The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
I do believe AMD said they were selling silicon, not IP. To me, this means they would hold the contract with TSMC, not Microsoft.
 
Would seem odd to me that ms would change back to a deal that seriously hurt them with the xbox instead of the deal that succeeded with the 36p
 
Would seem odd to me that ms would change back to a deal that seriously hurt them with the xbox instead of the deal that succeeded with the 36p

There's more to a deal than just who's in charge of manufacturing. There's no reason why there couldn't be a clause in the contract that compels AMD to shrink the APU down to 20nm and then 14nm when certain conditions are met (if that's what you meant by "hurt them with the Xbox").
 
Would seem odd to me that ms would change back to a deal that seriously hurt them with the xbox instead of the deal that succeeded with the 36p

It's likely Microsoft didn't negotiate the exact same deal that it did with Nvidia and Intel for the original Xbox.
If the problems were an inability to get regular die shrinks of the CPU and possibly graphics instruction licensing from Nvidia, the new agreement could have included a node transition schedule or shrink option along with some kind of understanding on GPU instruction set licensing.

It's not like going its own way for the 360 was without its challenges, and Microsoft wound up leaning on its chip design partners very heavily anyway.
 
http://techreport.com/news/25461/report-amd-to-introduce-arm-based-tablet-chip-this-year

Looking at how late their X86 CPU are I wonder if they could kill their X86 CPU line and develop their own ARM v8 CPU cores.
I read a couples of times that if ARM doesn't have the "power" advantage over X86 we used to think it actually easier to implement. D.Kanter stated that their least ISA (v8) should be pretty "easy" to implement. Against Intel they are outgunned though in the arm realm I would think that they have what it takes to compete against the best.
 
Some things are easier to implement, particularly in the case of AMD where their initial ARM products are using unmodified ARM cores.
I would hope that some of the real hard things that AMD has spent years on like power/frequency management, physical design, and system integration can be applied regardless of ISA. As has been found, the efforts of the ARM competition are not always that great.
It at least better be transferrable before competitors start to catch up.

It might be more that it's easier to implement up to the level of the competition when said competition isn't Intel x86. It may also be that since ARM hasn't traditionally had a presence in higher market segments with higher performance and RAS features, it's cheaper to build one in that segment than it would be to build an x86 core that fits into the mainline laptop/desktop/server mold, which has higher standards in terms of reliability and system validation.

While AMD has said at one point that an ARM core is much cheaper to design than an x86, I'm not convinced that the comparison was an apples to apples one.
 
http://techreport.com/news/25461/report-amd-to-introduce-arm-based-tablet-chip-this-year

Looking at how late their X86 CPU are I wonder if they could kill their X86 CPU line and develop their own ARM v8 CPU cores.
I read a couples of times that if ARM doesn't have the "power" advantage over X86 we used to think it actually easier to implement. D.Kanter stated that their least ISA (v8) should be pretty "easy" to implement. Against Intel they are outgunned though in the arm realm I would think that they have what it takes to compete against the best.

ARM field is "easier" in the sense that most use standard ARM cores, pretty much just Qualcomm and Apple do their own designs instead of just smashing together x ARM cores and GPU

edit: Yes, there's not as much competition on x86-side but at that side the competition is called Intel :p
 
Apple and the many former AMD employees employed by its CPU group show how important it is to get the implementation and physical design part done well, going by the A7.

The accumulated experience AMD has in processor design isn't fully leveraged if it doesn't at some point go for a more custom core.
 
