dirt 2 benchmarks dx9 vs dx 11

gareth170 · 25 Nov 2009 at 14:20

seems like people are to quick to say the 5x00 series is useeless for tesselation/dx11, 9.11 was the first non-beta driver that supports 5x00 which hasn't been optimized much. i really sure with fully optimized drivers we'll see these cards performance much better

Mr Paul · 25 Nov 2009 at 14:31

Cyber-Mav said:
thats the thing, they shouldnt run any slower with tesselation turned on for the ati cards, since the new 5 series cards are supposed to have a dedicated tesselation engine onboard. that means when a games doesnt use tesselation the tesselator unit on the gpu is inactive, and its active in games that use tesselation.

but it looks to me like ati are doing tesselation on thier gpu shaders thats why its taking a hit in performace. if it really was a dedicated engine on the gpu for tesselation then there would be no performance hit UNLESS the hardware implemented tesselation is so weak that it rest of the gpu ends slowing down in throughput due to the bottleneck caused by the tesselator.

either way wel will soon find out how things will pan out. but from the looks of it in those dirt 2 and uniengine tests these 5 series cards are useless for tesselation due to the massive performance impact they take. who knows it could be driver related but i guess only time will tell.

No it's because the hardware tesselator outputs shed loads more polygons and verticies which in turn will slow down the rest of the GPU as it's having to process much more data....

Secret_Window · 25 Nov 2009 at 14:33

I think I need to see more DX11 stuff to be more convinced the 58xx series is worth changing to at the moment. There's no doubt however that the future is DX11 though - it's just whether these cards are the best for it.

Owenb · 25 Nov 2009 at 14:42

Where are ye getting the idea that tessellation shouldn't cause a performance hit?

muon · 25 Nov 2009 at 15:37

People picking cards like the HD 5750/70 instead of the HD 4890 will be disappointed. People in this forum were claiming that in via dx11, it be faster than the hd 4890 even though the specs are worse. Common sense prevails.

Although, the difference between average fps and minimum fps isn't too much. So running the game at dx11 even at 30-40fps might not be that bad, if the graphical differences are significant enough.

gareth170 · 25 Nov 2009 at 15:51

u can't compare a 5770 with a 4890 in dx11 games because the 5770 would be using the full dx11 features where the 4890 would be using dx10.1 features

muon · 25 Nov 2009 at 16:00

gareth170 said:
u can't compare a 5770 with a 4890 in dx11 games because the 5770 would be using the full dx11 features where the 4890 would be using dx10.1 features

Didn't stop people claiming it would perform better.

The claim was HD5770 in dx11 faster than HD 4890 in dx9 (or dx10.1?).

Hairybudda · 25 Nov 2009 at 18:42

DirectX is a standard used to communicate with hardware, DX9 to 11 is like the move from DX6 to 8.1. The differences aren't as fundamental, it's not as revolutionary as moving from the fixed function pipeline to hw t&l and fragment shaders. 1st gen programming is always loose unoptomised garbage, especially when you're using it to tack useless eye candy onto an engine built for old hardware. Basically in summary dirt 2 is a useless benchmark. When you get DX11 engines being built from the ground up for DX11 hardware, that's when you'll see big numbers and pretty graphics, and by that point nv and ati will probably have a new generation of cards and mature drivers anyway.

gareth170 · 25 Nov 2009 at 19:03

ghost101 said:
Didn't stop people claiming it would perform better.

The claim was HD5770 in dx11 faster than HD 4890 in dx9 (or dx10.1?).

some people are saying the 4890 in dx11 is faster than a 5770 in dx11 which is totally crap because a 4890 can't do the full futures of dx11 it's just basicly running in dx10.1 when playing a dx11 game, so of course it looks faster but it isn't really if u think about it... if the 4890 had dx11 and could use the full features u would see a similar performance as a 5770.

craig3000 · 25 Nov 2009 at 19:51

Those screenshots are AWFULLL for making a decent comparison in this particular game, of course when a game fully utilizes DX11 and turns most of the level into tesselation only then will you see a good difference in a bad picture, I expect crysis 2 to deliver on using DX11 to its best

If these pictures showed off lighting more especially at night you'd see the difference more clearly between DX9 and DX11

phr4n · 25 Nov 2009 at 20:23

wow, so with my 5850 I can expect a MINIMUM of 40 fps playing a game with tessalation improved lighting and filtering effects - and that's with AA!!. That's pretty awesome in my opinion. I worry more about the fact that 3gb of ram is reccomended where as I am still using my XP elitist set up with 2gb

. It isn't something that has occured to me before as marketing benchmarks tend to focus on the GPU and CPU performance in games, nor is it something I have ever noticed as a "reccomended" amount in the specs. Fun times ahead.

HangTime · 26 Nov 2009 at 00:59

You need to remember that we are comparing DX11 with DX9 here, not DX10. DX11 might have comprable performance to DX10 (in titles where the latter is available).

Given that DX10 is (generally) regarded as being slightly slower than DX9, it's not really surprising that DX11 is slower than DX9.

LoadsaMoney · 26 Nov 2009 at 01:09

Yeah, but 30 frames slower

muon · 26 Nov 2009 at 01:09

gareth170 said:
some people are saying the 4890 in dx11 is faster than a 5770 in dx11 which is totally crap because a 4890 can't do the full futures of dx11 it's just basicly running in dx10.1 when playing a dx11 game, so of course it looks faster but it isn't really if u think about it... if the 4890 had dx11 and could use the full features u would see a similar performance as a 5770.

Where did anyone say the hd 4890 could run dx11? Everything I've seen has always been dx 10.1.

HangTime said:
You need to remember that we are comparing DX11 with DX9 here, not DX10. DX11 might have comprable performance to DX10 (in titles where the latter is available).

Given that DX10 is (generally) regarded as being slightly slower than DX9, it's not really surprising that DX11 is slower than DX9.

You should check the comparisons between dx9 and dx11 out. The differences seem to be just a subtle lighting effect and water effects.

http://www.pcgameshardware.com/aid,699998/Dirt-2-DirectX-9-vs-DirectX-11-graphics-compared/Practice/

LoadsaMoney · 26 Nov 2009 at 01:19

Its Nvidia's PhsyX all over again by the sounds of it, as they drop about 40 frames when thats switched on, and thats only a few subtle effects, e.g glass breaking, paper being blown about in a breeze etc....

muon · 26 Nov 2009 at 01:53

LoadsaMoney said:
Its Nvidia's PhsyX all over again by the sounds of it, as they drop about 40 frames when thats switched on, and thats only a few subtle effects, e.g glass breaking, paper being blown about in a breeze etc....

Well in this game, the effects seem to be subtle. But everyone saw what that tech demo showed with tesselation. The new stalker game also uses it pretty well to an extent and anandtech claim it performs better in dx11 than dx10.

jorzeh · 26 Nov 2009 at 02:11

i think you'll get a shock when the drivers are finished this is also early DX11 there will be much tweaking yet

wiak · 26 Nov 2009 at 07:27

LoadsaMoney said:
Oh dear, best ran in Dx9, as its about 30 frames faster than Dx11.

but did you SEE the freaking water effects? that cant be done EVER in DXcrap9

gareth170 · 26 Nov 2009 at 08:30

ghost101 said:
Where did anyone say the hd 4890 could run dx11? Everything I've seen has always been dx 10.1.

afew people on this forum

Martini1991 · 26 Nov 2009 at 08:32

gareth170 said:
afew people on this forum

4890's dx 10.1, which has a lot of DX11 features, but it can't run DX11.