• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

NVIDIA GeForce GTX 1080 Ti launch/review day - 2017/03/09

Pagusas

Elden Member
We didn't touch the voltages. Power limit is set to 125%. I am reading on overclock.net that there's a beta bios that increases the limit to 150%. I wondering if anyone here has tried it yet.

You need to in increase the voltage if you want to get anywhere. Unlock voltage controls and put it to 100% along with the increased power limit.
 
You need to in increase the voltage if you want to get anywhere. Unlock voltage controls and put it to 100% along with the increased power limit.

Sounds good. IIRC, the Aorus has 12+2 power phases, etc., which should definitely help. Tweaking the voltage on my FE doesn't do it any good, so I treated his card the same. My mistake! :p

I appreciate the input! =)
 
What she clocked at to hit that score? Surprised to see an FE out scoring all of our non-FE cards

I got +160core and +350mem. I never really messed around with it too much to be honest. No extra voltage either. Just right after I broke 10K

33865601822_d22b33b7db_b.jpg
 

Varna

Member
I got +160core and +350mem. I never really messed around with it too much to be honest. No extra voltage either. Just right after I broke 10K

What system you get that score on? Trying to get a sense of m own pc. Got 9636 w/ i7 3770k

edit: wait, i'm stupid. It's listed right there.
 

Jin

Member
Gamers Nexus posted a review of MSI Gaming X.

https://www.youtube.com/watch?v=ZxMuQu-m1iE&t=680s

TLDR - excellent card. What's interesting is that OCing the card even further than the already stock OC gives you higher frametimes.

1080ti-gaming-x-grw-frametimes-1080p.png


Gains from overclocking are largely insignificant, and introduce additional frametime variance as the card struggles to maintain even a small 75MHz offset. You can see in our frametime plots that, despite higher average FPS and more frame throughput overall, consistency goes down with the overclock.

It looks like these AIB cards is already pushing the OC headroom as much as they can on Pascal and it got me thinking of lowering the OC on my card as well.
 

riflen

Member
My out-of-the-box score was 9436. It took some serious juice to get near Remij's score for me.

This was with +70mv, +185 Mhz Core (2012-2075 Mhz) and +491 Mhz Memory (5994 Mhz / 11988 effective). This is an EVGA FE card and the CPU is at 4400 Mhz.

33899728481_231b51a3ca_o.png
 

Pagusas

Elden Member
Gamers Nexus posted a review of MSI Gaming X.

https://www.youtube.com/watch?v=ZxMuQu-m1iE&t=680s

TLDR - excellent card. What's interesting is that OCing the card even further than the already stock OC gives you higher frametimes.

1080ti-gaming-x-grw-frametimes-1080p.png




It looks like these AIB cards is already pushing the OC headroom as much as they can on Pascal and it got me thinking of lowering the OC on my card as well.

You know I've noticed this as well, especially with the memory overclocking.
 

laxu

Member
People posting Superposition benchmarks, just know that using the 4K optimized setting on any resolution will give you much higher scores than using the Extreme setting.

For example my 980 Ti gets something like 2302 at 1440p Extreme, with a miserable 17 fps average framerate. Set the shaders to 4K Optimized and that jumps to about 50-60 fps in the same resolution. At 4K I get about 50 fps with the 4K Optimized setting.
 

riflen

Member
People posting Superposition benchmarks, just know that using the 4K optimized setting on any resolution will give you much higher scores than using the Extreme setting.

For example my 980 Ti gets something like 2302 at 1440p Extreme, with a miserable 17 fps average framerate. Set the shaders to 4K Optimized and that jumps to about 50-60 fps in the same resolution. At 4K I get about 50 fps with the 4K Optimized setting.

Yes, if you want to really peg the GPU at 100%, extreme will do it. At QHD extreme I score 3425 and the GPU core holds 2075 almost constantly.
You can't submit to leaderboards with that test though, so I expect that's why people are using 4K optimised.
 

laxu

Member
Yes, if you want to really peg the GPU at 100%, extreme will do it. At QHD extreme I score 3425 and the GPU core holds 2075 almost constantly.
You can't submit to leaderboards with that test though, so I expect that's why people are using 4K optimised.

