[Unknown Soldier]

The Inquirer(yes your favourite hate site) is reporting that Nvidia says that the GeForce FX can do upto and over 64 GB/s.

http://www.theinquirer.net/?article=6330

US

[Psikotiko]

Originally Posted by Unknown Soldier

The Inquirer(yes your favourite hate site) is reporting that Nvidia says that the GeForce FX can do upto and over 64 GB/s.

http://www.theinquirer.net/?article=6330

US

Of course it can

[Nappe1]

its the chip's internal bandwidth figure.

Nvidia suggested that the internal bandwidth is actually 64 GB/s...

still, closest one I can come up with is:
( ( ( ( ( 512/8 )*500000000 )/1024 )/1024 )/1024 )*2= 59.604645 GB/s
and that would mean they have 512Bit internal DDR bus running at 500MHz.

well, it would not be first time when Inquirer proudly reports something that doesn't quite match....

[Kaizer]

Or it would be in best case scenarios where the 4:1 lossless colour compression works like a charm..

16GB/s*4=64GB/s

I wonder more about how they came up with the 48GB/s number..


With Regards
Kjetil

[Xmas]

Originally Posted by Nappe1

its the chip's internal bandwidth figure.

still, closest one I can come up with is:
( ( ( ( ( 512/8 )*500000000 )/1024 )/1024 )/1024 )*2= 59.604645 GB/s
and that would mean they have 512Bit internal DDR bus running at 500MHz.

Possibly. But a 512bit DDR (effective 1024bit) bus at 500MHz means plain 64GB/s for me

[Testiculus Giganticus]

The inquirer is NOT a premium source of information. They sometimes misunderstand info provided to them.

[g__day]

I seem to remember reading on tech hardware that both the Radeon 9700 Pro and the NV30 strangely enough have the same 64 GB/sec THEORETICAL bandwidth under IDEAL conditions. So what - you need to ask what bandwidth will be available under normal conditions.

Most reviewers seem to think the NV30 with its 16GB of raw bandwidth will beat the R300 19.8GB/sec of raw bandwidth by around 30% BEFORE a clear analysis of LMA 3 is evident.

Why don't we just wait and see?

[sancheuz]

You know what, i'm starting to believe theinquirer. If they say it it has to be true. I mean, pretty much everything they said about the nv30 was true. They were the only ones to say it would have 128 bit bus, they said it would have ddr2 at 1 ghz, all that came true plus much more. so if they say it will have 64 gb, it will have 64 gb.

[incandescent]

Originally Posted by g__day

Most reviewers seem to think the NV30 with its 16GB of raw bandwidth will beat the R300 19.8GB/sec of raw bandwidth by around 30% BEFORE a clear analysis of LMA 3 is evident.

Why don't we just wait and see?

or is that because that's all NVIDIA is telling them, seeing it as how no one but NVIDIA actually has working boards? Hmmm 8)

[Joe DeFuria]

You know what, i'm starting to believe theinquirer. If they say it it has to be true.

[Crazyace]

8 pixel shaders generating 128bit output ( 16 bytes ) at 500MHz gives 64GBytes/second - Maybe coincedence...

[T2k]

Originally Posted by Testiculus Giganticus

The inquirer is NOT a premium source of information. They sometimes misunderstand info provided to them.

Sometimes???

[Crusher]

I would tend to think Crazyace's formula is how NVIDIA would come up with such a number. Kind of like how they got the GeForce 256 to have an internal 256 bit bus, isn't it (4 pixel pipes, 2 textures per pipe, 32 bit textures)?

[Randell]

nope its (32bit colour+ 24 bit z+8 bit stencil) x 4 pipes

[megadrive0088]

yeah GeForce FX gets 64 GB/sec bandwidth and produces FF film quality
visuals in real-time. All with just a 128-bit bus, too!

[Nappe1]

Originally Posted by Joe DeFuria

LOL!
sorry, that actually was funny.

sancheuz: are you sure that you can see the difference between wanting to believe and believing?

