I no longer feel pressurised to go DX10

dizietsma

Banned
After reading Dereck's excellent piece on Anandtech I have suddenly realised that I do not have to go through the hassle of getting Vista and I do not have to upgrade to a top end video card on 90nm or 80nm that costs a fortune and has various compromises.

Which is a blessed relief to be honest. Instead I will play them at DX9 and then a year or so later play them at DX10 when the hardware is up to it, so getting 2x the value out of my games.

In the meantime I am going from 19inch to 24inch monitor and playing Far Cry yet again :) No matter how good the latest video card is nothing beats buying a bigger monitor and expensive comfy chair is at the end of the day.

Phew!
 
Well, I think "comprimises" could go both directions; ie DX9 vs DX10 tradeoffs. The G80 is a world-beater in DX9 apps, but there are a few drawbacks in DX10 (not huge ones, mind you, but still...) And you can make almost the exact opposite argument for ATI, but maybe with a bit more severity to the decisions.
 
What about R600 makes it a better DX10 architecture?

Well, GS capabilities, heavy HEAVY dependance on shader ALUs versus texel rate / fill rate (most "current" games are still quite texture-heavy), stuff like that. I don't mean to say the R600 is the bestest at DX10, but they're still a bit unbalanced -- even more so than the previous R5xx line. That balance appears to be pointed in the direction of "future" compatibility, even though (IMO) by the time that future comes long, the 2900HD will be at the bottom rung.
 
Well I don't think the verdict is in on the GS yet though we have some pretty compelling evidence of G80's limitations in that area. Drivers or design issue though? And in terms of shader power I don't consider that a DX10 specific attribute at all since DX9 games will benefit just as much. I'm sure Crysis in DX9 mode will still put the hurt on shader hardware.
 
Ok, so howabout texel and fillrate limitations? Not a huge deal in DX10 (need for multipass is much-diminished, need for textures should theoretically be dimished due to excess shader power). Those two traits aren't conducive to DX9 titles, and the G80 has both in considerably better supply...
 
Davros totally agrees with dizietsma

With maybe the exception of the high end cards which are out of my budget anyway i dont think the first gen of dx10 cards are up to to job of playing the first gen of dx10 games..
 
Every day I love my X1950 Pro purchase a little more. Its a wonderful card, it was a wonderful deal. I have been tempted by the 8800 GTS but it simply doesn't really make sense considering what I've been playing lately.
 
Ok, so howabout texel and fillrate limitations? Not a huge deal in DX10 (need for multipass is much-diminished, need for textures should theoretically be dimished due to excess shader power). Those two traits aren't conducive to DX9 titles, and the G80 has both in considerably better supply...

Don`t fall for the party line, it`s not quite like that. DX10 doesn't magically remove the need for texturing oomph, and we're not going to get procedurally generated everything...umm, probably ever. The need for textures will certainly not diminish, it will go up as well.
 
Don`t fall for the party line, it`s not quite like that. DX10 doesn't magically remove the need for texturing oomph, and we're not going to get procedurally generated everything...umm, probably ever. The need for textures will certainly not diminish, it will go up as well.

I agree and want to add that most of this “this game have a high shader calculation demandâ€￾ presentations are not fully correct. In most cases these slides compare shader ALU instructions with shader TEX instruction. It is not unlikely to see 10:1 numbers there. But this is not the value we need. Both sides need to multiply by the number of cycles that are needed to execute these instructions. With 16x AF a tex instructions can need up to 32 cycles but 4-5 seems a good mean value. On the other hand ALU instructions can often bind together. This will give us less than 1 cycle per instructions. If we assume a factor of 0.75 here we got a 7.5:5 or 1,5:1 usage. In this case even two ALUs per texture unit would be enough.
 
Actually the article had the exact opposite effect on me to what it intended. It improved my opinion of DX10!

I own a GTS and had thought that DX10 useless for me on the current games but it turns out that in fact I actually get a little more (or at least no less) speed in Lost Planet and I can run COH DX10 with no problems while having the game look quite a bit nicer. Sure its much slower but still well over 30fps and thus "fast enough". At over 30fps in an RTS I will take the eye candy any day! :D

I do take issue to the articles whining about Lost Planet though. Far from being a crappy console port its actually a very well transfered game if you use the 360 control pad. Some people may say thats not the point, its on PC so it should work well with the keyboard and mouse but I disagree. This game was designed from the ground up to work with the 360 control pad, all the movement, speed and balancing was designed with it in mind and it works well for it. Key/mouse control would screw that up so I don't see why the devs should make any special effort to optimise for it when all gamers need to do is go out and buy the 360 pad. It is a recommended requirement on the box afterall and like it or not, many more games will be using it in the future so it pays to have one.

Another issue I had with the article is the conclusion which reads:

"For now, AMD does seem to have an advantage in Call of Juarez, while NVIDIA leads the way in Company of Heroes and Lost Planet."

Yet go take a look at the CoJ benchmark and there's the 2900XT sitting in 3rd place behind the GTX and Ultra in every test... WTF?

Although I guess they are comparing to the GTS but its still a bit of a cop out.
 
I do take issue to the articles whining about Lost Planet though. Far from being a crappy console port its actually a very well transfered game if you use the 360 control pad. Some people may say thats not the point, its on PC so it should work well with the keyboard and mouse but I disagree. This game was designed from the ground up to work with the 360 control pad, all the movement, speed and balancing was designed with it in mind and it works well for it. Key/mouse control would screw that up so I don't see why the devs should make any special effort to optimise for it when all gamers need to do is go out and buy the 360 pad.
That would be fine if the game came with a free 360 pad in the box. But to oblige people to go out and buy non-standard PC hardware to play the game at all is not reasonable. It was the correct decision to optimise the Xbox version for the Xbox controller. It was the wrong decision to optimise the PC version for the Xbox controller.

What worries me about this Anantech article is this claim:

Relic didn't simply recompile their DX9 code for DX10; Company of Heroes was planned for DX10 from the start
If that is correct then it is particularly worrying that R600 takes more of a hit going from DX9 to DX10 than G80 does. ATI's only possible remaining defence for R600 is that is somehow "optimised for DX10" - in other words that, on games designed for DX10, it will have an architectural advantage over G80. Granted, this is only one game, but nonetheless this piece of evidence suggests the exact opposite, namely that G80's advantage over R600 is actually even bigger under DX10 than it is under DX9.

We'll see.
 
That would be fine if the game came with a free 360 pad in the box. But to oblige people to go out and buy non-standard PC hardware to play the game at all is not reasonable. It was the correct decision to optimise the Xbox version for the Xbox controller. It was the wrong decision to optimise the PC version for the Xbox controller.

I would argue that a decent pad should be a standard piece of kit for a gaming PC. At least a gaming PC thats being used to play games which were originally designed to be used with a pad and which work better that way (sports, driving, platform, console orientated TPS's etc...).

Its just my opinion but I am much happier with the game playing exactly like it was originally designed to play on the 360 than with them rejigging the controls for Key/Mouse, to give a new scheme which is inferior to the original but also hurts anyone who want to play the game on a pad. But thats just me!
 
to give a new scheme which is inferior to the original but also hurts anyone who want to play the game on a pad. But thats just me!

you can have both you know switching to the keyboard and mouse scheme should not have any effect on how the pad scheme works - and thats why its wrong

ps: i assume this game works great with any pad ? not just the 360 one ?
 
Back
Top