My experience whit 1080 TI to 2080TI

[Jordan Dion]

Hello guys!

So many of you are probably like me and are on the fence for buying the new high end and overpriced new nvidia's card..

So i finally made the move!

Here My spec :
Asus maximus x hero
CPU : intel 8700k
16gb ram
Ssd m2 + regular ssd
Oculus rift
Oled 55po 4k
Older card : strix 1080 Ti
New one : EVGA GeForce RTX 2080 Ti XC ULTRA

Playing 95% of time in Rift.

So after many test :

Performance gain difference
ACC : 30-40 % - SS 200 (all option ultra) run ****in smooth now

AC : - 5-20 % ????? - SS 140 big frame drop whit 2080ti . I dont understand why?

Iracing : 10% - SS 130 (but the new update whit dynamic weather is a really fps eater for me)the Iracing graphic engine is old but really clear and smooth. A little moment whit frame drop sometime on particular place on some tracks (sun reflection)...

Just as exemple :
BFV : 20-25 fps (all maximum whit rtx at medium)
By the way... Rtx is nice but.... Dont waste your money for this... Its cool but it is not a big deal.

Important fact :
My strix 1080 ti had 2 display port and 2 hdmi port. So i was ok whit this set up ...

But the evga 2080ti ultra have only one HDMI port... So dont forget to look at this because it is really a mess when you try to get hdr trough display port...

Good to know : Oculus is working fine through an hdmi>display port

So good look in your reflexion and your shopping!

 

[Durge Driven]

Nice card

amazes me studios have game bench tests, access to all hardware and don't give any info other then minimum and recommended

They should test at least test 3 hardware configs in popular resolutions and VR and supply full bench results

Benching sites barely use simulations

 

[RasmusP]

Nice post and good overview but you forgot to tell us what cpu you have

 

[luxxx797]

SS 200% all ultra? this hotlapping maybe....

 

[Jordan Dion]

Nice post and good overview but you forgot to tell us what cpu you have

Thanks!
I added my cpu in the list!

 

[Jordan Dion]

Nice card

amazes me studios have game bench tests, access to all hardware and don't give any info other then minimum and recommended

They should test at least test 3 hardware configs in popular resolutions and VR and supply full bench results

Benching sites barely use simulations

This is why i did this Post!
Hope this can help someone!

 

[Jordan Dion]

SS 200% all ultra? this hotlapping maybe....

This is working well whit 20 cars whit sunny day.
I didn't try whit rain at the moment

 

[davewilliams000]

Sorry... what's 'SS' stand for?

I've just gone from a 1070 to a 2080 after getting a good deal on a new one on ebay and am still fine tuning my settings at the moment

I might be a little CPU bound (especially in AC/ACC with a full grid of AI) as I only have an i7-4790K

P.S. if you do any recording/streaming then have you tried the new beta version of OBS that's just came out... they've improved the rendering performance (reduced the overhead) on all nvidia cards, but it works especially well on RTX hardware as there is improvements in quality for RTX cards only - most noticably any text is now a lot sharper as if it's been rendered using x264 software encoders!

 

[RasmusP]

Sorry... what's 'SS' stand for?

SuperSampling
So things are rendered in a higher resolution and then filtered down to the display's resolution.
Here's a quick example of how it works:
https://www.spec.org/gwpg/publish/fsaa_in_vp10.html
This would be the "real geometry", in infinite high resolution:

Without super sampling (or any anti aliasing) it looks like this:

And this is how it works:
The "steps" due to the too low resolution become "blurred" in relation to what the calculation in a higher resolution comes down to.

Summary:

Btw, "post process" anti aliasing like FXAA, SMAA, TAA etc are not rendering the actual geometry in a higher resolution and then blur the edges but instead just look at the finished picture that would be displayed, look at contrast changes and blur them.
Which is why they all look blurry and when forced via reshade or sweet fx, blur HUD texts etc too.
It costs way less performance though compared to true anti aliasing!

