Board and Chip Comparison Tables

NV35 used DDR2? my first reaction.

Core code seems to work--jeez, that will be handy.
 
R100 is 30 million transistors.

NV31 is listed twice, once as 2x2 and once as 4x1 - which one is it? And isn't NV34 PS/VS 2.0 'extended' ?
 
The Baron said:
NV35 used DDR2? my first reaction.

No, the chip chart is for the capabilities of the chip, not necessarily what is used at the board level. Click the Chip Codename link for further details on the chip and a list of boards - click on a board to see the details of the board (alternatively go to the board comparison chart and click on a board name for more board details).

The Baron said:
Code seems to work--jeez, that will be handy.

The code is mainly what I'm testing at the moment (although I do need input on the details), hopefully it will have a number of uses.

arjan de lumens said:
R100 is 30 million transistors.

Thanks.

arjan de lumens said:
NV31 is listed twice, once as 2x2 and once as 4x1 - which one is it? And isn't NV34 PS/VS 2.0 'extended' ?

Its listed twice because there is a Wirebond and Flipchip version. The 4x1/2x2 thing is because I entered them at different times and I'd changed my view on them during that time! They are slightly tricky to classify as much of the time they will operate as a 2x2, even though they are 4x1 in some circumstances.

digitalwanderer said:
What does "PS2.0 Extended" mean? What's the difference 'tween it and normal PS2.0?

PS/VS 2.0 'Extended' is the extended model MS added to expose more of FX's shader functionality - increased instruction lengths, branching etc.
 
If NV cards (or by the original pr spec: ps2.0+) are listed as ps2 extended, then shouldn't Ati R350 be listed as ps2 extended extended (ps2.0++ by original pr spec) :D
 
NV3x has the capabilities to expose the PS/VS2.0 Extended model - R3x0 exposes only PS/VS2.0 in terms of DX shader caps.
 
DaveBaumann said:
NV3x has the capabilities to expose the PS/VS2.0 Extended model - R3x0 exposes only PS/VS2.0 in terms of DX shader caps.
It makes the FX sound more advanced than the radeon in it's PS/VS capabilities though, and I thought that wasn't the case....am I wrong again?
 
The NV3x shader models do have a greater level of functionality than R3x0. There are other elements that DX9 exposes that R3x0 is capable of where NV3x is not, however these are not necessarily related to the PS/VS2.0 / 2.0 Extended shader models.
 
DaveBaumann said:
The NV3x shader models do have a greater level of functionality than R3x0. There are other elements that DX9 exposes that R3x0 is capable of where NV3x is not, however these are not necessarily related to the PS/VS2.0 / 2.0 Extended shader models.
Thanks, that's the nicest I've been told I was wrong in a loooong time! :LOL:
 
Great stuff!
I've bookmarked a few of the existing summaries on other sites. I knew I just should've waited for the B3D incarnation...! :D
 
DaveBaumann said:
NV3x has the capabilities to expose the PS/VS2.0 Extended model - R3x0 exposes only PS/VS2.0 in terms of DX shader caps.

Doesn't the R350(or is it 60?) support unlimited shader lengths?(at way reduced speeds)
 
If you sort by memory bit width it seems a bit confused by the (64xX).

Also I think the memory type sorts should put the ddrII/ddr chips at the top followed by ddrII then the ddr chips. But that might just be me. ;)
 
NV15 was also manufactured in 0,15 µm process for the Geforce 2 Ti. It had a core clock of 250 MHz and a memory clock of 200 MHz.

Another little correction: The chip in the GF2 Ultra was NV16, although it had no architechtural differences from (0,18 µm) NV15 (like NV35 and NV38). GF2 Ultra had a memory clock of 230 MHz, not 250 MHz.

Edit: And 3dfx Avenger had 8 million transistors
 
Fox5 said:
Doesn't the R350(or is it 60?) support unlimited shader lengths?(at way reduced speeds)

The R350 & R360 does have the capability in hardware for unlimited shader lengths via the F-Buffer but there is no way of exposing it through DX at the moment (and they haven't yet exposed it in OpenGL either).

snk said:
The RV200 is 30 million transistors as well.

Ta, updated.

AlphaWolf said:
If you sort by memory bit width it seems a bit confused by the (64xX).

Also I think the memory type sorts should put the ddrII/ddr chips at the top followed by ddrII then the ddr chips. But that might just be me. ;)

Bus width sorting should be fixed. Not sure what to do with the memory type sorting since its just done through MySQL alphabetical ordering - I could translate them to a number, but that would be extra processing.
 
About the code:

Two last chips seem to be always left out. Eg. When I sort by name, they are NV10 and Avenger, and when I sort by manufacturer, they are STG4500 and STG4800.
 
I would change the release date of the VSA-100 chips to Q2 of 2000. The V5 review unit, one of the initial 15 I believe were sent out, I didn't receive until mid- to late May IIRC.
 
snk said:
About the code:

Two last chips seem to be always left out. Eg. When I sort by name, they are NV10 and Avenger, and when I sort by manufacturer, they are STG4500 and STG4800.

Are you looking on the next page?
 
Back
Top