Monday, December 15th 2014
AMD Appoints New Global Consumer Sales Head
As part of recent AMD organizational changes, Mario Silveira has been promoted to head of worldwide consumer sales within the company's Computing and Graphics business unit.
Silveira, 40 years old, joined AMD in 2010 and most recently served as head of consumer go-to-market sales for the EMEA region. As he broadens his responsibility to a worldwide remit, the role of head of consumer sales for EMEA has been assigned to Gabriele Rentocchini, who will now be responsible for overseeing all sales and marketing activities related to the retail and consumer go-to- market segments in the region.
Silveira, 40 years old, joined AMD in 2010 and most recently served as head of consumer go-to-market sales for the EMEA region. As he broadens his responsibility to a worldwide remit, the role of head of consumer sales for EMEA has been assigned to Gabriele Rentocchini, who will now be responsible for overseeing all sales and marketing activities related to the retail and consumer go-to- market segments in the region.
6 Comments on AMD Appoints New Global Consumer Sales Head
Just my point, they are behind in processes and lithography is the biggest key to that, so they are suffering the lack of better lithographic processes, and moving this clown around has jack shit to do with anything that will ease their problems other than more PR spin.
Back when they were on top it was mainly due to having a great wet lithography process shared with IBM, not as they made better transistors, not as their silicon was blessed, not as they were smarter, just better technical specs in manufacturing.
So, AMD are either implementing a plan so cunning you could pin a tail on it and call it a weasel....or the BoD continue to play musical chairs to the accompaniment of Yakety Sax.
You ask if you can have 14nm FinFET tomorrow and the answer is yes. Some folks are already testing this process but it doesn't mean it will deliver great advantages nor that it is a current requirement for increased performance. Intel's node size claims are always dubious as is there performance claims so take that with a large grain of salt as they have been exposed numerous times for their marketing lies.
The problem is you assume that smaller node process = greater product performance when in fact that is not necessarily true. If you look at Intel's shrinks over the past few years as well as AMDs it's blatantly obvious that the days of large gains for node drops are gone, which should be obvious when you consider where we are with node sizes and the small percentage changes provided by each additional shrink vs. when we were say at 90nm and making significant node size/percentage drops. The primary gains in reduced nodes, i.e. "better litho" is reduce power consumption and wafer unit costs. Performance wise there is only minute gains in speed. Thus a refined node can often provide as good or better performance and value than a rushed out the door next shrink. GloFo's refined 28nm is now being used for the latest AMD's graphic products where once they were using TSMC's. The power consumption is reduced and the performance increased while still on 28nm. That's a good thing.
AMD's past success was based on having leading performance products, not some advanced litho. While gullible PC enthusiasts might be fooled by node drop claims, technically educated folks know that the node size alone is not that significant these days and that the products performance is more important than the node size. Just as many PC enthusiast are duped into buying high frequency DDR3 or now DDR4 RAM, which offers little to no tangible system performance gains over 1600 MHz. DDR3 on discrete CPU powered desktop machines, PC enthusiasts can also be duped into believing that the node size or process is more important than it really is. The reality is this is not true. Performance is more important than litho, period.
None of your beliefs however has anything to do with AMD's personnel appointment in this press release. LOL
While yes, the primary gain is lowering power consumption/voltage over increasing speed at higher voltage on post-32nm(28nm) nodes, that still can vicariously translate to better performance simply by the use of more transistors (or even a higher clock using less voltage on a smaller process). In fact, this has often been AMD's main approach to improving performance; throw more cores/bw at it. Give them a watt, and they will turn it into performance either through units or clockspeed. This may or may not be dictated by the fact their general improvements in core efficiency have paled/trailed in comparison to their competition.
Not only have their chips been rather large since falling behind in terms of process nodes, they have admitted to cutting (pertinent, like the ALUs in bulldozer) features to keep power down. Their main designs choke on heat/power...clockspeed capability is not a main concern (for current products). Lowering power on the core designs certainly has to be.
Why don't we have a full 285x on this fully mature 28nm node?
Perhaps because it would use over 225w (and full length), and hence be a joke in the age of GM204.
The market demands that level of performance be more efficient, hence 770 discontinuation (while 760 still exists). On a smaller node (like 20nm), this would not be an issue.
While Intel has one way of going about things (lowering power at reasonably similar cpu performance while sequentially improving GPU performance until it hits a similar threshold, further-more then scaling/disabling certain chunks to meet the needs of mobile markets), AMD is a very different company with a very different approach. Rather than scale a single design, they seem content on creating new designs (through flexible ip chunks like gpu CUs, 4 jaguar core blovcks, etc) focused towards a certain market. Jaguar. K12. Zen. Console apus. Generally this is a better approach as it's more granular/focused for final yields, but it's still dictated by power consumption and heat. Even if the design is efficient, it doesn't matter if process dictates it use too much heat at an efficient clockspeed.
While I certainly agree there are tangible cost/yield as well as 'known quantity' benefits to using a mature process versus risk on the next (TSMC has given us many examples of why...be it the lack of HKMG/vias and hence problems at 40nm, the transition to 20nm [which is actually similar to a 28nm optical shrink hence less voltage/speed tolerance...think 80/110nm...] without finfets and hence lower performance gains vs cost at non-mobile levels, etc) there is no arguing both that moving quickly to a new node has been a key part of amd's former strategies and successes, as well as the fact it is currently sorely needed, whatever the cost, if they actually wish to compete on an even level.
I hope, for their sake, their designs are well-engineered around the strengths of 14nm from top (k12/zen) to bottom (arm), as well as the process performing/yielding well, with an acceptable time to market. They are separate issues, to be sure, but they need them all to compete with the combo of what their competition has in those regards (be it through tsmc or Intel's own fabs).
_______________________________________________________________________
Actually on topic: Congrats to this very BoD choice-looking guy...It'll take quite a bit to replace John Byrne (from what I understand), good luck to him.