RX 7900 GRE

[humbug]

I just watched the HWB video that performance was tragic and that's me being generous. Considering the 7800xt is even more cut down than the GRE I don't even see it matching the 6800xt but see it as being worse. Considering you can buy the 6800xt for £500 I would agree with you that it needs to be about £400 at best, otherwise you have 0 peformance increase in 4 years and it's literally worse than a refresh.

AMD could save money by just putting 6800xt is in a new box that says 7800xt.

If its 60 CU's as rumoured it will not keep up with the 6800XT, it would only be a bit faster than the RX 6800.

 

[Raelgun]

Its more concerning looking at this just how big the gap between a prospective 7800XT will be vs the 7900XT, its like a similar story with the 4070ti and 4080, but even that isnt as massive as what looks to be happening here.

The 7900XT needs even more price cuts to be considered in line where GPUs have to be ( including Nvidias own )

 

[KompuKare]

I just watched the HWB video that performance was tragic and that's me being generous. Considering the 7800xt is even more cut down than the GRE I don't even see it matching the 6800xt but see it as being worse. Considering you can buy the 6800xt for £500 I would agree with you that it needs to be about £400 at best, otherwise you have 0 peformance increase in 4 years and it's literally worse than a refresh.

AMD could save money by just putting 6800xt is in a new box that says 7800xt.

It's crazy.
Taking Navi 32's 200mm² GCD and the four ICDs are pretty much the same price as Navi 21 in the first place:

Die	Node	Area	defect rate	Dies	Good	Defect	Yield	QTY	Wafer cost	Cost
Navi 32 CU die	TSMC 5nm	200.0​	0.07​	275​	239​	36​	86.9%​	1​	$17,000	$71.13
Navi 32 IO die	TSMC 6nm	37.0​	0.07​	1613​	1572​	41​	97.5%​	4​	$6,000	$15.27
Navi 21	TSMC 7nm	520.0​	0.09​	98​	62​	36​	63.3%​	1​	$6,000	$96.77​

Almost the same and 7nm might be even cheaper now (although TSMC were keen to push customers to 6nm).

Navi 21 yields are probably good enough that they could even do a full die and run it a bit slower to narrow the perf/watt difference with Navi32.

A wasted GPU generation.