Tomb Raider Benchmarks!

Ok guys sorry 4 the delay.

P170HM
2820
6990m

12.11 beta 11 drivers

Ultimate
Min 3.6
Max 21.4
Average 17.6

Ultimate with tressfx off
Min 6
Max 31.1
Average 26.0


Edit: results with 13.2 beta 6
Ultimate
Min 6
Max 21.4
Average 17.5

Ultimate with tressfx off
Min 7.5
Max 31.1
Average 26.0

and here with post processing off

Ultimate with tressfx off
Min 10.2
Max 46.1
Average 37.2

About to inject SMAA instead of the crapy FXAA and ssAo in game anti-aliasing.

yup the smaa works beautifully with only a 3fps hit averaging 34.6 fps. Best AA yet along with post processing off and tressfx off and everything else ultimate and it looks and plays stunningly well. time to play the game

 

Anyway, the question here is Tomb Raider's Benchmark, it's not AMD vs Nvidia. People buy what they want and/or what they can afford. "Nvidia games" will run better on Nvidia GPUs, the same for AMD. And remember FPS isn't all!!!

2013 Benchmarks (Single Cards, not SLi or CF):

- Battlefield 3 on PC: GeForce GTX 680M vs. Radeon HD 7970M Frame-Rate Tests - YouTube
- Skyrim on PC: GeForce GTX 680M vs. Radeon HD 7970M Frame-Rate Tests - YouTube
- Crysis 2 on PC: GeForce GTX680M vs. Radeon HD 7970M Frame-Rate Tests - YouTube

I only bought Nvidia because of PhysX, TXAA, FXAA and CUDA.

 

Well I have gone more deeper into the game and at times it slows down so benchmark is not a true factor of extreme graphics you will encounter in game.
I found tessellation and post processing are the biggest fps factors. Turn these off and it works great now.

 

Truth to be told, I read that Nvidia received the source for the game just a couple of weeks before the release and didn't have time to polish a properly working/optimized driver for it. Future drivers should improve the situation but tresfx should always remain a slightly bigger performance hit on nvidia cards until they release a "complete" consumer mobile card again (akin to TITAN).

 

Nice comparison as we have similar systems.

Ultimate with tressfx and tessellation off
Min 30.2
Max 39.5
Average 35.5

With everything on including tressfx
Average was around 21.

Looking forward to playing this game but I won't be able to use tressfx

 

7970m CrossFire
Ultimate settings with tressFX and tesselation

Min. 51
Avg. 62
Max. 70

TR with new patch.

 

hm wonder if i can get playable framerates with everything set to max, since my oced 7970m gets around 60-70% of CF stock bench scores...

Sent from my Galaxy Nexus using Tapatalk 2

 

^^
No way to tell except to try wut happens

 

It's how i've been playin so far, everything maxed (including tresfx hair) on stock clock 7970m @30-40 fps with the occasional dip down to 20-25fps for 1-2 seconds but thats only during transition scenes. @ 52% completion on the game and so far so good. I dont want to finish it, it's a really good game lol

 

Great game so far a true adventure. About 3 levels to go.

Feels like Far Cry 3 but reversed roles. More catered for women but men will also enjoy the eye candy. I feel this title is > Far Cry 3.
Some of the controls are a challenge but otherwise a great game.
Square Enix with Hitman:Assassin, Sleeping Dogs and now Tombraider make quality tested games.
Can't wait for their Thief game next year.

 

Has anyone else noticed very low cpu usage when playing TR?
Not sure if its just my system and that gpu drain is so high that its throttling cpu. Anyway my cpu didn't get above 30-40% usage on any core during 30 minutes of gameplay.

 

Games these days are not being made to take advantage of the i7 processor and balance the workload. Skyrim and countless other games do the same thing.

 

Processing occurs in burst. Meaning that it is quite possible that one (or more core) are heavily taxed for a few miliseconds and then get to rest for a couple of more (usually awaiting for a rendering cycle of a game to complete).

Games are usually formed by two major cycles (not gonna count input interruption mechanisms): Logic and Graphics.

