**The New Official THIEF-Incorporating Bench mark Thread**

[Kaapstad]

I think you missed this one I posted earlier too.

Two card 1600p entry

2 x 290Xs @1230/1625

4930k @4.8

Spoiler

 

[LtMatt]

Finally i can beat Kaap at something, yay.

 

[humbug]

I think you missed this one I posted earlier too.

Two card 1600p entry

2 x 290Xs @1230/1625

4930k @4.8

Spoiler

Sorry. i'm on it...

 

[humbug]

Done. i think...

 

[Kaapstad]

How do your DX titans compare to your Mantle 290's Kaap in Thief in terms of fps averages? Has Mantle tilted the performance of Unreal Engine (always a Nvidia favourite) back towards AMD? There's a shortage of Nvidia dual/tri gpu runs to compare to. The only one i could find were Scottys and Gregs runs from the old thread.

The 4 card 1600p Titan score is on the scoreboard and bottlenecked badly. I will need to turn them loose at 4K to really do a comparison. It is difficult for me to do 2 and 3 card Titan runs as the side of my LDV8 case is bolted on and I need to get in there to change the SLI connectors.

 

[LtMatt]

The 4 card 1600p Titan score is on the scoreboard and bottlenecked badly. I will need to turn them loose at 4K to really do a comparison. It is difficult for me to do 2 and 3 card Titan runs as the side of my LDV8 case is bolted on and I need to get in there to change the SLI connectors.

Ah that's a shame. I will ask Greg and Scotty to submit some entries then. It would be nice to have something to compare with to see if Mantle is offering equal or better performance than Nvidia DX in SLI in this game.

 

[Kaapstad]

Finally i can beat Kaap at something, yay.

Yes there IB-E CPUs are really crap at overclocking, I am not very hopeful for Haswell-E with the 5960X coming in at 3.0ghz lol.

 

[LtMatt]

Yes there IB-E CPUs are really crap at overclocking, I am not very hopeful for Haswell-E with the 5960X coming in at 3.0ghz lol.

It's odd that i can score higher than you with a technically slower cpu, memory, motherboard and gpu's.

 

[Kaapstad]

Ah that's a shame. I will ask Greg and Scotty to submit some entries then. It would be nice to have something to compare with to see if Mantle is offering equal or better performance than Nvidia DX in SLI in this game.

Hazard a guess DX11 wins with NVidia cards on the single GPU 1080p scores and Mantle will win with 2, 3 and 4 GPU setups.

 

[Kaapstad]

It's odd that i can score higher than you with a technically slower cpu, memory, motherboard and gpu's.

This game is all about raw CPU clockspeed even when using Mantle. Fast RAM does not really help in games either.

Having said that it will be interesting to see how things go at 4K where there is less fps but more data per frame.

 

[Kaapstad]

Entry

4 x 290Xs @1200/1625

1600p

4930k @4.8

Spoiler

Score is not much different to the 3 card score as there is a serious bottleneck.

QuadFire is working (check the max fps) but until I run at 4K there wont be much difference between 3 and 4 cards scores.

 

[LtMatt]

Thanks for the feedback Kaap. I will do what i can through official feedback to get crossfire performance improved.

 

[Wilko_Dog]

Entry

4 x 290Xs @1200/1625

1600p

4930k @4.8

Spoiler

Score is not much different to the 3 card score as there is a serious bottleneck.

QuadFire is working (check the max fps) but until I run at 4K there wont be much difference between 3 and 4 cards scores.

Nice score , how does it look at 4k ? I think the art style in this game is really nice bet it looks sweet at that resolution !

 

[Kaapstad]

Thanks for the feedback Kaap. I will do what i can through official feedback to get crossfire performance improved.

There is not much wrong with quadfire, it just needs to be used at 4K to show the difference. I also think that the scaling at 4K will be around 90% or better.

If you check the difference between the max fps on my 3 and 4 card runs there is a good indicator as to how quadfire will perform at 4K once the CPU bottleneck is removed.


3 cards max fps 158.4
4 cards max fps 194.2

Increase in max fps 35.8

 

[Kaapstad]

Nice score , how does it look at 4k ? I think the art style in this game is really nice bet it looks sweet at that resolution !

I still have to get a monitor first.

 

[LtMatt]

There is not much wrong with quadfire, it just needs to be used at 4K to show the difference. I also think that the scaling at 4K will be around 90% or better.

If you check the difference between the max fps on my 3 and 4 card runs there is a good indicator as to how quadfire will perform at 4K once the CPU bottleneck is removed.


3 cards max fps 158.4
4 cards max fps 194.2

Increase in max fps 35.8

It's up to the developer by the way to improve this. There is very little AMD can do as the responsibility of the driver is now minimal in Mantle. I feel bad for pestering Thracks. I've signed up to the Eidos forum but what are the chances of them listening to some newb from the internet who says the potential is there, but you need to work on 3/4 card scaling/performance under Mantle, or use Mantle more efficiently to lower than cpu bottleneck in 3/4 card setups? Will let you know if they respond to me.

 

[Wilko_Dog]

I still have to get a monitor first.

Ah I see ... Get one ordered haha

 

[Kaapstad]

It's up to the developer by the way to improve this. There is very little AMD can do as the responsibility of the driver is now minimal in Mantle. I feel bad for pestering Thracks. I've signed up to the Eidos forum but what are the chances of them listening to some newb from the internet who says the potential is there, but you need to work on 3/4 card scaling/performance under Mantle, or use Mantle more efficiently to lower than cpu bottleneck in 3/4 card setups? Will let you know if they respond to me.

I don't think there is anymore Mantle could do, it is already noticeably quicker than DX11 in Thief. I don't think there is anymore overhead for Mantle to remove.

Mantle can not make up for a badly written game, it can only make it as easy as possible for the CPU to run it.

Mantle is already about 10% faster than DX11 running multi GPU setups in Thief.

 

[LtMatt]

I don't think there is anymore Mantle could do, it is already noticeably quicker than DX11 in Thief. I don't think there is anymore overhead for Mantle to remove.

Mantle can not make up for a badly written game, it can only make it as easy as possible for the CPU to run it.

Mantle is already about 10% faster than DX11 running multi GPU setups in Thief.

Maybe im expecting too much. Still it will be interesting to see what they say and if they respond. It would be magically to see Mantle showcase good three to four gpu scaling at 1080P/1440P/1600P in a game that runs so horribly (scaling wise) on like minded three to four way DirectX gpu setups.

 

[Kaapstad]

Maybe im expecting too much. Still it will be interesting to see what they say and if they respond. It would be magically to see Mantle showcase good three to four gpu scaling at 1080P/1440P/1600P in a game that runs so horribly (scaling wise) on like minded three to four way DirectX gpu setups.

Wait until I get a 4K monitor and then you will see some good scaling.