Yeah, I just wanted to point out that even though 4K optimized seems like it would be a very high quality setting, it's really not and most high end GPUs can run it well.
 

riflen

Member
Yeah, I just wanted to point out that even though 4K optimized seems like it would be a very high quality setting, it's really not and most high end GPUs can run it well.

Agreed. Even the 8K optimised test doesn't crush the GPU as well as extreme runs.
 

Pagusas

Elden Member
Im very suprised with the score fluctuations we are seeing in the test, even the optimized one, given that this test is suppose to be purely GPU from my understanding. I'm scoring a 9550 on a small OC, others are hitting low 9000's then we have some hitting 10k. a 10% difference is pretty huge when we are all using the same card.
 
A bunch of comparison numbers tempt me to overclock beyond the manufacture overclock but I know my core won't go much higher, maybe 50Mhz, the memory could potentially go a bit more but in the end real world performance between 2000Mhz and 2050Mhz isn't worth the time to make sure all is stable.

My out-of-the-box score was 9436. It took some serious juice to get near Remij's score for me.

This was with +70mv, +185 Mhz Core (2012-2075 Mhz) and +491 Mhz Memory (5994 Mhz / 11988 effective). This is an EVGA FE card and the CPU is at 4400 Mhz.

33899728481_231b51a3ca_o.png

You must be on a custom water loop.
 

laxu

Member

dr_rus

Member
Do they think it's because it's a great custom card or because they got lucky on the silicon lottery?

Judging from their noise measurements this is mostly because AMP Extreme is running at a much higher air flow than the rest of 1080Tis.

So far I'd say that Gaming X, Strix OC and Aorus are all pretty close to each other in key characteristics so it's mostly a question of what vendor you prefer which is a good situation for all of us.

But it's almost as loud as the FE cards on load so running at slightly higher speed comes at a huge noise. The ASUS Strix runs almost as fast at a much lower volume and interestingly lower temps as well. If the Strix can be BIOS-modded for higher power limit it should perform just as well.

Yep. I'm still undecided on whether I'll even get one but Strix OC is certainly one of possible options for me. Will wait for Vega and see how things will be there.
 

Orion514

Member
Gamers Nexus posted a review of MSI Gaming X.

https://www.youtube.com/watch?v=ZxMuQu-m1iE&t=680s

TLDR - excellent card. What's interesting is that OCing the card even further than the already stock OC gives you higher frametimes.

1080ti-gaming-x-grw-frametimes-1080p.png




It looks like these AIB cards is already pushing the OC headroom as much as they can on Pascal and it got me thinking of lowering the OC on my card as well.

That sounds great for me. If it comes OC'd to the most stable max out of the box, I am perfectly okay with not having to fiddle with anything. Being the newb I am, haha.
 

Pagusas

Elden Member
That sounds great for me. If it comes OC'd to the most stable max out of the box, I am perfectly okay with not having to fiddle with anything. Being the newb I am, haha.

It's 100% chip dependent, every single chip will behave slightly differently, the reviewed chip just happen to already be at its max.

What this should bring attention to though is this: just because your benchmark score increased does not mean your gpu's overclocking is stable or is actually giving you a better experience. I know I can kick my core up to +150 and memory to +500 and see big benchmark gains, with no artifacts, but my frame times go crazy.

So people in here who are just running benchmarks and going off them for testing stability, make sure your also comparing your actual frame times. You might not be as stable as you think.

Edit: I wonder if the lack of artifacts but crazy frame times means there is some sort of error correction going on? Anyone have any insight into this?


This is the first GPU I've ever had perform this way. Most would perform show artifacts when they hit there limits, these seem to develop frame timing issues way before artifacting.
 

Yaari

Member
I didnt know the Zotac AMP was louder than most, leaning towards the Aorus Xtreme then. Hopefully I can pick that up here soon.
 
It's 100% chip dependent, every single chip will behave slightly differently, the reviewed chip just happen to already be at its max.