The name pretty much says it all. In the logic cycle, physics and mathematical equations are solved in order to update the game to its new state (player A moved from X to Y, box B collided with box C and must be rotated W degrees, you get the gist). This means this cycle is pretty much CPU heavy. It is run in a loop as long as the game is going. However, it must wait for the Rendering cycle to complete, otherwise you may get superimposed game states.

Graphic cycle is where you draw the game objects, HUD and apply post processing effects (these things may also tax a wee bit of your CPU, but usually the GPU will be strained here).

What does this mean? Well, a faster GPU allows for rendering cycles to take less time, meaning more FPS. Likewise, a faster CPU accomplishes computational tasks faster, hence a quicker logic cycle, meaning more FPS. The thing is, both one and the other aren't taxed at maximum all the times! And guess what, in order to run a game at 60 fps it means that each pair of Logic-Rendering cycle must take 17 miliseconds at max (reeeaaaaaally quick!). Add to that fill rate speeds (for instance, a Tegra 2 can only touch a full hd screen around 200 times each second) and you can have quite a few moments where your CPU isn't working at full speed .

TL;DR . You will hardly find a game where CPU is used above 90% (unless it is not a very graphic game, such as sim games, chess, etc).

Edit:

Source - I'm a Game Development and Computer Graphics teacher at my Uni ^^.

 

Yeah something like what the guy above me said. Basically the CPU power is rarely required. Some games use more CPU power depending on what is being done/the type of game. But mainly, your CPU is almost never going to be a direct bottleneck.

On another note, Tomb Raider without TressFX is really easy to run, apparently. A friend of mine can max it and stream it at 60fps on a 550Ti with a Phenom II x6. That's nothing compared to most of these laptops' CPU and GPUs.

EDIT: There's something I'm forgetting in these settings... I just don't remember what. I think there's a graphics setting that does nothing. High precision or tessellation or something. They also offer a massive fps boost when turned off.

 

Yeah pretty weird stuff. I am curious what engine it is using for graphics.

 

High precision fits what you describe : better fps with little to no improvement, visually.

 

I must say though I'm starting to see a good trend in PC gaming lately. More and more games are being graphically optimized for the PC. I think this generation may last us the longest, as we won't need to have WAYYYYY too much excess power to run just decent visuals. I'm glad that tomb raider can be run so prettily so easily. Previously they just made the graphics and figured PC gamers would toss enough money at their machines that their hugely excess power would just be enough =(.

 

Ok finished the game @ 61%
So far GOTY very comprehensive and well made the way a game should be made.. highly highly recommended.

 

I ran the benchmark on high, ultra and ultimate and actually I found that the biggest step in quality was from high to ultra rather than ultra to ultimate.
To improve framerates (and possibly reduce gpu usage and therefore temperatures) turn off tressfx, high precision and maybe also tessellation? Not sure what difference tessellation makes and if it uses a lot of graphics power.
I'd like to keep textures etc all on ultra. What else would reduce gpu usage whilst keeping ultra quality?

 

System:

P170EM
Intel Core i7-3630QM 2.40GHz
SMD 7970M
8GB Corair MEM
Samsung 840 250GB SSD

Everying to max but - SSAA x 2 & Tessfx off:

Benchanmrk test gives me max 52FPS but using FRAPS Im getting on average 38FPS.

No stuttering and plays like a dream! One thing i did notice, when you play the mountain range (the same bit as the benchmark when looking over the cliff) was the sun light peering from the clouds. I wasnt getting it when i had Tessilation and Post processing off but when i set everything to Ultra (barring tressfx) it seems to switch it on.

Very happy with the game and the way its running on my card! Now if only AMD sent me my CD key so i can give Crysis 3 a go... its been 48 hours now since and nothing

 

Yeah I noticed that too.
My setup I found is to have high precision off, post processing off, tressfx off and everything else ultimate. I disable fxxa and I inject SMAA(default settings) and it played and looked beautiful all the way until the end.

Enjoy

 

You should check if it's tessellation or post processing that gives those sun rays then decide accordingly =P

 

Sun rays do sound like post processing. Depending on the effects used, post-processing can be very light (say, sepia or B&W frames) or reasonably heavy (Bokeh filtering, AKA Depth of Field, etc).

