LAS VEGAS – A number of reputable chip companies, including ST-Ericsson, Renesas Mobile, NVidia, Marvell, have been developing LTE baseband chips for more than a year. So far, there is little to show for it.
All claim they already have their modem chips on the tarmac. But that’s it. No takeoff. Nothing about actual design wins. CES last week was no exception. We heard zilch.
We can certainly blame the virtually non-existent LTE market for that. Except in the U.S., the roll out of LTE has been slow. If you want volume for your modem chips, you need to play not in LTE but the 3G market dominated by Qualcomm, MediaTek, Spreadtrum, Broadcom, Intel and others.
Additional blame goes to Samsung and Apple. The two companies have pretty much sewn up the high-end smartphone market. Samsung designs its own LTE baseband chips while Apple uses Qualcomm’s chips. Not much room is left for anyone else.
But the real story lies in the very nature of a baseband chip, which we in the media apparently don’t appreciate very much.
Developing a functional baseband chip is a never-ending job, according to CEVA CEO Gideon Wertheizer. When it comes to apps processors, chip suppliers know their job is done. It happens when they see tape out. In contrast, baseband chips, after tape-out, need to be tweaked, tested, modified, certified and/or tweaked again before they can move to mass production.
Wertheizer said, “I’ve heard that you’d need to dispatch something like 400 engineers to Samsung in order to have your chip tested in their handsets, certified by different operators and tweaked to work well on different bands, and finally getting designed in.”
As far as I know, The last time it was in a phone was the CDMA variant of Galaxy Nexus and I havent heard of it since. To me it sounds like its just not ready for prime time. All other SSG LTE phones have a Qualcomm chip. One of my friend working at http://movingangels.com has got this phone.
When I think about 400, its not a huge number, there is tremendous effort involved with validation. The field testing involves travelling across the target market identifying operational issues, fixing them, retesting.
Tweaking the stack for performance even if it means non-conformance to 3GPP is also not very uncommon.
Imagine the number of engineers required for traveling across Europe for testing, reporting, identifying the root cause, fixing and retesting?
This is because the complexity of the designs is increasing so much that soon companies won't be able to keep up with the pace of new standards. Just like it doesn't make sense to develop and maintain your own CPU, you use an ARM. It will soon be the same in the baseband business, you will let a company make the effort of designing and maintaining a scalable, software programmable IP, and just buy a license that fits your needs, in terms of processing power, size, etc. It doesn't mean that you lose a part of your business, it means you focus on what makes the difference for your product. This is just what this company does: www.simpulse-dsp.com
I disagree with the 'No takeoff'. One of the four companies mentioned in the first paragraph already has a LTE chip in the market, so thats a major leap and proves maturity. Lets see how the rest catch up.
The word on the street is that Samsung is indeed working on the LTE baseband chips for the merchant market. That, probably, is different from the one they are using in their own handsets. We just don't know when that will reach the market.
Its true that Samsung makes its own LTE baseband chips but it hasnt been commercialised like one would think. The last time it was in a phone was the CDMA variant of Galaxy Nexus and I havent heard of it since. To me it sounds like its just not ready for prime time. All other SSG LTE phones have a Qualcomm chip.
Europe is aggressively rolling out LTE and it will severly hurt the aforementioned 3G volume suppliers who are far from launching LTE chips
Reading the article, I had a feeling of deja-vu. If I'm correct in the early times of computing, there were different makers for CPU (Intel, Zylog, Motorola are the first names coming to my mind). After a few years, expect for specific servers, only Intel and AMD remains (the last blow was when Apple started using Intel in its Mac series...)
Do you think the mobile business is evolving the same way and that in five years or so only two will remain...?
This RF wireless modem business seems so difficult to me. The designs are becoming more and more extraordinarily difficult and clever, and then in the end, only a few survive and quickly become commodity items. It almost hurts to see so much effort expended, when you know that a lot of that work might never see the light of day.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.