What this should bring attention to though is this: just because your benchmark score increased does not mean your gpu's overclocking is stable or is actually giving you a better experience. I know I can kick my core up to +150 and memory to +500 and see big benchmark gains, with no artifacts, but my frame times go crazy.

So people in here who are just running benchmarks and going off them for testing stability, make sure your also comparing your actual frame times. You might not be as stable as you think.

Edit: I wonder if the lack of artifacts but crazy frame times means there is some sort of error correction going on? Anyone have any insight into this?


This is the first GPU I've ever had perform this way. Most would perform show artifacts when they hit there limits, these seem to develop frame timing issues way before artifacting.

Strange. What temps are you seeing? Are you on stock voltage?
 

Pagusas

Elden Member
Strange. What temps are you seeing? Are you on stock voltage?

When my media center exhaust fans are on, max temp is 73c at 60% fan, Voltage at +100%, Power at 115%.

When the media center exhaust fans are off, max temps hit 78c at 66% fan speeds (so I tend to leave my media centers fans off, lol)

By media center I mean this is an HTPC and sits inside a TV stand that has built in venting systems to keep components inside cool.


Its not just me who sees this, that review saw the exact same behavior, and I bet most users here would see the same if they examined their frame times.
 
Sounds good. IIRC, the Aorus has 12+2 power phases, etc., which should definitely help. Tweaking the voltage on my FE doesn't do it any good, so I treated his card the same. My mistake! :p

I appreciate the input! =)

Update: My dad's Aorus won't take any extra volts without crashing seconds into any benchmarks and/or games. I've tried both the stock, and F4 Beta bios. It looks like a max boost of 1,949Mhz is really all it can do, so I'll adjust the fan curve to make sure it spends as much time as possible at that speed.


Thanks for linking that. 1,949Mhz max boost clock doesn't seem out of the ordinary for some of these cards. I feel much better now. :p
 

paskowitz

Member
Gamers Nexus posted a review of MSI Gaming X.

https://www.youtube.com/watch?v=ZxMuQu-m1iE&t=680s

TLDR - excellent card. What's interesting is that OCing the card even further than the already stock OC gives you higher frametimes.

1080ti-gaming-x-grw-frametimes-1080p.png




It looks like these AIB cards is already pushing the OC headroom as much as they can on Pascal and it got me thinking of lowering the OC on my card as well.

$1 says this won't happen on a Hybrid or custom loop. Certainly an improvement over reference though. Pretty much max out of the box, which is great for the vast majority of people.
 

The Dude

Member
So tinkered a bit and leveled off at running ME at 2025, 60 locked and Temps 61 to 64ish, I can barely hear the fans.

Really impressed with the Aorus at this point.

This is at 4k, looks pretty awesome.
 

Sky Chief

Member
Been playing a bunch of VR with my Strix but going to take a little break. What are the best and most gorgeous traditional games out right now to throw my Strix and 3440x1440 ultrawidescreen at?
 

Kadey

Mrs. Harvey
Team Strix. So glad I got this over the Gaming X. It would look so out of place with my build. Would get a Galax white version in a heartbeat if it came out.
 
Just a heads up: I got very different results OC'ing with EVGA Precision and Afterburner.

Yesterday I ran EVGA Precision with +130 and +495. Firestrike completed with a graphics score of 28,431 (http://www.3dmark.com/fs/12306030).

Today I tried to run it at the same clocks and Firestrike wouldn't even load. In fact, with stock clocks, even if I simply upped power limit to 120%, there were immediate artifacts during the loading screen. Da fuq.

Install Afterburner and run +130 and +500. Not only did Firestrike finish with no artifacts, I got a graphics score of 29,400 (http://www.3dmark.com/fs/12331678). Almost 1000-point swing just switching applications.

In summation: goodbye EVGA Precision.

This card is going on water so I'm encouraged by what I'm seeing on air.
 

longdi

Banned
You need to in increase the voltage if you want to get anywhere. Unlock voltage controls and put it to 100% along with the increased power limit.
I experienced this with my FE card too. Put 100% voltage and reduce the +core clocks, is more stable than pushing core clocks.

Dont worry, 100% voltage will only go as high as Nvidia lock down 1.093v, so i won't burn my GPU.