Tesselation means two things:
- To create a triangle based mesh (Delauney triangulation, or similar Delaunay triangulation - Wikipedia, the free encyclopedia )
- And in this context, it means to apply a displacement map upon a flat surface in order to create the effect of volume (check tesselation with displacement mapping).

Obviously, the last one costs a LOT of processing power. However, complex shapes such as stones or pavement should look way better with it on. You guys should also check Bump Mapping and, my personal favourite Parallax Mapping (used in recent DX11 games).

Cheers.

 

Is FXAA better than SMAA??

 

SMAA produces way better results visually, but is also way heavier than FXAA.

SMAA was developed by these guys (one of them is Portuguese, Tiago, he works for Crytek, a really cool guy).

Here you can see the video about the advantages of SMAA. Notice particularly the advantage in aliasing texels in extreme oblique views of a plane (a problem that most AA algorithms handle poorly, and even with Anisotrophic Filtering you still see a lot of flicker).


SMAA: Enhanced Subpixel Morphological Antialiasing

 

If you guys want the ultimate candy for TombRaider install SweetFX

See here: SweetFX Shader Suite release and discussion thread - Guru3D.com Forums and goto page 42 and use marcel's settings.
Wowza..

EDIT: just did some more testing. No further visual enhancement with the following options so leave them off:
Post processing - water actually looks worst with this on.
Tesselation - leave off could not see any improved visual enhancement before and after with this on.
High precision( cant remember what its called but last option in advanced settings menu) - again leave off no further visual improvement.

With SweetFX, everything ultra(TressFX normal and FXAA should be off) and the last 3 options off you are set for gaming graphics nirvana.
Enjoy.

 

The beautiful shafts of light during the benchmark come from post processing. Turning off tessellation didn't do much in my opinion.
I really want to run TressFX because its so awesome but its a significant step down in performance.

Its ashame the cpu couldn't handle tressfx seeing as how its not doing much.

 

Somehow, I believe using the CPU to render all the hair strands would max it out before you could blink twice.

 

Pretty much true. Physics with the CPU is very limiting:

Using bullet physics engine, simulating the colision of 380 boxes on a single thread running of the CPU I get about 15 FPS, and between 0 and 2% of GPU utilization
Using the same engine, simulating the collision of 1000000 (yes, one million boxes) on the GPU, i get almost 30 FPS, and GPU utilization close to 90%.

All in all, physics processing will have to shift from CPU to the GPU eventually. TressFX running in the CPU would completely hog resources...

 

Finally...

[New] Nvidia Driver 314.21 (BETA)

--------------------------------------------

Stock Clocks

Ultimate + TressFX ON:

MIN: 40
MAX: 80
AVG: 60.4

-----------

Ultimate + TressFX OFF:

MIN: 84
MAX: 122
AVG: 101.8

--------------------------------------------

OC +135 / +500

Ultimate + TressFX ON:

MIN: 50.4
MAX: 84
AVG: 65.1

-----------

Ultimate + TressFX OFF:

MIN: 90
MAX: 134
AVG: 110.3

--------------------------------------------

Another thing I noticed is that now TressFX is more accurate, I mean, before, in some points the hair just went crazy, and now is more real, or at least not crazy...

 

Has anyone run benchmarks with a 675mx? I just got a new np9170, but don't have the game yet.

 

wow how can u achieve such a great result? mine with almost same hardware cant get that result

 

did u make sure to also apply the most current tomb raider patch? IIRC those beta drivers only come to full play when u also have that patch installed.

 

I'm running Tomb Raider 1.00.718.4 with Nvidia Driver 314.21 (BETA). My system specs are in the signature.

-------

I'll update to Tomb Raider 1.0.722.3 and post a new Benchmark soon. :thumbsup:

 

> Tomb Raider 1.0.722.3
> Nvidia Driver 314.21 (BETA)

Stock Clocks

Ultimate + TressFX ON:

MIN: 44.6
MAX: 76
AVG: 62.2

-----------

Ultimate + TressFX OFF:

MIN: 82
MAX: 122
AVG: 101

--------------------------------------------

OC +135 / +500

Ultimate + TressFX ON:

MIN: 48
MAX: 84
AVG: 65.7

-----------

Ultimate + TressFX OFF:

MIN: 88
MAX: 130
AVG: 108.9

 

[JC] turn the shadows to the highest setting and re-ran the benchmark. They added much softer shadows in 722.3 update, you can't turn them on with the ultimate setting have to do it manually) The game is super eye candy now!!!

