According to AT's review, the Fury Nano is a full Fiji die albeit downclocked and (likely?) power limited. Almost like a "mobile" version, so it'd likely be in the 800ish range. Makes sense, and they should be able to hit the power and performance targets just fine. It's just going to be expensive; probably at least $499. I have to wonder what the availability is going to be as well considering I'm sure they would rather use the good dies for Fury X instead of the cheaper Nano.
Hopefully AMD has a mobile version, Fury Nano would make for a beastly DTR laptop.
Why would they put full Fiji into a cheaper product? Why not cut it down? Do they have enough rejects from Fury X that still run fine at lower speeds?
How can Nano be full Fiji and not cost more than Fury, which is a cut down Fiji and costs $549?
AMD would be extremely generous and yields are very good for lower tier products to be full Fiji.
Agreed. It's not in the business model for such a large GPU to be full up at such a low price point. Yields would have to be extremely good or, as you said, AMD would need to be extremely generous. The former is good for them and their future, the latter is not so much.
It seems to me that if Nano will be at 800mhz stock, then it could well be faster than a cut down Fury, which makes no sense with the given pricing. Plus if Nano can overclock a little, it will be close to X performance.
This comes straight from AMD. And while the rep I spoke to could be mistaken, I don't believe they are.I'm not convinced that Nano will actually end up being a fully enabled die, but even if it is (assuming the 800 MHz estimate is correct), you get approximately a 24% performance reduction from Fury X. That leaves plenty of room for a Fury at 1050 MHz (or even 1000 MHz) to slide between the two with reduced core count.
I think everyone is/was hoping for something that competes with the 970's price range, but isn't a 290x/390x, only because of the nice form factor they showed, and the potential that it offers.
It would be a no-brainer against the gimped 3.5GB+.5GB 970.
Maybe a Nano light
As to the OP, it would be nice to see AMD aiming for the efficiency sweet spot rather than just adjusting their clocks to meet/beat the competition. The Hawaii launch in particular smelled like a card pushed out of it's comfort range. Personally I've got a lot more respect for a card which stands on it's own qualities rather than stretching to reach someone else's benchmark.
I assume the reason we don't get full Tonga is because it's creeping up towards 290/390(X?) performance & would cannibalise their sales. Perhaps this is the card you're looking for?
Fiji seems to be ROP bottlenecked, which hurts its geometry performance. I think the same is true of Tonga, since Tonga is basically 1/2 of Fiji in all stats (aside from the memory interface). 64 ROPs aren't enough to back up 4096 SPs, and I think 32 ROPs aren't enough for 2048 SPs. If full Tonga only offers a minor performance improvement over the cut-down version, then it might not be worth putting out a separate SKU for it. It's worth pointing out that the 7970 only outperformed the 7950 by a few percent when clock speeds and RAM speeds were matched, and those Tahiti-based chips had the same shader/TMU/ROP counts as Tonga XT and Tonga PRO, respectively.
Apple cares more about OpenCL performance, so the Retina iMac can still benefit from full Tonga in apps like Final Cut Pro, where the ROPs aren't an issue.
If I'm right about this, then the corollary is that we could expect the cut-down Fury card to perform very close to the full Fury X at equivalent clocks (at least in gaming).