Rise of the Tomb Raider : Benchmark and comparaison
Rise of the Tomb Raider was released as an Xbox exclusive last November, a decision that took some serious flak from gamers ever since it was announced. Microsoft was out looking for a new flagship game for its console, a game that could be as significant as PlayStation’s Uncharted. So far it’s proven to be a winning decision. Rise of the Tomb Raider has been widely praised for its gameplay and visuals and now that the PC version is out, we are taking it for a spin.
Developed by Crystal Dynamics and published by Square Enix, the second installment in the Tomb Raider series reboot has been ported to PC by Dutch studio Nixxes, the same folks that ported Square Enix’s Deus Ex: Human Revolution, Sleeping Dogs, and 2013’s Tomb Raider. With those credentials, in addition to participation from Nvidia, Rise of the Tomb Raider fell into capable hands.
It has to be said the game already looked surprisingly good on the Xbox One. Powered by an updated version of Crystal Dynamics’ in-house Foundation Engine, it boasts several next-gen visual features such as image-based lighting, physically-based materials system, deformable snow, enhanced hair simulation, tessellated terrain, and more.
With such a solid base to work with, Nixxes has been able to deliver a number of PC exclusive options that help to make an already beautiful game look truly spectacular. Rise of the Tomb Raider is arguably the best looking game to hit the PC yet, and without question the cut-scenes are the best I have seen.
Nvidia had a hand in the game’s adaption to PC, too, though that might not be the best way to word it. Yes, Tomb Raider is now a GameWorks title and in spite of featuring Pure Hair which has its roots deeply embedded in AMD’s TressFX, we can assume Nvidia has made every effort to optimize this and other visual features for its own hardware.
For testing Rise of the Tomb Raider we will be using the Acropolis Expedition stage of the game (score attack mode) which is about an hour into the main single player campaign. This test features a good blend of outdoor and indoor environments with a few in-game cutscenes, the test lasts 90 seconds at three resolutions: 1920×1080, 2560×1440 and 3840×2160 using the Very High and High graphics quality presets. We also tested with Pure Hair disabled and as usual we’ve included a number of CPU tests to see how the game scales.
Test System Specs
CPU : Intel Core i7-6700K (4.00GHz)
RAM : 4GBx2 Kingston Predator DDR4-2400
MB : Asrock Z170 Extreme7+ (Intel Z170)
PSU : Silverstone Strider 700w PSU
HDD : Crucial MX200 1TB
OS : Microsoft Windows 10 Pro 64-bit
Drivers : Nvidia GeForce 361.75 WHQL / AMD Crimson Edition 16.1
|Radeon R9 Fury X (4096MB)||Radeon R9 Nano (4096MB)||Radeon R9 390X (8192MB)||Radeon R9 390 (8192MB)|
|Radeon R9 380X (4096MB)||Radeon R9 380 (2048MB)||Radeon R9 290X (4096MB)||Radeon R9 290 (4096MB)|
|Radeon R9 285 (2048MB)||Radeon R9 280X (3072MB)||Radeon R9 270X (2048MB)||Radeon HD 7970 GHz (3072MB)|
|Radeon HD 7970 (3072MB)||Radeon HD 7950 Boost (3072MB)||Radeon HD 7950 (3072MB)||Radeon HD 7870 (2048MB)|
|GeForce GTX Titan (6144MB)||GeForce GTX 980 Ti (6144MB)||GeForce GTX 980 (4096MB)||GeForce GTX 970 (4096MB)|
|GeForce GTX 960 (2048MB)||GeForce GTX 950 (2048MB)||GeForce GTX 780 Ti (3072MB)||GeForce GTX 780 (3072MB)|
|GeForce GTX 770 (2048MB)||GeForce GTX 760 (2048MB)||GeForce GTX 750 Ti (2048MB)||GeForce GTX 680 (2048MB)|
|GeForce GTX 660 Ti (2048MB)|
There are five quality presets: lowest, low, medium, high and very high. Very high doesn’t actually apply the maximum quality settings as features such as shadow quality, sun soft shadows and Pure Hair can all be turned up another notch. That said, few gamers are going to have the required hardware to harness these quality settings as we are about to demonstrate with the very high preset at 1080p.
As you can see, using the very high preset at 1080p yields pretty low numbers on the R9 390 and GTX 970 — both were well south of 60fps with frame dips around 30fps. Still, for the most part performance was smooth so these frame rates are acceptable in Rise of the Tomb Raider. Interestingly, the Nano and Fury X ran into frame buffer issues and as a result minimums dropped below 20fps, oddly this didn’t impact the GTX 970. It would appear that Nvidia’s optimized drivers are working well here so we hope AMD can work similar magic with its high-end Nano and Fury graphics cards.
Due to the extreme demands of the very high preset we have conducted the majority of our testing using high instead and it has to be said the IQ difference is almost indistinguishable. Features such as tessellation and Pure Hair are still enabled. The biggest performance difference comes from the ambient occlusion mode which has been changed from HBAO+ to ‘on’ which enables Crystal Dynamics’ own in-house SSAO technique dubbed ‘Broad Temporal Ambient Obscurance’ (BTAO).
