Overclock.net banner
181 - 200 of 329 Posts
So I have been playing Cyberpunk 2077 for awhile which is absolutely brutal on the GPU loading. 90 - 100% GPU loading for the entire game and look no more green! So it worked. I think that it takes a day or two for the silver to fully cure and when it does the resistance drops even further because now I am maxing out at 235W whereas before it was 330W so that tells me the resistance has decreased over the last 24 hours.

now the next task is to figure out how to get more than 1.08v to the core. Would be nice to do at least 1.15v. Anyone know a mod for that one?

View attachment 2474745
Yes it takes 24 hours to cure MG842AR on ambient temps.

And:
You don't want to do that (exceeding max voltage) on a board with fuses. You have power draw limits because of the fuse. And the GPU will go nuclear hot at higher than 1.10v. Another user had a single stage (phase change/peltier?) cooler and setting 1.13v overwhelmed the cooler.

But you would have to buy an evc2 and solder wires to the proper controller points. And I don't know if that board even has marked spots for the pinouts (the Strix does, and I think one other model/brand has headers marked for an evc2 (not just the Kingpin)).
 
Discussion starter · #182 ·
Replace the Vref resistor on the up9511 with a potentiometer,
Increase resistance so the chip sees less voltage returning to it then its expecting. So then it increases the voltage to get that vref back to expected value causing voltage to shift above what the VFR is demanding.
 
Yes it takes 24 hours to cure MG842AR on ambient temps.

And:
You don't want to do that (exceeding max voltage) on a board with fuses. You have power draw limits because of the fuse. And the GPU will go nuclear hot at higher than 1.10v. Another user had a single stage (phase change/peltier?) cooler and setting 1.13v overwhelmed the cooler.

But you would have to buy an evc2 and solder wires to the proper controller points. And I don't know if that board even has marked spots for the pinouts (the Strix does, and I think one other model/brand has headers marked for an evc2 (not just the Kingpin)).
Well I can solder over the fuses if they blow, so it's not the end of the world. As far as cooling goes, I have a 360, 280 and 120mm rad and 6 fans pushing air through them. I feel like there really isint anything that I could power with my 860W PSU that would overwhelm those radiators.

So what is the next step then for more performance? I am at 2070 Mhz locked solid. I can go higher in benchmarks, but 2085 MHz. crashes after awhile in Cyberpunk 2077. I have already maxed out the VRAM. Any higher and again it crashes the game.
 
Well I can solder over the fuses if they blow, so it's not the end of the world. As far as cooling goes, I have a 360, 280 and 120mm rad and 6 fans pushing air through them. I feel like there really isint anything that I could power with my 860W PSU that would overwhelm those radiators.

So what is the next step then for more performance? I am at 2070 Mhz locked solid. I can go higher in benchmarks, but 2085 MHz. crashes after awhile in Cyberpunk 2077. I have already maxed out the VRAM. Any higher and again it crashes the game.
Isn't really much more you can do at this point besides keeping the temps down. The next step is water cooling and possibly a volt mod. But if you really wanted a volt mod, you should have a 3090 card, not a 3080 if you're trying to get the absolute most performance out of the card. Look at the 3090 thread where someone has both a KPE + classified tool+hybrid cooler and a Strix. You can see the mods help his bench scores but games reach a diminishing returns point.

If you're just doing it for a daily, just keep it where it's at now and throw a block on it. A volt mod on a 3080 isn't going to do anything except give you higher benchmark scores. And just increasing core voltage won't do very much except add on heat. You would need access to BOTH MSVDD (Uncore) and MVDDC (memory voltage--default is 1.35v, 1.40v is safe if cooled well).
 
Isn't really much more you can do at this point besides keeping the temps down. The next step is water cooling and possibly a volt mod. But if you really wanted a volt mod, you should have a 3090 card, not a 3080 if you're trying to get the absolute most performance out of the card. Look at the 3090 thread where someone has both a KPE + classified tool+hybrid cooler and a Strix. You can see the mods help his bench scores but games reach a diminishing returns point.