So guys try to increase voltage before core clocks and let Boost3.0 take over
 

Pagusas

Elden Member

Sounds about right. 1y/8760h at full capacity. So, if you're running benchmarks/playing games at 100% GPU utilization with an extra 100% overvoltage allowance for 24 per day, the average lifespan of the chip will be about one year. Bring it down to 8h per day and you arrive at the 3-year warranty offered by most board partners these days.
 

longdi

Banned

Hmm, is he referring to hardware mod which will bypass Nvidia driver limit?

I back off to +25% now, which gives me a nice solid 1.062v on load clocks, where 0% will drop to 1.05v while +100% is 1.093v. All three voltages are within Pascal defined range afaik.

Oh well, if mine dies within 18months, i wont mind getting it replaced with Volta 1180. :D
 
One more reason to buy from a company that gives you a long warrenty like MSI or evga.

And isn't that in reference to the stock pcb? All the custom pcb variations up the quality of the voltage and power systems, likely to support the demands of overclocking wirhout getting tons of 1-2 year warrenty claims.

No idea how/if non reference boards change it

Hmm, is he referring to hardware mod which will bypass Nvidia driver limit?

I back off to +25% now, which gives me a nice solid 1.062v on load clocks, where 0% will drop to 1.05v while +100% is 1.093v. All three voltages are within Pascal defined range afaik.

Oh well, if mine dies within 18months, i wont mind getting it replaced with Volta 1180. :D

No hardware mod, just the standard voltage slider you can unlock in oc software
 

Weevilone

Member
If the slider truly allowed "1 year" then we'd see a lot more dead GPUs. They've built fudge factor in there for sure.

One more reason to buy from a company that gives you a long warrenty like MSI or evga.

And isn't that in reference to the stock pcb? All the custom pcb variations up the quality of the voltage and power systems, likely to support the demands of overclocking wirhout getting tons of 1-2 year warrenty claims.

They are talking about overvolting the GPU die, so the power delivery shouldn't matter. I mean it's great that it's cleaner power or whatever, but you are still pumping excess voltage to the die.
 
If the slider truly allowed "1 year" then we'd see a lot more dead GPUs. They've built fudge factor in there for sure.



They are talking about overvolting the GPU die, so the power delivery shouldn't matter. I mean it's great that it's cleaner power or whatever, but you are still pumping excess voltage to the die.

He should've clarified it's one year in terms of hours of use at full load/max allowed voltage. An idling chip at a lower power state won't suffer the same effects. As I mentioned before, that type of usage for 8h every day will net you about 3 years. In any case, you're right about them allowing for all sorts of variances when arriving at the "1 year" average. You're also correct to say that more efficient power delivery has a negligible effect on accelerated degradation due to electromigration, etc.

I'd wager most enthusiasts (who will tend to overvolt their GPUs) will have upgraded to a new card long before degradation becomes a problem for them. Those buying >1-year-old cards secondhand had better be careful, though.
 

Kevin

Member
Anyone else having issues with Doom on their 1080ti? As soon as I start the game it immediately crashes. I even tried to fully reinstall it and still nothing. All other games still seem to be working (upgraded from a 980 ti)

It crashes upon entering the game menu, a white error box allowing you to file a crash report comes up (but you can't actually get to it) and that's it.

EDIT: I am hearing the Windows 10 Creator's Update may have broken the game? Anyone else experience this issue? Know a fix? I tried to do a clean install with the video driver with no luck.
 

Pagusas

Elden Member
Anyone else having issues with Doom on their 1080ti? As soon as I start the game it immediately crashes. I even tried to fully reinstall it and still nothing. All other games still seem to be working (upgraded from a 980 ti)

It crashes upon entering the game menu, a white error box allowing you to file a crash report comes up (but you can't actually get to it) and that's it.

EDIT: I am hearing the Windows 10 Creator's Update may have broken the game? Anyone else experience this issue? Know a fix? I tried to do a clean install with the video driver with no luck.

its broken alot of other games, I'd either roll back or wait for nvidia to issue new drivers.
 
Top Bottom