[H] Thief 4 Performance Preview - No Mantle for now

[OCGuy]

On graphics without Mantle (Mantle delayed until next month)

In our first graph, we are comparing the Radeon R9 290X and the GeForce GTX 780 Ti. There seems to be a big difference in performance between the two, with the GTX 780 Ti delivering better performance by 20%. However, both are insanely fast averaging near 100FPS and performance well above 60FPS at all times. Remember, this is the highest settings possible in the game, all graphics options turned on.

On Mantle potential:

BF4 is a good case scenario for Mantle. However, a game like Thief isn't like BF4. Thief, being a singleplayer game, is more about the experience, and less about needing the highest possible FPS for competitive gaming. Given the fact it already produces performance that is more than acceptable, what more can Mantle bring to the table to improve upon the experience?

http://hardocp.com/article/2014/02/26/thief_video_card_performance_preview/3#.Uw4MBmd3uUk

 

[Attic]

More from the article which confirms my fears about the game.

Thiefs Graphics

Thief's Graphics
We could make a joke here about the title of the game being "Thief" and the feeling that someone "stole" the graphics from this game. This game suffers from a problem we've experienced before, notably in another game, Dishonored. The difference between the two games though, Dishonored was actually fun to play.

The problem is that it is using a very old, by today's definitions, graphics engine with Unreal Engine 3. Sure, it is heavily modified, but the fact is, it is rooted in an old engine. This leads to the game number one performing extremely well, but also leaves the game not able to compete with modern games in terms of graphics. More importantly, to video card makers, it does not give you a reason to need to spend money on a new graphics card to upgrade. It does not push the graphics card or gaming industry forward.

Thief has subpar textures, subpar dynamic world environment, subpar physics, and a subpar graphical experience compared to modern games. Yes, it has contact hardening shadows, and those look nice, yes it has parallax occlusion mapping, and that looks nice, yes it has tessellation, and that looks nice, but all of these things we've seen before, and we've seen these used in greater detail and quantity in other games. These are "tacked on" features in Thief, rather than being built from an engine ground up with these things in mind. While Thief is just catching up with these graphical features, other games are far surpassing it graphically.

Thief looks and feels old for a 2014 game, and this is not a good selling point for new graphics cards. Now, maybe TruAudio and Mantle can add something appealing to the game. However, we feel that even with those two additions, the game is still going to feel overall very stiff and ancient. Inserting new technologies also cannot improve the gameplay, which is certainly an important factor for any singleplayer game.

I was hoping the screenshots I had seen were just not finalized, but it looks certain that the game just won't look very good graphically for a modern PC game.

Sucks, and some of the character models are just terrible.

The Bottom Line

The Bottom Line
...
It will sit on the side for us as we patiently await TruAudio and Mantle support to test the only two aspects of the game that are appealing. We were hoping to find a new game to replace one of our older games in our gaming suite, so far, this game is not it. Quite frankly, you will likely be better off saving your hard earned cash till this title hits the bargain bin.

Interesting, and I have some interest in TruAudio as something to add to gameplay experience outside of pure graphics power.

But as for now, this game is squarely off my purchase list and it was squarely on it before reading this. Hard generally calls things as they are.

 

[Gloomy]

Huh, they didn't use the benchmark? I've heard the benchmark is much more demanding than the actual game.

I haven't played it yet though I did run the benchmark before updating the driver and then again after, and the driver boosted performance quite a lot for me. Presumably Nvidia will be able to do the same with a driver update.

As for graphics, it's a mixed bag... I've heard nice things about POM, the shadows, the reflections, and some of the textures. But also a lot about how some textures are really bad close up...

I really do wonder what AMD is doing with Mantle in this game.

 

[OCGuy]

Huh, they didn't use the benchmark? I've heard the benchmark is much more demanding than the actual game.

A benchmark that does not reflect real-game performance is fairly useless.

 

[biostud]

I ran the benchmark with max settings 1440p

with SSAA high I got 31.7 fps
without I got 43.3 fps

I played a little without SSAA and it felt faster than 43.3 fps, but sluggish if I enabled SSA on high.

 

[Gloomy]

A benchmark that does not reflect real-game performance is fairly useless.

Right, I agree. I was just mentioning there is a benchmark, not really vouching for it.

 

[VulgarDisplay]

What is with the stupid settings for their runs? Cool, you get 150+ fps with FXAA. How about we actually use some IQ settings to see what is actually happening when the cards are pushed.

