• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Any solid evidence on R420?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Originally posted by: Acanthus
I never claimed that they redesign the GPUs every 6 months. They Design a GPU, refresh it. Then design a new one, refresh it. Its a 1 year cycle.

I dont think just because a new core comes out once a year and a refresh between that each new core is a complete redesign. They just fix things that were learned to be broken or lacking, and add on new things. Think of along the lines of how banias was certainly a new and great core, but it wasnt designed from the ground up, they started with a p3, fixed broken stuff, added new stuff, make on smaller process. Same with K8 was K7+some stuff, and K7 was K6+some stuff. I could go on and on, it is just not economically feasible to redesign any silicon part from the ground up once a year.
 
Originally posted by: rgreen83
Originally posted by: Acanthus
I never claimed that they redesign the GPUs every 6 months. They Design a GPU, refresh it. Then design a new one, refresh it. Its a 1 year cycle.

I dont think just because a new core comes out once a year and a refresh between that each new core is a complete redesign. They just fix things that were learned to be broken or lacking, and add on new things. Think of along the lines of how banias was certainly a new and great core, but it wasnt designed from the ground up, they started with a p3, fixed broken stuff, added new stuff, make on smaller process. Same with K8 was K7+some stuff, and K7 was K6+some stuff. I could go on and on, it is just not economically feasible to redesign any silicon part from the ground up once a year.

I guess you dont understand that even a die shrink is a total redesign.
 
Originally posted by: Acanthus
Originally posted by: rgreen83
Originally posted by: Acanthus
I never claimed that they redesign the GPUs every 6 months. They Design a GPU, refresh it. Then design a new one, refresh it. Its a 1 year cycle.

I dont think just because a new core comes out once a year and a refresh between that each new core is a complete redesign. They just fix things that were learned to be broken or lacking, and add on new things. Think of along the lines of how banias was certainly a new and great core, but it wasnt designed from the ground up, they started with a p3, fixed broken stuff, added new stuff, make on smaller process. Same with K8 was K7+some stuff, and K7 was K6+some stuff. I could go on and on, it is just not economically feasible to redesign any silicon part from the ground up once a year.

I guess you dont understand that even a die shrink is a total redesign.

I think we are misinterpreting each other on context, I am thinking a redesign from technological standpoint, and I think you are refering to a redesign from a mechanical standpoint. Also when you say "total redesign" its sounds like your saying they have to reinvent the wheel, and while yes die shrinks do involve some tweaking, they dont necessitate a total redesign. Look here http://www.rojakpot.com/ on page 2 of the GPU guide, NV15 went from .18micron to .15 micron while retaining same number of transistors and same code name.

I think I saw a while back that ATI had licensed technologies from a company to automate most of the basic die layouts, maybe this could help them focus more on the technological side?
 
Acanthus, a die shrink is a question of engineering, whereas I'm talking about architecture. And, no, I wasn't referring to all V8s, just those made by the same company. GM still uses pushrod V6s, no? I think we're just thinking of different things when we refer to GPU "design."
 
Originally posted by: Pete
Acanthus, a die shrink is a question of engineering, whereas I'm talking about architecture. I think we're just thinking of different things when we refer to GPU "design."

Hey thats the same thing I said!😀
 
Originally posted by: rgreen83
Originally posted by: shady06
Originally posted by: rgreen83
Originally posted by: keysplayr2003
Originally posted by: remagavon
No, the initial new design (r400) got scrapped for reasons unknown. AFAIK r420 is supposed to be about 2x the performance of rv360 (the 9600xt), but that's not confirmed.

It better do a helluva lot better that that.

Agreed, the 9800xt is already darn close to being 2x the 9600xt right now.

i think you are overestimating the performance difference between the 9600XT and the 9800XT

http://www.pcstats.com/articleview.cfm?articleid=1522&page=4[/Q

nope, as your link shows, at higher resolutions(and they dont even have AA/AF on) the frames are near double. If I wanted to play games at 10x7 with no eye candy I wouldnt care about R420 or have a 9800xt because those are the only situations which stress these cards.


Right, but remember the 9600xt has less ram as well which does help in higher resolutions. I'm pretty sure that it's been said it'll be the performance of two rv360's but highly tweaked of course. What I mean is that I'm sure it will be faster than just twice that performance, especially given that the card will have more pipelines (which will help in higher resolutions) and may have new aniso/FSAA algorithims which would improve performance in those areas. I'm not by any means saying that the card is going to be slow, just that I highly doubt it'll be something like twice the 9800xt, unless it's under weird circumstances like 6xfsaa (the r420 should support 8x, so 6 may be playable at higher resolutions), or the like. Just my opinion though, and I could very well be wrong. We'll see in a week or so 🙂
 
Ok it seems we weren't talking about the same thing then.

When i was referring to a "total redesign" im referring to a radically different chip, not the same design just reoganized in a die shrink or with added pipelines or featuresets.
 
Kind of OT but isnt 6xAA a little much?

I would think 1280x1024 with full 4xAA would have no aliasing at all.

I know for sure 1600x1200 4xAA has none.
 
There is *slight* ailising with 4xfsaa, but at 1600x1200 it is indeed hardly noticable. With 6x on a radeon 9700 or higher, I have to look extremely hard even at 1024x768 to see any sort of ailising or crawling. It's very nice, I hope that it's more usable in the future 🙂
 
rgreen83,

Feh! You can never have too many analogies, I say. 🙂

Acanthus,

This: "I guess you dont understand that even a die shrink is a total redesign."

Doesn't square with this: "When i was referring to a "total redesign" im referring to a radically different chip, not the same design just reoganized in a die shrink or with added pipelines or featuresets."

I guess I'm waiting for you to admit you were wrong, at least in your terminology. 😛 RV350 is a die shrink of R350 (150nm -> 130nm), but I don't think you can call it a "total" redesign" (it's basically half the pipes and vertex shaders and memory controller).
 
Back
Top