Overclock.net banner

12081 - 12100 of 12608 Posts

·
Registered
Ryzen 7 3800xt on MSI Meg Ace x570. RTX 2080 Ti K|NGP|N for when I'm feeling insecure and small.
Joined
·
2 Posts
I had no intentions on water cooling my newest 2080Ti but damn is it a good overclocker. First let me say, I got lucky with samsung memory. Today was 40 degrees and decided to try the pc in my garage bench (with the garage door open to see how far I could push it. Pretty freaking far, this card is unmodified and is running a non-a chip. Got a graphics score of 16800 in timespy. That is with an offset of 475 on the core and 1000 on the mem.

Like I mentioned this was suppose to be a place holder for either a 6900XT or 3090 (or 3080Ti if released) but if it can hold those same clocks watercooled I might go that route.
16800 graphics in TS is solid on air, but what frequency is core clock sitting at on load? I'm clueless so I don't know if it's intentional or just something that works, but early when I was testing my 2080, I set the core boost to something like +250 or +300, and it boosted to 2145mhz at peak. Worked great. But if I dropped to +125, it would then also peak at the same frequency with (as I recall) the same behavior. It's been a while now and I don't recall being able to replicate it on another BIOS later, so it might have just been a temporary thing or BIOS-specific bug. And maybe I just haven't seen enough solid OC results out of non-A 2080ti, which might make up a bit of that offset (since A chips would come with some stock OC, I assume). Do you remember how much power the card was pulling in that test? Closest result I see in my chart to 16800 is 16793 graphics in TimeSpy, and it looks like that was on reg BIOS at +50/1200 (2115mhz peak core clock)
 

·
Registered
Joined
·
861 Posts
16800 graphics in TS is solid on air, but what frequency is core clock sitting at on load? I'm clueless so I don't know if it's intentional or just something that works, but early when I was testing my 2080, I set the core boost to something like +250 or +300, and it boosted to 2145mhz at peak. Worked great. But if I dropped to +125, it would then also peak at the same frequency with (as I recall) the same behavior. It's been a while now and I don't recall being able to replicate it on another BIOS later, so it might have just been a temporary thing or BIOS-specific bug. And maybe I just haven't seen enough solid OC results out of non-A 2080ti, which might make up a bit of that offset (since A chips would come with some stock OC, I assume). Do you remember how much power the card was pulling in that test? Closest result I see in my chart to 16800 is 16793 graphics in TimeSpy, and it looks like that was on reg BIOS at +50/1200 (2115mhz peak core clock)
This was my air cooled 2080Ti FE. It broke 17K timespy graphics on just air! That’s with 380 watt Galax bios and shunts soldered. All inside of a closed case in normal ambient temps.




My GPU is watercooled now. but, of all (4) 2080Ti’s ive owned. I have never seen over 17K in air. I thought it was amazing!
 

·
Old to Overclock.net
Joined
·
1,683 Posts
16800 graphics in TS is solid on air, but what frequency is core clock sitting at on load? I'm clueless so I don't know if it's intentional or just something that works, but early when I was testing my 2080, I set the core boost to something like +250 or +300, and it boosted to 2145mhz at peak. Worked great. But if I dropped to +125, it would then also peak at the same frequency with (as I recall) the same behavior. It's been a while now and I don't recall being able to replicate it on another BIOS later, so it might have just been a temporary thing or BIOS-specific bug. And maybe I just haven't seen enough solid OC results out of non-A 2080ti, which might make up a bit of that offset (since A chips would come with some stock OC, I assume). Do you remember how much power the card was pulling in that test? Closest result I see in my chart to 16800 is 16793 graphics in TimeSpy, and it looks like that was on reg BIOS at +50/1200 (2115mhz peak core clock)
You are correct, since it is a non a chip the offset needs to be greater. This is the biggest scoring rtx 2080ti. My first one was a zotac, second was unknown brand, third was another zotac, fourth was a evga (water-cooled this one, also non a and scored around 16500 on timespy on stock bios and no shunt mid), and this new evga. Now those have been my personal ones. I have built at least 5 high end rigs for flips with 2080tis. Out of all the ones I played with this non a scores the best. This thing has the stock BIOS and as far as I can tell no shunt modified. I texted the person I bought from to see if he modified it because it performs extremely well especially being a non a chip. No response, I even specified not looking for a refund just really impressed with this card.