If you're just doing it for a daily, just keep it where it's at now and throw a block on it. A volt mod on a 3080 isn't going to do anything except give you higher benchmark scores. And just increasing core voltage won't do very much except add on heat. You would need access to BOTH MSVDD (Uncore) and MVDDC (memory voltage--default is 1.35v, 1.40v is safe if cooled well).
Lol I have a waterblock. I said that in the last post. There is no way I could do this mod without waterclooling. The stock cooler cant even really handle OEM temps let alone these. I actually modified my EWKB block to allow for cooling of the coils. I added some aluminum bar and TIM to bridge the gap. It was a PITA because I had to be really precise with how thick the bar is so it just lines up perfectly. I dont know why EWKB dident machine the block to allow for coil cooling. Those coils get absurdly hot. Like so hot it burns my hand on the backside of the PCB even with this mod. The coils get hotter than the core even with this mod.

2474754
 
Lol I have a waterblock. I said that in the last post. There is no way I could do this mod without waterclooling. The stock cooler cant even really handle OEM temps let alone these. I actually modified my EWKB block to allow for cooling of the coils. I added some aluminum bar and TIM to bridge the gap. It was a PITA because I had to be really precise with how thick the bar is so it just lines up perfectly. I dont know why EWKB dident machine the block to allow for coil cooling. Those coils get absurdly hot. Like so hot it burns my hand on the backside of the PCB even with this mod. The coils get hotter than the core even with this mod.

View attachment 2474754
Sorry I wasn't paying attention. I'm replying to posts inbetween deaths in Overwatch so yeah.
I would just call it a day then. There's no point volt modding at all unless you're actually going sub-ambient. The chip heats up like no tomorrow if you go past 1.10v.

You get bigger gains by changing MVDDC and MSVDD tweaking, but you don't have access to that on that card.
The EVC2 allows access to those rails if you can actually find the correct solder points.
 
Lol I have a waterblock. I said that in the last post. There is no way I could do this mod without waterclooling. The stock cooler cant even really handle OEM temps let alone these. I actually modified my EWKB block to allow for cooling of the coils. I added some aluminum bar and TIM to bridge the gap. It was a PITA because I had to be really precise with how thick the bar is so it just lines up perfectly. I dont know why EWKB dident machine the block to allow for coil cooling. Those coils get absurdly hot. Like so hot it burns my hand on the backside of the PCB even with this mod. The coils get hotter than the core even with this mod.

View attachment 2474754
so inductors can't be left uncooled this gen? on pascal most do it to reduce coil whine. could this be why my bykski block (up to 57c at 360w) is so much hotter than my 1080ti was on ek (40c at 320w)?
 
so inductors can't be left uncooled this gen? on pascal most do it to reduce coil whine. could this be why my bykski block (up to 57c at 360w) is so much hotter than my 1080ti was on ek (40c at 320w)?
IDK, but the OEM heatsink included cooling for the coils. All I know is they get absurdly hot. Like, it would literately melt your skin if you put your finger on one bare and held it there for a bit. I would bet internal temps on those coils are over 100C. I supposed it depends on how many phases you have. If you have a 20 phase card you're probably okay. I have a 16 phase card so the load on each coil will be higher than if it was a 20.
 
Discussion starter · #190 ·
The inductors are rated for 50C and normal operating temps 80-100C so dont touch them.
vRel = max voltage reliablity. So your card isnt clocking the core any higher because it cant get enough voltage for the next bin up.
vOp = max operating voltage, Cards not allowed to operate at higher voltage triggered when its at the nvidia drivers lock of 1.1V for ampere
 
  • Rep+
Reactions: Falkentyne
3080 xc3 hybrid
5 layers (842ar) on each resistor
This is about 8 hours after reinstalling. Thoughts?
Heaven Benchmark 1080p fullscreen, no overclock but power and voltage sliders maxed

2474970


Outside that, ran Port Royal a couple of times, have set a new PR with a fair bit lower OC... boost staying higher and more stable, although still get the "spike and crash" to 2220 mhz on occasion.

