• 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.

Discussion RDNA 5 / UDNA (CDNA Next) speculation

Page 120 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I think an issue with AMD is and why releasing AT0 doesn't make sense unless it outright has the performance crown is it provides just as much competition for its own cards.

If AT0 was released, it has to be priced at more than twice the price of the AT2 cards or it starts to provide competition for these cards.

Think of the RTX 4090 vs 4080. The 4080 cards initially launched at 1199 and sold terribly vs the 1599 4090. People wanted the far less produced 4090 and demand outstripped supply. The 4080 super was trying to correct for this but it still wasn't enough. The RTX 4080 super was not a great seller and could be found at MSRP regularly. The 5080 to 5090 was an even further correction of this 2000 vs 1000. Even here the 5090 demand was still far outstripped(some related to gaming and mostly AI). This is with a 50% difference in performance.

With AT2 vs AT0 having a spec difference where there could be 70%+ difference in performance and most if not all of AT0 being supplied to business. You could see a situation where AT0 does more harm than good if it isn't getting first place.

E.g AMD wants to sell AT2 at 999 which is realistic considering pricing of GDDR7 at the moment, Nvidia raising their prices next generation and pricing of 3nm. AT0 tries to be sold at 2000 but it doesn't move because no one wants to buy a 2nd place card for that price(particularly an AMD one). The price drops to 1400 since it doesn't sell vs flagship rtx 5090. All of a sudden AT2 needs to drop to 599 for it to make any sense since there is such a massive difference in performance. This isn't terribly compelling business position considering GDDR7's price and the cost of 3nm wafers.

This has happened before with AMD card generations and has nearly destroyed the division. It happened with tahiti series, the Vega series and the hawaii series where the top end card ended up getting cuts which eroded the rest of the lineup and caused sweeping price drops to near cost or below cost levels. e.g the 7970 launched at 550 but could be found at 300 when on sale(prior to Hawaii's launch).

It may be different with this AI market but if this memory crisis end before then this could happen again. I think this is why AMD has taken the Nvidia performance minus 50-100 dollars strategy permanently and has largely abandoned the high end.
 
Last edited:
tajoh111 said:
I think an issue with AMD is and why releasing AT0 doesn't make sense unless it outright has the performance crown is it provides just as much competition for its own cards.
Click to expand...
There is little crosshopping between sub $1k and halo buyers.
tajoh111 said:
If AT0 was released, it has to be priced at more than twice the price of the AT2 cards or it starts to provide competition for these cards.
Click to expand...
If AT0 is priced low enough to be better perf/$ than AT2 then that is probably a good thing for AMD if margins are there and it actually sells. Refer to point 1, AT2 budget guys ain't gonna find the money for even aggressive AT0.
AT2 is for Nextbox primarily.
tajoh111 said:
Think of the RTX 4090 vs 4080. The 4080 cards initially launched at 1199 and sold terribly vs the 1599 4090. People wanted the far less produced 4090 and demand outstripped supply. The 4080 super was trying to correct for this but it still wasn't enough. The RTX 4080 super was not a great seller and could be found at MSRP regularly. The 5080 to 5090 was an even further correction of this 2000 vs 1000. Even here the 5090 demand was still far outstripped(some related to gaming and mostly AI). This is with a 50% difference in performance.
Click to expand...
xx80 parts are purely price anchors now, making both the xx90 and the 203 cutdown to be more desirable.
NV wants top 203's in laptops, any excess can be sorted out later in the generation.
tajoh111 said:
With AT2 vs AT0 having a spec difference where there could be 70%+ difference in performance and most if not all of AT0 being supplied to business. You could see a situation where AT0 does more harm than good if it isn't getting first place.
Click to expand...
AT2 is for Nextbox gamer cattle, AT0 is for rich people who actually intend to make money off them.
Xbox Cloud is lead customer.
tajoh111 said:
E.g AMD wants to sell AT2 at 999 which is realistic considering pricing of GDDR7 at the moment, Nvidia raising their prices next generation and pricing of 3nm. AT0 tries to be sold at 2000 but it doesn't move because no one wants to buy a 2nd place card for that price(particularly an AMD one). The price drops to 1400 since it doesn't sell vs flagship rtx 6090. All of a sudden AT2 needs to drop to 599 for it to make any sense since there is such a massive difference in performance. This isn't terribly compelling business position considering GDDR7's price and the cost of 3nm wafers.
Click to expand...
Even if memory prices are inflated still, AT2 ain't selling for more than like $699.
9070XT is basically MSRP today vs inflated NV comp and sales are still tragic.
If AMD cannot move even good AT0 bins for $2k then there is no reason to release them to DIY.
tajoh111 said:
This has happened before with AMD card generations and has nearly destroyed the division. It happened with tahiti series, the Vega series and the hawaii series where the top end card ended up getting cuts which eroded the rest of the lineup and caused sweeping price drops to near cost or below cost levels. e.g the 7970 launched at 550 but could be found at 300 when on sale(prior to Hawaii's launch).
Click to expand...
AMD tried to prep people for Hawaii but they had to desperation price cut, which hurt Hawaii's value and also forced NV to price accordingly.
tajoh111 said:
It may be different with this AI market but if this memory crisis end before then this could happen again. I think this is why AMD has taken the Nvidia performance minus 50-100 dollars strategy permanently and has largely abandoned the high end.
Click to expand...
There is the nuclear option but why bother for a pretty small TAM?
Unlike Ryzen which is just mini EPYC, you can't really scale GPUs between GPGPU accelerators and client gfx cards in the same way.
Which is exactly why RDNA5 is the way it is, every part has a primary use in markets NV hasn't completely ruined, and the secondary use of AIB cards.

AT3/4 are APUs with LPDDR but are effectively laptop dGPU compete, they have the shakiest foundation but I think they will do well.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top