I am going to water cool my main rig again I water-cooled it half assed 2 months ago but took everything apart and used the cpu pump block combo for a client rig and sold my old 2080ti with the block. I just assumed I was going to get a 5900x and 6800xt or 3080, boy was I wrong.

Just bought a new barrow block from eBay for 70, owner claims he just opened to inspect it and 145 to titanrig for a barrow cpu block, alphacool reservoir ddc pump combo, 12 compression fittings, and the silicone tube that goes inside hard line tubing for easy bending. I already have the radiators and hard line tubing. Just going to push my 3950x and 2080ti as far as I can.
 

·
Registered
Joined
·
861 Posts
You are correct, since it is a non a chip the offset needs to be greater. This is the biggest scoring rtx 2080ti. My first one was a zotac, second was unknown brand, third was another zotac, fourth was a evga (water-cooled this one, also non a and scored around 16500 on timespy on stock bios and no shunt mid), and this new evga. Now those have been my personal ones. I have built at least 5 high end rigs for flips with 2080tis. Out of all the ones I played with this non a scores the best. This thing has the stock BIOS and as far as I can tell no shunt modified. I texted the person I bought from to see if he modified it because it performs extremely well especially being a non a chip. No response, I even specified not looking for a refund just really impressed with this card.

I am going to water cool my main rig again I water-cooled it half assed 2 months ago but took everything apart and used the cpu pump block combo for a client rig and sold my old 2080ti with the block. I just assumed I was going to get a 5900x and 6800xt or 3080, boy was I wrong.

Just bought a new barrow block from eBay for 70, owner claims he just opened to inspect it and 145 to titanrig for a barrow cpu block, alphacool reservoir ddc pump combo, 12 compression fittings, and the silicone tube that goes inside hard line tubing for easy bending. I already have the radiators and hard line tubing. Just going to push my 3950x and 2080ti as far as I can.
I could get around 17,250 with my old NON-A 2080Ti on watercooling shunt modded and flashed, this may not seem like a whole lot more. But being it is watercooling, this is sustained performance.

My air cooled Gigabyte non A would manage 16,800 on air. But, it had shunt mods and 310 watt bios.

The card must have something done to it. 16,800 timespy graphics on air cooling sounds near impossible without a lot of extra power.

Breaking 17K with my A bin FE on air cooling proved quite difficult. It had “Up to” 532 watts of available power that it could use if needed.
 

·
Registered
Joined
·
98 Posts
Hi guys. I arrived late to the party. I now own a 2080ti FTW3 (air) and I'd like to know what are some typical overclock values for 24/7 stability.
For now I went with an undervolt of 2040 core / 8000 memory @ 1V. Temps are 72-73c at continuous (~60 scenes) of FireStrike Ultra, TimeSpy Extreme and Port Royal, and ~70c after 2 hours of Superposition stress.
From my experience with overclocking 2080ti, there isn't much point going to high voltages (anything above 1.062V) because the power limit will kick too often and even if you think you have 2200 @ 1.093V, in reality you will have something like 2000. The solution to that was shunt mod but I don't want to shunt this card as it is air cooled.
So, I am interested about what to expect when I go to the stock voltage of 1.05V. Is it typical to achieve 24/7 stable 2100? What about memory?
Thanks!
By the way, I have to say I'm a bit disappointed with EVGA, first of all the card is pretty loud, especially compared to my previous cards, the Palit 2080s Gamerock and Aorus 1080ti. Second, the Precision software isn't as good as MSI Afterburner - the UI more difficult to control, the monitoring isn't as good as Afterburner, and the voltage curve is ehhh. Third, the fan control is so weird with this card. MSI AB can't control the fans at all, and in fact when I try to do it weird things happen and I have to open Precision to somehow "re-configure" the fans. I wonder if this is a bios issue - anyone ran into it?
 