Even with the slightly dialed down quality settings, Rise of the Tomb Raider is still extremely demanding and mid-range to low-end graphs cards really struggle. For a minimum frame rate of 30fps gamers will require an R9 380 or GTX 770 at 1080p. For an average of 60fps, you’re looking at a R9 390X or GTX 970!
Turning Pure Hair off only affords graphics cards such as the R9 380 an extra 3fps on average which equates to an 8% performance boost. However, if we look at the minimum frame rates they are much improved as the R9 380 is now 14% faster.
Disabling Pure Hair will buy those with lower-end hardware a small but much needed performance boost, though this rendering feature looks so impressive that I would be more inclined to look for performance gains elsewhere.
At 1440p, gamers will require no less than an R9 290/390 or GTX 970 just to break the 40fps barrier and we are of course talking about the average frame rate here. These GPUs only just kept the minimum frame rate from dipping into the 20s.
At the top of the food chain the Fury X wasn’t a great deal slower than the GTX 980 Ti and we are confident with an optimized driver the Fury X will end up alongside its green competitor.
As for mid-range graphics cards such as the R9 380 and GTX 960, they aren’t really designed for AAA gaming at 1440p using high quality settings, so their sub-30fps average shouldn’t be a huge shock in this stunning title.
Given how demanding the game is at 1440p using the high quality preset we weren’t expecting to find playable performance at 4K even with the GTX 980 Ti. Although the card did manage an average of 31fps, we couldn’t enjoy the gameplay with 24fps minimums.
Even with SLI scaling at 100%, the GTX 980 Ti would be good for just a 60fps average at 4K using the high preset, the same is true for Crossfire Fury X cards as well.
The above CPU results have been arranged by the minimum frame rate result rather than the average. Rise of the Tomb Raider is surprisingly demanding on the CPU and while not as extreme as some of the games we have tested in recent times it still requires a decent processor to get the most out of a high-end GPU.
There is a noticeable performance jump from the Core i7-4770K to the new 6700K. Rarely in games do we see a margin like this but keep in mind the 6700K is clocked 14% higher. When comparing minimum framerates, the 6700K was 17% faster.
Looking to AMD’s processors, we find the FX-series producing some rather impressive average frame rates. Even the minimum frame rates aren’t that low, though we do see that the FX-8350 is comparable to the Core i3-6100.
Once we get down to the dual-core Pentium G4400 performance really starts to drop away and surprisingly it is the Athlon X4 860K that provides our lowest minimum frame rate result.
Initially I had attributed the rather large performance gap between the Core i7-4770K and the 6700K to the fact that the 6700K was clocked 500MHz faster — that and the updated CPU architecture. However, once we underclocked the 6700K all the way down to 2.5GHz, the frame rate performance when paired with the GTX 980 Ti goes unchanged.
Although we have seen this kind of thing in other games the scaling performance almost always goes unchanged when the general CPU performance is the same. The results seem unlikely but having re-tested multiple times we haven’t seen a drop in performance with the 6700K, with more time I plan to run some scaling tests using the Haswell Core i7 as well.
The average frame rate performance of the AMD FX-9590 isn’t greatly different when comparing the 2.5GHz and 4.5GHz results as the higher clock speed is just 14% faster. That’s an insignificant margin for an 80% increase in clock speed but if we look at the minimum frame rate we find that the FX-9590 became 35% faster when going from 2.5GHz to 4.5GHz.
Finally let’s look at how the fan favorite dual-core Pentium G3258 Anniversary Edition processor scales. This chip comes clocked at just 3.2GHz by default but overclocking it to 4.5GHz offers a substantial boost in average frame rate performance. Oddly the minimum frame rate almost maxes out at just 3.5GHz where it dipped to 44fps. The dual-core processor seems to be creating some kind of choke point that boosting the frequency can’t overcome.
Still with a minimum of 47fps and an average of 77fps the Pentium G3258 still enables playable performance.
Image Quality Comparison
For the most part, Medium graphics quality doesn’t look bad compared to High and Very High, with the biggest difference seen in the textures that look a bit ghastly in some comparisons. Keep in mind that comparing the various quality presets in still imagery doesn’t the game justice and many of the visuals aren’t fully realized. Moreover, even the highest quality YouTube footage just doesn’t come close to what you will experience firsthand.
The most noticeable difference in the above screenshots is the difference between Nvidia’s HBAO+ (Horizon Based Ambient Occlusion) and the developers’ in-house SSAO technique dubbed ‘Broad Temporal Ambient Obscurance’ (BTAO).
In my opinion HBAO+ delivers a more realistic looking image and in the top left side of the above images you can notice a tree and a snowy rock face, this is where the difference can be easily seen.
From High to Medium there doesn’t seem to be a huge difference to the world but textures are certainly more detailed in High.
