[SOLVED] Big Navi RX6000 to be announced on Oct 28 - What does it mean in regard to Ampere?

mjbn1977

Distinguished
AMD's launch event for big Navi RX6000 cards will not happen until October 28.

What do you guy's think?

Does that mean that they not coming up to RTX 3080 performance? If I would be AMD and have something really competitive to those recently introduced Nvidia cards, I would introduce, or at least tease their specs before the Nvidia cards start shipping in order to put a little bit of hesitation into the mind of all the waiting Nvidia shoppers....

Or does it mean that they are tweaking their design last minute in order to achieve RTX 3080 similar performance specs. And that is the reason for this late launch?

Let's discuss....
 
Last edited:
Solution
Supposedly, based on some rumors, Nvidia originally planned the GA 103 with 60 SM units for the RTX 3080 and changed that last minute to the GA 102 (same as RTX 3090) with 68 SM units (of 86). It seems almost like Nvidia thought that they wouldn't be able to compete with big Navi using the GA 103 for the 3080 (which is now rumored to eventually become a RTX 3070 Ti or something like that. Is this telling us something about the possible Big Navi performance?
A different GPU number like that implies a different implementation of the architecture. i.e., having three different implementations (GA102, GA103, and GA104). That to me makes no sense. It makes more sense from a manufacturing point to 1. limit the number of implementations...
If they have something to sell soon, like on that date, I don't think they can change the specs all that much. The board manufacturers designed and validated based on the original specs. Changing them up now means they have to change/revalidate their designs, or release cards with the original specs. Remember the fiasco with the 5600 XT? Manufacturers weren't exactly happy with that situation.
 

mjbn1977

Distinguished
Supposedly, based on some rumors, Nvidia originally planned the GA 103 with 60 SM units for the RTX 3080 and changed that last minute to the GA 102 (same as RTX 3090) with 68 SM units (of 86). It seems almost like Nvidia thought that they wouldn't be able to compete with big Navi using the GA 103 for the 3080 (which is now rumored to eventually become a RTX 3070 Ti or something like that. Is this telling us something about the possible Big Navi performance?
 
Supposedly, based on some rumors, Nvidia originally planned the GA 103 with 60 SM units for the RTX 3080 and changed that last minute to the GA 102 (same as RTX 3090) with 68 SM units (of 86). It seems almost like Nvidia thought that they wouldn't be able to compete with big Navi using the GA 103 for the 3080 (which is now rumored to eventually become a RTX 3070 Ti or something like that. Is this telling us something about the possible Big Navi performance?
A different GPU number like that implies a different implementation of the architecture. i.e., having three different implementations (GA102, GA103, and GA104). That to me makes no sense. It makes more sense from a manufacturing point to 1. limit the number of implementations to roll out and 2. make a bunch of the big one anyway and whatever doesn't make the cut of the top-tier, sell it as a lower one. Considering the difference of SMs between the 3080 and 3090, I think that's what they did rather than react to what AMD claims they can do. Especially since the time to react to that would've been before the initial tape-out which is in the "let me get my crystal ball" territory.

Also NVIDIA doesn't do odd numbers in their GPU code names until they're past 6. So I don't also see a reason to start breaking that.
 
Solution

mjbn1977

Distinguished
A different GPU number like that implies a different implementation of the architecture. i.e., having three different implementations (GA102, GA103, and GA104). That to me makes no sense. It makes more sense from a manufacturing point to 1. limit the number of implementations to roll out and 2. make a bunch of the big one anyway and whatever doesn't make the cut of the top-tier, sell it as a lower one. Considering the difference of SMs between the 3080 and 3090, I think that's what they did rather than react to what AMD claims they can do. Especially since the time to react to that would've been before the initial tape-out which is in the "let me get my crystal ball" territory.

Also NVIDIA doesn't do odd numbers in their GPU code names until they're past 6. So I don't also see a reason to start breaking that.
Well, two years ago when RTX 2080Ti and RTX 2080 and RTX 2070 came out, all three had a different chip. But this year two share the same, one has a different one, and one is not even used.....
 
Well, considering AMD is already on its way to lower prices, they certainly don't have anything nvidia killer after all, odds are that they may have something to compete with a 3080 like the 5700XT vs 2070S.

what worries me the most is the hype train and a possible crypto currency mining BS, those folks could wipe clean the stocks ramping the price up, so im gonna get a 3080 as soon as MSI have those available, not getting FE, nor im waiting for this said big navi, I know AMD have made a fantastic comeback with ryzen but nvidia isn't as sloppy as intel,
 
Well, two years ago when RTX 2080Ti and RTX 2080 and RTX 2070 came out, all three had a different chip. But this year two share the same, one has a different one, and one is not even used.....
You're right, however, the 2070 was a 6 GPU, which usually targeted the 60 level SKU. NVIDIA has never made an in-between to the 2 or 4 GPU.

Either way, if there is a "3070 Ti", it'd likely come from a cut down GA-102 with fewer shaders than the 3080. Looking at what the 3070 has, it's looks like a slightly cut down GA-104