·
Registered
Joined
·
861 Posts
Hi guys. I arrived late to the party. I now own a 2080ti FTW3 (air) and I'd like to know what are some typical overclock values for 24/7 stability.
For now I went with an undervolt of 2040 core / 8000 memory @ 1V. Temps are 72-73c at continuous (~60 scenes) of FireStrike Ultra, TimeSpy Extreme and Port Royal, and ~70c after 2 hours of Superposition stress.
From my experience with overclocking 2080ti, there isn't much point going to high voltages (anything above 1.062V) because the power limit will kick too often and even if you think you have 2200 @ 1.093V, in reality you will have something like 2000. The solution to that was shunt mod but I don't want to shunt this card as it is air cooled.
So, I am interested about what to expect when I go to the stock voltage of 1.05V. Is it typical to achieve 24/7 stable 2100? What about memory?
Thanks!
By the way, I have to say I'm a bit disappointed with EVGA, first of all the card is pretty loud, especially compared to my previous cards, the Palit 2080s Gamerock and Aorus 1080ti. Second, the Precision software isn't as good as MSI Afterburner - the UI more difficult to control, the monitoring isn't as good as Afterburner, and the voltage curve is ehhh. Third, the fan control is so weird with this card. MSI AB can't control the fans at all, and in fact when I try to do it weird things happen and I have to open Precision to somehow "re-configure" the fans. I wonder if this is a bios issue - anyone ran into it?
People own 2080Ti’s and they can sustain 2,200 or even 2,250Mhz. It’s all about how cool you can keep the card. The only reason you are undervolting is to get your card cooler to run a higher frequency, by consuming less power and in turn creating much less heat.

The colder a GPU runs the less power it consumes. Or certain bios allow nearly unlimited power draw, or much higher voltages. But no matter what bios you have, or how much you can send you must keep the GPU cold.


I ran timespy extreme last night, my card was at 2,160Mhz and pulled 463 watts of juice. It certainly wasn’t running out of power. And it held the frequency Without clocking down.


I guess what I’m saying is, if your card is running very cool like 42C you don’t even need to undervolt it at all. Just overclock it, and it’ll lock down 2,160–2,175Mhz.


^ This is how my FE model is.
 

·
Registered
Joined
·
98 Posts
People own 2080Ti’s and they can sustain 2,200 or even 2,250Mhz. It’s all about how cool you can keep the card. The only reason you are undervolting is to get your card cooler to run a higher frequency, by consuming less power and in turn creating much less heat.

The colder a GPU runs the less power it consumes. Or certain bios allow nearly unlimited power draw, or much higher voltages. But no matter what bios you have, or how much you can send you must keep the GPU cold.


I ran timespy extreme last night, my card was at 2,160Mhz and pulled 463 watts of juice. It certainly wasn’t running out of power. And it held the frequency Without clocking down.


I guess what I’m saying is, if your card is running very cool like 42C you don’t even need to undervolt it at all. Just overclock it, and it’ll lock down 2,160–2,175Mhz.


^ This is how my FE model is.
Hi, thanks for the comment!
My card is on air. When I run things like FireStrike Ultra it will power limit even at 1V, although it happens only at 65c+.
At what voltage are you running 2160MHz?
 

·
Registered
Joined
·
861 Posts
Hi, thanks for the comment!
My card is on air. When I run things like FireStrike Ultra it will power limit even at 1V, although it happens only at 65c+.
At what voltage are you running 2160MHz?
Yeah reducing the temperature would greatly reduce power consumption. My card on the previous bios would run 2,130Mhz at 1.062MV and it pulls around 275-285 watts playing death stranding. On air cooling and undervolted, and a much lower frequency it Pulled 80-100 more watts easily.

I run 1,100MV for 2,160Mhz.

