• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Alan Wake 2 Performance Benchmark

Default script from W1zz
That's not the one I used for testing. That's the one you get when you start the game for the first time. I've clarified that in the text
 
It's quite evident that, although these VRAM numbers were measured, they aren't strictly necessary for great performance, it seems it can be overshot by 2GB+
On cards with good bandwidth, most likely. That also relates to the 3080's place in the charts.

That's because they're lower powered GPUs - Not a VRAM limitation
Every GPU is about balance. The 3080 has a mere 10GB, but covers that up nicely with a royal amount of bandwidth, comparatively to Ada it has a LOT more relative to its core power.
This is how that GPU can come side by side with the 4070ti 12GB (below 4K) that has lower bandwidth, and clearly a stronger core, but can't properly use all of its core perf because it combines insufficient VRAM with lower bandwidth. Exactly as one could have predicted looking at the relative hardware of Ada's 4070ti vs its 4070, or the 4080 that makes a notable jump since it covers the VRAM requirement and can therefore use all of its core power.

Another aspect that cannot be understated is that Alan Wake 2 is clearly an Nvidia-first title so it'll be heavily optimized/tailored to cover a VRAM deficit, something implicitly noted by the position of AMD cards relative to Nvidia's, even if the gap isn't huge with numbers elsewhere, its there. For those (not you) that have trouble with comprehension: I'm not saying Nvidia crippled AMD cards here, it just made sure its own don't run into issues.

10GB 3080 beats 16GB 7800XT & 6900XT at 1440p/4K (rasterization) in BOTH Average fps and minimum fps

10GB 3080 beats 20GB 7900XT in 1080p/1440p (Ray Tracing) in BOTH Average fps and minimum fps

Anything above that is useless....... Even 7900XT runs below 30fps when using path tracing.


IN OTHER WORDS, 10GB is not an issue....

I love how you guys twist things up..... nvidia is clearly doing better in this game, even if has lower VRAM......at any realistic settings.



Ha ?? Did you even look at benchmarks and fps ????!


10GB 3080 beats 16GB 7800XT & 6900XT at 1440p/4K (rasterization) in BOTH Average fps and minimum fps

10GB 3080 beats 20GB 7900XT in 1080p/1440p (Ray Tracing) in BOTH Average fps and minimum fps

Path tracing is pointless... The fps is so poor even on top AMD GPU

8GB 3060 Ti is beating 12GB 6700XT at all resolution (rasterization) in both average fps and minimum

Also, 8GB 3060 Ti beats 6700XT at 1080p with RT but both cards had poor fps.... Any higher resolution is slide show in both cards



First of all, people will not be running this game on extreme setting with path tracing on midranged GPU anyway....

Second of all, 8GB nvidia GPU is performing well in this game compared to 12GB AMD GPU with same generation.... For example 3060 Ti beats 6700XT 12GB in rasterization (all resolutions)........ and in 1080p with normal RT (not path tracing)...........Anything above that runs below 20fps on 6700XT.... It does not matter who wins when both cards are below 20fps or 30fps because nobody will play at those setting anyway.

Are you guys even look at the benchmarks ???!
You made a discussion on VRAM a discussion between different GPU architectures & brands. I'm looking exclusively at the order of cards on the Nvidia side.

This isn't a pissing contest or camp battle, its a performance analysis.

And if you are really still making a case for sub 12GB cards... all I can say is, you do you, I've moved on ;) The requirements don't lie.
 
Last edited:
Charts need to be updated indicating the type of upscaling used and in general that upscaling is on. Already seen four people in the comments miss that detail because it's only in the test setup page. It's also important for when those charts eventually get used in other comparisons.
Added a note in front of every chart. Not sure why people have such a hard time understanding

Really? :eek: That's scummy AF if true! :shadedshu:

Edit: The review also says that you can set the render resolution to your in-game resolution which will basically run without upscaling.
It will still run with DLSS/FSR algorithms
 
I don’t see a big difference in the graphics between the first part. I don’t know why the game is so praised, I see a naked king. The first part ran at a speed above 60fps at maximum settings in FHD on gtx480!
 
It will still run with DLSS/FSR algorithms
I see. Does it "upscale" from 1080p to 1080p for example when the render and output resolution match? If it's a quick "100%=100%" calculation, then I suspect the visual and performance impacts are minimal, if any.
 
I see. Does it "upscale" from 1080p to 1080p for example when the render and output resolution match? If it's a quick "100%=100%" calculation, then I suspect the visual and performance impacts are minimal, if any.
100% = 100% is definitely not a null-op in DLSS/FSR

NVIDIA calls this "DLAA". And that definitely runs a lot of magic algorithms to solve the temporal anti-aliasing problem
 
Man Native with no fancy tech looks nice but PT+DLSS+RR+FG sure elevate Alan Wake II to another level in term visual/performance

4K Native vs 4K DLSS.B+PT+FG+RR
 
Does it "upscale" from 1080p to 1080p for example when the render and output resolution match?
It just substitutes native TAA with DLAA/AMD TAA, at least from my understanding. Seems to me we will have more and more games without any TAA other than DLAA/AMD or Intel equivalents.

Performance is a little bit trashed but it usually looks way better than any other TAA.
 
