Sniper Elite 3 Benchmark with Quad R9 290x (Mantle vs DirectX)

[KaRLiToS]

This is my first hand experience with the new Sniper Elite 3 benchmark.
I tested Multi-GPUs and also made a comparaison between Mantle and DirectX for single card.

I have to say, I'm kind of disapointed that Sniper Elite 3 doesn't have CrossFire support with Mantle yet. Like PCPer mentionned in their review, only One card utilization for mantle in that game.

From PCper review

AMD's R9 290X actually scales a bit better in DirectX 11 mode, 89% or so. However, Mantle doesn't scale at all with our two Radeon R9 290X cards and the second GPU is more less a bust when it comes to performance increases. Despite the benefit we saw in the single GPU performance results the Mantle API does still struggle with CrossFire implementation.

http://www.pcper.com/reviews/Graphi...-Performance-Maxwell-vs-Hawaii-DX11-vs-Mantle

I also tested stock and OC on the GPUs (up to 1225/1525, could go higher but I only used 1600watts instead of the 2600watts that is available to me but requires the installation of another PSU and balancing the power accros the Lepa +12 rails. The other PSU is an EVGA SuperNova 1000G2)

(Note that with Tri and Quad, I Oced the 1st and 3rd card at 1225/1525 and the 2nd and 4th at 1175/1475, I did this because I was too lazy to install my second PSU and balance the power accross the rails)

I experienced a rig shutdown, a kill-a-watt beeping and I saw it pull 1900watts from the wall. Poor little Lepa G 1600 watts.

________________________________

Test Setup:

1. CPU: i7 4930k @ 4.5 Ghz (1.23v)
2. Asus Rampage IV Extreme (BIOS 4901)
3. Corsair Dominator GT (16GB) 2133mhz 9-11-10-23
4. Adata SX900 SSD and OCZ Vertex 4 SSD
5. Quad R9 290x (Reference models with stock BIOS)
6. Windows 8.1
7. Catalyst Driver: 14.9.1 (Because 14.9 doesn't work for me)
8. Lepa G 1600 PSU (1600watts)
9. EVERYTHING WATERCOOLED
10. 2 x CrossOvers 27Q Led-P
11. 1 x Samsung U28D590D
12. Me

VIDEO OF THE BENCHMARK ON MY SETUP (Vsync ON)

_____________________________________________

Graphs and such

NORMAL SETTINGS

_________________________________________________
_________________________________________________



ULTRA SETTINGS ( Anti-Aliasing High , supersampling off )

And with SuperSampling 2,25

_______________________________________________
_______________________________________________



Now the scaling between each GPUs (This is very good scaling)

NOTE: I had to force Optimize 1x1 for the crossfire profiles for the 3rd and 4th card to be utilized. If I didn't do that, it was max 2 cards utilization.

 

[davie jambo]

I tested Mantle on this game myself the other day

I5 2500k & R9 280x
Ultra in dx11 was exactly 60fps average and on mantle it was 64.9

5 extra frames for free. Can't complain

I've already completed it so it's pointless , wish they had this on release

 

[NTMBK]

Wow, those minimum frametimes are impressive for Mantle. From 10fps to 18fps, that's a good improvement.

Sucks about the broken Crossfire though.

 

[KaRLiToS]

Wow, those minimum frametimes are impressive for Mantle. From 10fps to 18fps, that's a good improvement.

Sucks about the broken Crossfire though.

Yes, I wished we could compare Quad GPUs on DirectX VS Quad GPUs on Mantle.

Not for now I guess.

 

[wand3r3r]

CrossfireX will be interesting. How stable is the game?

 

[Wild Thing]

Thanks for your quality testing and results Karl.
You are a class act.:thumbsup:

 

[KaRLiToS]

CrossfireX will be interesting. How stable is the game?

The game is very stable, actually, I haven't seen any sign of instability whatever configurations I used.

And I used
DirectX: Single, dual, tri and quad cards. I tried 1000/1250 and up to 1225/1525

Mantle: Single card stock and 1225/1525

Thanks for your quality testing and results Karl.
You are a class act.:thumbsup:

Man, thank you so much. :$

 

[VulgarDisplay]

Wow, those minimum frametimes are impressive for Mantle. From 10fps to 18fps, that's a good improvement.

Sucks about the broken Crossfire though.

The raw max fps numbers for mantle aren't the big draw. Frame delivery is just ridiculously consistent in any of the games I have tried, and you can definitely feel it while playing.

 

[el etro]

Karlitos, do a BF4 mantle x DirectX comparison with your rig! BF4 may allow the 4-way CFx to work properly.

 

[NTMBK]

Karlitos, do a BF4 mantle x DirectX comparison with your rig! BF4 may allow the 4-way CFx to work properly.

Does BF4 let anything work properly?!

