Overclock.net banner
1 - 11 of 11 Posts

·
Registered
Joined
·
389 Posts
Discussion Starter · #1 · (Edited)
I picked up this Dell 1080ti for $400 right before the 3000 series release. In response to my 1070ti just not cutting it. Honestly, I had no idea about any of the differences between any of the 1080ti's. But I did know a blower cooler (barring water) would be best for my severely inadequate case. Well, I've learned a lot. And honestly that was kind of the whole idea.

It was thrown into a Dell T3610 workstation with a 1680v2, and quad channel rdimm cl11 1800 samsung server memory (no idea of die). This worked great, but I think the stock cooler held me back considerably. And eventually the cpu power circuit on the motherboard ate it. I decided to salvage, and build an AM4 rig.

Computer Personal computer Circuit component Computer cooling Passive circuit component

Fast forward to today. I've fine tuned my cpu, and ram. Under-volting, and dialing in latency. I've spent a ton of time adjusting and testing. But I never really gave my GPU the same treatment. As I had it set up, it was choking down around 300 watts, and I could overclock the core 150mhz, and ram 250.

Today, I spent time adjuting the voltage curve editor, and adjusting power limits to see where it's efficiency range was. I ended up using factory (240 watt) power limit, and +115mhz core (2015mhz), and +237mhz on the ram (5742mhz). I also picked up 9 FPS on Ghost Recon Breakpoint running 1440P ultra, max settings but with blur and bloom disabled. 111fps average. I use this game alot because of any games I run, this one will crash when no other one will. Old x79 rig would run average of ~75fps

60 watts less, and 9fps, I'll take it. Also pretty pleased with overall performance, and boost considering it's a Dell card.

Ultimately, I was able to run 2050mhz on the core, but she was using more juice, and performance was actually dropping off below how it is configured now. With ambient temp of 83F, it's settling in just below 50C with 2 120mm rads.

Sky Water Rectangle Plant Personal computer
Purple Entertainment Audio equipment Electronic instrument Music venue
 

·
Registered
Joined
·
468 Posts
your curve is set up for like air cooled card. Mine does 2000mhz stable at 0,975mV. Maybe do OC scanner for it to fine tune ur GPU curve and not overvolt it so much. Also drop temp limit to 65c, in case ur pump dies or something, so that your card start to throttle as fast as possible.
 

·
Registered
Joined
·
389 Posts
Discussion Starter · #3 ·
Thanks for the tips, the curve is actually set to where the core is stable. I tried less voltage, and it just wouldn't do it. I slowly adjusted across until it was happy. I initially did start at .980mv. As that seems to be the norm with a lot of AIBs. This is as low as I could get it, with a tiny bit of wiggle room for margin of error.

OC scanner refuses to complete as well. At any rate I spent a lot of time testing, and feeling out where the gpu was happy at what voltage, speed, and current. So, basically what the scanner does, but manually.

I've tried flashing other bios as well, and it was not happy with anything I did. Maybe a bios edit would help, but I havn't gone down that rabbit hole.
 

·
Registered
Joined
·
773 Posts
I don't know if it was due to CPU limitations (8700k @5.0) or its an architecture think but my old 1080 TI (gigabyte Aorus Extreme Waterblock) didn't see much of a difference overclocked. On the stock bios it clocked to 2106 @ 1.03v and using the XOC bios and 1.2V core, it clocked up to 2226. Memory ran about +550 but I usually ran it +500.

HOWEVER,

Performance wasn't much more than a handful of FPS at 1440p but it dumped A LOT of heat. For games/settings that didn't run smoothly on the 1080ti, the overclock wasn't a game changer. It did help with overall smoothness but I found lowing render details was still needed so I opted to do about what you are doing; running undervolted as it kept my room from heating up and still performed.