Man Native with no fancy tech looks nice but PT+DLSS+RR+FG sure elevate Alan Wake II to another level in term visual/performance

4K Native vs 4K DLSS.B+PT+FG+RR
Because you have... um... better reflections, I guess? :wtf: That's so many abbreviations that I can't even process which one means what, yet, the visible difference is negligible.

100% = 100% is definitely not a null-op in DLSS/FSR

NVIDIA calls this "DLAA". And that definitely runs a lot of magic algorithms to solve the temporal anti-aliasing problem
It just substitutes native TAA with DLAA/AMD TAA, at least from my understanding. Seems to me we will have more and more games without any TAA other than DLAA/AMD or Intel equivalents.

Performance is a little bit trashed but it usually looks way better than any other TAA.
Ah, now I understand. Thanks both for the explanation. :)
 
I don't know what mushrooms you need to see a new level of visuals there honestly.
Things off screen being rendered correctly in reflections?
 
That's so many abbreviations that I can't even process which one means what,
PT = Path Tracing.
DLSS = Deep Learning Super Sampling.
RR = Ray Reconstruction.
FG = Frame Generation.
 
Things off screen being rendered correctly in reflections?

Well maybe there are people who just refuse to acknowledge something that are so obvious to others :D

Native vs PT
non rt.png
rt.png
 
Page 5 has a typo showing ridiculous numbers with Ray Tracing, much lower than Path Tracing in two of three of the comparison quality screenshots as you can see here:

YycoXmk.jpg
Retested and fixed
 
Well maybe there are people who just refuse to acknowledge something that are so obvious to others :D

Native vs PT
View attachment 319093View attachment 319094
The difference is obvious. Its also way too pronounced, every reflective surface is polished to perfection, the glass is stainless, and yet, people just had dinner. Contrast is higher, too.

Still this is better than Cyberpunk, where reflective surfaces are utterly ridiculous in RT. But a 'new level' of visuals... I guess if you really want to call it that, it's true.
 
The difference is obvious. Its also way too pronounced, every reflective surface is polished to perfection, the glass is stainless, and yet, people just had dinner. Contrast is higher, too.

Still this is better than Cyberpunk, where reflective surfaces are utterly ridiculous in RT. But a 'new level' of visuals... I guess if you really want to call it that, it's true.
Although I agree with you that PT in this game is not a transformative experience, we are still just looking at screenshots. I think the main issue RT is trying to solve isn't better image quality, cause that can be achieved with tradiniotal raster as well. What RT can do is not have reflections and shadows disappear right in front of you while you are walking around or just moving the camera, which is a huge issue with raster.

It doesn't help the discussion that both parties, but let's not kid ourselves usually it's the people with amd cards, just poop on everything with nvidia logo on it. They dislike both dlss and rt - even when their implementation is great. A little bit more impartiality would go a long way I think. M
 
The difference is obvious. Its also way too pronounced, every reflective surface is polished to perfection, the glass is stainless, and yet, people just had dinner. Contrast is higher, too.

Still this is better than Cyberpunk, where reflective surfaces are utterly ridiculous in RT. But a 'new level' of visuals... I guess if you really want to call it that, it's true.
"Contrast" is higher because you have accurate bounce lighting, I've yet to see any game that implements tables getting greasy. You can see the red leather of the couch lighting the window frame next to it.

Had a wonderful photo from my office where an orange bag left on the mesh chair was lit by light passing through the window which resulted in orange bounce light on chair, table and wall behind it. There's no way you could pull something like this with baked lighting.
 
I've yet to see any game that implements tables getting greasy.
It would be quite a feat, though. The biggest problem within game graphics these days, imo, is that everything looks clean and shiny. Even dirty things, and non-shiny things like wood, fabric, or skin.
 
What's reason that 5700XT is slower than 1660 Ti at 1080p? Does 1660ti support DX 12_2?
 
What's reason that 5700XT is slower than 1660 Ti at 1080p?
One (or multiple) of these:
• Mesh shaders (unsupported by GTX 10xx, RX 5xxx and older GPUs);
• Incomplete or imperfect DX12 implementation in the AMD GPU;
• Game devs not trying to adapt the game for the RDNA1 architecture since it's "blacklisted." GTX 1660 Ti belongs to Turing which is exactly the same generation as RTX 2060, the minimum requirements GPU, minus a couple percent raw performance, minus RT and minus tensor cores.
• "Just because."
 
One (or multiple) of these:
• Mesh shaders (unsupported by GTX 10xx, RX 5xxx and older GPUs);
• Incomplete or imperfect DX12 implementation in the AMD GPU;
• Game devs not trying to adapt the game for the RDNA1 architecture since it's "blacklisted." GTX 1660 Ti belongs to Turing which is exactly the same generation as RTX 2060, the minimum requirements GPU, minus a couple percent raw performance, minus RT and minus tensor cores.
• "Just because."
also I think the new driver wasn't out yet
 
also I think the new driver wasn't out yet
The most massive driver caused improvement I've seen was about 6 or 8 percent which would definitely count as a boost but it's not enough to outplay the 1660 Ti. New drivers usually help games crash less and have more stable performance (Phantom Liberty had like 60 FPS before GRD and has about 62 FPS with GRD so it's not a big deal, yet the game stopped crashing in Dogtown which is massive).
 
right now, maybe we can waiting, .../rtx 4080 ti 24gb on next year........
No, RTX 4090 serie(s) is(are) required for 24GB VRAM!
 
Back
Top