methylenedioxy wrote:
http://www.driverheaven.net/articles...rk3/index2.htm

If you look at the differences in jumps between the various nvidia card
drivers you really do have to wonder what they are cutting back on. Look at
Ati, it jumped only 1000 points between cat 3.6 and 3.7, now look at the
equivalnet nvidia card driver jumps, between 44.03 drivers and 45.23 it
jumped by 11,000 POINTS!!! theres no way that they aren't skimping on
features, you just can't do that without really "optimising" drivers to such
an extent that something is bveing cut back on. Ansd then the jump between
45.23 and 51.75 drivers is another 8,000 points so in total between 2 driver
releases an astonishing jump of 19,000 points, sureley this speaks volumes
in itself????



Well ... as I have understood nVidia has done the FX GPU-s with
different pixel shader code than the DX9 spec says. That would mean that
when raw DX9 code path is used the card will perform not too well. Now
if they write a wrapper that will rewrite the pixel shader data to the
nVidia specific format then the hardware will run the code a lot faster.
Now if the nVidia written code is superior to the DX9 code (just let's
make an assumption). Then it would be possible to actually increase the
performance of games just by driver upgrades and not just 5% but 50% or so.

So why do you always assume that nVidia has to downgrade some other
options to optimize some game specific code in the drivers? Maybe they
just have to write in some converters that will give you the same output
(quality way) but with some additional work from the driver. Now if
there would be a game that uses directly the nVidia supported paths then
it might be a lot faster than using standard DX9 paths with ATI. You
don't know that. Noone knows that (excpet nVidia and some spies maybe).

So just keep an open mind regarding driver optimization and don't bitch
around.

Mario