to 600W going by my kill-a-watt calibration by the end!). So yes you can reach or even exceed 600W, but something still causes a power limit event under certain load conditions.

Edit take a look at this post.


There are three "mini" 005 shunt resistors on the FE backplate side. One is connected to the LED thingy (Don't shunt that one), and the other two seem to be connected to "Chip Power" and "PCIE Slot". @olrdtg tried shunting those two but he saw no difference in power draw/TDP change. However it's very possible that shunting those two might change the "strange" throttling you get before hitting the actual power limit or rail limits. If it's the GPU Chip signalling a power event or PCIE slot something, who knows.....I'm not going to test that out though (it would be absolute hell to try to remove paint from those, and olrdtg already broke one trying to mod it with solder or a wire, I think, and had to create a rework to fix it...(??)

GPU Chip Power shunt/mini shunt is the one on the bottom by the lower part of the GPU X-bracket on the backplate side (the one closer to the V is the LED cable one).
The PCIE Slot mini one is by the I/O bracket area by the PCIE shunt one.

If you guys want to test that to see if it addresses the "hidden" throttling--I'm not responsible if something goes wrong.
 

·
Registered
Joined
·
861 Posts
fixed

2467912



I should have put my KAW when I had the pcie slot bridged with solder to see the difference in wattage usage....because I run a timespay extreme stress test (20 runs) and nothing happened, nothing fried or melted.
 

·
Facepalm
Joined
·
9,518 Posts
fixed

View attachment 2467912


I should have put my KAW when I had the pcie slot bridged with solder to see the difference in wattage usage....because I run a timespay extreme stress test (20 runs) and nothing happened, nothing fried or melted.
So you have that strange throttling during Timespy Graphics Test #2 just like everyone else, huh?

Here's mine. The only difference between my picture and yours is my graph is shorter, so test #1's VREL isn't on the screen, and yours went into VREL+VOP after the throttling, while mine just went to "Utilization" (Grey).

2467913


BTW do you want to live dangerously?


Check those two shunts--the MINI ONES. The tiny 005 ones (NOT the one going to the LED cable).

Take a #2 or #4 HB pencil (#4 is probably the best) and try penciling them down and bridging them (do NOT apply pressure, do it lightly). That is easily reversible by just erasing them LIGHTLY.
Olrdtg already shunted those but saw no difference in power draw/TDP, however he did not test for the "micro throttling" issue on Timespy Test #2 (i've seen at least two other people with mods have that same pattern happen on TS#2).

See if that helps remove that micro throttle.
 

·
Registered
Joined
·
313 Posts
I think I am done experimenting with my card at this time. 500W is more than enough to enjoy gaming TBH, definitely better than the 366W limit at stock. I just couldn't live with that limit, after spending over $1800 on this card. Until someone figures out how to defeat the power limit completely, or one of the XOC guys leaks the 1000W bios (heard all the stealth power limiters are disabled on this bios), I will just chill and enjoy this card for now. December 10th cannot come soon enough, lol.
 

·
Registered
Joined
·
861 Posts
So you have that strange throttling during Timespy Graphics Test #2 just like everyone else, huh?

Here's mine. The only difference between my picture and yours is my graph is shorter, so test #1's VREL isn't on the screen, and yours went into VREL+VOP after the throttling, while mine just went to "Utilization" (Grey).

View attachment 2467913

BTW do you want to live dangerously?


Check those two shunts--the MINI ONES. The tiny 005 ones (NOT the one going to the LED cable).

Take a #2 or #4 HB pencil (#4 is probably the best) and try penciling them down and bridging them (do NOT apply pressure, do it lightly). That is easily reversible by just erasing them LIGHTLY.
Olrdtg already shunted those but saw no difference in power draw/TDP, however he did not test for the "micro throttling" issue on Timespy Test #2 (i've seen at least two other people with mods have that same pattern happen on TS#2).

See if that helps remove that micro throttle.
I have this here, would that work?

2467916


can you test superposition with these settings here and see if your normalized tdp hits also the limit? I think I might go back to 5mOhm instead of 8mOhm

2467915
 

·
Facepalm
Joined
·
9,518 Posts
Don't use any silver or pen to shunt the tiny two resistors. You want it to be easily removeable. That is not going to be a permanent mod. A #4 pencil would be best, since you can remove it easily if you don't like the results. Also the circuitwriter pen didn't work at all on a FE board for another user for the big shunts (only the MG Silver paint pen worked).

I can do superposition after I finish my ranked game.

Ok here you go.

2467921
 

·
Registered
Joined
·
2,783 Posts
So you have that strange throttling during Timespy Graphics Test #2 just like everyone else, huh?

Here's mine. The only difference between my picture and yours is my graph is shorter, so test #1's VREL isn't on the screen, and yours went into VREL+VOP after the throttling, while mine just went to "Utilization" (Grey).

View attachment 2467913

BTW do you want to live dangerously?