I think I could get by with a little less voltage. But I haven’t messed with it too much.

Watercool that card and you’ll get much lower power consumption and much higher frequencies that don’t down clock in games.


You can squeeze roughly within 3% of a RTX3080FE, once you setup a 2080Ti properly. And you’ll obtain superior ray tracing performance to something like a 6800XT.

The GPU market is a mess these days. That’s why I grabbed another 2080Ti solely to watercool it for maximum performance. I sold my last one to try and grab next gen, and regretted it.
 
  • Rep+
Reactions: yoadknux

·
Village Idiot
Joined
·
2,421 Posts
Hi guys. I arrived late to the party. I now own a 2080ti FTW3 (air) and I'd like to know what are some typical overclock values for 24/7 stability.
For now I went with an undervolt of 2040 core / 8000 memory @ 1V. Temps are 72-73c at continuous (~60 scenes) of FireStrike Ultra, TimeSpy Extreme and Port Royal, and ~70c after 2 hours of Superposition stress.
From my experience with overclocking 2080ti, there isn't much point going to high voltages (anything above 1.062V) because the power limit will kick too often and even if you think you have 2200 @ 1.093V, in reality you will have something like 2000. The solution to that was shunt mod but I don't want to shunt this card as it is air cooled.
So, I am interested about what to expect when I go to the stock voltage of 1.05V. Is it typical to achieve 24/7 stable 2100? What about memory?
2040 @ 1.000v 72-73c load sounds pretty good. That would be somewhere around 2100-2130 at water cooled temps, which is quite good for that voltage.

"stock" voltage is up to 1.068v, although the curve will typically max out at 1.05v. With a 373w (FTW3) or 380w (Galax BIOS) power limit, ~1.025v is about the max you can consistently run for gaming in my experience without power throttling. Maybe a bit lower on air cooling.

This all varies by game and resolution of course. Running games at 4k will typically result in way higher power consumption than if you were running at 1080p. And every game/engine varies wildly. For example, in AC Valhalla or Watch Dogs Legion (with RTX on), my card will average a measly 300w even at 1.093v. But in somethung like Metro Exodus, average power draw is in the low 400s with peaks nearing 500w.

Memory is all over the map as well. My card has Micron, and topped out around +750mhz on stock cooling (EVGA XC Ultra in a case with ~800 RPM radiator fans for airflow). After throwing a water block on it, I can run up to +1100mhz. Samsung cards seem to be capable of +1400mhz and beyond if you get a good one.
 

·
Registered
Joined
·
98 Posts
2040 @ 1.000v 72-73c load sounds pretty good. That would be somewhere around 2100-2130 at water cooled temps, which is quite good for that voltage.

"stock" voltage is up to 1.068v, although the curve will typically max out at 1.05v. With a 373w (FTW3) or 380w (Galax BIOS) power limit, ~1.025v is about the max you can consistently run for gaming in my experience without power throttling. Maybe a bit lower on air cooling.

This all varies by game and resolution of course. Running games at 4k will typically result in way higher power consumption than if you were running at 1080p. And every game/engine varies wildly. For example, in AC Valhalla or Watch Dogs Legion (with RTX on), my card will average a measly 300w even at 1.093v. But in somethung like Metro Exodus, average power draw is in the low 400s with peaks nearing 500w.

Memory is all over the map as well. My card has Micron, and topped out around +750mhz on stock cooling (EVGA XC Ultra in a case with ~800 RPM radiator fans for airflow). After throwing a water block on it, I can run up to +1100mhz. Samsung cards seem to be capable of +1400mhz and beyond if you get a good one.
Thank you as well for the comment. So it seems the overclock I'm getting is quite typical for this card (on air). Memory can bench at 1500 too, but for 24/7, not really. I think that at some point the memory modules overheat and that causes stability issues. It's actually very weird, the FTW3 has three memory sensors, and one is very cool (~50c stressed), while the others are hot (~80c stressed). When I changed thermal paste, I noticed that only one side of the memory is connected through a thermal pad to the heatsink, while the rest of the components seem hidden below some black front plate. I think this is a design flaw, because it traps heat and it isn't connected to the main heatsink. This is my first EVGA card since the 970 and I'm actually a bit disappointed, it's a very loud card. Precision is also a very buggy thing and MSI AB can't control the fan curve on those cards (and even worse, when I try to do that, it disables two fans until I run Precision). Perhaps this card was meant to be "fully unlocked" under water, but at stock configuration, I wouldn't call it great.
 