 

[BrentJ]

What is with the stupid settings for their runs? Cool, you get 150+ fps with FXAA. How about we actually use some IQ settings to see what is actually happening when the cards are pushed.

The game was running at the highest quality settings, with all the graphics options enabled. This allows a nice baseline test to start from, and give you an idea how the graphics perform in the game without artificially bringing down performance due to SSAA, which is always very demanding. Those that have the power to run SSAA, will, those that don't, will have to use FXAA. This wasn't a "Highest Playable" article, but a Preview of Performance, in which we pick the same setting, ap2ap, and compare all cards so you can compare each card to each other. The goal here is, to see how much of a demand the graphics themselves entail in the game, of which, they were maxed out.

 

[ASM-coder]

But as for now, this game is squarely off my purchase list and it was squarely on it before reading this. Hard generally calls things as they are.

Interesting reaction to a game performing very well on a wide range of graphics cards.

From the article:
"This means you do not need to spend a lot of money on a graphics cards for Thief. This game does not give us a compelling reason to recommend upgrading video cards at this time. Quite frankly, you will likely be better off saving your hard earned cash till this title hits the bargain bin."

Hmmm, I better look for a game that doesn't run well on my current rig, pay $60, become disappointed in the frame-rates, update my rig, wait for the next game that exceeds my specs, and then start the cycle all over again.

There are a lot of good games running on the U3 engine.

 

[BrentJ]

The game was running at the highest quality settings, with all the graphics options enabled. This allows a nice baseline test to start from, and give you an idea how the graphics perform in the game without artificially bringing down performance due to SSAA, which is always very demanding. Those that have the power to run SSAA, will, those that don't, will have to use FXAA. This wasn't a "Highest Playable" article, but a Preview of Performance, in which we pick the same setting, ap2ap, and compare all cards so you can compare each card to each other. The goal here is, to see how much of a demand the graphics themselves entail in the game, of which, they were maxed out.

To add: When you enable SSAA, at that point, you are pretty much testing only the performance of SSAA, because it is the most demanding thing in the game. Before you do that, it is important to know how the graphics in the game burden the GPU. Do the hard contact shadows cause a GPU performance burden, does the parrallax occlusion mapping cause a burden, does tessellation decrease performance. It is important to know how much the graphics options themselves impact the game performance first. In this game we found out they do not, and performance is very high even with everything enabled. That is very important information.

 

[f1sherman]

Huh, they didn't use the benchmark? I've heard the benchmark is much more demanding than the actual game.

I haven't played it yet though I did run the benchmark before updating the driver and then again after, and the driver boosted performance quite a lot for me. Presumably Nvidia will be able to do the same with a driver update.

As for graphics, it's a mixed bag... I've heard nice things about POM, the shadows, the reflections, and some of the textures. But also a lot about how some textures are really bad close up...

I really do wonder what AMD is doing with Mantle in this game.

Yes, mixed bag describes it (graphics) best.
It definitely has its charms, but I think they went overboard with that "olden times" color filter.
Can be tweaked with FXAA/SweetFX, but I'll play with shipped settings - turns out sharpening does not suit my taste as it nullifies SSAA, and I want to keep pixel crawling at minimum.
Softer image works pretty well in urban enviroment which by the nature of geometry has huge amount of horizontal/vertical lines that just prey on player with pixel crawling.

Too bad there is no Rotated Grid SSAA like in Splinter Cell Blacklist, so turn on SSAA if you can afford it.

As for game-play, I suggest turning off HUD and Focus, and not reading game reviews.
Which ofc does not include hw reviews like [H].
I feel they were a bit harsh with Thief graphics - UE3 is stone-age, but Thief is not a clear case of subpar graphics IMHO.

 

[sushiwarrior]

I would consider the game generally good looking, I'm a bit confused at all the backlash. The lighting in the game is really good IMO, and it's also not crazy hard to run. Not ever game has to be like Crysis 1, it's nice you can play it at max settings with a reasonable card.

 

[OCGuy]

This is how an ancient engine is supposed to look/run on modern desktop hardware. The fact that people were making "Zomg upgrade my 780 ti to Titan SLI for Titanfall?" threads actually made me lose IQ points. And I need every one.

 

[Ventanni]

Honestly, I thought Thief looked pretty darn good considering how smoothly it ran.

 

[Attic]

Interesting reaction to a game performing very well on a wide range of graphics cards.

From the article:
"This means you do not need to spend a lot of money on a graphics cards for Thief. This game does not give us a compelling reason to recommend upgrading video cards at this time. Quite frankly, you will likely be better off saving your hard earned cash till this title hits the bargain bin."

