Shadow of the Tomb Raider: RTX and DLSS Update
Click here to post a comment for Shadow of the Tomb Raider: RTX and DLSS Update on our message forum
Dj_ALeX
My dear friend even if someone tries super duper hard he would never produce those numbers in this game with "specific hardware", its impossible. The difference that my numbers have with gurus is huge! You cannot gain that much even after tweaking/updating everything to MAX! I know this game has some inconsistencies but to that degree its simply out of question! And all that with gurus test system being much better than mine for sure! Finally it is always better hearing that you're not alone so.. i'm glad others have seen this too!
Dragam1337
Maxwell scales alot with gpu core clocks - hillbert uses a founders edition 980 with stock clocks, so if you are running a custom 980, which is OC'ed, you could easily get 20% higher performance from your gpu 🙂
Dj_ALeX
Yeah thats true Maxwell is an excellent overclocker and you gain a lot out of it but despite that the difference that i have is pretty huge..!
I was wondering what cards were used and i made an assumption about the use of reference cards! In that case the results cannot be taken as legit! Who uses reference cards anyway..! Furthermore i suppose no optimization has been made anywhere in their whole test system.. oh well.. ! End of story!
EDIT: I just downclocked my card to reference design clocks @1220MHz and i hit 58fps. There is still 6 fps difference there and i have a worse test system than guru's..!
Dragam1337
I think the stock version also downclocks during heavy load, due to the very restrictive stock power target on the reference cards.
Dj_ALeX
The upcoming RT update has nothing to do with Maxwell & Pascal performance. What do you think is going to happen exactly?? Furthermore i don't need numbers any more from gurus gaming database knowing that their using reference cards that they cannot even maintain the original boost clock.. I have my own realistic numbers for both Maxwell & Pascal and soon for Turing too.
Denial
His argument is that Hilbert is benching with reference Nvidia cards - for Pascal reference Nvidia cards would downclock under sustained load after 5 minutes, with Maxwell the reference clocks were pretty low compared to AIB/User overclocks. With Maxwell, Nvidia completely stopped selling reference designs after several months. Probably paints a bit of a skewed window. Probably a skew that Nvidia deserves for designing such a crappy reference card then marking it up $100 (FE).
I'm not really sure how he expects Hilbert to solve this though. In a perfect world I guess you'd weigh the sales of ref vs AIB for each SKU and create some kind of target clock average or something but I don't know how you'd expect hilbert to do that. I personally prefer reference benching because I know on average how much more performance I can pull from reference on an OC for various cards. I look at like Hilbert's 980Ti numbers and I go "my 980Ti would be 20% faster than that" for example. Seems easy enough.
waltc3
I cannot see a bit of difference from these screen shots and my rx-590/480 8GB system @ 3840x2160--other than the fact that some sort of green post-processing filtering seems to have been applied to these images, imo. I routinely get >100fps peaks in this game--my average is ~62 fps, with the in-game benchmark. I surely hope those who paid $800-$1300 for these cards feel good about what they purchased. No further comment required.
waltc3
I've noticed the game itself is much better under D3d12--and of course the multi-gpu support only adds to that, for my setup.
Caesar
https://i.ibb.co/VSj1LmY/Untitled.jpg
Both are same.
Need a small correction.
--------------------------------------------
Nice review.
But not a big differrence with RTX ON and DLSS ON
Kaarme
I still think just to get those softer shadows with dedicated hardware, there shouldn't be any fps drop whatsoever. If it's just those shadows, not a bunch of water reflections, glass reflections, mirror polished metal reflections, etc. Back when I was modding an fps game a long, long time ago, we always had to be very careful with reflective surfaces because it basically meant rendering the current scene twice, so potentially you could halve the fps. You could only use it when you were in absolute control of the scene and knew there was fps to spare (not too much to render at the time). But that's clearly a different situation compared to making shadows with dedicated hardware.
Netherwind
I had a hard time seeing a color difference due to the lines being too thin.
Caesar
[youtube=BMVIMZZosnI]
toyo
Looks better with RTX off in those pictures.
MonstroMart
Watching this on a 4k Samsung QLED TV and I assume this was heavily compressed because man it doesn't look great at all. Lot of aliasing. Very blurry. Game itself looks stunning but that is definitely blurry.
MBTP
Am i missing something? Is not possible to use RTX ON + DLSS OFF?
Michael Chen
Like the post, I see no FPS difference between RTX high and ultra.
Using ultra gives more shadow, but sometimes that makes the scene harder to navigate...
For example, on many covered jungle scene, RTX ultra tends to make the whole scene darker. Guess the RTX was added after the scene was designed.
But I think that's more real / natural for such scenes.
Also, there's a new release of GRD driver today, that mentioned optimization on SOTTR.
Just see this while I'll be away from the rig for few days...
I'm quite curious what the new release will do on the game.
Hopefully someone will share their insight soon.
Mufflore
NVidia have dropped stereoscopic support.
The driver you have supports it but it wont be fixed because newer drivers wont have it.
Mufflore
Ah, k.
