Should DXx have minimum performance requirements?

[nelg]

Should DXx have minimum performance requirements?

The issue of releasing features first, with sometimes anemic performance, has both advantages and disadvantages, IMHO. Using SM3.0 as an example we had NV releasing the first SM3.0 capable GPU. This is certainly good for developers as the sooner they have new hardware the more quickly they can experiment and develop for it. OTOH when we have large disparities in performance between different IHV's implementations (PS branching for example) does this effect or even delay when wide scale usage will occur? If so and if practical do you think that MS should have minimum performance requirements for such features?

 

[SugarCoat]

no of course it doesnt. Look at the NV30 and DX9

 

[neliz]

Leveraging performance and do WHAT to the customer base?
Price over performance is the way it goes and as long as DX9 hardware gets a "Ready for Vista" Sticker, no one will shell out $25 extra on a graphics processor that is more capable.. it's just not going to work at the low end..

 

[Geo]

Entirely impractical. You're effectively asking that MS enforce performance standards for other developers apps. . .

Prolly the most they could do is enforce some kind of performance standard for accelerating the GUI.

But generally, "this is what reviews are for".

 

[JHoxley]

With D3D10 having fixed-caps the IHV's are pretty much left with only the performance angle to work on. So I'd imagine it'll become less of an issue in the future.

I personally blame the GeForceFX line of products for the late adoption of SM2 as a baseline/common target. Sure, they had SM2 support but a lot of the common variants just plain sucked on the performance front. I've known more than enough developers who treat the FX5200's as SM1 parts.

Jack

 

[ERK]

In other words, should FPS be part of WHQL?

 

[nelg]

I personally blame the GeForceFX line of products for the late adoption of SM2 as a baseline/common target. Sure, they had SM2 support but a lot of the common variants just plain sucked on the performance front. I've known more than enough developers who treat the FX5200's as SM1 parts.

Jack

That is the point I am trying to address. Seeing that these design decisions can live into following generations, is the rush to get a check mark feature a hindrance in later generations? If so would mandating a minimal performance level (if possible) alleviate the problem.

 

[Deathlike2]

That is the point I am trying to address. Seeing that these design decisions can live into following generations, is the rush to get a check mark feature a hindrance in later generations? If so would mandating a minimal performance level (if possible) alleviate the problem.

The best overused phrase I've seen a lot best explains it...

"It depends."

The NV3x series had practically unacceptable PS performance... of course it was worse on the low end of things. The 5900 had usable performance (in limited circumstances) whereas the 5200 was complete trash when it came to supporting SM2...

I have no idea the performance characteristics would be if dynamic branching was "hacked into the R3xx or R4xx series"... we have still yet to figure out how good or bad the NV4x/G70 is at dynamic branching.. it's still up in the air...

 

[Ailuros]

In other words, should FPS be part of WHQL?

I vote for SPF, since we should be talking about unlimited resources

 

[Fox5]

It's very possible MS will impose performance requirements. Intel's GMA900, despite being DX9 capable is not considered Aeroglass compatible I believe. In addition, Microsoft is trying to push a more focused platform on the PC, sort of an open environment style xbox, so minimum performance requirements could help that. (though most pc games will probably be using xbox 360 hardware as the base requirement anyway)

 

[AndrewM]

The ideal situation should be just to scale resolution.