AMD: Navi Speculation, Rumours and Discussion [2019-2020]

Status
Not open for further replies.
And the multiple models just doesn't sit with how slowly AMD's been doling out its different bins as of late. The 5600 has waited half a year till after the 5700 series despite being a bin of the same chip.
To be fair, when was the last time AMD had 3 different cut bins for consumer performance/highend parts(+$300)?
Not counting a different bin for a specific OEM/region, or just a slight speed bin or a different form factor(Nano)?

I believe it was Tahiti. New architecture and new node. Plus that was doled out over an entire year.

You have to stockpile a sufficient supply for that lower bin, which isn't exactly a main goal, especially once you start targeting a lower MSRP ($249-$279) therefore significantly increasing your TAM.
I'm actually surprised they are able to do so in ~6months, historically it seems to be something you see after ~10months. Which could point to two vastly different conclusions, Navi 10 is getting close to EOL (they have something similar on N7+) or Navi 10 has a long life ahead and they foresee sufficient stock of this bin.
Edit- Guess there could be two other reasons for a fast release of a 3rd bin, higher production than previously seen for a +$300 part, and/or worse yields.
 
Last edited:
Last edited:
Rumor puts RDNA2 at a 50% improvement in performance per watt: http://www.redgamingtech.com/rdna-2...vements-to-take-on-nvidia-analysis-exclusive/

Seems high, 7nm+ is only claimed to be up to 10% better power usage and I'm not sure their 7nm modified works with the + process, might have overlapping layers there. Still it connects up with the double the performance high end anomoly. That'd give that chip enough power headroom, assuming the board power is like 300 watts.
What do you mean with "7nm modified"? RDNA2 has always been 7nm+
 
What do you mean with "7nm modified"? RDNA2 has always been 7nm+

Their... gah whatever it was, I can't remember what they called it. But it's a modification of 7nm that replaces a few layers and provides better potential energy usage, a lot of stuff is built using it now. "N7P" https://www.anandtech.com/show/14687/tsmc-announces-performanceenhanced-7nm-5nm-process-technologies

Anyway it's minor, and they announced it'd carry over to 5nm. But I didn't see any announcement for 7nm+, so I just wondered if they didn't bother with that or the EUV layering overlapped with it somehow.

Also, as stated 7nm+ doesn't offer much potential improvement in energy usage over regular 7nm, at least as originally stated by TSMC. That'd mark the arch improvement in performance per watt at 36%, a really great result year over year. If the above is applicable though, that's a potential of around 24% improvement in arch, very good and better than most year over year, but obviously not quite as remarkable.
 
Last edited:
Rumor puts RDNA2 at a 50% improvement in performance per watt: http://www.redgamingtech.com/rdna-2...vements-to-take-on-nvidia-analysis-exclusive/

Seems high, 7nm+ is only claimed to be up to 10% better power usage and I'm not sure their EDIT N7p modifications work with 7nm+ process, might have overlapping layers there. Still it connects up with the double the performance high end anomoly. That'd give that chip enough power headroom, assuming the board power is like 300 watts.

Rdna is not very efficient, so a big big boost here is not impossible.
 
Their... gah whatever it was, I can't remember what they called it. But it's a modification of 7nm that replaces a few layers and provides better potential energy usage, a lot of stuff is built using it now. "N7P" https://www.anandtech.com/show/14687/tsmc-announces-performanceenhanced-7nm-5nm-process-technologies

Anyway it's minor, and they announced it'd carry over to 5nm. But I didn't see any announcement for 7nm+, so I just wondered if they didn't bother with that or the EUV layering overlapped with it somehow.

Also, as stated 7nm+ doesn't offer much potential improvement in energy usage over regular 7nm, at least as originally stated by TSMC. That'd mark the arch improvement in performance per watt at 36%, a really great result year over year. If the above is applicable though, that's a potential of around 24% improvement in arch, very good and better than most year over year, but obviously not quite as remarkable.
N7P is different from N7+ and like said in the Anandtech article, it's design compatible with N7 (meaning that probably all N7 products will move over to N7P as fast as TSMC can update their fabs) and not a replacement node.
Regardless of that, RDNA2 has always been "7nm+" aka N7+, which is completely different process from N7/N7P and uses EUV instead of DUV lithography
 
Regardless I'm just hoping we'll see more information sooner rather than later. With nVidia bound to launch their own 7nm this year, and the 5000 series basically coming close to their current 14nm parts, they certainly have their work cut out for them. It's an exciting dance between competitors. And now with Intel joining the fray this year is going to be very interesting.

