GTX 1080 volt issues past 1.25v

[Kinda Bottlenecked]

7 hours ago, Hunched said:

I'm curious, when was the last time Gigabyte and MSI completely botched a GPU design?

Cause in my eyes they seem to be doing a consistently better job than the rest of the competition.

 

Look up 390/x G1

[boboman342]

8 hours ago, App4that said:

The benchmarks don't show async. I think the async in Pascal is like the driver update to enable it in Maxwell.

Well in games that use async like the new Hitman, you can see the fury x dominate the 980ti, where the 980ti doesn't have proper async management or whatever. The 1080 spanks the fury x in that same game, so they did something other than just an increase in raw power to deal with async. I assume it's something like how the implemented SMP, just adding a new compute block on the gpu to do the extra work. Thus preventing the improved async improvements seen on pascal being brought to maxwell.

[Bhav]

Nvidia are getting beyond lame for overclocking with their voltage constraints.

 

 

[DXMember]

10 hours ago, Hunched said:

Hopefully MSI can deliver, 8+6pin is good.

I believe it was revealed today, but the Gigabyte GTX 1080 G1 has a single 8-pin, so it's shit nobody buy it.

At least there's the Gigabyte GTX 1080 Xtreme, has dual 8-pins... I wish it looked better though.

It's going to be silly if they can't solve the bios voltage limit but are slapping down dual 8-pin connectors on their card.

"Wow I'm so glad I can do literally nothing with this extra power!"

 

the power plug is not a problem

you can overdraw the 8-pin as much as two times and a little bit more it's rated delivery as long as the wire gauge is adequate and PSU is rated accordingly - have you seen the single 8+8 PCI-E power cables?

there should be no reason for it to be power throttled under LN2 with a custom BIOS and especially with custom power delivery

the boo-boo is that apparently nVidia has further locked the voltage in driver or on die that is preventing any reasonable overclocks 

[M4st4M1nd]

Looks to me like they've gimped the card. Hardly an unusual thing for Nvidia to do. Honestly if AMD's offerings are good enough I'm switching ASAP -.-

[App4that]

4 hours ago, boboman342 said:

Well in games that use async like the new Hitman, you can see the fury x dominate the 980ti, where the 980ti doesn't have proper async management or whatever. The 1080 spanks the fury x in that same game, so they did something other than just an increase in raw power to deal with async. I assume it's something like how the implemented SMP, just adding a new compute block on the gpu to do the extra work. Thus preventing the improved async improvements seen on pascal being brought to maxwell.

Yeah, it looks like raw power since the 1080 still looses fps in DX12 applications.

[Arc_Jester]

17 hours ago, Hunched said:

I'm curious, when was the last time Gigabyte and MSI completely botched a GPU design?

Cause in my eyes they seem to be doing a consistently better job than the rest of the competition.

Every single Gigabyte AMD GPU is a botched job but MSI is good all round. 

[DarkBlade2117]

On ‎5‎/‎30‎/‎2016 at 2:43 PM, DXMember said:

pfail, who's rebranding now? Pascal = 10x Maxwell my ass

400-500 series are the same

600-700 series are the same

900-1000 series are the same.

They've always rebranded, if I'm correct the 680 and 770 are the EXACT same thing.

[Hunched]

I feel this is what happens when Nvidia keeps chasing the power efficiency/low heat dream.

 

Everyone that actually wants good overclocks has to undo all that bullshit, or GPU Boost 2.0 throttles you at 68c no matter what.

900 Series required a custom BIOS to disable that.

 

Nobody gives a fuck how cool and efficient you can get your cards by placing limits on hardware, BIOS, drivers, etc. Nvidia...

I can BIOS lock my 970 at 135mhz with minimum voltage, 30c full load, wow what progress, I could be an Nvidia engineer...

[Lays]

18 hours ago, Hunched said:

I wouldn't go as far as saying a single 8-pin is acceptable because of a voltage lock, no.

Especially if the issue is ever resolved, you would regret the decision of getting a single 8-pin even more. You'd miss out on some unlocked performance.

When it's cold in here I can get 30c load on my 980 ti, the magic of watercooling with big radiators   (Shameless brag )

 

But yeah I completely understand why they limit all this stuff, they don't want boatloads of people killing their cards.  But at the same time, they're giving all serious / passtime overclockers the shaft

 

I love overclocking, and being able to push 1.6v through my GPU on dry ice if I desire, am I going to run that voltage 24/7? hell no, but I'd like the option to at least be there!

[DXMember]

8 hours ago, DarkBlade2117 said:

400-500 series are the same

600-700 series are the same

900-1000 series are the same.

