The First DX9 "Game" Benchmark... Well, Nearly

[Reverend]

Also, can you guys quit with the bad [H] stuff and about Brent in particular? I don't like seeing that on our site.

 

[John Reynolds]

Also, can you guys quit with the bad [H] stuff and about Brent in particular? I don't like seeing that on our site.

Agreed, though I'd like to add that there's nothing wrong with timely and constructive criticism of a site or particular reviewer's words (including B3D), and I'd still like to see those relegated to the General Discussions forum. Habitual pot-shots at other sites, though, just aren't necessary here.

 

[Typedef Enum]

OK, does that include Anand/Tom too?

 

[Ostsol]

In any case, it gets old rather fast. . .

 

[Hellbinder]

To give your readers another glimpse of the technology behind Gun Metal, and to generate a little buzz for Gun Metal in the process, Yeti Studios is pleased to announce the availability of the Gun Metal benchmark. The benchmark utilises our the game engine to generate 3D scenes that utilise a broadly-overlapping set of DirectX 9.0 features. As Gun Metal is a real game, this benchmark presents real-world demands of users' hardware for an accurate performance measurement. We hope you decide to incorporate it into your test suite. Obviously the Benchmark includes a number of features and functions that would not have been sensible to include in the retail version of the game without sacrificing compatibility. We do believe however that the Benchmark really pushes today's latest hardware with real game requirements. Our intention was to build a Benchmark that included the functionality of our dreams with no regard to installed base.

Ok what broad set of overlaping DX9 features????

What features here could not get included in the retail version???

Look I dont see anything in there that looks even REMOTELY Dx9ish. Low rez Textures, what looks like simple EMBM to make the Gun Metal Shiny, and flat out DX7 looking explosions.

Further, how can they proudly present this as a valid benchmark to Reviewers when its obvious they either did not test it on a Radeon card OR REleased it knowing its all screwed up on the radeon card. Heck isnt it a little toasty on Nvidia cards to???

 

[Moloch]

Not much of a benchmark. Should have used ps 1.4 min, and pref. 2.0!. Though I don't know what they could do extra, or do more effiently.

 

[Typedef Enum]

I get a nice GPF everytime I run it. It never gets beyond a small pause, and then...a MessageBox.

This is on a 9800 Pro + WinXP.

 

[Typedef Enum]

On a sidenote, this is really starting to remind me of DroneZ.

That game was one of the alltime worst games that I've ever played...I couldn't believe how pathetically bad it was. And because it was sporting a couple of new technologies, everybody had to include it in their benchmark suite.

 

[Kristof]

Hmmm, had a quick peek at the vertex shaders used and quite a few are declared 2.0 for no real reason (@ first glance no unsupported instructions and instruction count should be ok as well). There is one large shader though which has an instruction line count of 175 which would be above the 128 instructions supported by VS1.1. So I guess that makes it a VS2.0 test... oh and there is also a shader containing sincos which is not in 1.1

But don't know why this gets a VS2.0 header :

vs_2_0

def c5 , 0.250000, 0.500000, 0.750000, 1.000000

def c6 , -24.980801, 60.145802, -85.453796, 64.939400

def c7 , -19.739201, 1.000000, -1.000000, 0.159155

def c4 , 0.001000, 2.000000, 0.000000, 0.000000

dcl_texcoord1  v2 
dcl_texcoord0  v1 
dcl_position0  v0 
mov oT1.xy , v2.xyxx 
dp4 r0.x , c3 , v0 
mov oPos , r0.xxxx 
dp4 oPos.x , c0 , v0 
dp4 oPos.y , c1 , v0 
dp4 oPos.z , c2 , v0 
mov oPos.w , r0.xxxx 
mov oT0.xy , v1.xyxx 
mul r0.x , v0.yyyy , c4.xxxx 
mul r1 , c7.wwww , r0.xxxx 
frc r1.y , r1.xxxx 
slt r4 , r1.yyyy , c5 
add r0.xyz , r4.yzww , -r4.xyzz 
mov r4.yzw , r0.xxyz 
dp3 r1.z , r4.yzww , c5.yyww 
dp4 r1.w , r4 , c5.xxzz 
add r0.xy , r1.yyyy , -r1.zwww 
mul r1.xyw , r0.xyyy , r0.xyyy 
mul r0.xyz , r1.xyyy , r1.xyyy 
mov r3.xz , r1.xyyw 
mov r3.yw , r0.xxzy 
dp4 r2.x , r4 , c7.yzzy 
dp4 r2.y , r4 , c7.yyzz 
mad r0 , c6.xyxy , r3.yyww , c6.zwzw 
mad r1 , r0 , r3.yyww , c7.xyxy 
mad r0.xyw , r1.xzzz , r3.xzzz , r1.ywww 
mov r1.xz , r0.xyyw 
mul r2.xy , r2.xyyy , r1.xzzz 
mul r0.x , r2.xxxx , r2.xxxx 
mul r0.y , v0.xxxx , c4.xxxx 
mul r1 , c7.wwww , r0.yyyy 
frc r1.y , r1.xxxx 
slt r2 , r1.yyyy , c5 
add r0.yzw , r2.yyzw , -r2.xxyz 
mov r2.yzw , r0.yyzw 
dp3 r1.z , r2.yzww , c5.yyww 
dp4 r1.w , r2 , c5.xxzz 
add r0.yz , r1.yyyy , -r1.zzww 
mul r1.xyw , r0.yzzz , r0.yzzz 
mul r0.yzw , r1.xxyy , r1.xxyy 
mov r3.xz , r1.xyyw 
mov r3.yw , r0.yywz 
dp4 r0.y , r2 , c7.yzzy 
dp4 r0.z , r2 , c7.yyzz 
mad r1 , c6.xyxy , r3.yyww , c6.zwzw 
mad r2 , r1 , r3.yyww , c7.xyxy 
mad r1.xyw , r2.xzzz , r3.xzzz , r2.ywww 
mov r2.xz , r1.xyyw 
mul r0.yz , r0.yyzz , r2.xxzz 
mad r0.y , r0.yyyy , r0.yyyy , r0.xxxx 
rcp r0.x , c4.yyyy 
mul r0.xyw , r0.yyyy , r0.xxxx 
mov oT0.z , r0.xyxw

Then again it is before 8am in the morning 8)

The PS's seem to be quite sad for a "DX9" benchmark given that they seem to implement basic blend modes, then again not all that much more you can do while sticking to PS1.1

Oh and the app crashed halfway through for me as well... sigh... so I might not have seen all the shaders even though you normally create all your shaders up front and not mid-game.

