AMD: Pirate Islands (R* 3** series) Speculation/Rumor Thread

The GTX 980 price was cut from US$549 to US$499 with the launch of the GTX 980 Ti.
Hadn't noticed that, thanks.


Im a bit intriguated about the Nano, a cut down version with less SP ?

Well nVidia did manage to bring the GM204 down to some measly 100W for the mobile cards, so I guess this should be a cut-down card.
Regardless, even if they do reduce the chip down to 44 CUs at 1GHz, it should still be faster than a 290X.
 
Im a bit intriguated about the Nano, a cut down version with less SP ?

FuryX 1.5x perf/W Nano 2x perf/W

And possibly much lower clocks. 800MHz comes to mind.

Who wants to bet, people are gonna slap on a different cooler and OC that thing up to 1GHz ;)

Unless of course it's limited with power delivery.
 
Last edited:
The FuryX will be considerably more expensive to make than the GTX 980 Ti, so they can't undercut it in price. Performance will probably be very similar.

They say that it will be the fastest card in the market and given the pure specifications, it should be without any doubt.
I highly doubt that the BOM cost will be more. The PCB itself should be less expensive with this size.

If it'd beat it, they'd ask a higher price

No. There are numerous reasons why this thinking is not applicable in the case.
 
They say that it will be the fastest card in the market and given the pure specifications, it should be without any doubt.
And whenever they were talking about fastest, it was in the context of their dual Fiji thing.

But I agree: given the specs it should be the fastest.

I highly doubt that the BOM cost will be more. The PCB itself should be less expensive with this size.
The only thing that's cheaper on the FuryX is the PCB. Unfortunately, PCBs don't cost a whole lot. And everything else is significantly more expensive.
 
The smaller PCB means that the BOM itself should be shorter or with less components. Otherwise, if the components are the same quantity, how would you fit them when it's already with high density?
 
Could they have streamlined away one of the memory VRM phases compared to Hawaii?
Granted, for the FuryX they'd probably jack the count back up again.
 
Other bits of info:

* Typical board power 275W for Fury-X and WC cooling capacity of 500W
* Illuminated RADEON logo
* Fully user configurable LED's on the edge of PCB in default mode showing GPU utilization
* Build for overclocking
* High quality materials used for cooler
* ~50C running temperature for GPU when stock [if I heard it right]
* 400A 10 Phase PWM's (hinting at at least 375W, if not more power headroom!) [if I heard it right]

notes: if I heard it right - thanks to my wife yapping ... :)

I'm slightly exited, now let's wait for some real performance numbers!
 
The number of layers is more critical to the PCB's BOM than the area size alone. AFAIK, the most expensive reference board (512-bit) had ~10 layers.
 
The smaller PCB means that the BOM itself should be shorter or with less components. Otherwise, if the components are the same quantity, how would you fit them when it's already with high density?
BOM stands for bill of materials. By moving the memory to the GPU, the GDDR5 gets removed from the BOM and the area gets reduced.

What other BOM reductions do you have in mind?
 
For what it's worth, if AMD's methodology for typical board power is consistent with other boards its given figures for, the stock consumption for FuryX appears to be the same neighborhood as Tahiti and Hawaii instead of being higher.

edit: found a more clear figure for the 290X's typical power, so Fury is slightly higher.
 
Last edited:
BOM stands for bill of materials. By moving the memory to the GPU, the GDDR5 gets removed from the BOM and the area gets reduced.

What other BOM reductions do you have in mind?

Capacitors, resistors, coils, etc.... I do not know exactly what, perhaps elements of the power delivery circuitry, smaller radiator, less heat pipes. You name it.
 
The number of layers is more critical to the PCB's BOM than the area size alone. AFAIK, the most expensive reference board (512-bit) had ~10 layers.
Correct. For an 10 layer board, you're looking at something like at something like a $0.40/sq.inch in mass production. For a full size R9 290X, that's $18 for the PCB. Same layer board of 6 instead of 10in is $10.8. Make that $8 for a few layers less. So best case, that's a $10 savings on PCB, but since AMD must have better pricing, the difference is probably going to be quite a bit less.

Peanuts compared to all the rest.
 
Supreme common sense. Other than the smaller PCB, which part do you think will not be more expensive?

Shorter PCB, less layers because there's a lot less components, no VRMs dedicated for VRAM, less components to solder on the board so fewer assembly costs.
Also, while Fiji has 900M more transistors than GM200, AMD designs have traditionally been quite a bit more dense, so die area may not even be that different, making the chip just as expensive to produce as a GM200.
And then there's the fact that we don't know if 4GB of the slowest HBM are, in fact, much more expensive than 6GB of the fastest 7000MT/s GDDR5 there is.



You really have no clue, do you?
There are a few people here who actually have a clue as to either the Fiji cards are cheaper or more expensive than the GM200.

It's just that you're obviously not one of them.
So why not ease down on the false-truth and flame-baity posts like that one?
 
Last edited by a moderator:
Back
Top