Going from Medium to Low there are noticeable differences when playing the game though you wouldn’t think so when looking at this screenshot. The depth of field effect is virtually gone and textures are noticeably lower now. The game still manages to look quite good overall.
The difference between the Low and Lowest quality preset is like night and day. Lowest looks very poor and should be avoided (dropping resolution would be a better trade-off).
In the above screenshot, which compares Very High to Medium, take note at the quality of the rock ledge to the right of the screen. There is a significant difference between the presets and this is very noticeable when playing the game. The lack of ambient occlusion and texture quality can also be seen on the much darker side of the image as well.
Here we see very subtle differences between High and Very High, but overall I feel Very High does look much better.
In this instance the difference between Very High and Medium is significant. Granted the Medium quality is still very nice, but if you have an eye for detail then it would be hard to reduce the settings. The textures and lighting effects on the ground are greatly reduced and the city in the distance looks much more realistic with graphics maxed out.
For a more complete showcase of what each setting will do visually, Nvidia’s GeForce portal has published a detailed image comparison. We recommend checking out the Ambient Occlusion comparisons (here, here and here) which might depict best case scenarios for Nvidia’s HBAO+ usage but it’s an impressive showing on a really great game nonetheless.
Not everyone will agree but I think Rise of the Tomb Raider looks even better than other visually stunning games, including prodigious The Witcher 3.
In the past few days we’ve seen forums filling up with comments about how unoptimized the game is and this just isn’t true. There might still be room for further optimization, but it’s an excellent port on the whole. Rise of the Tomb Raider is a lot like the original Crysis in that it’s a bit ahead of its time, which is a good thing.
If more games were like this the PC gaming world would be more on the cutting edge and enthusiasts who have heavily invested in high-end multi-GPU setups would be able to enjoy an experience worthy of that investment more often.
Rise of the Tomb Raider is a lot like the original Crysis in that it’s a bit ahead of its time, which is a good thing.
The graphics are so mind-blowing that I’ve nearly played the game twice. The first time I was so taken in by the incredible graphics that I often lost focus of the storyline and objective. There isn’t a world of difference between the Medium, High and Very High presets (for a detailed side-by-side comparison, go here). Having spent quite a bit of time playing around with the various graphics settings it became clear that the big killer for performance was Nvidia’s HBAO+. This ambient occlusion method reduced performance of Radeon GPUs by 15% or more while Nvidia’s GPUs took a much smaller hit.
Turning off Pure Hair allowed for an additional 6 – 8% performance on both AMD and Nvidia GPUs so the impact of this hair technology isn’t significant in Rise of the Tomb Raider and certainly nothing like Nvidia’s HairWorks in The Witcher 3.
Gamers using AMD hardware, namely current-gen flagships such as the Fury and Nano series, we recommend avoiding the Very High quality preset for now as performance gets a bit messy due to brutal frame drops. These same settings don’t hammer the 390X and 390 nearly as much and they are able to deliver quite nice performance at 1080p. The issue has to be VRAM-related but for whatever reason Nvidia’s own GTX 970 doesn’t suffer the same performance issues, though it’s a good bit slower than the 390 here.
This game gobbles memory — both VRAM and system RAM. In our test it would allocate up to 7GB of VRAM when using the R9 390 or 390X. With the 390 or 390X installed the game would consume 4.2GB of system memory and that jumps to 7.1GB with the Nano or Fury. It’s worth mentioning that for these benchmarks our test system was only configured with 8GB of DDR4 memory rather than the usual 16GB. Not realizing this could be an issue and it’s something I will look into shortly using the Fury X and Nano.
When running at 1440p on ‘High’, we didn’t suffer any frame rate stuttering with the Nano or Fury X like we did at 1080p using the ‘Very High’ preset. In fact, AMD’s flagship GPUs performed well and provided noticeably better performance than the 390X and GTX 980. This is short of saying, a driver update for Radeon cards should address some of these issues.
Gamers looking to take advantage of their shiny new 4K panels will want at least $1,300 worth of GPUs in their rig as we found dual Fury Xs or 980 Tis to be a must here.
On the CPU performance front we found some interesting results, most notably of which were found when comparing the Core i7-6700K and 4770K. The 6700K was quite a bit faster and yet when downclocked to 2.5GHz performance wasn’t degraded. As usual, gamers running a Core i5 or Core i7 processor released in the last four years shouldn’t have any worries with this title.
Heavily clocked AMD FX processors will also get by without much trouble. We were surprised by how well the Core i3 processors performed, in particular the i3-6100 which matched the FX-8350. Despite their dual-core designs, the Pentium G3258 and G4400 also managed to deliver more consistent performance than the Athlon X4 860K.
Playing on the Core i7-6700K showed a CPU load of around 20-30% and this shot up to around 60-80% on the FX-9590.
Overall, Rise of the Tomb Raider looks amazing and I believe the gorgeous visuals warrant its steep demand for graphics hardware
Tweet with us
Windows 10? A little education goes a long way. Let https://t.co/BLB1VRuiJN help.Follow