I wonder how long we'll see RDNA in one form or another? AMD had a very successful run of iteration on GCN. If RDNA is even half as iterrable(?) I'd be very impressed.
 
I wonder how long we'll see RDNA in one form or another? AMD had a very successful run of iteration on GCN. If RDNA is even half as iterrable(?) I'd be very impressed.
I'd certainly hope RDNA wouldn't be so "iterable". GCN was kept on life support with minimal updates.
 
Considering AMD was seriously low on cash for most of GCN, and especially come GCN 4, and especially 5, I'd say they did an admirable job under the circumstances. Had GCN 5 been afforded a bit more time and resources for polishing I'm sure it would have turned out quite a bit differently. But at the time AMD had little choice. Zen had to be prioritised. I think it speaks volume to the soundness of the underlying architecture that it adapted and performed as it did over the years. Heck, considering the efficiency gains in Vega on Renoir and the rumoured efficiency improvements in store on RDNA it seems AMD had left a lot of fine tuning wiggle room in their products historically. Lack of time perhaps? If RDNA is even half as adaptable as an architecture, and with more money going in to develop it as profits increase thanks to ZEN, it's going to be interesting to see it evolve in turn I think.

Especially if nVidia and Intel picks up their pace RTG will need more resources to compete. nVidia will not let the crown go easily, and Intel has everything to win. Having a flexible enough architecture to counter the competition is key.
 
GCN's biggest downfall was the fact that it was compute first and compute on the game side of things got never as big as AMD thought it would (not the first time they thought pure compute power would be more important that it ever was during the time)
 
I don't think AMD will be selling a 500mm^2 GPU that performs 30% above the 2080 Ti at rock bottom prices.
Not only would that would be detrimental to AMD's Radeon brand, but they now have enough mindshare to get closer to the profit margins practiced by nvidia. Which they already did with the Navi 10 cards.


You are saying that, without understand everyone was saying the same thing about Dr Su when she released Zen, then Zen+, then Zen2. And AMD can easily sell @ 500mm^2 gpu die for $999..

What I said was, that cut down/faulty version of that die, will sell at rock bottom prices and still offer Ti like performance.
 
Not sure if this is old news but first time I’m seeing it. Dr. Su says Big Navi is coming this year. Whoop!!

https://www.techspot.com/news/83615-amd-lisa-su-doubles-down-big-navi-zen.html

"I’ve heard a little bit through Twitter and Reddit that people are wondering about, you know, Big Navi. I can say you're gonna see Big Navi in 2020. There might be few people wondering about Zen 3 as well, and I can tell you that Zen 3 is doing really well, we are excited about it, and I look forward talking about that later in 2020," said Su
 
I just hope "Big Navi" comes courtesy of RDNA2. If not I have this horrible vision of AMD uncovering their 2080 killer to the hoops and cries of long waiting fans, only for nVidia to release their RTX 3080 moments later and steal AMD's glory. To an offended cacophony of shrieks wailing "It's Vega all over again, It's Vega all over again!"

Actually they have nothing.

I dunno about that. Intel needs a GPU in their lineup for HPC applications primarily, leveraging the same architecture for consumer side products is just marketshare and ROI left on the table unless they take it. So in that sense they've got everything to win. I especially see how it at least initially can become a competitor to nVidia and AMD's low power efforts in the consumer space.
 
The whole talk was about 2020 and it was specifically mentioned for Zen 3 and RT acceleration in Radeons.

Right but this is the first time I've seen big Navi and 2020 explicitly mentioned in the same sentence. RT support on its own does not necessarily confirm big Navi.
 
Status
Not open for further replies.
Back
Top