David E. Orton's talk
- Thread starter 991060
- Start date
geo said:
I don't mean that the R420 won't be a big step up or that it isn't classified as a new generation of chip. I mean that I don't think we will have to wait 18-24 months for the R500 (as we would through a regular product cycle).
I get the impression that the R420 is bridging the gap between the R3XX and the R500, taking technology from both.
Just to add to the confusion and speculation here is the picture of ATI's mystery chip taken at Computex (September). Notice that the name of the chip is Viper.
http://www.theinquirer.net/?article=11742
MuFu said:
I wasn't here when the pictures were discussed, so I was unaware of any conclusions.
I added the picture to illustrate that there are many rumours and heaps of speculation regarding future ATI products and this makes it hard to draw any conclusions of future graphics chips.
geo said:overclocked said:
From B3D current review, an R360 is 107M. An RV360 is 75M. Assuming that the difference between the two is almost entirely the 4 extra pipes, that gives 32M for the 4 extra pipes, and thus R420 would weigh in at about 139M (107+32)? This assumes (big assumption) they didn't add anything else of significance, like whatever it would take to go from FP24 to FP32. The other assumption is that the "R420=3 x RV350" rumor is more or less correct.
Gee, that's damn near substantive for me; wonder where I screwed it up.
You forget that R3x0 as 2 times more VS units than RV3x0. I also think that you're probably doing an error by assuming that RV3x0 is 75M
Its not an assumption - its what we're told.
Tridam said:
Okay, I'll bite. Given that it has been reported that way by ATI, which you acknowledge, what makes you think it isn't correct? Free floating anxiety complex? Or something more substantive? (I'm hoping for the latter, btw, but am often accused of being much too optimistic and not nearly cynical enuf).
Sorry for this. It's not really nice to say this without some explanations. Unfortunately, I've not the time to talk about every thing that let me think that RV3x0 is less than 75M.
The first thing is that ATI said M10 is 60M. How is it possible ? M10 and RV350 is the same die. Marketing guys of mobile and desktop products have different numbers ? ATI always said me that RV350 was officially 75M.
I can approach the transistor count in every way I know and I always see 60M more probable than 75M. I was thinking that maybe some part of the RV350 could be broken (like the 10-15M broken in NV31). But it doesn't seem that ATI has fixed anything in RV360. For my part, I think that RV3x0 is not 75M and that ATI said 75M to have a transistor number in the same range as the NV31. Of course, maybe I'm just false. But I have enough elements to think that RV3x0 is less than 75M.
The first thing is that ATI said M10 is 60M. How is it possible ? M10 and RV350 is the same die. Marketing guys of mobile and desktop products have different numbers ? ATI always said me that RV350 was officially 75M.
I can approach the transistor count in every way I know and I always see 60M more probable than 75M. I was thinking that maybe some part of the RV350 could be broken (like the 10-15M broken in NV31). But it doesn't seem that ATI has fixed anything in RV360. For my part, I think that RV3x0 is not 75M and that ATI said 75M to have a transistor number in the same range as the NV31. Of course, maybe I'm just false. But I have enough elements to think that RV3x0 is less than 75M.
Tridam--
Ah. So you're pointing at reports like this: http://www.hothardware.com/hh_files/S&V/m10preview.shtml
"First, let's focus on the 9600 series that derives from the desktop part bearing a similar name. At its core, the Mobility RADEON 9600, codenamed M10, sports four pixel pipelines, each equipped with a single texture unit. Like its desktop counterpart, the Mobility RADEON 9600 core is comprised of 60 million transistors. That's right, for the first time we're seeing the same processor make its way into desktop and mobile products at the same time with the same capabilities and very similar performance."
Ah. So you're pointing at reports like this: http://www.hothardware.com/hh_files/S&V/m10preview.shtml
"First, let's focus on the 9600 series that derives from the desktop part bearing a similar name. At its core, the Mobility RADEON 9600, codenamed M10, sports four pixel pipelines, each equipped with a single texture unit. Like its desktop counterpart, the Mobility RADEON 9600 core is comprised of 60 million transistors. That's right, for the first time we're seeing the same processor make its way into desktop and mobile products at the same time with the same capabilities and very similar performance."
Tridam makes sense. This thread indicated that a single vertex shader is on the order of 7-9M transistors. Just adding the two extra VS engines to the RV350's purported 75M transistors brings us up to 89-93M transistors. Now, I'm no hardware engineer by any stretch of the imagination, but it seems hard to believe R350 can fit four more pipes and some extra HyperZ hardware into an extra 14-18M transistors.
Sure, ATi was a bit approximative with inital specs putting the R300 at 110M trans., and later ones shaving that to 107. But I can't imagine ATi would be so liberal with the RV350's trans. count simply to compete on paper with the NV31. I mean, honestly, who factors in transistor count into their buying decision? Strange.
Sure, ATi was a bit approximative with inital specs putting the R300 at 110M trans., and later ones shaving that to 107. But I can't imagine ATi would be so liberal with the RV350's trans. count simply to compete on paper with the NV31. I mean, honestly, who factors in transistor count into their buying decision? Strange.
MuFu said:Hanners said:
What, like 200 million?
MuFu.
ATI did say their margins were going to be smaller because they are going for performance.
If they can make a 200 million transistor chip with clocks like the 9600XT we are going to need some new games because hl2 won't be enough.