Hmmm, I better look for a game that doesn't run well on my current rig, pay $60, become disappointed in the frame-rates, update my rig, wait for the next game that exceeds my specs, and then start the cycle all over again.

There are a lot of good games running on the U3 engine.

A bit over the top.


I loved Dishonored. Which is why this excerpt, in the quote you responded to, stuck out to me.

Thief's Graphics
We could make a joke here about the title of the game being "Thief" and the feeling that someone "stole" the graphics from this game. This game suffers from a problem we've experienced before, notably in another game, Dishonored. The difference between the two games though, Dishonored was actually fun to play.

Good gameplay and poor graphics is a different beast from mediocore/poor gamplay and medicore/poor graphics.

Need to excell in one or the other with at least decent gameplay for a game to get my $$$.

 

[ASM-coder]

A bit over the top.

Need to excell in one or the other with at least decent gameplay for a game to get my $$$.

And that's the way it should be. I just don't see any reason to complain because it runs well on low priced cards.

 

[Unoid]

Unless the shadows are real shadows, like in thief 3, and not just slightly darkened spots like dishonored, I'll enjoy the game I'm sure.

 

[VulgarDisplay]

The game was running at the highest quality settings, with all the graphics options enabled. This allows a nice baseline test to start from, and give you an idea how the graphics perform in the game without artificially bringing down performance due to SSAA, which is always very demanding. Those that have the power to run SSAA, will, those that don't, will have to use FXAA. This wasn't a "Highest Playable" article, but a Preview of Performance, in which we pick the same setting, ap2ap, and compare all cards so you can compare each card to each other. The goal here is, to see how much of a demand the graphics themselves entail in the game, of which, they were maxed out.

I wasn't necessarily calling for 8xssaa testing to bring every card to its knees. More like 4xMSAA because it looked like most cards could handle it easily. Unless the game doesn't support it?

 

[BrentJ]

I wasn't necessarily calling for 8xssaa testing to bring every card to its knees. More like 4xMSAA because it looked like most cards could handle it easily. Unless the game doesn't support it?

The game's SSAA options are "Low" and "High." Other than that, there is FXAA. That is all.

 

[tential]

Is this comment serious?

Given the fact it already produces performance that is more than acceptable, what more can Mantle bring to the table to improve upon the experience?

Oh, so you mean if I have a R9 290X that what more can Mantle do? Clearly the reviewer isn't thinking straight. He acts as if everyone has an R9 290X and as if it's an entry level card or something.

Also, the review says that you should wait for the game to hit the bargain bin based on graphics alone?

I find that part also to be ridiculous.

 

[SPBHM]

SSAA looks nicer, and the overall framerate is not as bad as I expected, but I get some ugly stutters when it's on, also I noticed some input lag with it on, so better play without it,

Mantle could be quite good if you are playing the game with an i3 or similar CPU,

it looks like i5s can keep the game at 60Hz quite easily,

 

[Red Hawk]

Obviously Mantle will help low-end Radeon 260 or 250 users, and users with weak CPUs. But for enthusiasts looking for a good showcase of Mantle's ability to get the best out of even high-end GPUs, this is not the game to do that.

 

[el etro]

BF4 is a good case scenario for Mantle. However, a game like Thief isn't like BF4. Thief, being a singleplayer game, is more about the experience, and less about needing the highest possible FPS for competitive gaming. Given the fact it already produces performance that is more than acceptable, what more can Mantle bring to the table to improve upon the experience? http://hardocp.com/article/2014/02/26/thief_video_card_performance_preview/3#.Uw4MBmd3uUk

Heavily GPU bound, like most of AAA titles. A Very Good test for Mantle.
Reviewers should use configurations with GPUs costing twice the used CPUs.

 

[ShintaiDK]

Obviously Mantle will help low-end Radeon 260 or 250 users, and users with weak CPUs. But for enthusiasts looking for a good showcase of Mantle's ability to get the best out of even high-end GPUs, this is not the game to do that.

Mantle doesnt really boost any GPU performance. Just CPU. And considering even an i3 2100 runs this game pretty flawless...

 

[Silverforce11]

It will be very surprising if Mantle can boost performance in this game..

Also: "Thief has subpar textures, subpar dynamic world environment, subpar physics, and a subpar graphical experience compared to modern games."

Ew. Textures are critical. Hence the love of ultra HD texture mods where available.