·
Registered
Joined
·
3,240 Posts
User Imprezzion reported some success with Galax HOF XOC BIOS - but not without problems.
You lose a DP Port and you can't use Afterburner as usual...

And as you already use the 380w BIOS with shunt mod – there isn’t IMHO really a point in trying.

As I said – above a certain point there are really diminishing returns – somewhere beyond 420W or so…and you really need very very good cooling.
Using XOC BIOS only really makes sense for LN2-OC.

Best regards,
Medizinmann
Yes, XOC works fine if you can live without profiles in MSI AB and having to do your OC every boot manually.

Also, DP/HDMI depending on your specific model of card may or may not work. My Gainward Phoenix GS doesn't work with all ports. Only HDMI and 2 DP's work. And it doesn't wanna show a UEFI BIOS screen over DP. Only HDMI so I have to run Legacy GOP in the BIOS for that to show on my main monitor.

I can run 2145Mhz on 1.125v perfectly stable on XOC. Power draw is about 400-420w and temps around 48-50c under AIO water. On a normal BIOS I am limited to 2085Mhz on 1.093v and some games like GTA V on 4K let it hit throtttling even with 373w EVGA FTW3 Ultra BIOS.
 

·
Registered
Joined
·
861 Posts
Yes, XOC works fine if you can live without profiles in MSI AB and having to do your OC every boot manually.

Also, DP/HDMI depending on your specific model of card may or may not work. My Gainward Phoenix GS doesn't work with all ports. Only HDMI and 2 DP's work. And it doesn't wanna show a UEFI BIOS screen over DP. Only HDMI so I have to run Legacy GOP in the BIOS for that to show on my main monitor.

I can run 2145Mhz on 1.125v perfectly stable on XOC. Power draw is about 400-420w and temps around 48-50c under AIO water. On a normal BIOS I am limited to 2085Mhz on 1.093v and some games like GTA V on 4K let it hit throtttling even with 373w EVGA FTW3 Ultra BIOS.
Yeah I tested the Galax XOC bios. It’ll pull some serious juice. Luckily my water loop keeps it very very cool. I ran timespy GT2 at 4K and pulled over 520 watts of power.

With my 7980XE at 4.8Ghz and my 2080Ti on XOC bios combined I am pushing the capacity of my 1200 watt PSU In death stranding. And I am getting power shut downs!



If I run the normal Galax 380 watt bios, I can easily manage 2,115Mhz sustained at 1.050V.


This thing will walk a 6800XT in ray tracing performance guys!


 

·
Registered
Joined
·
861 Posts
Anyone running a 2080Ti watercooled with the 380 watt Galax bios and running "Soldered Shunts", I highly recommend locking your clocks at 1.093MV and 2,160Mhz. Use "SHIFT+L in MSI AfterBurner. This fixed all sorts of crashing issues I was getting.


The issues with running an offset like +150Mhz is the GPU will try to run 2,160Mhz or even 2,130mhz and then, the GPU will slowly lower voltage as the GPU heats up even only running 40-42C. And this is what causes the crashing. You may be running 2,130Mhz in a game 100% stable and your GPU flickers the voltage down to 1.043MV and it crashes your game.

When you hit SHIFT+L you literally LOCK that voltage so it'll never lower during game play. The one downside is that MSI AB locks 2,160Mhz even the desktop. So, just save the profile. And click profile "1" right before you play a game.

I have been going back and forward with the Galax 380 watt bios, and a Galaxy XOC 2K watt bios trying to find the best stability with the absolute maximum performance. And I have finally found the best option for my 2080Ti.
 

