Review: Red Dead Redemption 2: PC graphics benchmark analysis

nonsense and foolishness.

there is nothing to fix.

We have been spoilt by a minimally-shifting feature set for too long and the tech community has gotten ignorant to what actually happens when new graphical features are introduced with a new game.

 

No it is not insane, NVIDIA is doing it, to force us to upgrade from Pascal to Turing and makes tons of money!!
I remember when Kepler GPU's were running very bad in Witcher 3 comparing to Maxwell. (780ti was same performance as 960!) and there was a post on NVIDIA forum talking about this issue. NVIDIA listened to us and released a new driver for Kepler GPU's for witcher 3 that fixed Kepler cards performance in witcher 3 and gained 10-15fps, I had GTX 780 and tested it.

 

It happened in witcher 3 when NVIDIA released a driver for Kepler cards that fixed the performance.
Anyway, I will switch to AMD and I don't trust NVIDIA anymore.

 

Irrelevant, come up with a new argument but get yourself clued in on the vulkan capabilities improved upon with Turing and Vega/RDNA capabilities first.

Pascal has architectural limitations that mean Vulkan doesn't work as well as it can on Turing.

DX9 and 11 are not Vulkan and D3D12, there is not alot in the way of Driver optimizations you can introduce which aren't actually optional features of the API in the first place.

Turing VK features such as
FP16 math
Turing Mesh acceleration (VK / DX12) 3x faster than Pascal
Concurrent Execute
Variable Rate Shading

 

Kepler was two gens old at the time and had abysmal sp32 compute compared to Maxwell and latter. Kepler was not a priority for optimization therefore it was put off so that the newer GPU architectures got optimizations first. At least Nvidia addressed optimizing Kepler, AMD on the otherhand dropped Terascale optimizations immediately following GCN1 release and bug fixes stopped 6 months later.

 

These posts are honestly cancer.

Not only do Hilbert's tests not show what your saying but in other benchmarks, for example GN, they have the 1080Ti 10-14% faster than 2060 Super - techpowerup has the 1080Ti ~18% faster than a RTX2060. So how in the world are you saying your 1080Ti is slower than a 2060 non-super?

Either your computer is completely screwed up and you should consider setting it on fire or you're intentionally lying to propagate fanboy nonsense.

The frustrating part about all this is I know like 2 years from now I'm going to see posts using this game as an example of Nvidia downgrading old cards. @Glottiz These posts are perfect examples as to why you wait for overwhelming evidence before you jump on the bandwagon. It's just another example in many where people are jumping to conclusions and now when more information comes out the initial results aren't even indicative of the overall games performance.

 

They're upset because the 2060 with its improved feature set in this quite heavy title isn't as far behind the 1080ti as say.... COD:MW was.

 

Yeah I mean the architecture differences (of which I think Turing is the largest departure in architecture since Kepler) are going to factor a lot into these performance differences.. but in this case no other site even shows what people here are claiming. Hilbert's testing area, for whatever reason, has the 1080Ti/Super cards closest. There are multiple other sites that reviewed performance now and all them have the 1080Ti outperforming the 2060 and even the 2060 Super by a significant margin. I don't know why the settings/area Hilbert tested showed these results but they clearly aren't indicative of the relative performance of the cards throughout the game. I think he tested the snow area? Perhaps some shader there just runs better on Turing? Not sure but Techpowerup/Hardware Unboxed/Gamers Nexus all show radically different results in respect to the 1080Ti compared to the 2060 Super. I didn't look into any of the other cards relative comparisons.

In this guys case he's saying his 1080Ti is worse than 2060.. which even Hilberts test don't show despite them already being an outlier in favor for Turing.

Maybe I'm reading these posts wrong but it sounds like you guys are both arguing that the game looks good and runs well but you're both being hostile to each other? Lol

 

I imagine that the specific snow scene is heavy in "near" volumetric effects to achieve the billowy air movements that come up.

I'm glad this game released, and im glad it has the heavy taxing max settings that it has, it gives both current vendors an idea of where to tune Ampere/Arcturus and aim for besting this beast.

 

I'd say there's a huge difference between not doing further optimizations, and actually gimping the performance by deliberately writing bad code. According to the latter camp, release drivers from Kepler should outperform the current drivers.
And Nvidia did address the Kepler issue, though it took 1-2 weeks for them to do so.

I still have Kepler. They've been updated to support WDDM 2.6 and Shader Model 6.4. How much further room there is for Kepler-optimization in new games is up for debate and can only be answered by Nvidia, but they have at least not stopped it from supporting new DirectX features.

This test is two years old now but I still can't find a more recent one. All cards got an improvement with later drivers, sometimes marginal or minimal, sometimes a clear improvement of 5-10 FPS.
https://www.computerbase.de/2017-01/geforce-treiber-test/2/#diagramm-f1-2015

 

Hi Hilbert, please look at this video:

https://youtu.be/2UPXld_xlDs

Async compute is disabled by default and delivers better frametimes on AMD cards at least, needs to manually change the config file to enable it.
Very strange omission in the graphics menu, i wonder why.

 

I'm just wondering why my gtx 1060 is in recommended category/slot pre-release?

was/is 720p now the recommended resolution....?....to con me into buying it?

 

This is why I wait now. Watchdogs got me and that was the last time.

 

Again be aware it's a vocal minority (always the loudest) complaining about issues. For me, and many (probably most) others, the game runs amazingly well for how good it looks, even with most settings on high ultra.

 

So according to Digitalfoundry, the settings on consoles are a mix of low to high, but some settings are lower than lowest


So, despite what all elite graphics programmers on the forums say, there's no bad optimization at work. Just a port that in general goes above and beyond the console visuals.

 

People seem to forget that the 10 series cards launched all the way back in May 2016, just like previous generations of cards from both vendors they can’t stay relevant forever, there’s always a point when their weaknesses vs newer architectures starts to show in newer titles.

It’s not Nvidia gimping performance, they don’t develop the game, it’s more like developers actually taking advantage of what the newer hardware can do.

It sucks if you’re on older cards but this isn’t new, you should be happy that the 10 series has lasted as long as it has, older generations weren’t always so lucky.

 

You remember when 1080ti was 2080 performance level?

 

im amazed by the Polaris performance. Rich even said 580 is even running better than 1660.

https://i.**********/pV1SNtfS/RZOMO0Q.png

 

Was expecting a nightmare with this after reading all the stuff online, but haven't had a single problem with it myself. No launcher issues and it doesn't appear to be as much as a performance hog as has been made out. Running it a steady 60 without fluctuating 3440 ultra wide/Medium settings and it's definitely one of the best looking games I've ever seen.

 

I'd say that Pascal is gimped by good 10-20%. The question is who is to blame here? Is it NVIDIA? Is it R*? Personally, I'd blame R*. They probably didn't have "time" to optimize for specific architectures on PC, they were just fixing visual bugs. It's kinda sad, considering Pascal (and 1080TI specifically) is the most widespread discrete PC GPU architecture...