• 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 Reports 2016 First Quarter Results + Conference Call [3 semi-customs this year]

tuxfool

Banned
There are a lot of ways to mitigate leakage. FinFET by design helps. Lower gate voltages (which FinFET enables) help. Advances in doping help.

If this is all it took, then the likes of Intel and AMD wouldn't all look at things like having more fine grained power gating and more complex throttling mechanisms, but what you said is all true. I'd speculate that the 20nm HP processes at TSMC failed because they couldn't get power under control.
 

MuchoMalo

Banned
The CPU in Neo would not be a design win. However, as the new GPU contains a higher CU count that does not exist in the generation of GPUs from which the PS4 GPU that would definitely be a new design and not counted in the original contract for PS4 APU.

That doesn't necessarily mean that it's Polaris though, even though the specs seem to match up. I don't know how they would count it if it's just the same arch with some added features and double the CPU count.

But as I said, if that's the case then it makes it extreme unlikely that one of the wins is a new XBox, and extremely likely that Nintendo's win is ARM and very low-power.

Edit: Also keep in mind that Apple is rumored to have wins for both 2017 and 2018 iMacs. I doubt that Apple is going to use a single chip for two entire generations of iMacs, unless another win is announced by AMD soon.
 
If this is all it took, then the likes of Intel and AMD wouldn't all look at things like having more fine grained power gating and more complex throttling mechanisms, but what you said is all true. I'd speculate that the 20nm HP processes at TSMC failed because they couldn't get power under control.

Of course, I was over-simplifying and focusing on the device physics level. There's plenty of stuff to do at the arch level as well.
 

Crono27

Member
What if two of chips are for nx handheld and console. Doesn't have to be one for each of big three. Obviously third is for vita 2.0 (;
 

AmyS

Member
Actually NX won't be one of the semi-custom wins...

...because it's fully custom. #specialsauce

:p

Hell yeah.

Then with NX not being one of the semi-custom wins (but still very much coming soon) we get even more new consoles overall :p
 

SRG01

Member
They have a semi-custom ARM server win? Ha!

That's actually a really surprising one. Wasn't SeaMicro spun down last year because of the lack of demand for ARM servers? Or did they just roll it into AMD instead of a separate division?

edit: Also, anyone who was holding AMD today just made 20% :D Currently at around 3.2 after hours, which is a level that hasn't been seen since 2014.
 

RandSec

Neo Member
1. One of the projects is supposed to be ARM: How do we know that NX is x86 and not ARM?

2. One of the projects is supposed to be "beyond gaming" but not VR, which complicates things. Without that, we would assume PS4K this year, then XB2 next year. The issue with both of these, of course, is platform compatibility and consequences thereof, since the most obvious technical improvements could not or should not be made if backward compatibility is needed. And Microsoft may be unhappy with the way things turned out from their last deal with AMD.

3. These are "Semicustom" designs, in which AMD supplies existing technology and the customer pays as NRE (Non-Recurring Engineering) for customization work as it occurs. The customer thus owns the resulting chip, but not the original IP. Since AMD has only a minimal project investment, with a guaranteed long-term contract and no marketing at all, the small margins involved still represent a substantial ROI.

4. How is a "shrink" NOT a "design win"? First of all, going from 28nm to 14nm is not classic shrink territory; it is certainly not automatic. The required healthy payments, and the decision to start, essentially mean a new project. But how is that project structurally different from the original "design win," and why would it not be described in the same way?
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
That's actually a really surprising one. Wasn't SeaMicro spun down last year because of the lack of demand for ARM servers? Or did they just roll it into AMD instead of a separate division?
From what I recall, they spun it down. Of course, the Freedom Fabric IP stays with AMD, and will likely continue to bring revenue.

1. One of the projects is supposed to be ARM: How do we know that NX is x86 and not ARM?
If the surprise of the report - the ARM semi-custom server win, is correct (I'm going from posts in this thread), then that would imply the NX is not an ARM SoC, as AMD have a single ARM semi-custom design win.

2. One of the projects is supposed to be "beyond gaming" but not VR, which complicates things. Without that, we would assume PS4K this year, then XB2 next year. The issue with both of these, of course, is platform compatibility and consequences thereof, since the most obvious technical improvements could not or should not be made if backward compatibility is needed. And Microsoft may be unhappy with the way things turned out from their last deal with AMD.
The 'beyond gaming' win is most likely the semi-custom ARM server win (I still can't wrap my head around who might be the customer). My curiosity rises, though, with regard to the rumored win with apple. Of course the apple connection could be just a stand-alone Polaris for high-end macs, but (a) that would hardly be semi-custom, and (b) given how Intel are extorting serious moneys from the PC industry with the CoreM/Y, I can't imagine Apple not wanting to test waters with an alternative x86 vendor. At least before they go full metal ; )

4. How is a "shrink" NOT a "design win"? First of all, going from 28nm to 14nm is not classic shrink territory; it is certainly not automatic. The required healthy payments, and the decision to start, essentially mean a new project. But how is that project structurally different from the original "design win," and why would it not be described in the same way?
Full nodes are never automatic shrink territory - only half-nodes are. Anyhow, shrinks are not a win per se, because they don't bring a new revenue stream - just a re-negotiation/adjustment to an existing one.
 
Top Bottom