·
Registered
Joined
·
3,240 Posts
Doesn't have to be shift L at all. Just make sure your 1.093v is 2 clicks higher (so 2x15Mhz) then your next lowest point. So, if you wanna run +150 set it to a +120 offset, hit apply and watch the curve adjust, then raise ONLY the 1.093v point to the desired clocks, hit apply again, Save that as a profile and you never have to touch Afterburner again and it will always run 1.093v.

I run +105 offset with 2 clicks higher 1.093v point for example
 

·
Village Idiot
Joined
·
2,421 Posts
Anyone running a 2080Ti watercooled with the 380 watt Galax bios and running "Soldered Shunts", I highly recommend locking your clocks at 1.093MV and 2,160Mhz. Use "SHIFT+L in MSI AfterBurner. This fixed all sorts of crashing issues I was getting.


The issues with running an offset like +150Mhz is the GPU will try to run 2,160Mhz or even 2,130mhz and then, the GPU will slowly lower voltage as the GPU heats up even only running 40-42C. And this is what causes the crashing. You may be running 2,130Mhz in a game 100% stable and your GPU flickers the voltage down to 1.043MV and it crashes your game.

When you hit SHIFT+L you literally LOCK that voltage so it'll never lower during game play. The one downside is that MSI AB locks 2,160Mhz even the desktop. So, just save the profile. And click profile "1" right before you play a game.

I have been going back and forward with the Galax 380 watt bios, and a Galaxy XOC 2K watt bios trying to find the best stability with the absolute maximum performance. And I have finally found the best option for my 2080Ti.
As long as your 1.093v point is set +15mhz higher than everything else on the curve, you shouldn't need to "L" that point on the curve.

Also, if you leave all the other points alone and only increase 1.093v (or whatever voltage point you're aiming for), the card will consistently perform worse. Why this is, I have no idea. But after shunting my card, I started jacking up only the 1.093v point on the curve, and my scores sucked. After some Google-Fu, I had found a post (I think in this thread) saying to set the other points -1mhz below your voltage point of choice. Bottom line is that the offsets at other points on the curve have an affect on SOMETHING that impacts performance...

Just to make sure I'm not totally insane, I tested it again just now in TS Extreme GT1:

1.093v +180, all other points +165 = 49.47 fps (2160mhz actual clock)
1.093v +180, all other points +0 = 48.26 fps (2160mhz actual clock)
1.093v +180, all other points +105 = 48.81 fps (throttled to 2145mhz half way through due to higher water temps)
+165 offset = 49.36 fps (ran 2145mhz @ 1.043v, jumped to 2145mhz @ 1.093v towards the end)

My biggest issue overclocking this card is the damn temperature/load dependent shifting of the v-f curve. I've run 2130mhz or 2145mhz (depending on game) for HOURS without a hitch. But it always ends up crashing because either water temps have dropped just a fraction of a degree enough, or load has temporarily dropped off just long enough, that the card decides it's going to clock up 15mhz at the same exact voltage. And boom, insta-crash.

It seems the only way to get consistent maximum clocks is to keep your peak load temps exactly at 39-40*C. Hopefully with an extra 2x D5s I can get there...
 

·
Registered
Joined
·
861 Posts
As long as your 1.093v point is set +15mhz higher than everything else on the curve, you shouldn't need to "L" that point on the curve.

Also, if you leave all the other points alone and only increase 1.093v (or whatever voltage point you're aiming for), the card will consistently perform worse. Why this is, I have no idea. But after shunting my card, I started jacking up only the 1.093v point on the curve, and my scores sucked. After some Google-Fu, I had found a post (I think in this thread) saying to set the other points -1mhz below your voltage point of choice. Bottom line is that the offsets at other points on the curve have an affect on SOMETHING that impacts performance...

