-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Current State and Future of AMD Radeon Graphics: Teleconference Transcript
- Thread starter Final8y
- Start date
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Couple ex Nvidia marketing guy, with an unknown product map process slips and and a lot of work going into consoles and I think the message is simply unclear because things are changing day to day. Next gen, as in 20nm, is iffy because the process can slip, and Apple might hog all the capacity. IE if 20nm is due for December, its not worth releasing refreshes now, but if 20nm slips to mid next year... it is worth doing refreshes now.
If they are planning on moving to Glofo for 20nm, its not worth investing in new masks for a different process at TSMC. My theory is that the CI/SI product line might be mixed up because AMD might make 28nm CI stuff at Glofo, for a little bit of practice making big GPU's at GloFo, before they switch to 20nm. Meaning, they continue making HD7xxx parts at TSMC as they are now, they make a few parts at 28nm at Glofo that are CI, for practice on their processes with a different type of part, but there isn't much point investing huge money doing new products in every range. So they end up making 7970/50's at TSMC, and a new 7980, or 8970 with a marginally tweaked core at GloFo, and maybe they keep the 7870 at TSMC, but do a part closer to a 680gtx in size at Glofo, so you'd have up and down the range, different parts being CI/SI, and hence the marketing issue... because the average buyer won't know whats what.
So you call all the new stuff 7xxx parts, then people get upset that no one knows what part is what architecture, if you call them 8xxx parts and they are basically clock for clock 3-4% faster due to a few minor fixes, people get angry over a percieved rebrand. There isn't really any transistor/die size space to go bigger and better, so there isn't much new performance to come, but people randomly, for no reason, want new parts on the same process for the same cost, and get upset if they can't have a new card every year.
Like I said, thats a theory of mine, but AMD, Intel, just about every player has been hit by moving to a new process, with a new architecture at a new foundry, with massive problems. Hence Intel's tick tock, and Llano being a dog to manufacture and Trinity being a different architecture, but same process that they had practice and experience on, not being a problem.
Ultimately, it doesn't matter, the 6970 wasn't a massive step up, the 580gtx was ONLY a bigger step up in performance because the 480gtx was screwed... if the 480gtx worked, the 580gtx would have been a 5-10% bump. Refresh = boring, new gen on new process = wooooo.
20nm was never going to come this year in full production, there was never some 80% faster part coming from AMD, or Nvidia, people getting excited over nothing. Why 90% of this forum can't put new process = new levels of performance together, and realise neither AMD or Nvidia has the slightest influence over when new processes come, I really don't know. This has been obvious for 2 decades in the silicon industry.
If they are planning on moving to Glofo for 20nm, its not worth investing in new masks for a different process at TSMC. My theory is that the CI/SI product line might be mixed up because AMD might make 28nm CI stuff at Glofo, for a little bit of practice making big GPU's at GloFo, before they switch to 20nm. Meaning, they continue making HD7xxx parts at TSMC as they are now, they make a few parts at 28nm at Glofo that are CI, for practice on their processes with a different type of part, but there isn't much point investing huge money doing new products in every range. So they end up making 7970/50's at TSMC, and a new 7980, or 8970 with a marginally tweaked core at GloFo, and maybe they keep the 7870 at TSMC, but do a part closer to a 680gtx in size at Glofo, so you'd have up and down the range, different parts being CI/SI, and hence the marketing issue... because the average buyer won't know whats what.
So you call all the new stuff 7xxx parts, then people get upset that no one knows what part is what architecture, if you call them 8xxx parts and they are basically clock for clock 3-4% faster due to a few minor fixes, people get angry over a percieved rebrand. There isn't really any transistor/die size space to go bigger and better, so there isn't much new performance to come, but people randomly, for no reason, want new parts on the same process for the same cost, and get upset if they can't have a new card every year.
Like I said, thats a theory of mine, but AMD, Intel, just about every player has been hit by moving to a new process, with a new architecture at a new foundry, with massive problems. Hence Intel's tick tock, and Llano being a dog to manufacture and Trinity being a different architecture, but same process that they had practice and experience on, not being a problem.
Ultimately, it doesn't matter, the 6970 wasn't a massive step up, the 580gtx was ONLY a bigger step up in performance because the 480gtx was screwed... if the 480gtx worked, the 580gtx would have been a 5-10% bump. Refresh = boring, new gen on new process = wooooo.
20nm was never going to come this year in full production, there was never some 80% faster part coming from AMD, or Nvidia, people getting excited over nothing. Why 90% of this forum can't put new process = new levels of performance together, and realise neither AMD or Nvidia has the slightest influence over when new processes come, I really don't know. This has been obvious for 2 decades in the silicon industry.
Last edited: