• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

AMD Radeon RX 9070 expected to start at $479, benchmarks now released (UP: RX 9000 series reveal on Feb 28; hits shelves in early March)

Wolzard

Member
Names don't matter. The chip is just slightly less complex than Navi 31 and is using a more advanced N4 node than Navi 31. Expecting it to cost as Navi 32 is fantasy land.

The Navi 48 is literally the successor to the Navi 32.
The successor to the Navi 31 was the cancelled Navi 4C.

It doesn't matter if the chip uses more advanced nodes, there are different segments of boards, performance is always expected to increase without the price going up too, or it becomes meaningless.

If it were by that logic, a 3060 Ti should cost the same as a 2080 Super, just because it has a more advanced node and is more complex.
 

SolidQ

Member
Furmark
8c9e0c637f4e2b18373801e2396739ea.png
 

dgrdsv

Member
The Navi 48 is literally the successor to the Navi 32.
The successor to the Navi 31 was the cancelled Navi 4C.
We don't know what Navi 4C was supposed to be.
From complexities alone Navi 48 is essentially a monolithic Navi 31 made on N4 instead of a mix of N5+N6. This isn't a recipe for a cheaper chip.

It doesn't matter if the chip uses more advanced nodes, there are different segments of boards, performance is always expected to increase without the price going up too, or it becomes meaningless.
It does matter A LOT for pricing which node a chip is using.

If it were by that logic, a 3060 Ti should cost the same as a 2080 Super, just because it has a more advanced node and is more complex.
This is a completely different scenario, and yeah, it would've likely been this way if Nvidia would use TSMC for Ampere.
 
Last edited:

Wolzard

Member
We don't know what Navi 4C was supposed to be.
From complexities alone Navi 48 is essentially a monolithic Navi 31 made on N4 instead of a mix of N5+N6. This isn't a recipe for a cheaper chip.


It does matter A LOT for pricing which node a chip is using.


This is a completely different scenario, and yeah, it would've likely been this way if Nvidia would use TSMC for Ampere.

Rumors (according to AMD Linux driver) are that this chip would have 9 shader engines.
Navi 31 had 16 CU per Shader Engine (6), for a total of 96 Compute Units.
If AMD continued with the same Shader Engine configuration as Navi 31, this means we would've seen Navi 4C with 9 Shader Engines, 20 Compute Units, and 180 CUs in total.
It would be about 50% above the RX 7900 XTX.

Your argument favors the idea that Nvidia currently increases price with increased performance, which contradicts the whole story. It makes no sense at all.

The 4nm node is no longer advanced, it is the standard. TSMC is already working with 3nm, with 2nm coming soon, as the advanced node. In addition, the RNDA 4 uses GDDR6 memory. Overall, the entire construction of these GPUs is much cheaper.
 
Top Bottom