ATI responses about GeforceFX in firingsquad

[andypski]

So, are you saying that on a long, compute bound shader NV30 is getting high utilisation from its memory bus resources?

Since it's compute-bound one would suspect not.

Hu? I've understand the contrary...

Then I would suggest you've misunderstood

Democoder's comment was that on long shaders R300 is unbalanced because it gets no use of its memory bus, with the implication that NV30 is better balanced for this case, but the fact is that on long shaders NV30 would also be unbalanced because it would also get no utilisation of its bus.

In the very short shader case if we assume (probably safely) that NV30 cannot get full use of its core because it is completely memory bound, then R300 with nearly twice the memory bandwidth per pixel available could theoretically get much better utilisation.

Excuse me, a silly question: More bandwidth to do what?

Fill pixels at the fastest possible rate. There are plenty of applications for filling pixels with short pixel shaders (every legacy application in existence, for starters...)

 

[no_way]

I've previously mentioned this (I can't remember which thread).
Multipass shaders are not slower than single pass shaders IF bandwidth is not a limitation. In fact, under some circumstances a multipass shader can be faster than a single pass shader.

.. As stated by JC, with 8500 being slower in single-pass rendering than GF3 doing multiple passes in Doom.

 

[Evildeus]

Then I would suggest you've misunderstood

Democoder's comment was that on long shaders R300 is unbalanced because it gets no use of its memory bus, with the implication that NV30 is better balanced for this case, but the fact is that on long shaders NV30 would also be unbalanced because it would also get no utilisation of its bus.

Fill pixels at the fastest possible rate. There are plenty of applications for filling pixels with short pixel shaders (every legacy application in existence, for starters...)

Hehe, so we have understood the same way, but we have different approach. I think he says better utilisation, because there's more waste with the R300. Thanx anyway

 

[T2k]

I've previously mentioned this (I can't remember which thread).
Multipass shaders are not slower than single pass shaders IF bandwidth is not a limitation. In fact, under some circumstances a multipass shader can be faster than a single pass shader.

.. As stated by JC, with 8500 being slower in single-pass rendering than GF3 doing multiple passes in Doom.

According to his .plan?

Check the date: February 11, 2002

Hehe, it's pretty outdated.

EDIT: typo

 

[Evildeus]

ATI said it is using 8.8 to 1 compression method on the Radeon 9700 PRO cards that can deliver an astonishing 174.2GB per second in the best case. A rare bird in practice, we'd suggest.

http://www.theinquirer.net/?article=6424

 

[RoOoBo]

ATI said it is using 8.8 to 1 compression method on the Radeon 9700 PRO cards that can deliver an astonishing 174.2GB per second in the best case. A rare bird in practice, we'd suggest.

http://www.theinquirer.net/?article=6424

That's nothing compared with this:

http://www.theinquirer.net/?article=6429

will have an extra connector at the rear part of the printed circuit board (PCB).

This may be the fruit of the integration of Nvidia and 3dfx, as the NV30, now known as the Geforce FX, appears to have elements that were part of the Rampage technology.

 

[Evildeus]

Well, it's the Inq 8)