[Tagrineth]

Originally Posted by Randell

nope its (32bit colour+ 24 bit z+8 bit stencil) x 4 pipes

So how is Parhelia 512-bit?

[Crusher]

I think the 512-bit number with parhelia is a reference to the memory bus (256 bit DDR).

[T2k]

Originally Posted by Crusher

I think the 512-bit number with parhelia is a reference to the memory bus (256 bit DDR).

And to the chip (4 VS x 128bit), of course.

[WaltC]

Originally Posted by Tagrineth

So how is Parhelia 512-bit?

Heh-Heh...the nVidia description of "256" has nothing to do with current conventions in naming things like this--it was entirely a nVidia invention designed primarily to foster the impression of a "256-bit internal bus" when in fact none existed. At least, that's what I think. This all takes me wa-a-a-a-ay back in the way-back machine....

nVidia marketers have been clever in the past. They never "lie" outright (usually.) What they do is mislead, and then sit back and wait for the person they are trying to manipulate to reach an erroneous conclusion, and then they don't bother to correct it...

We saw a smidgeon of that in Dave's interview with Geoff. Dave did not allow himself to be led. Anyway, Dave asked Geoff whether or not the FX still took only 8 samples for AF. Geoff did not say, "yes." Instead, he said:

"Eight trilinear filtered samples." (Working from memory here, so pardon me if I'm not exact on the wording.) I thought that was a very strange answer considering that Dave hadn't asked anything about Trilinear samples.... Then it hit me: the hope was that Dave would say something like, "So you mean the ATI 9700 Pro does 16 bilinear samples?" (Because what else could Dave's question really concern since only the 9700 does > 8 AF samples?) If Dave had said this I think Geoff would just have shrugged, grunted, or winked--and let Dave merrily go on to repeat the error to the whole world. Dave's not that shallow--thank God!....

Seriously, I don't know how many times I've seen nVidia PR people work that same kind of trick on less knowledgeable people--and my stomach churned as the people fell for it and glibly repeated whatever fundamental error it was the nVidia PR person had led them into. It was amazing to watch them do this sort of thing when 3dfx was trying to tell everybody that AGP x bus speeds weren't a big deal for 3D performance--and some of the "pundits" were screeching about AGP x4 as though it was a critical feature. The beauty of it is that nVidia never directly repeats or states the errors--it's the people they deign to grant interviews who do all of that (except for Dave B....) All nVidia PR does is skillfully lead them into the errors--they do all the rest.

(Sorry if this is a bit strong for some--but I've honestly seen it many times. Perez was a master.)

[DemoCoder]

WaltC, it's getting old now.

[tamattack]

Originally Posted by DemoCoder

WaltC, it's getting old now.

Doesn't make it any less true. If anything, it seems to be getting worse (IMO of course.)

[tamattack]

Back on topic, I just noticed this over at Rage3D:

"As far as NVIDIA’s bandwidth claims of GeForce FX’s 48GB/sec memory bandwidth, ATI states that the color compression in their HYPERZ III technology performs the same thing today, and with all of the techniques they use in RADEON 9700, they could claim bandwidth of nearly 100GB/sec, but if they did so no one would believe them, hence they’ve stood with offering just shy of 20GB/sec of bandwidth."

More fuel for the fire...

[Sage]

[I] WRONTE: nVidia is claiming an effective 48Gb/s bandwidth and everyone has been assuming that its because of their 4:1 colour compression. remember much about Mojo? It was going to be based on gigapixel's defferred renderer. with the ability to cull TO (not with) a depth complexity of one then:

uncullable info (shader progs, geometry, etc) = UI
bandwidth = 16
avg depth complexity = 4
ineffeciency = %20 (?)

((16 - (.2 * 16)) * 4) - UI = effective bandwidth
(16 - 3.2) * 4 = effective bandwidth
12.8 * 4 = effective bandwidth
51.2 - UI = effective bandwidth
48 = estimated effective bandwidth