Check those two shunts--the MINI ONES. The tiny 005 ones (NOT the one going to the LED cable).

Take a #2 or #4 HB pencil (#4 is probably the best) and try penciling them down and bridging them (do NOT apply pressure, do it lightly). That is easily reversible by just erasing them LIGHTLY.
Olrdtg already shunted those but saw no difference in power draw/TDP, however he did not test for the "micro throttling" issue on Timespy Test #2 (i've seen at least two other people with mods have that same pattern happen on TS#2).

See if that helps remove that micro throttle.
Be carful with that kind of analysis. The beginning of time spy #2 is not very demanding on the gpu. Just because the power is there to use does not mean it will use it. You should in no way expect to be pegged at the power limit 24/7 under every load. That is not how load balancing or power usage works for anything. It's the same reason you can peg you cpu at 100% in r20 but you aren't going to pull the same power as p95.

Power usage will vary depending on the load given, even if you are at max voltage/clocks. Same reason you can sit at desktop at max voltage/clocks and not draw the same power. It's the same reason it's #1 and ts#2 draw different power amounts.
 

·
Facepalm
Joined
·
9,518 Posts
can you test superposition with these settings here and see if your normalized tdp hits also the limit? I think I might go back to 5mOhm instead of 8mOhm
Only Superposition 4k shows that kind of Throttling.
Timespy Test #2 only has those two bumps and I'm well below the 114% normalized point

2467923



Port Royal: Zero throttling anywhere.

2467924



So Superposition is hitting something hard.

Do you have a #2 or #4 pencil? (DO NOT USE THE CIRCUITWRITER PEN).
Can you try drawing over the two 005 resistors linked to GPU Chip power and PCIE Slot and try Superposition again?
(No I don't have a pencil. I'd have to order one on Amazon).

Just managing to lower the resistance a tiny bit and see if it influences superposition is better than nothing.
Unfortunately, those baby shunts looked to be "stepped" just like the big shunts :( Going to be hard to do those...that pencil is going to be doing some work for sure...

As always, blow/clean the area around it when you're done.

2467925


2467926
 

·
Facepalm
Joined
·
9,518 Posts
@dante`afk
Take a look at this.

Superposition 1080p extreme.
No throttling.

2467927


Normalized below 114% cleanly.
What the hell is happening at 4k? Is it hitting a memory bandwidth related power rail limit somewhere?
Besides the SRC being higher (90W) at 4k on GPU-Z, the other rails don't seem to be that much different.
 

·
Facepalm
Joined
·
9,518 Posts
i have pencils, but don't know which # they are.
If you want to go through with it, try penciling the two small resistors very carefully. They are smaller thus the connection is more fragile than the large ones. Take your time and be careful. Bridge the pencil through the top of the resistor to connect the lower silver part. It's going to be hard because of that dip since the 'vertical' part of the dip going to the middle also has to be penciled.

To reverse, just use an eraser very lightly on the top of the resistor so its no longer bridged with the lower silver sides. Do not touch the PCB of course.
BTW did you notice my 1080p extreme preset result? No throttling. What do you get on 1080p extreme?
 

·
Registered
Joined
·
861 Posts
I'll try that later.

what I just noticed while benchmarking, with bridged pcie slot shunt just with solder, I was able to go 265mhz on gpu, now with 8mOhm only 235mhz.
 

·
Registered
Joined
·
861 Posts
sooo, I just wanted to switch the PSU to see the results with a KAW and I see this; it melted one of the pins from the psu cable and the 12 plug nvidia cable, not sure which caused which.

2467956


2467958


too much draw from pcie plugs or from pcie slot?

I didn't notice anything while it ran melted. could also be just a cable issue as I'm looking currently for custom cables for replacement, and people experience the same thing with new cables. I'm using cablemods cables.
 

·
High Clocker
Joined
·
3,662 Posts
Pin must of been loose inside the metal contact causing more resistance. Resistance causes heat.
If your PSU is modular id just cut off those adapter bits and soldering something up directly from 12pin to 2X 8pins.
 
  • Rep+
Reactions: dante`afk

·
Registered
Joined
·
313 Posts
sooo, I just wanted to switch the PSU to see the results with a KAW and I see this; it melted one of the pins from the psu cable and the 12 plug nvidia cable, not sure which caused which.

View attachment 2467956

View attachment 2467958

too much draw from pcie plugs or from pcie slot?

I didn't notice anything while it ran melted. could also be just a cable issue as I'm looking currently for custom cables for replacement, and people experience the same thing with new cables. I'm using cablemods cables.

H*ly ****!
 

·
Registered
Joined
·
861 Posts
Pin must of been loose inside the metal contact causing more resistance. Resistance causes heat.
If your PSU is modular id just cut off those adapter bits and soldering something up directly from 12pin to 2X 8pins.
Yea, I'll throw those cablemods cables to the trash and use just the original corsair ones.
 
421 - 440 of 637 Posts
Top