"TAA" - Temporal Anti Aliasing like used in Assetto Corsa Competizione is post processing too, and therefore looks blurry. It also takes future frames, in example the directions of movements into consideration, which is why you'll see some ghosting.

 

[Jordan Dion]

SuperSampling
So things are rendered in a higher resolution and then filtered down to the display's resolution.
Here's a quick example of how it works:
https://www.spec.org/gwpg/publish/fsaa_in_vp10.html
This would be the "real geometry", in infinite high resolution:

Without super sampling (or any anti aliasing) it looks like this:

And this is how it works:
The "steps" due to the too low resolution become "blurred" in relation to what the calculation in a higher resolution comes down to.

Summary:

Btw, "post process" anti aliasing like FXAA, SMAA, TAA etc are not rendering the actual geometry in a higher resolution and then blur the edges but instead just look at the finished picture that would be displayed, look at contrast changes and blur them.
Which is why they all look blurry and when forced via reshade or sweet fx, blur HUD texts etc too.
It costs way less performance though compared to true anti aliasing!

"TAA" - Temporal Anti Aliasing like used in Assetto Corsa Competizione is post processing too, and therefore looks blurry. It also takes future frames, in example the directions of movements into consideration, which is why you'll see some ghosting.

Nice article man!

 

[davewilliams000]

SuperSampling
So things are rendered in a higher resolution and then filtered down to the display's resolution.
Here's a quick example of how it works:
https://www.spec.org/gwpg/publish/fsaa_in_vp10.html
This would be the "real geometry", in infinite high resolution:

Without super sampling (or any anti aliasing) it looks like this:

And this is how it works:
The "steps" due to the too low resolution become "blurred" in relation to what the calculation in a higher resolution comes down to.

Summary:

Btw, "post process" anti aliasing like FXAA, SMAA, TAA etc are not rendering the actual geometry in a higher resolution and then blur the edges but instead just look at the finished picture that would be displayed, look at contrast changes and blur them.
Which is why they all look blurry and when forced via reshade or sweet fx, blur HUD texts etc too.
It costs way less performance though compared to true anti aliasing!

"TAA" - Temporal Anti Aliasing like used in Assetto Corsa Competizione is post processing too, and therefore looks blurry. It also takes future frames, in example the directions of movements into consideration, which is why you'll see some ghosting.

Nice article man!

Awesome thanks! I'd heard the term supersampling and roughly got the concept, but didn't fully understand it until now.

So are you saying that in ACC you set the 'resolution scale' to 200, or is it done in the nvidia control panel? and also does this mean you've turned of AA alltogether in ACC if your running SS?

Would you mind posting a screenshot of your graphics settings page for ACC? Although of course I don't have the Ti version of the 2080 so won't be able to run it quite as well.

 

[RasmusP]

Awesome thanks! I'd heard the term supersampling and roughly got the concept, but didn't fully understand it until now.

So are you saying that in ACC you set the 'resolution scale' to 200, or is it done in the nvidia control panel? and also does this mean you've turned of AA alltogether in ACC if your running SS?

Would you mind posting a screenshot of your graphics settings page for ACC? Although of course I don't have the Ti version of the 2080 so won't be able to run it quite as well.

Only got a 1070 for 3440x1440, 60 fps locked, lowered resolution scale to 83%.

Now that you know the difference between "true AA" and "post processed AA", here's a little overview:
MSAA: only renders "real objects" in a higher resolution, not transparent stuff.
SSAA/SS/FSAA: true supersampling, renders everything in a higher resolution. Kills performance.
SGSSAA: Sparse-Grid Supersampling, has a different filter algorithm. Sometimes blurry, mostly awesome

All of them are not really "doable" with newer rendering processes (so you'll see MSAA in most DX9 titles but barely for DX11 titles!).
A workaround is to just increase the resolution, instead of applying it to the actual geometry etc, you just tell the whole renderer to do his job in a higher resolution.

