PDA

View Full Version : R420 have 160M transistors


Pages : 1 [2]

SurfMonkey
12-22-03, 09:45 AM
Chances are that the die size will be increasing just in order to have a larger surface area for sinking that heat production.

And, as I think the NV3x has proved, transistor count isn't everything. It's how they are put together that counts. I think the NV4x and R42x are going to be quite evenly matched. It all depends upon the whistles and bells (PPP etc) that are included and whose engineers deliver the best design optimisations.

Paul
12-22-03, 01:05 PM
Originally posted by MUYA
Which brings me to this....any ideas or thoughts if the new gen q1 2004 GPUS will have 8 pipeline bretherens in the mainstream market?

Yes, definitely. Not Q1 though.

Paul
12-22-03, 01:10 PM
Originally posted by Ninja Prime
If NV40 is 175million, that doesn't exactly mean its gonna be better/faster. Look at NV30... 15million more (~12%)than R300, but waaaay slower. 175million is 9% more than 160million of R420... R420 might still be twice as fast like NV30-R300. Especially since NV40 is supposed to use a similar tech as NV30-35.(or so I hear)

He was hinting that nVidia found out that R420 was sitting pretty at 160million transistors, whilst NV40 was only at 150million. They then decided to use some extra transistors, to just over the amount in use on the R420, to get them some extra performance/compatibility. Regardless of whether it's faster/slower than the R420, extra speed is always good.

nobie
12-22-03, 01:26 PM
Especially since NV40 is supposed to use a similar tech as NV30-35.(or so I hear)

Funny, I heard the exact opposite! (that R4xx is heavily based off of R3xx)

I think most likely they'll both have some things in common with/reused from with the previous generation, though.

MUYA
12-22-03, 01:40 PM
Originally posted by Paul
Yes, definitely. Not Q1 though.

Dude I mean 8 pixel pipeline as is rigid none of this 8X and 4X2 type thing u get with the curent series?

Any word on the memory interface? 128 bit or 256 bit? Is putting a 256 bit memory interface ona mainstream card in q2 2004 going to be that expensive? Ya please look into your crystal ball and find out mate? ;)

MUYA

Paul
12-22-03, 02:04 PM
Originally posted by MUYA
Dude I mean 8 pixel pipeline as is rigid none of this 8X and 4X2 type thing u get with the curent series?

Any word on the memory interface? 128 bit or 256 bit? Is putting a 256 bit memory interface ona mainstream card in q2 2004 going to be that expensive? Ya please look into your crystal ball and find out mate? ;)

Yes to all of this. 8 Pipeline, 256bit memory interface cards will surface in the mainstream during 2004, although only at the higher end of that segment.

Budget offerings are still going to be stuck at 4 pipelines and 128bit though, although that's no suprise.

MUYA
12-22-03, 02:24 PM
Originally posted by Paul
Yes to all of this. 8 Pipeline, 256bit memory interface cards will surface in the mainstream during 2004, although only at the higher end of that segment.

Budget offerings are still going to be stuck at 4 pipelines and 128bit though, although that's no suprise.

Its sounding cool. :D Maybe Doom III and HL2, Far Cry or Salker may actually be playable with them mainstream cards which may not be the case right now. Ie with all visuals and aa and af on etc

Cotita
12-24-03, 08:15 PM
Originally posted by Paul
Yes to all of this. 8 Pipeline, 256bit memory interface cards will surface in the mainstream during 2004, although only at the higher end of that segment.

Budget offerings are still going to be stuck at 4 pipelines and 128bit though, although that's no suprise.

nvidia already has mainstream cards that have 256bit memory and 8 pipelines, ever heard of the geforce FX 5900 SE?

The Baron
12-24-03, 08:49 PM
how about mainstream chipset then, because like it or not, the 5900SE exists only because of lots of extra NV35s, and that it's cheaper to produce NV38 (according to Uttar).

Hanners
12-25-03, 03:56 AM
Originally posted by AthlonXP1800
When ATI developed R400 and they got some information about NV40, decided that R400 will not good to compete NV40. ATI do not want to go into NV30 disaster path if they launched R400 and they realised that R400 project is very high risk strategy, they moved R400 up to R500 team and started a low risk strategy, R420 project based on the most successful R300 technology.

You were right about everything there apart from the original R400 not being able to compete with NV40 - If anything, the opposite was true and R400 was overkill compared to NV40.

By the way, while we're talking R420 it appears (according to MuFu) that it will be an eight pipeline design.

Cotita
12-25-03, 01:55 PM
Originally posted by The Baron
how about mainstream chipset then, because like it or not, the 5900SE exists only because of lots of extra NV35s, and that it's cheaper to produce NV38 (according to Uttar).

You mean the single channel nforce2? Its been out for months

If you mean GPU, then we should congratulate nvidia for slipping its highend GPU to the mainstream, like with the geforce 3ti200 and Geforce 4Ti 4200.

The ti200 practically killed the geforce2MX400 and the geforce4ti4200 did the same thing to the geforce4mx 460, so either nvidia lowers the 5700 price to sub 150dlls or it will have the fate of its predecessors.

rwolf
12-28-03, 04:29 PM
Originally posted by Hanners
By the way, while we're talking R420 it appears (according to MuFu) that it will be an eight pipeline design.

Which explains the 160M transistors.

I would imagine that ATI has improved the internal parallelism of the existing pipes rather than add more pipes.

aapo
01-21-04, 02:40 PM
Originally posted by Uttar
Your 120-130 number is for the NV3x, not for the R3xx. Thus, 8x1->12x1. Not 6x2 though. Anymore questions? :p

I've been thinking that 12x1 seems a little low for the ATi's comments about sacrificing some profits for the R420 performance. Could it be that they are planning for some kind of dual core R420 MAXX? That would explain low profits and extreme performance even with 12x1 configuration. Is it possible to make dual cores with relatively large chips like R420?

nobie
01-21-04, 04:36 PM
you could, but it would probably require two power connectors on the board, or even an external power source like Voodoo 5 6000.

Paul
01-21-04, 04:54 PM
Appo - They were talking about margins, i.e. core/memory speed.

bkswaney
01-22-04, 04:33 PM
Well, I might be the only one thinking this but I think that both the R420 and NV40 are both going to be killer cards.
There specs are going to be very close. So will performance.

Nvidia is not going to let ATI way out in front this go around.

This time around Nvidia will bring up there lacking DX9 speeds
to meet ATI's.

I will bet on it. :)