ARM field is "easier" in the sense that most use standard ARM cores, pretty much just Qualcomm and Apple do their own designs instead of just smashing together x ARM cores and GPU
Not what I meant, I think this quote from D.Kanter's article on the ARM v8 ISA might do a better job:
The ARMv8 architecture is classically British; a clean and elegant 64-bit instruction set, with backwards compatibility for existing 32-bit software. The new AArch64 is certainly an improvement over ARMv7, with many improvements above and beyond simply extending the virtual address space to 48-bits.
The most notable additions in ARMv8 are the larger and highly regular integer register file, double precision vectors with IEEE support, and new synchronization primitives with a well-defined memory ordering model. In some respects though, the more significant changes came not from adding features, but removing them.
Like x86, ARMv7 had a fair bit of cruft, and the architects took care to remove many of the byzantine aspects of the instruction set that were difficult to implement. The peculiar interrupt modes and banked registers are mostly gone. Predication and implicit shift operations have been dramatically curtailed. The load/store multiple instructions have also been eliminated, replaced with load/store pair. Collectively, these changes make AArch64 potentially more efficient than ARMv7 and easier to implement in modern process technology.
There are no ARMv8 implementations available to judge the merits of the architecture in practice. But overall, ARMv8 is clearly a sound design that was well thought out and should enable reasonable implementations.
I also remember discussions on realworldtech (may be here too) on the matter with a consensus about how tricky X86 is to implement with its memory model(s ?), etc.

Unrelated to that is the fact that X86 in my view no longer qualify as a "single /monolith /searching for a bettter wording" ISA. Intel keeps implementing news instructions and features and AMD can't keep up (it's been a while actually). X86 is more and more Intel and Intel only ISA, the loop is close to be complete.

I believe that, for the reason 3Dillentante detailed, AMD could do great in the ARM realm. Now they use ARM vanilla cores but I think, it is about time to market.
AMD has designed some of the most complex CPU in the world, they have top of the class GPU and the matching software, their strength would better displayed against competitors that play by the same "rules" (same ISA, same SIMD extension, same process, etc.).
 
It is probably better to keep a few small and brilliant design teams for ultimate performance while using ARM for the mainstream. In AMD's position I'm not sure they have any choice anyway. They just can't afford x86 CPU design at 20nm and below but ARM cores give them a huge relief. It's the right move but probably 2 years later than it should have been made.
 
Dont forget AMD is working with ARM on the design level of ARM architecture (( not only on HSA collaboration, this make many years now they working together ), they dont really need to redesign the ARM cores, when they are working on their developpement ( their design choice is allready there )..

Anyway, they will offcourse keep x86 cpu, ARM are great for low level performance, SOC, and low power server.. but not for other tasks. ( not yet )
 
any news on when the desktop verisons of whatever replaces the E series actually hits ?

I've been looking all over
 
AMD is in the Green again
Okay, that sounded a little narly. But good news for AMD as they have managed to achieve a net income of $48 million on revenue of $1.46 billion, this would be the company's first profit since Q2 2012. A good chunk of profit came from a sale-and-leaseback operation of AMD's Singapore facility that netted the company $22 million.

Non-GAAP net income came in at $31 million. AMD's computing solutions (aka processors) sales fell from $841 million to $790 million versus the previous quarter, while graphics and visual solutions revenue soared from $320 million to $671 million in the same timeframe. The large increase in GPU sales is thanks to AMD's deals with Microsoft and Sony for the next-gen consoles, AMD stated that that sales of GPUs for PCs fell sequentially and year-over-year.

AMD (NYSE: AMD) today announced revenue for the third quarter of 2013 of $1.46 billion, operating income of $95 million and net income of $48 million, or $0.06 per share. The company reported non-GAAP operating income of $78 million and non-GAAP net income of $31 million, or $0.04 per share.

"AMD returned to profitability and generated free cash flow in the third quarter as we continued to successfully execute the strategic transformation plan we outlined a year ago," said Rory Read, AMD president and CEO. "We achieved 26 percent sequential revenue growth driven by our semi-custom business and remain committed to generating approximately 50 percent of revenue from high-growth markets over the next two years. Developing industry-leading technology remains at our core, and we are in the middle of a multi-year journey to redefine AMD as a leader across a more diverse set of growth markets."
 
Status
Not open for further replies.
Back
Top