For this there are different "positions" where you can select it:
- inGame resolution scale sliders (with whatever filtering method!)
- DSR/VSR in the graphics card drivers and then selecting higher resolutions ingame, GPU driver will filter it down to your Monitor.
- VR software settings: Oculus software, steam VR settings etc etc
- Game VR settings: some games have additional settings when starting their VR-Mode.

This leads to the problem that depending on how you combine them, the filtering will mix/mash, the resolution multipliers will "stack" or whatever. That's why people suggest and use different combinations.
The root problems are:
- FXAA/SMAA/CMAA/TAA will look blurry and you need to sharpen them. This leads to artifacts and jaggies. So you go with TAA+120% resolution scale+sharpening. Still better than no AA and still better performance than similar image quality with 4x MSAA or 200% resolution scale for example

- VR headsets have a too low resolution to look good. So you need to supersample the crap out of them

Now "stacking" examples:
nvidia DSR from 4k down to 1080p. You select 4k in the game, GPU filters it down to your 1080p display.
Now you go with 200% resolution scale in the game, which makes it 4k*200%!

So you think: ah, that's 8k then? And I'd go with "yeah more or less".
Nobody knows if "200%" means:
(1920x1080)*2 = 4147200 pixels -> 16:9 -> 2715x1527
OR:
1920*2 x 1080*2 = 8294400 pixels -> 16:9 -> 3840x2160

And this is not standardized as far as I know so you'll get a huge mess and chaos when choosing "150%" in DSR or ingame or in the VR software etc.

Summary:
xx% resolution scale/supersampling might be hugely different depending where you set it and also the filtering algorithms are different. So depending on the combinations you'll get different image qualities/looks and different performance hits.

 

[Durge Driven]

Also went 1070 to 2080

10% ( $130AU) off with sale at my store ( they opened a new store and gave 10% off everything ! )
2 games I registered for ( BF5 and Anthem ) cost $130AU on instant gaming or $200AU on Steam

Effectively $260-$330AU worth of savings

So don't forget 2080Ti or 2080 you get both titles before they all gone
https://au.aorus.com/event-detail.php?i=896

 

[Turk]

Here My spec :
Asus maximus x hero
CPU : intel 8700k
16gb ram

I have a similar setup but I have a 1070 GC. It must be a big bottleneck because I can't run ACC above medium settings with AI.

Don't think I can really justify the cost of a 2080 any time soon

 

[Rob Fitness]

thanks for sharing your results.
im praying my 1080ti never dies as the cost for a 2080ti over what i paid for the 1080ti is a bit rediculous when looking at the performance increases or lack thereof really for the huge pricing differences

 

[Terry Rock]

I have a similar setup but I have a 1070 GC. It must be a big bottleneck because I can't run ACC above medium settings with AI.
Don't think I can really justify the cost of a 2080 any time soon

You guys really need to dig into those systems to find out why you can't run ACC with more than medium setting.
I ran a much, much less capable CPU than the 8700...with a GTX-1070 non-Ti...at well above medium settings.
It ran very well....even at night and in rain.
Here's a video to prove it.
You should not be encountering those issues with a better CPU.

 

[Turk]

You guys really need to dig into those systems to find out why you can't run ACC with more than medium setting.
I ran a much, much less capable CPU than the 8700...with a GTX-1070 non-Ti...at well above medium settings.
It ran very well....even at night and in rain.
Here's a video to prove it.
You should not be encountering those issues with a better CPU.

VR.

 

[Terry Rock]

VR.

I also ran in VR on the Oculus CV1 with higher than medium settings.
Again....very well.

 

[Turk]

I must look up a guide, I only set up at the weekend so my time is very limited.

 

[RobertR1]

OC the crap out of the 2080ti. the PCB and VRM's are bullet proof so go for it.