Yes, because I owned a Geforce FX, and I played HL2 in DX9 on it, in both partial precision and full precision! The game was quite playable even in full precision even though my FX card was only a 5900 XT.

Those who have never personally owned a GeForce FX should refrain from commenting on their capabilities since 99 time out of 100, they have no clue what they are talking about!

http://img836.imageshack.us/img836/6320/dscf0005wj.jpg


http://img600.imageshack.us/img600/5236/dscf0006t.jpg

There is and never was no such thing as a DX9 24-bit standard! That is pure fanATIc FUD! There is only full precision and partial precision.

If 24-bit was "the standard", how come ATi changed from "the standard" to 32 bit with r5xx?

Hint: Because 32 bit was always the 32 bit standard, despite what ATi would have you believe. The DX9 shader model roadmap was laid out long before there were any actual DX9 GPU's in existence. In other words the specs for SM 3.0 were known years before nv40 was manufactured and by all DirectX graphics vendors. nvidia designed their entire DX9 GPU line to use the SM3.0 spec as the basis for all of their FP formats, ATi didn't.

nv40 came as such a rude shock to ATi because although they knew about SM3.0, they didn't think anyone would bring it to market as rapidly as nvidia did. They had barely started to layout their own SM3.0 designs. Serves them right, for thinking nvidia was incompetent and wouldn't be able to catch them up.

The initial DX9 spec (not 9.0c/SM3.0) called for at least 24-bit, allowing for 16-bit precision for low-detail stuff in whatever areas, but to make it stand out, 24-bit was good enough for the time being, and VERY effective. Your FX 5900XT could barely do it at 1024x768 playable, with 0xAA, 0xAF, while a 9700 Pro released 2 years earlier could do it at over 60fps, at a higher rez with higher settings. True, HL2 did look fine in DX8.1 for the most part, as my brother played it nicely on his Geforce 4 Ti.. On an FX card, it would've played just fine with 16-bit precision in 95% of the parts if the game was properly coded for it (while 5% actually needed full precision), but Carmack himself said that it was a disaster coding a proper rendering path for the FX series for Doom 3 so Gabe Newell wasn't the only one complaining.. I didn't care about the water in HL2 anyways.

It was Far Cry that was the true showcase of R300 graphics, from the beginning.

Well, I don't care about it really anymore.. Such a big deal is being made about some silly "specific" visuals - 20 years down the road, we'll look back and think how ridiculous it was to make such a fuss over having a tiny bit better-looking water reflection or marble tile/glass reflections.

I just want a curved screen like this:

in 3D, with Kinect 10.0 and DirectX20, plus improved eagle eyes with 20/3 vision (7 times more keen than human 20/20), and implanted eyes on the sides of my head for 360-degree vision, plus the ability to fly and to feel everything out there to the maximum pleasure.
http://www.mycockpit.org/forums/showthr ... ved-screen

Here you go again, making crazy assumptions! I never played any game that needed DX9 shaders on my 5900 XT as low as 1024x768! Which game are you referring to?!

FP16 was smoother in HL2 than FP32 was, but not by a really huge amount, and aside from the giant TV screens there were no IQ issues that I can think of caused by FP16. The whole FP16 reduces IQ thing was a giant lie unless you were using FP16 to calculate texture addresses when you should have been using full precision, then it could cause damage. For color data you couldn't tell the difference.

Somewhere in amongst all my data files there are some comparison pics I did between "normal" HL2 on 5900XT and FP16. If I find them I'll post them, though they are probably linked in a B3D thread somewhere assuming they didn't get deleted with all the other "embarrassing" stuff that occurred soon after Baumann left to work full-time for ATi.

Did find a couple of pdf clippings though (no images in them, sadly).

That was a trip down memory lane


When will Kepler launch?

Game developers conference (GDC) is March 5-9 and OBR is saying that's where NVIDIA will launch Kepler, and it won't be a mainstream part but high performance.



I'm starting to think apoppin is right, kepler will launch sooner then we think.

http://www.obr-hardware.com/2012/02/gdc-2012-at-san-francisco-will-be.html
http://www.gdconf.com/

i just bit my tongue

http://www.obr-hardware.com/2012/01/exc ... epler.html

this is funny as heck

They had better launch it Pronto then! It the 27th here in Australia, only two days left of February!
So this guy says new wonder architecture, nvidia says part of Fermi family/architecture. Hmmm, who should I believe?