 

[KaRLiToS]

Karlitos, do a BF4 mantle x DirectX comparison with your rig! BF4 may allow the 4-way CFx to work properly.

I would but there is no benchmark utility for that game. I don't want skewed results.

 

[wand3r3r]

Thanks for the comparison. Clearly the mantle (amd) hate is unwarranted as it delivers.

I'll have to try it out.

 

[KaRLiToS]

Thanks for the comparison. Clearly the mantle (amd) haters will have to find something else to complain about.

I'll have to try it out.

Sniper Elite 3 with Mantle is not mature enough to show the gains, but still, there is an increase in performance and consistency. When it will support multi-gpu, it is gonna be even more awsome.

On the other side, Thief is really mature and I'm gonna do a lot of tests. If someone wants me to make a particular test, I will.

Honestly, I never really paid attention to Mantle or tested it. In fact I was sceptic of its capacity. The other thread about it forced me to do the testing myself. Hell, I never thought it was that awsome.

 

[wand3r3r]

I agree, it's a surprisingly big difference. Enough to make me curious and want to test it out some more.

I'm neutral on the whole concept but all of the hate on it is just pointing out it's benefits when investigated a bit, ironically enough. If it doesn't open up eventually I think it will be relegated to die (rightly so imo, proprietary doesn't help me as a bang/$ consumer) , but it has so many things going for it such as the console porting and extremely wide support for something that requires developers to get behind. Either way, it's something that I can test out so I will, just like I did for physx.

 

[Gloomy]

Hey Karlitos, the developer of this game mentioned in their Mantle announcement that because they get full control over what is in GPU memory, the Mantle renderer is more likely to have higher quality textures in memory more often compared to DX.

Do you notice a difference in IQ, or is it one of those things where they say it's a possibility but it's random and rarely happens?

 

[el etro]

Sniper Elite 3: AMD Mantle vs AMD DX11 vs Nvidia DX11.

source: Videocardz


EDit: Mantle didn't lowered the R9 285 FPS.

 

[Wild Thing]

WhyCry's quote from the VideoCardz review.

WhyCry VideoCardz.com Uncore • 3 days ago Thank you. I hope I will be able to test Mantle on low-end system some day. It's an amazing technology that should be implemented in all games, and it will at some point.

 

[KaRLiToS]

I will redo single card test with the CPU at 2.5 GHz and card at stock.

 

[MisterLilBig]

Interesting results on the 285, it goes to show how its dependent on the application/game. Just how its supposed to be.


(On a side note, anyone care to PM me about what happened to the other thread "Has Mantle failed to deliver"? I gave some information which I was looking upto some feedback on and it seems like the thread got deleted...? Would have preferred a lock. Why can't I find it? Why was it deleted?)

PM sent.
-- stahlhart

 

[AtenRa]

Interesting results on the 285, it goes to show how its dependent on the application/game. Just how its supposed to be.


(On a side note, anyone care to PM me about what happened to the other thread "Has Mantle failed to deliver"? I gave some information which I was looking upto some feedback on and it seems like the thread got deleted...? Would have preferred a lock. Why can't I find it? Why was it deleted?)

PM sent.
-- stahlhart

Heh, I was trying to find that thread as well today. PM me pls thx

 

[Granseth]

I really like how some of you fellow forum users shares your experiences. If it is benchmarks like this, or build(s) like Lava.

 

[KaRLiToS]

Hey guys, I tested the same bench as the normal settings one, with a resolution of 8044x1440 with my i7 4930k @ 2.4 ghz and the results are exactly the same.

Same percentage difference.

When I get the time Thursday or Friday I will bench Thief in every possible way but it's hard during the week cause I work 9.5 hours every day (4 days a week) and need to take care of my son during the evening.

I could find the time during the week but I prefer spending that time with my boy.

Thanks for your comprehension guys

 

[monstercameron]

Hey guys, I tested the same bench as the normal settings one, with a resolution of 8044x1440 with my i7 4930k @ 2.4 ghz and the results are exactly the same.

Same percentage difference.

When I get the time Thursday or Friday I will bench Thief in every possible way but it's hard during the week cause I work 9.5 hours every day (4 days a week) and need to take care of my son during the evening.

I could find the time during the week but I prefer spending that time with my boy.

Thanks for your comprehension guys

pfft you are so unreasonable, we are so much more important than your beloved son. :sneaky:

 

[KaRLiToS]

pfft you are so unreasonable, we are so much more important than you beloved son. :sneaky:

Hehehe.

And the funny thing is ,everyday, I just want to arrive home and do the bench because it's fun to do to finally realize that I really don't have the time.

 

[ocre]

I think Rebellion Developments did an amassing job implementing Mantle on sniper elite 3. I think one of the best parts is that I didnt see a bunch of prehype before hand,

I think they are really talented and made great use of the API