Thoughts? Looking at SPL's posts, may need a fair bit of time to "cure"
 
hmm
On the plus, gained about 200 pts in PR. Still looks off in GPU-Z though.
View attachment 2475123 View attachment 2475124
Anything standing out? (I did also paint the pcie on the back)
Yea, you want to get to the point where all the green in the perfcap is gone. As long as there is green, you're being power throttled. Also looking at the total power draw, it says 330W which means you're throttled. If you get to the point where you're maxing the card out and you're under 300W being reported in GPU-Z, then that's a good sign. Your stuff does look weird. It looks ceramic in color. My stuff was more silver and sparkley. Did you scrape the coating off the solder before you applied it?
 
hmm
On the plus, gained about 200 pts in PR. Still looks off in GPU-Z though.
View attachment 2475123 View attachment 2475124
Anything standing out? (I did also paint the pcie on the back)
What kind of paint did you use? Brand?
And as asked above, did you scrape off the conformal coating off the edges of the original shunts? (not sure if evga has it but Gigabyte and FE do).
 
MG Chemicals 842AR Super Shield (linked earlier in the thread) is what I used, in the little jar. (shaken pre application each time) this exactly

conformal coating scraped off each shunt

*edit
I have shaken and stirred, but I don't see much if any sparkle.

I'm going to go back, remove all and re-scrape the coating, maybe I didn't go deep enough.

Here's a picture of what was underneath. Scrape more?
2475271
 
MG Chemicals 842AR Super Shield (linked earlier in the thread) is what I used, in the little jar. (shaken pre application each time) this exactly

conformal coating scraped off each shunt

*edit
I have shaken and stirred, but I don't see much if any sparkle.

I'm going to go back, remove all and re-scrape the coating, maybe I didn't go deep enough.

Here's a picture of what was underneath. Scrape more?
View attachment 2475271
Damn, you scraped the hell out of those shunts. At this point it's obvious you modified the card so you're better off just soldering shunts overtop or replacing them entirely.
 
Yeah wasn't worried about it looking modded or not.
Not against soldering it, but while I'm not too concerned about it, I'm also not just looking to fry the card.

Scraped, and almost done with new layers of paint. Will see how it goes

*got it reinstalled... scraped all the shunts again, 6 layers on each with ~20' dry time between layers. All pretty even.
Seems like if anything it's a little worse. Maybe the paint I got is bad or old. Has already cured overnight, but I guess I'll let it run a while and see if anything changes. May indeed end up soldering
 
Alright so this is really weird.... I downloaded the beta version of MSI Afterburner so I could change the voltage. I added +100mv, which in reality is actually only going from 1.06 to 1.09, but whatever. Anyway, I went from all vrel to now all PWR limited. However, without changing the clocks at all I am seeing the card is automatically boosting one bin higher with the voltage maxed vs stock (see left vs right side of MSI graph - two separate benchmarks).

What is weird is there is only a 9W difference between 1.06v and 1.09 so I dont know why I am seeing so much green!

2475695


2475696


2475698
 

Attachments

Alright so this is really weird.... I downloaded the beta version of MSI Afterburner so I could change the voltage. I added +100mv, which in reality is actually only going from 1.06 to 1.09, but whatever. Anyway, I went from all vrel to now all PWR limited. However, without changing the clocks at all I am seeing the card is automatically boosting one bin higher with the voltage maxed vs stock (see left vs right side of MSI graph - two separate benchmarks).

What is weird is there is only a 9W difference between 1.06v and 1.09 so I dont know why I am seeing so much green!

View attachment 2475695

View attachment 2475696

View attachment 2475698
What are you running to get that power limit flag?
Sounds like you ran into MSVDD / NVVDD2 limit. Can't bypass this without a bios mod. Shunts don't have access to this.
You can verify this by looking at "requested" clocks vs "effective" clocks. If you are NOT power limited and requested clocks >70 mhz higher than effective clocks, you may be at NVVDD2 limit and its throttling effective clocks slowly first without PWR flag (go any higher it will trigger power and throttle requested clocks also).
 
181 - 200 of 329 Posts