K-

 

[Randell]

The PS's seem to be quite sad for a "DX9" benchmark given that they seem to implement basic blend modes, then again not all that much more you can do while sticking to PS1.1 K-

rubbish, theres nothing you can do with PS 1.4 you can't do with PS 1.1

I know two things about this game;

i) The demo was nVidia exclusive
ii) There are loads of used xbox ones for sale at my local Game store.

 

[LeStoffer]

both games runs fine on my 9700 pro, pretty ironic

Why do you think you know how the games was really meant to be played? Maybe stuttering, crashes and random reboots is that special flavor that are built into the games to really get you in an intentional state of madness?

Who want a game these days that runs plain old borring 'fine'? Blah!

 

[Ante P]

Here are my scores. Game at 1024x768 with 4x FSAA and no AF

Benchmark 1
Min 10.46
Average 17.92
MAX 43.01

Benchmark 2
Min 9.89
Average 17.92
MAX 48.66

According to the FAQ Aniso is always forced in this benchmark.

Could anyone with a 9800 Pro try out this benchmark at the default settings please? (which is 2x FSAA chosen in-game and whatever AF mod the game selects)

Strange that you got so low scores when the game actually runs better on a 9700 than my 5900

 

[just me]

Hats off to Brent for his honesty 8) & Rev & JR for their integrity.

Interested in how well the bench stresses the cards, becuz it sure looks ugly to me > based on screenshots [9700Pro = RIP]

 

[Dave Baumann]

Sent this in reply:

Could I ask if you’ve done any compatibility testing on this benchmark at all? As it stands there are numerous graphical errors on Radeon boards. Given that the earlier “NVIDIA only†demo could be made to run on Radeon’s without these graphical glitches there doesn’t appear to be anything to suggest that Radeon boards should not be able to run without these errors.

Anyway, I hope you have already been in contact with ATI’s developer relations if you are unable to resolve these graphical glitches internally, however if not I’ve copied this to Richards Huddy’s email address and I’m sure he’d be happy to assist you if you are having trouble. As it stands I would not be comfortable using this as a benchmark when it doesn’t render correctly on the majority of DX9 boards presently in use.

On another note, could I ask exactly what the requirement for VS2.0 is? Given this was an Xbox title, what has changed from the Xbox to the PC that requires VS2.0? Also if this is being billed as a DX9 benchmark why does it not utilize PS2.0? With only PS1.1 in use I would categorize this more as a DX8 benchmark given that VS2.0 can be achieved at reasonable rates across a CPU.

Let see what the responce is...

 

[Fuz]

As it stands I would not be comfortable using this as a benchmark when it doesn’t render correctly on the majority of DX9 boards presently in use.

Good call.... I wonder what the figures actually are, now with nV shipping the FX5200 series, I am sure nV are catching up.

About the benchmark... this is how I see it:
An Nvidia exclusive title that has now become a benchmark that has rendering problems on ATI cards.... doesn't quite sit well with me. Jmho of course.

 

[CorwinB]

Personnaly, I refuse to buy or download any game that is part of the "The way it's meant to be played" program : this thing is destroying the PC gaming market, which is based on non-proprietary APIs and vendor independance. Note that would ATI do a similar program, I would refuse buying any game part of it too.

And even if the game runs fine on non-NV hardware (like UT2K3), I still hate getting stupid advertising (splash logo) in a product I've paid for.

Just my 0.02Euros anyway...

 

[Ante P]

Personnaly, I refuse to buy or download any game that is part of the "The way it's meant to be played" program : this thing is destroying the PC gaming market, which is based on non-proprietary APIs and vendor independance. Note that would ATI do a similar program, I would refuse buying any game part of it too.

And even if the game runs fine on non-NV hardware (like UT2K3), I still hate getting stupid advertising (splash logo) in a product I've paid for.

Just my 0.02Euros anyway...

Have fun not playing HL2

I dunno why but I also have a hunch that cardmack also might go the TWIMTBP way.

 

[Dave Baumann]

NVIDIA have paid activision for co-marketting of Doom3, not for a small amount either ($5Mill is the figure I hear).

HL2 will probably go the opposite direction, although not to the same degree.

 

[John Reynolds]

Jason Cross has an editorial on page 99 of the July issue of Computer Games magazine expressing his concerns on this issue.

It is my sincere hope that the EA/NVIDIA partnership amounts to little more than a cross-marketing campaign. I don't want to see menu items on my EA games that turn on features only for NVIDIA GPUs. It's a step backwards for the whole graphics industry, and splits apart a market that has struggled long and hard to make meaningful standards everyone can agree on. If it is successful from a marketing perspective, it starts us down a slippery slope toward even more confusing hardware requirements, incompatibilities, support only for individual brands, and all the other ills of the PC market that increasingly push gamers toward console systems.

Couldn't agree more.

 

[demalion]

Hey, Kristof, I'm just curious...please forgive if it's a silly question...

Were you running this on an nVidia or ATI card?