Shadow Warrior 2 benchmarks

[Head1985]

Pcgameshardware tested 20gpu from 2010-1016
http://www.pcgameshardware.de/Shado...als/Benchmarks-Test-Gameplay-Release-1211483/
Kepler is very good-780TI is faster than 390x and GTX970.
They using fastest new gigabyte xtreme 1070 at 2025/8160mhz(in past they use evga 1070 at 1880Mhz).Still its slower than aftermarket GTX980TI at 1340Mhz LOL...
1080P

1440P

 

[Mercennarius]

Looks like AMD needs to rework their drivers for this...not the best AMD performance here.

 

[railven]

There is actually a nifty tech used in this game where the center of the screen is the focus, thus in full resolution, while the edges are lower resolution creating a blur of some sort.

Eurogamer did a video on it.

 

[Face2Face]

Wow, that 780 Ti is killing it. Too bad it's performance is not this good for all newer games...though, it does pretty well in BF1.

 

[Bacon1]

Hey very nice they have Ultrawide 1440p benchmarks too, hope they keep that up!

Yeah its a very pretty game and pretty well optimized too for the looks. Its no Frostbite but they have made good improvements to their engine.

If they could multi thread it better they'd probably get another good boost to perf.

Only 3% difference between 2 core and 8 core.

 

[Arachnotronic]

Hey very nice they have Ultrawide 1440p benchmarks too, hope they keep that up!

Yeah its a very pretty game and pretty well optimized too for the looks. Its no Frostbite but they have made good improvements to their engine.

If they could multi thread it better they'd probably get another good boost to perf.

Only 3% difference between 2 core and 8 core.

It's a crying shame that more games don't utilize more cores. If the software could take advantage of more cores, devs might be more willing/able to include more CPU-intensive effects/features to improve realism/user experience.

 

[tviceman]

Side question: are there any PC exclusive titles that noticeably favor AMD gpu's? I know AOTS did when it first came out, but not anymore. Any other PC exclusive titles that currently favor AMD over Nvidia?

 

[Triloby]

Side question: are there any PC exclusive titles that noticeably favor AMD gpu's? I know AOTS did when it first came out, but not anymore. Any other PC exclusive titles that currently favor AMD over Nvidia?

No. Even titles that favor AMD such as Hitman (2016), Doom with the Vulkan API, and SW: Battlefront are all multi-plats. I dunno if Battlefield 1 favors AMD, since I've been hearing about terrible minimums and frame times for Radeon GPU's under DX12.

 

[railven]

Pretty sure any PC-Exclusive outside of AOTS is performing better on NV hardware at this point in time. But with that list being basically RTS/Strategy/MMO games it's easy to imagine why.

 

[tviceman]

Pretty sure any PC-Exclusive outside of AOTS is performing better on NV hardware at this point in time. But with that list being basically RTS/Strategy/MMO games it's easy to imagine why.

Civ 6, AOTS, Shadow Warrior 2 (for now), a whole slew of VR titles, X-Com 2 was for awhile, and a few others I'm sure. VR titles are all PC exclusive and are showing to do much better on Nvidia hardware time and again.

There are also titles that are on PC and only 1 of the 2 consoles (Nintendo excluded) like No Mans Sky or that MMO Dinosaur game that either do well on both camps or favor Nvidia.

Just an observation, that's all.

 

[psolord]

I made a couple of benchmarking videos myself a few days back. I found some weird things and I'm still trying to find what I did wrong. If I actually did something wrong that is.

First things first. Here are the bench vids for anyone that needs a visual, but I will post the interesting stuff below anyhow.

Shadow Warrior 2 1920x1080 Ultra GTX 970 @1.5Ghz Core i5 2500k @4.8GHz - 82fps

Shadow Warrior 2 1920x1080 Ultra 7950 @1.1Ghz CORE i7-860 @4GHz - 48fps

At fist glance results seem to be ok. Actually they are. The delta seems to be in accordance to what the pros found. The weird thing I found has to do with frametimes.

Have a look at the 970's framerate and frametimes graphs from MSI Afterburner at first



Frametimes are twice as long as they should be for the respective framerate. The jagged gpu load is not a good indication either, although it's nearly maxed. No card throttling or power limit reached however. It does the same thing at stock clocks.

The analysis of FLA Calculator from the whole run also showed something quite remarkable





I have never witnessed such frametimes and stuttering in a game before. The bad thing of course, is that you can actually feel that ingame.

For comparison here are the 7950's respective graphs.







As you can see the 7950's graphs are a lot more consistent. Less stuttering and you don't see the blue frametimes graph jumping all over the place. This results in the game being only marginally more enjoyable on the 970, despite the seemingly huge framerate advantage.

I did try unloading MSI Afterburner just in case, but it didn't help. I also retested after these initial tests, after a couple of patches came out and with the latest Forceware too, but I see no improvement.

I am still very puzzled as to why the game does not like the 970, but I will keep testing and come back if I have any breaking news. This must the be the only time I have seen such bad occurrences on the 970.

This is what a typical gaming scenario looks like on the 970 (from my ROTTR benchmark runs)



Beautiful. Absolutely beautiful!

On the bright side, even with such a shortcoming, the game is immensely fun to play. I cannot stress that enough.

 

[Bacon1]

Shadow Warrior 2 1920x1080 Ultra GTX 970 @1.5Ghz Core i5 2500k @4.8GHz - 82fps

Shadow Warrior 2 1920x1080 Ultra 7950 @1.1Ghz CORE i7-860 @4GHz - 48fps

At fist glance results seem to be ok. Actually they are. The delta seems to be in accordance to what the pros found. The weird thing I found has to do with frametimes.

What happens if you turn on vsync on the 970?

Thanks for testing!

 

[psolord]

What happens if you turn on vsync on the 970?

Thanks for testing!

Well frametimes go up with vsync. Up to 30ms most of the time.



However the game feels smoother. It does not feel for 30ms frametimes.

I don't know, maybe the engine is weird and confuses the monitoring programs.

 

[littleg]

Could it be the 500MB of slow vram on the 970? If the driver hasn't been optimised for this game and it's simply seeing 4GB of usable vram and filling it up you'd expect to see some spikes when that data was accessed.

 

[dark zero]

Looks like AMD needs to rework their drivers for this...not the best AMD performance here.

Heck, even Maxwell and Pascal are not doing well there too.. a 780 TI is defeating a 970 and trading blows with supposedly a 980 with 1GB less? Even a 1060 which is supposed to be a 980 is being defeated... with the double of memory!
This game is Kepler based or what?

 

[Headfoot]

Could it be the 500MB of slow vram on the 970? If the driver hasn't been optimised for this game and it's simply seeing 4GB of usable vram and filling it up you'd expect to see some spikes when that data was accessed.

Certainly sounds plausible - don't know how to produce evidence to support it though

 

[psolord]

Could it be the 500MB of slow vram on the 970? If the driver hasn't been optimised for this game and it's simply seeing 4GB of usable vram and filling it up you'd expect to see some spikes when that data was accessed.

I am not saying it's impossible, but video ram usage maxed at only 3.3GBs.





On the other hand I have seen vram usage on the 970 having gone up to 4GBs, with no issue. For example in my Rise of the Tomb Raider test.



and the frametimes were great

Now I understand that it's different to use that 0.5GB space as storage or use it as rendering space, but I would guess even if there is no driver optimization for a specific game, Nvidia would have described this 0.5GB in their driver as a generic non rendering memory space, so all apps would know where to render.