 

Stock Clocks

Ultimate + Ultra Shadow + TressFX ON:

MIN: 38.8
MAX: 76
AVG: 57.8

-----------

Ultimate + Ultra Shadow + TressFX OFF:

MIN: 66
MAX: 114
AVG: 90

--------------------------------------------

OC +135 / +500

Ultimate + Ultra Shadow + TressFX ON:

MIN: 42
MAX: 78
AVG: 61.6

-----------

Ultimate + Ultra Shadow + TressFX OFF:

MIN: 74
MAX: 124
AVG: 97

--------------------------------------------

:thumbsup:

 

I'm also running Nvidia Driver 314.21 (BETA) and latest Tomb Raider patch as on steam up to date (Tomb Raider 1.0.722.3).
my spec on my signature
maybe u do some tweaks or something can u share with us here? or mine is underperformed?

 

how far off are u from his results?

Sent from my Galaxy Nexus using Tapatalk 2

 

with Ultimate preset + TressFX ON and all at stock :



dunno whats wrong with my setup

 

- My specs are in signature;
- Game without tweaks, clean;
- Running the game alone, without other softwares running in background (only, Kaspersky AV, GeForce Experience, Logitech Gaming Mouse, Adobe Application Updates, THX TruStudio Pro, Clevo Hotkey);
- Game installed on Seagate Momentus XT (750GB), I always install my games on HDD instead of SSD;

--------------

Ultimate preset + TressFX ON + Stock Clocks:



--------------

Go: " Control Panel > NVIDIA Control Panel > Set SLI and PhysX Configuration", and verify if " Maximize 3D performance" is checked, and " Physx settings" is Auto-select (recommended).

 

very similliar setup and config but different result. Well something not right with my system, dunno what it is

Edit :
OMG....whats wrong with my eyes, sorry guys i just realize that v-sync still on!!!

here my result now :



:thumbsup:

thx and sorry guys :hi2:

 

lol ok, glad u figured it out

 

Strong numbers I´m JEALOUS!!!

 

I've already finished and deleted the game, so I can't run more benchmarks, I wanted, because nVidia released new Drivers, so the performance and beauty of the game on GTX 680M SLi would be even better!

The next Benchmark: Bioshock Infinete on NP9370, stay tuned...

 

Nice numbers u got m8

Sent from my awesome Nexus 7 using Dirty Root Box v2.5 & M-Kernel a46

 

These are some seriously nice results.

I hope that this shows that Direct Compute (TressFx) is really something worth looking into. It's easy for developers to use (a bit like a High Level Shader Language) and it is adequate for throwing parallel tasks to the GPU (albeit it is only for DX11 GPUs).

We really need to push OpenCL/Direct Compute like technologies forward. These are technologies that are not proprietary (okay DC is not exactly cross-platform, buuuuut). I for one abhor the idea of Physx. If you are not on the green team, you are pretty much screwed. I have been pushing my fellow researchers in using OpenCL, MPI, and the likes in parallel processing to be used in simulations and games rather than Physx.

I mean, Tomb Raider is a game that runs great maxxed out, both on a 680m or a 7970m. Borderlands 2 on the other hand, is a totally different story (with Physx on high for both).

TL;DR - So happy to see a game with GPGPU running great on both red and green teams

 

TressFX is a bit overkill in my opinion. Is it really worth to take a hit like that to your FPS just to have "better hair"? I personally think not. But if I had dual GTX680m or dual 7970m, then hell yeah i would have that going! Sadly for me, my laptop is having overheating problems with any games in DX11 mode, so im forced to play TR in DX9 mode. Still looks pretty but it hurts to not be able to use my 7970m to it's full potential

 

TressFX, is just like any heavy graphical option, such as SSAO or HBAO (or MSAA):if you use it, it will take quite a toll, but it will result in a higher quality experience. Is it worth playing the game at 20 fps to have it? I agree it is not. But if it is choosing between 50fps or 60fps (with the setting off) I'd go with 50fps