Minion: The point is the GeforceFX is a crock of sh*te.
In summary;
You: already made a hasty judgment.
- "GeforceFX=surprisingly slow"
- "GeforceFX is a crock of sh*te"
Me: wait for the final release, before making a judgment.
I don’t think “sitting on the fence” = defending nVidia in this case. A real Nvidiot would say "nVidia rulez” at every opportunity .
I wonder who is the fan boy now, when the final card was not even release yet.
Minion: 20 odd days aint goona get a huge performance increase in all likelyhood, is it? (and yes I know its been done before)
AND
Minion: I know, but thats not the point. This is the best Nvidia can come up with.
Not in this case, when a revised GeF FX(refer to
www.hardocp.com) and newer Detonator drivers exist after majority(v42.6x) of the sample (engineering release) reviews.
Via guru3D.com
Detonator 42.86 was dated at 2/1/03 6.12/10
Detonator 42.81 was dated at 2/1/03 5.25/10
20 days is quite alot for nVidia's case.
The asserted claim for “This is the best Nvidia can come up with” is simply false.
We don’t know the inter-workings of nVidia labs.
PS; Note the date on the revised GeF FX.
The processor has more transistors than the R300, yet it's slower clock for clock than the R300.
That kind of argument doesn’t stick with Intel’s Pentium 4 btw. The packaged overall performance is more important.
There are reasons for the increase amount of transistors. Which I have given.
But you responded by;
I checked your link - personally I dont care if the R300 ONLY has 96 bit pixel shader precision, that still allows 7.9x10^28 different positions. I think thats plenty.
Sounds like “640kb is enough for everybody” statements…
I see you haven't not taken this into an account.
GeF FX for the following;
1. 1024 Texture address operations per pass
2. 1024 Color instructions per pass
They would need extra transistors IF they support that feature in hardware.
Can you repeat your rhetoric IF ATI includes their own "128bit pixel shader precision" and support for "1024 Texture address operations per pass"/"Color instructions per pass" features in their next Rxx0 release?
I wonder who is the fan boy now.
Refer John Carmack's related statements regarding "maximum
instruction count" and "program limits on the R300".
Trevor Wilkin also echos a similar statements as with John Carmack.
(Trevor Wilkin is Lead Programmer for
Microsoft - Salt Lake City group)
I suspect this will be the same for DX9+ games aswell, driver optimisations or not.
Do you have basis for this?
As for anandtech.com's review...
As
www.hardocp.com has indicated, Anandtech just embarrassed itself by presenting the sample GeF FX as the final review. The rush to be the first review for a particular product hardly equals quality review…
