|
Marvell unravels XScale mystery! |
|
|
|
Nov 29, 2006 at 02:26 PM |
Intel has made some significant attempts in the last ten years to spread its wings to enter related markets that require microprocessors, especially in the communications field. The recent troubles at the site of the largest chip manufacturer are well documented and the subsequent sell-off of divisions that never quite got to where Intel wanted is also in the public domain.
However, the recent Marvell Technology Group's $600 million acquisition of Intel's
application and communication processor division, better known for using the XScale architecture, has highlighted some very interesting aspects of Intel's operations that may not be widely known!
The vast majority of news stories have highlighted that Intel was shedding underperforming divisions with the understanding that Intel had not been able to compete well enough against incumbent competitors such as Texas Instruments in the mobile comms space.
That may be true at the macro level, but during Marvell's recent quarterly conference call with financial analysts, Sehat Sutardja, Marvell's Chairman and Chief Executive Officer, spent no time at all ploughing into the reasons why Intel wasn't competitive in XScale markets and stating that Marvell had a very strong case to turn things around.
"As we started the business, it is apparent that most of the factors that have led to these high product costs are centered around manufacturing-related issues, such as the products being built in low volume, captive manufacturing facilities that utilize 8-inch technology. Also, a variety of other manufacturing positions remain a focus on product performance, and sacrifice of cost."
Sutardja's simple plan was...
"One of the main tasks at hand is for us to quickly port the existing as well as the new designs over to our foundry partners to help drive down product costs. For a number of different reasons, the cost structure of these products is currently too high."
He went on to state that shifting production to foundry partners of whom TSMC is biggest meant moving everything to 300mm and the 65nm node. He felt that the product was actually significantly superior to any other - including TI's - and felt that there was a considerable amount of new business potential going forward. In fact, after a detailed review of the division, Marvell had decided to bring over more of the Intel employees than they had originally thought they would need as the opportunities were that good, they would need the people sooner rather than later!
But I am getting ahead of myself: the key point here is related to manufacturing - or more to the point, the lack of it!
Sutardja also noted that he had spoken to key customers and that the only complaint he had repeatedly heard of was lack of capacity. If Intel had allocated the manufacturing lines then the division would have sold more chips! Simple when you know how!
Now, that may have been a problem you would have expected from a small chip manufacturer that was caught between new fab build or poor cash flow or something, but you don't expect the largest chip manufacturer on the planet not to realize that capacity issues were at the core of the problem. Furthermore, Intel could easily gain access to foundries like TSMC to share production, especially since that is what TI does anyway.
Again, if this were related to a smaller company, they may not get what they need from a foundry, especially a in period of high utilization.
Xscale chips are primarily produced at Intel's ex- DEC campus in Hudson, but the 200mm fab is also home to chipsets. As Sutardja saw, lot runs allocated to XScale were not optimized or, as I see it, not prioritized sufficiently amongst other chips being produced at the fab.
Surely Intel could have sorted this out?
Anyway, that's history now and Sutardja is wasting no time in transferring all production possible over to TSMC. Unfortunately for Marvell, it will take about a year to qualify the first wave of products on a foundry process, and so they are compelled to use the Hudson facility and the subsequent limitations of this strategy into 2008.
So rather than product competitiveness it would seem from Marvell's perspective that it was manufacturing competitiveness that was Intel's problem. We will have to wait to see if Marvell is right - but this surely highlights how Intel can get things so wrong!
|
| Senior Technical Program Manager (Dublin, Ireland, 02/04/2008) | | Senior Embedded Design Support Engineer (Shannon, Co Clare, Ireland, 19/03/2008) | | Embedded Support Engineer (Intel Shannon, Co Clare, Ireland, 19/03/2008) | | Senior Component Design Engineer (Intel Shannon, Co Clare, Ireland, 19/03/2008) | | Product Marketing Manager (Silicon Valley , 14/09/2007) |
|
|
|
Blogs 
