Tomb Raider Benchmarks On Linux With NVIDIA Graphics

[faldzip]

So you like playing on low instead of high, don't like TressFX, as it looks worse then without it, and spends lot of time to tweak the OS. So I guess you are far away from a typical gamer and the devs, publishers and so on don't really care about you as you are the minority of minority. But must admit your use case is still better then someone's who said playing in window is better (then FS) as you can play and do other things at once... How many ppl do this? I guess it's one.

[Passso]

So you like playing on low instead of high, don't like TressFX, as it looks worse then without it, and spends lot of time to tweak the OS. So I guess you are far away from a typical gamer and the devs, publishers and so on don't really care about you as you are the minority of minority. But must admit your use case is still better then someone's who said playing in window is better (then FS) as you can play and do other things at once... How many ppl do this? I guess it's one.

I see your point : every fps helps, every micro details even non visible helps... Really?

Here is mine :

- I have cheap hardware and the game runs perfectly
- I see no significant difference when I increase options and take a screenshot
- I am happy with this port, whatever how they did it
- I do not need Wine or tweaks or whatever. I clicked "install"
- I can play have fun on my favorite desktop environment

[faldzip]

LOL, guys, you just see black or white, gray does not exist, right?

I have a hard time telling if the game runs on 45fps or 60fps without fps counter, but it doesn't matter - one likes cucumbers, other gardener's doughters (hmm... that rhymes in Polish... nvm). To be honest the minimal fps is more important than average as the dips can be noticed easily and can be annoying. But nvm. The point is, that poor ports makes Linux look like gaming platform for masochists, no one who would like to just play won't even look at Linux. I know that I can tweak the system, compile new kernel, new mesa, and so on. But it doesn't matter as on Windows or consoles you have much less problems (or just amount of work - you probably do not have to tweak anything) and still much better performance. In my opinion - this is not going to increase the Linux usage in gaming market share as the people do not see any reason to migrate to linux. I really appreciate the SteamOS - installation is simple and then it boots up straight to Big Picture mode - that's the big step forward to popularize linux gaming. But now we need games that run at least as good as on Windows, because otherwise - linux does not seem to be atractive to gamers in any way.

[faldzip]

You can find tests that shows that windows games do run faster in linux than in windows. use wine-staging 1.9.5 csmt enabled or wine nine dx9 perfomance boost enabled. We play here at good fps fullhd every linux or windows game we have with x4 860k and 750ti. The 17 year old kid who owns the pc, do not use dualboot. He is former win7 and win10 user and knows that Debian testing Xfce is much better os.

So how is Witcher 3, Rise of the Tomb Raider, Fallout 4, The Division, Hitman, DiRT Rally, Assasins Creed Syndicate, Ashes of the Singularity running on Debian testing Xfce? I am really interested in such results.

[juno]

Just a note: With a 60Hz monitor makes it is physically impossible for the eye to see 45 frames per second. The eye can only see these: 60 FPS, 60/2=30 FPS, 60/3=20 FPS, 60/4=15 FPS, etc.

But, you know, it averages out to 45, if your card renders 45 fps and therefore every second frame is displayed twice on the panel. FPS are always averaged values, not very correct and not at all meaningful to measure the actual feeling. And when someone tells you about how it looks, it's about his feeling at this frame rate the GPU is producing. However, of course if your card renders constantly 45 FPS you either have terrible v-sync stuttering or heavy tearing. Not to see that is a real challenge

It is unlikely the ports are poor - it is likely the graphics drivers are poor (notably the open-source Radeon drivers).

It would be true that the ports are poor if it was the responsibility of the game developers doing the porting to optimize the graphics drivers.

Well, then why is the performance on Windows twice as fast for Nvidia, too? Are their ogl drivers also so much worse compared to the d3d ones?
Sure, you can argue that the game/engine is not developed with that in mind but only d3d etc. but to get the performance with the available APIs and drivers is an essential part of porting, imho. Otherwise, they could as well just use a "simple" wrapper




The German site ComputerBase did some testing, too. Direct comparison Linux vs Windows: http://www.computerbase.de/2016-04/t...tx-970-szene-1
Up to 3.5x the performance, it is so heavily cpu-bound
Image quality differs, too

[juno]

They write about lighting/shading, esp. ambient occlusion, not the brightness in the picture in general.
Hard to tell based on this few shots but we can see that the images look similar for the bright parts and differ for the darker parts. The Windows version goes darker, so there is more overall contrast. I'm not sure a simple gamma adjustment would fix it. Maybe, Feral also did this on purpose.
However, the claims about AO seem legit to me, while it's still hard to tell as these scenes are not the best to see the effect

I don't know why - Nvidia hasn't contacted me to optimize their Linux drivers

Hmmm, did AMD?

[wdb974]

Any idea how to convince them to use Vulkan? I have one: do not buy ports with DX-to-OpenGL-wrapper crap and leave the comment on Steam that you would buy it if it was good Vulkan port.

Do you have a list of games that use such wrappers?

[bridgman]

Hmmm, did AMD?

We called, but he was out.

[bridgman]

This was in response to the question "did AMD contact you about optimizing our drivers ?".

You would obviously have responded "no" (because we didn't), thought I would just have a little fun with the question.

My response meant "we phoned you but nobody answered the phone", or "someone answered the phone and said you weren't there". Makes more sense if you think "home phone" rather than "cell phone".

(yes I know an increasing percentage of people don't bother with land lines these days but until all the cell networks get per-tower backup generators and logistical support to keep them all fuelled during a long power outage I'm staying with my land line )

[bridgman]

So, the sentence "We called, but he was out" was a lie.

As you say below, we don't know for sure.

Furthermore, you don't have enough information to decide (as far as I know) whether I am good at optimizing software or not.

Correct, but we don't call people randomly just in case they might be good at it.

Just a note: It is impossible for any single AMD employee to be fully aware of what other AMD employees do (unless you have microphones and cameras everywhere and spend 90% of work time listening to the sounds and watching the videos).

... proving that something doesn't exist has much higher complexity than proving that something exists.

Yep. I read a study somewhere about what the overhead would be in order to achieve close to 100% surveillance - not just people watching the employees but other people watching the watchers, and so on up a pyramid with someone you can't trust anyways at the top. Once you factor in the parallel org required to look out for collusion between the watchers you end up around 40% overhead and you still can't trust the person at the top.

We don't do that so no, I can't say whether we called or not with certainty.

EDIT - for clarity, when I said "you obviously would have responded no" that was in reference to someone asking you if we called, not what you would have said to us if we had called.