They've always rebranded, if I'm correct the 680 and 770 are the EXACT same thing.

the entire 500 is exactly the same as 400 series, they just tuned the clock speeds

[jjohnthedon1]

13 hours ago, Lays said:

When it's cold in here I can get 30c load on my 980 ti, the magic of watercooling with big radiators   (Shameless brag )

 

But yeah I completely understand why they limit all this stuff, they don't want boatloads of people killing their cards.  But at the same time, they're giving all serious / passtime overclockers the shaft

 

I love overclocking, and being able to push 1.6v through my GPU on dry ice if I desire, am I going to run that voltage 24/7? hell no, but I'd like the option to at least be there!

What is ur core clock in ur loop ? 

[ybriK]

So much for overclockers dream...

[Lays]

4 hours ago, jjohnthedon1 said:

What is ur core clock in ur loop ? 

I usually run 1520/8200 or around there 

[bandage106]

It's not voltage locked, I talked to the guy who created the original mod for the GTX 1080.I can screencap the message if you like but this is what he said when I asked him about the "1.25 limit"

 

"I'm going to post details in my guide regarding that and few other items. It's bit early to jump into any conclusions yet, as drivers still bit buggy. There is no hardware/PCB limit to go any voltage you want, problems are related to running settings as expected (and thats not only "1.25v limit"). Just have some patience. "

 

So it's good news. 

 

Here is the mod I'm talking about of course. 

 

https://xdevs.com/guide/pascal_oc/ 

 

 

 

[Lays]

2 minutes ago, bandage106 said:

It's not voltage locked, I talked to the guy who created the original mod for the GTX 1080.I can screencap the message if you like but this is what he said when I asked him about the "1.25 limit"

 

"I'm going to post details in my guide regarding that and few other items. It's bit early to jump into any conclusions yet, as drivers still bit buggy. There is no hardware/PCB limit to go any voltage you want, problems are related to running settings as expected (and thats not only "1.25v limit"). Just have some patience. "

 

So it's good news. 

 

Here is the mod I'm talking about of course. 

 

https://xdevs.com/guide/pascal_oc/ 

 

 

 

As I said previously the card can go above 1.25v. But the driver crashes when  this happens. People that have done the mod (der8auer) has said the driver crashes when going past 1.25v.

[bandage106]

2 minutes ago, Lays said:

As I said previously the card can go above 1.25v. But the driver crashes when  this happens. People that have done the mod (der8auer) has said the driver crashes when going past 1.25v.

Right which is a consequence of running an unstable overclock, it happens to my GTX 970 too. For instance my GTX 970 G1 gaming I've OC'ed to 1700 stable boost everyone assumed the "limit" for the GTX 970 G1 gaming was 1.275v it's actually around 1.312 but OCing software will display it running at 1.275v even as I'm going overvoltage, there is also the issue with perf caps on the GTX 970 I don't believe those exist on the EVGA clasified but I do not own a EVGA classified. 

 

It might be the same thing here and in that case bios modding it might prove highly effective. If it's not a perf cap then it's a voltage cap and in that case it's figuring out how much headroom we can push before it crashes and hard modding it and sending more mv to the PCB manually, someone did point out knocking out the shunts might not be able to increase mv to the PCB because there are more compotents throughout the PCB that regulate voltage. 

 

I don't own a GTX 1080 yet but when I do I'll definetely play around with it a bit, or I'll wait until some bios mods are out, keep in mind this is a 16nm node so I'm not sure how safe 1.25v is though, some people on reddit noticed they were only able to get 1.08v but still get 2050mhz I think if someone were to get 1.25v they may be able to push the limit quite hard 2300-2400. 

[Lays]

1 hour ago, bandage106 said:

Right which is a consequence of running an unstable overclock, it happens to my GTX 970 too. For instance my GTX 970 G1 gaming I've OC'ed to 1700 stable boost everyone assumed the "limit" for the GTX 970 G1 gaming was 1.275v it's actually around 1.312 but OCing software will display it running at 1.275v even as I'm going overvoltage, there is also the issue with perf caps on the GTX 970 I don't believe those exist on the EVGA clasified but I do not own a EVGA classified. 

 

It might be the same thing here and in that case bios modding it might prove highly effective. If it's not a perf cap then it's a voltage cap and in that case it's figuring out how much headroom we can push before it crashes and hard modding it and sending more mv to the PCB manually, someone did point out knocking out the shunts might not be able to increase mv to the PCB because there are more compotents throughout the PCB that regulate voltage. 

 

I don't own a GTX 1080 yet but when I do I'll definetely play around with it a bit, or I'll wait until some bios mods are out, keep in mind this is a 16nm node so I'm not sure how safe 1.25v is though, some people on reddit noticed they were only able to get 1.08v but still get 2050mhz I think if someone were to get 1.25v they may be able to push the limit quite hard 2300-2400. 

So you're telling me these guys using LN2 are having instability at over 1.25v with a custom bios with limits removed and all these mods on the card?  The people experiencing these problems are literally the best overclockers on the planet, if they say there is something wrong, they would of done extensive testing to try to figure out what the issue is.  The resistor shunt thing is irrelevant if they soldered an external VRM board, bypassing the default power connector and VRM. 