Just to make sure I'm not totally insane, I tested it again just now in TS Extreme GT1:

1.093v +180, all other points +165 = 49.47 fps (2160mhz actual clock)
1.093v +180, all other points +0 = 48.26 fps (2160mhz actual clock)
1.093v +180, all other points +105 = 48.81 fps (throttled to 2145mhz half way through due to higher water temps)
+165 offset = 49.36 fps (ran 2145mhz @ 1.043v, jumped to 2145mhz @ 1.093v towards the end)

My biggest issue overclocking this card is the damn temperature/load dependent shifting of the v-f curve. I've run 2130mhz or 2145mhz (depending on game) for HOURS without a hitch. But it always ends up crashing because either water temps have dropped just a fraction of a degree enough, or load has temporarily dropped off just long enough, that the card decides it's going to clock up 15mhz at the same exact voltage. And boom, insta-crash.

It seems the only way to get consistent maximum clocks is to keep your peak load temps exactly at 39-40*C. Hopefully with an extra 2x D5s I can get there...
So there’s a work around by setting a voltage curve that doesn’t reduce performance? Could you elaborate more please? I really appreciate it.

Ive always known that setting a curve lowers performance slightly vs running a offset. It is weird I know trust me, and I really don’t know exactly why it does it either, it is really stupid and frustrating lol. From how derbau8er explained it, it is due to the silicon it’s self literally reacting differently with different frequencies and different voltages or different straps or something like that. But, from what I’ve gathered It makes no sense why locking 1.093 at 2,160 or 2,175 wouldn’t perform just as good as running an offset. I wish I knew..


It’s almost like there no such thing as a perfect overclock with a reference board “A” Bin 2080Ti... The Galaxy XOC bios allows massive power draw and 1.125MV which is awesome, but it is super unstable. I get power shut downs for no reason, and you can’t save a profile in MSI AB, and it isn’t my PSU at all as even being the issue. The Galax 380 watt bios is only 1.093MV and it likes to step down voltages on its own when you run a standard offset on the core clock, and it’s only 380 watts so soldering the 8 ohms is still required.

So, I really wish I had an actual Galaxy HOF WC 2000watt 2080Ti lol. Maybe that would solve the problem. I could just send 1.093MV or more and 2,145Mhz or more and be happy.


I wish there were some
better bios out there available for use with reference boards and “A bin TU102 boards”

I am seriously considering that water chiller lol. My current custom loop is sufficient, it’s just not good enough to stay under 40C with such high power draw, and my power hog 7980XE being in the mix too. That old beast sits on the desktop idle at 200+ watts.
 

·
Village Idiot
Joined
·
2,421 Posts
So there’s a work around by setting a voltage curve that doesn’t reduce performance? Could you elaborate more please? I really appreciate it.
Yes, sort of. The boost algorithm tends to level off the curve from 1.043-1.050v and beyond once the card gets up to load temps/power. And it will always try to run the highest clocks possible at the lowest voltage possible based on the curve.

So on my card, with the 380w GALAX BIOS, a +180 offset will usually result in the curve calling for a clock speed of 2160mhz at every point on the curve from 1.043v through 1.093v. So naturally, the boost algorithm will set the card to run 2160mhz @ 1.043v, because why run 1.093v if the curve says it can run the same clock at 1.043v? Only problem is, 2160mhz @ 1.043v = insta crash.

The "work around" is to set a +165 offset, then open the curve editor and drag up the 1.093v point to +180. So now once you get the card up to load temps, the curve should be calling for 2145mhz from 1.043v through 1.087v. But the 1.093v point will still be 2160mhz since it's offset is still +180. So the card will sit at 1.093v since that voltage point corresponds to the highest clock according to the curve.

It’s almost like there no such thing as a perfect overclock with a reference board “A” Bin 2080Ti... The Galaxy XOC bios allows massive power draw and 1.125MV which is awesome, but it is super unstable. I get power shut downs for no reason, and you can’t save a profile in MSI AB, and it isn’t my PSU at all as even being the issue.
If your computer is literally shutting off, I wouldn't be so quick to dismiss your PSU. When I got my card, still on stock cooler, maxing out the 338w stock power limit would trip the overcurrent protection on my 1000w Seasonic based PSU. AC power draw was only ~500w, the same as with my old GTX 1070 SLI setup. But the 2080 Ti would shut off my computer. So I ended up with a 1600w PSU for my 600w rig. Good times.

