Shadow of the Tomb Raider quick performance analysis

[Lathlaer]

Since I have the game from early access, I ran some tests both in DX11 and in DX12. My loose remarks:

 

1. Is about 20% more demanding than Rise of the Tomb Raider at least from official benchmarks.

 

RotTR: 4213x1764, everything on ultra, SMAA: about 112 average FPS with 1080ti in SLI

SotTR: same settings, avg. FPS 92

 

It should be noted though that SotTR has some graphics options that RotTR doesn't. It might very well be the difference.

 

2. SLI scales close to 100% in DX12. As of now, no profile for DX11 SLI exists. Changing SLI bits to those from Rise of the Tomb Raider or from Tomb Raider (2013) changes GPU usage to about 70% each but the FPS won't budge. Avg FPS is about 45-50 with the above settings.

 

3. DX12 is, for me at least, very unstable. Haven't tried the game itself yet but whenever I tried to change the AA method (while game is launched, not via options from load menu), the game crashed and sometimes even forced me to restart PC. Dunno if this is new driver issue or just DX12 in the game but from what has been reported during early DX12 days of the previous game, I'd say it's the DX12 mode. I hope real gameplay will be more stable.

 

If gameplay is stable then for SLI users I'd recommend DX12 (duh) and for everyone else maybe DX11 for stability.

 

BTW. no VXAO even in DX11 for some reason.

 

I may update this after trying the game itself but have to go out now ;-)

[Lathlaer]

OK, my game seems to be stable in DX12 - it's just the options whenever they need to adjust the screen (AA or resolution), for some reason it crashes. A minor inconvenience because those are set and forget.  Also I've seen opinions that it actually looks better in DX12 than in DX11. Can recommend DX12 without any problems.

 

All things considered, I am very pleased with how the game looks and plays. Vsually stunning, FPS so far actually even higher than the benchmark shows by about 5-10 and extremely good SLI support from day 1. 

 

WTB that kind of performance from every AAA game ;-)

 

 

[Rackooo]

Must agree apart from the ocassional stutter (caused by optimal performance mode in nvidia control panel, and fixed by setting max performance) the game performs exceptionaly well and the graphics are very nice. Very emersive and must say one of very few games that makes me wanna complete the story

[Frankenburger]

5 hours ago, Lathlaer said:

As of now, no profile for DX11 SLI exists. Changing SLI bits to those from Rise of the Tomb Raider or from Tomb Raider (2013) changes GPU usage to about 70% each but the FPS won't budge.

A user has found 0x080000F5 to work with DX11 and reportedly nets about a 50% improvement.

 

 

How does DX12 MGPU handle SMAAT2X and SMAA4X?

[Lathlaer]

2 minutes ago, Frankenburger said:

How does DX12 MGPU handle SMAAT2X and SMAA4X?

There you go:

 

Settings: resolution 4213x1764 via DSR, everything on Ultra/maxed, V-Sync off. Additional notes:

 

1. There are 3 scenes during benchmark. GPU usage varies depending on the scene.

2. There is a discrepancy between FPS shown by MSI Afterburner and ingame FPS counter during benchmark

 

Now the results

 

SMAA: avg. FPS 93. GPU usage in scenes: 1. 100%, 2. 100%, 3. 85-100% (least demanding scene)

SMAAT2X: avg. FPS 74 GPU usage in scenes: 1. 100%, 2. 67-84% (most between 70 and 80%) 3. 53-100% (most between 60-80%, 100% only couple of seconds)

SMAA4X: avg FPS 55 GPU usage in scenes: 1. 100%, 2. 80-90%, 3. 55-100% (most between 60-75%, 100% only couple of seconds)

 

Extra notes: the disparity between MSI Afterburner FPS counter and ingame is huge in the case of SMAAT2x and SMAA4X. We are talking 20FPS difference in extreme situations (50+ as shown by MSI vs. 70+ shown by the game). The difference in normal SMAA is relatively small (never more than 5-6 FPS).

 

Both SMAAT2X and SMAA4X suffer from extreme flickering and stutters during scene 3.

49 minutes ago, Frankenburger said:

A user has found 0x080000F5 to work with DX11 and reportedly nets about a 50% improvement.

Hahaha

 

When I used that one and saw about 50 fps I just thought that SLI simply doesn't work. I would never have assumed that the 50fps I was getting was ALREADY improved from a 37fps with single GPU. Geez.

[Frankenburger]

3 minutes ago, Lathlaer said:

There you go:

*Snip*

I'm glad to see that SMAAT2X and SMAA4X is working better in DX12 MGPU than in SLI. Flickering aside, it's definitely steps in the right direction. Hopefully other devs start catching on. Thanks for looking into it

 

5 minutes ago, Lathlaer said:

When I used that one and saw about 50 fps I just thought that SLI simply doesn't work. I would never have assumed that the 50fps I was getting was ALREADY improved from a 37fps with single GPU. Geez.

Lol, yeah. From what I've seen in user benchmarks, DX11 seems sub-optimal regardless if using a single GPU or multi GPU system.

[Lathlaer]

7 minutes ago, Frankenburger said:

Lol, yeah. From what I've seen in user benchmarks, DX11 seems sub-optimal regardless if using a single GPU or multi GPU system.

Which is strange because I don't remember Rise working so poorly in DX11. I mean, sure, DX12 was still better but it was about 10-15 fps in my case tops.

 

Side note: just tested with that shadow option turned down from high to normal (like the reddit post suggested) and turned off AA altogether and got avg. 98 FPS here. So at this point it can be safely assumed that using standard SMAA is basically free if someone wants.

 

Oh and CPU usage was never higher than 30% in my case.

[Frankenburger]

34 minutes ago, Lathlaer said:

Which is strange because I don't remember Rise working so poorly in DX11. I mean, sure, DX12 was still better but it was about 10-15 fps in my case tops.

 

Side note: just tested with that shadow option turned down from high to normal (like the reddit post suggested) and turned off AA altogether and got avg. 98 FPS here. So at this point it can be safely assumed that using standard SMAA is basically free if someone wants.

 

Oh and CPU usage was never higher than 30% in my case.

I think it's because Shadow uses a modified engine. Digital Foundry briefly mentioned the development process in their PS4 vs Xbone overview.

[JZStudios]

8 hours ago, Frankenburger said:

I think it's because Shadow uses a modified engine. Digital Foundry briefly mentioned the development process in their PS4 vs Xbone overview.

Digital Foundry has now done an analysis video.

Seems like I'm going to have to upgrade at some point... damn this new generation!