[Dackzy]

My idea why they did this lock is because else we would see Fermi types of cards and it kinda makes sense if that is why, because as we have seen they run a bit hot and then if you push over the 1.25 then who knows what might happen, maybe if they didn't have that lock GPU boost would just push over the 1.25 and then we would have Fermi back....

[bandage106]

7 minutes ago, Lays said:

So you're telling me these guys using LN2 are having instability at over 1.25v with a custom bios with limits removed and all these mods on the card?  The people experiencing these problems are literally the best overclockers on the planet, if they say there is something wrong, they would of done extensive testing to try to figure out what the issue is.  The resistor shunt thing is irrelevant if they soldered an external VRM board, bypassing the default power connector and VRM. 

I doubt they have a working BIOs mod, infact a lot of those experts backtracked saying they'd wait until custom V-bios came out including the guy who originally made the 1080 mod in the first place. The GTX 1080 just came out, we should wait until either there is a working bios modding tool for pascal to come out or start dissecting the Custom AIB's before we start jumping to assumptions, just what the issue is. You can look but a custom bios modding tool for the GTX 1080 doesn't exist. 

 

If they did extensive testing they'd of also shown their test results and given the data to the public, that'd of been extremely helpful in distinguishing the actual issue. 

 

[Lays]

27 minutes ago, bandage106 said:

I doubt they have a working BIOs mod, infact a lot of those experts backtracked saying they'd wait until custom V-bios came out including the guy who originally made the 1080 mod in the first place. The GTX 1080 just came out, we should wait until either there is a working bios modding tool for pascal to come out or start dissecting the Custom AIB's before we start jumping to assumptions, just what the issue is. You can look but a custom bios modding tool for the GTX 1080 doesn't exist. 

 

If they did extensive testing they'd of also shown their test results and given the data to the public, that'd of been extremely helpful in distinguishing the actual issue. 

 

 

Der8auer has exclusive access to things we don't have, of course he has a working BIOS mod, he's most likely been in contact with Nvidia since he got his review samples for testing on Ln2. I'm fairly sure you can just use a hex editor to edit the BIOS, no need for a custom tool like "maxwell bios tweaker". He does a lot of stuff behind the scenes with modding cards, testing BIOS, making Ln2 pots and such. He's #3 in the world for pro overclocking, and has invented boatloads of stuff that's been crucial in the scene. He's the one behind Thermal Grizzly, and the Delid Die mate. He also bins CPU's on LN2 and sells them for a very hefty premium. (He sold a few i7 6700k that did 6500+ mhz on hwbot prime on every core for like 1500 euro)

 

He says he's tested multiple cards, and he also has a review sample of a 1080 Strix, which I assume he also tested on, so aftermarket PCB still has this issue.

 

 

 

[TheGamingBarrel]

28 minutes ago, bandage106 said:

I doubt they have a working BIOs mod, infact a lot of those experts backtracked saying they'd wait until custom V-bios came out including the guy who originally made the 1080 mod in the first place. The GTX 1080 just came out, we should wait until either there is a working bios modding tool for pascal to come out or start dissecting the Custom AIB's before we start jumping to assumptions, just what the issue is. You can look but a custom bios modding tool for the GTX 1080 doesn't exist. 

 

If they did extensive testing they'd of also shown their test results and given the data to the public, that'd of been extremely helpful in distinguishing the actual issue. 

 

If there is an issue with the core, a BIOS Will not help, even installing new VRMs does not help, the core just refuses more than 1.25V, TiN Of EVGA Said that he had the VRms pushing a lot more, but the core wouldnt take it.

K|NGP|N And der8auer have both made custom BIOS And it does not fix it whatsoever.

[bandage106]

8 minutes ago, TheGamingBarrel said:

If there is an issue with the core, a BIOS Will not help, even installing new VRMs does not help, the core just refuses more than 1.25V, TiN Of EVGA Said that he had the VRms pushing a lot more, but the core wouldnt take it.

K|NGP|N And der8auer have both made custom BIOS And it does not fix it whatsoever.

I've been talking to TiN already, that's the message I posted up top after the 1.25v rumor started circulating I got in contact with him and sent him a message he sent me one back. 

 

"I'm going to post details in my guide regarding that and few other items. It's bit early to jump into any conclusions yet, as drivers still bit buggy. There is no hardware/PCB limit to go any voltage you want, problems are related to running settings as expected (and thats not only "1.25v limit"). Just have some patience. "

 

Bolded the part that's really most important. 

[r3h4f]

So as far as i can tell, a 980 ti with an oc is better than the 1080 because of the voltage limit, atleast for now. what id like to know is whether i should buy a 980ti or the 1080 is im not going to liquid cool it, just air (for now). is an oc 980 still better?

 

thanks

 

[IcEWoLF]

A stock 1080 is much better than the 980 gtx ti overclocked, if you are thinking about getting a plain 980 don't, get the 1070 which is very good.