Hmm... OK, it's been almost three years since Intel sold the XSCALE line to Marvell. Back just after the sale, there was a lot of noise about the Monahans part (PXA-320) and how power efficient it was, etc. We all waited, but the PXA-320 never really made the splash or the inroads into the cell phone market that it was supposed to.

In the mean time, TI has come out with multiple OMAP variants that are making news. One only needs to look at the BeagleBoard or the Gumstix Overo to see that OMAP is doing well. FreeScale has also made good progess with the i.Mx series. Good low-power processors that have a reasonable support infrastructure. So what happened to XSCALE?

My theory is that XSCALE suffers from Marvell's paranoia. The same paranoia that caused so many people to avoid Broadcom parts like the plauge. It's just too difficult to get information, data sheets etc. out of Marvell. Every time I've dealt with Marvell, it's been like pulling teeth to get any information out of them. I've got to get V.P. level signatures for NDAs that only last for 30 days so I can download data sheets with so many erratta that I have to keep re-upping the NDAs just to stay on top of it.

As a small unit production developer, I feel like Marvell is afraid that I'll steal their proprietary IP and produce a copy of their chip if they tell me what registers exist so I can port and O/S or a driver to their silicon. I just want to use the silicon. I need information to make that happen. If it's difficult to get the information, I'll use someone else's silicon who will provide me the info like TI or FreeScale.

OK, I've heard that for huge volumes, Marvell will provide great service. I don't doubt that for million unit quantities they will assign a dedicated resource to me. But, the majority of business volume in the U.S. is done by small businesses. And, if it's difficult to get data sheets as a small business, engineers who move to large business will still avoid the parts because of past experience.

So, I guess my point is that I'm not really seeing much buzz about XSCALE in the past couple of years since Marvell took it over. It's certainly not because it wasn't a good architecture. Look at all of the PXA-270 based phones that where produced when Intel ran the line. I believe that XSCALE is dying because of paranoia about data sheets falling into the hands of the open-source community.

My therory is that if Marvell wants XSCALE to stage a comeback, they need to change their data access policies. They need to embrace open-source and create a low-cost platform for the open-source developers to be able to port and experiment with. Without cooperation with the community, XSCALE will become a backwater eddy in the processor world and eventually die out due to lack of interest. I'm going to play with my BeagleBoard now...