These cards can have some crazy high transient current draws, and combined with your power monster of a CPU, it's very possible a 1.125v unlimited power BIOS is more than your PSU can handle.
 
  • Rep+
Reactions: J7SC

·
Invalid Media
Joined
·
2,695 Posts
Yes, sort of. The boost algorithm tends to level off the curve from 1.043-1.050v and beyond once the card gets up to load temps/power. And it will always try to run the highest clocks possible at the lowest voltage possible based on the curve. (...)

These cards can have some crazy high transient current draws, and combined with your power monster of a CPU, it's very possible a 1.125v unlimited power BIOS is more than your PSU can handle.
Your comments on brief voltage spikes are appreciated. I run 2x 2080 Ti with their own big water loop which usually keeps them below 40 C, and never bothered with max-v or fixed voltage MSI AB curves etc ...Because with two cards at max 380 W each (stock Aorus bios) plus an oc'ed Threadripper and various peripherals, the 1300 W PSU is getting quite a workout already. With short spikes, which happen with RTX as you posted, I rather not go any higher via base oc to avoid PSU OCP.

My top card typically hits and stays at 1.043V even well above 2205 (when cold), the second one hits 1.062 V at 2175 (when cold). Those voltages do not change when for example benching PortRoyal, just the temps and with it the clocks do ( https://www.3dmark.com/pr/466677 starts at 2160 for both, along with 2056 VRAM).

Maxing GPUv from the start makes no sense to me, rather the opposite. Minimizing temps still has the best overarching results, imo.

MSFS2020-gpuv.jpg
 

·
Registered
Joined
·
2,021 Posts
Your comments on brief voltage spikes are appreciated. I run 2x 2080 Ti with their own big water loop which usually keeps them below 40 C, and never bothered with max-v or fixed voltage MSI AB curves etc ...Because with two cards at max 380 W each (stock Aorus bios) plus an oc'ed Threadripper and various peripherals, the 1300 W PSU is getting quite a workout already. With short spikes, which happen with RTX as you posted, I rather not go any higher via base oc to avoid PSU OCP.

My top card typically hits and stays at 1.043V even well above 2205 (when cold), the second one hits 1.062 V at 2175 (when cold). Those voltages do not change when for example benching PortRoyal, just the temps and with it the clocks do ( https://www.3dmark.com/pr/466677 starts at 2160 for both, along with 2056 VRAM).

Maxing GPUv from the start makes no sense to me, rather the opposite. Minimizing temps still has the best overarching results, imo.

View attachment 2467329
380w seems high. My KPE pulls 240-300w at load on stock 520w bios and +0 on voltage slider. I can make it pull over 400 watts by flipping both voltage dip switches on the back of the card, but this lowers the core speed due to increased temperature. Have you tried lowering the voltage?
 

·
Invalid Media
Joined
·
2,695 Posts
380w seems high. My KPE pulls 240-300w at load on stock 520w bios and +0 on voltage slider. I can make it pull over 400 watts by flipping both voltage dip switches on the back of the card, but this lowers the core speed due to increased temperature. Have you tried lowering the voltage?
...nope, I let them drink their stock allotment, below 'only' 375.7 W per card (Aorus stock bios XTR WB), but I have seen 379.9 W before. I just don't touch the voltage slider at all - may be I should try to undervolt after all...

...my biggest challenge has been to lower the total system power budget from 1150+ W to about 900 W with my current fav app: MS FS 2020 on SLI-CFR...PL down to 112 %, GPU oc down to 2100 MHz and VRAM down to about 2000 MHz (16,000 'effective) works fine w/o any appreciable loss in visuals / fps

Aorus_2235_8417_375-7_B.jpg


2467343
 
12081 - 12100 of 12608 Posts
Top