So good job and don't sweat anymore performance as, in my experience, there wasn't much more to be gained (y)
 
  • Rep+
Reactions: Insidious Supra

·
Registered
Joined
·
389 Posts
Discussion Starter · #5 ·
Your factory settings sound about identical to what I wound up at. The card seems very happy there, so I am too. 60 watts is a pretty good bit to shave, and mine did put off a ton of heat. I'm sure it still will, but like you observed, the gains from pushing harder are little if any. In exchange for a lot more power draw and heat output. I did pick up 1fps at 2025mhz, but watts increased by 15 or so and mv by .05

But there is a large difference versus the 1900mhz original clock. And apparently the card was essentially choking prior on how much power I was feeding it to be stable at 2050.... Maybe a larger cooling setup would change that, but I can't really see the point. Once you start dumping power and heat for a fraction of a percent of performance gain.... It's time to take a step back, lol.
 

·
Registered
Joined
·
788 Posts
Have you tried leaving power limit at max, then using just the V/F curve to limit the power by setting the max voltage?

That might give you a bit more consistent frametimes, since you can avoid hitting PL which would cause the card to downclock / decrease voltage.

GPU-Z sensors tab is really handy for tuning, since you can watch the "Perfcap" flags and all the sensors.
 

·
Registered
Joined
·
389 Posts
Discussion Starter · #7 · (Edited)
I don't think I tried exactly that, but adding more headroom for power draw basically meant that the card was going to use it.... And from what I gathered, it seemed like a small amount was needed for 2025mhz and I needed %120 (300 watts) for 2050mhz. But ran even better at 2015mhz @ factory 240 watts... I'll give it a look over again in case I missed something 👍

Voltage settings seemed to have no affect on wattage
 
  • Rep+
Reactions: T.Sharp

·
Registered
Joined
·
788 Posts
I don't think I tried exactly that, but adding more headroom for power draw basically meant that the card was going to use it.... And from what I gathered, it seemed like a small amount was needed for 2025mhz and I needed %120 (300 watts) for 2050mhz. But ran even better at 2015mhz @ factory 240 watts... I'll give it a look over again in case I missed something 👍
If it used more power when you increased the PL, that means the PL is the primary "Perfcap" limit. So it's not reaching your max V/F point at 1050mV in normal use. That means it's probably always hunting for the right voltage / frequency to stay within the PL.

If you find the voltage that the core uses when under load (GPU-Z sensors is the best imo), then you can set that as the max voltage point on your V/F curve and max out PL. It will still maintain similar power draw, but the core clock will stay locked at your chosen frequency.

Voltage settings seemed to have no affect on wattage
The voltage slider only lets you use up to the max VBIOS allowed voltage, should be 1093mV iirc. If the card is limited below that voltage, the slider does nothing.
 

·
Registered
Joined
·
389 Posts
Discussion Starter · #9 ·
Oh I got what you mean. I was thinking of the power limit slider
 

·
Registered
Joined
·
788 Posts
Oh I got what you mean. I was thinking of the power limit slider
lol yeah I was talking about the power limit slider. 😄

You can limit the power using just the slider, which automatically chooses the max core voltage it can, while staying within the power budget. This causes the card to constantly hunt for the max V/F point as the load changes.

or you can use the V/F curve editor and limit the power by limiting the voltage directly.

Here's what my settings look like :
2519695


PL slider at max but the V/F curve is limiting the card to 950mV. (setting my voltage slider to +100 is pointless but w/e)

Power draw will vary a bit depending on load if you do it this way, but the advantage is that the core clock stays locked (as long as voltage is capped low enough so you never hit the power limit).


Hope I made it more clear :)
 

·
Registered
Joined
·
389 Posts
Discussion Starter · #11 · (Edited)
Yep, I tried this, and apparently my gpu doesn't like having the power adjusted. My effective clock was absolutely falling on it's face. Wish I hadn't deleted the log, but it looked ROUGH.

I set it back how it was. Looks like effective clock now fluctuates 15mz every minute or two and when it changes, it stays there. 1985-2000mhz effective. Clock speed stays dead at 2015, if clock goes higher with that voltage curve, the card will crash.
 
  • Rep+
Reactions: T.Sharp
1 - 11 of 11 Posts
Top