NVIDIA to acquire Icera

Arun! Your mastermind is behind this, isn't it?

 

Arun who ?

It's a necessary move from Nvidia. If you want to play a role in connected devices (tablets etc) or smartphones you need a base-band supplier. Some customers prefer a "kit".

Now.. what is Nvidia going to do with Bluetooth, NFC, WiFI, GPS etc ? CSR is around the corner in the UK I assume ?

 

So this means future Tegras will have integrated baseband processor, like today's Qualcomm's offerings?

So this is one more step for nVidia's "world domination" attempt, and at the same time one more step away from the x86 PC market (as the guy clearly indicates during the blog entry).

I wonder how much better this can be, as it requires the company to have two manufacturing lines (CDMA and GSM) for the same SoC (well, technically not the same SoC anymore).

 

They have no choice. PC market is at its max, the future is in portable devices. Just a Tegra is not sufficient, no way you can compete with Qualcomm.

As I already wondered, what's next for them ? Connectivity ? For this, they would have to splash another 300-500M million on e.g. CSR.

 

I'm sure we all hope this will come to fruition soon.

 

Good God, what have you done?

 

Lots of SoCs have done well in the market w/o integrated baseband, although I don't know how much this will change in the future with more companies moving to it.

 

http://newenterprise.allthingsd.com/20110509/nvidia-acquires-icera-the-internal-memo/

 

Given the fact my last question for Rayfield at MWC was "So what are the chances of seeing a bunch of joint smartphone design wins with Icera?" (paraphased), and that we had a short discussion on Icera at the press dinner before that (where I told him their architecture was at least as efficient as the SDR arch from CEVA or Coresonic) along with an argument on whether it was viable to focus exclusively on the high-end... (before even considering all our previous short discussions about Icera in the last 2.5 years)

You could say I played a very very small part. But more seriously, the single most important factor by far was the fact that Icera's revenues went down noticeably starting in around March as the USB dongle market went down (see: http://seekingalpha.com/article/267...sses-q1-2011-results-earnings-call-transcript). I'm sure NVIDIA (and TI, and others) must have talked with Icera about this for many months (see: http://www.informilo.com/20101207/building-europes-next-billion-dollar-businesses-iceras-story-327) but I suspect the price was just too high for anyone to accept.

When that unexpected revenue shortfall happened, Icera would no longer have been in a position to get to break-even & IPO without a fair bit more VC funding, which would have been very hard to get at this point. And so they probably reduced their asking price significantly (if I had to guess, I'd say they wanted upwards of $500M at the very least before, maybe as high as $750M, or perhaps they considered becoming a joint venture with Samsung like Infineon was proposed last year). It wouldn't have made any sense for NV to have bought them at $600M, but at $350M it's suddenly very appealing. And I can't say I'm unhappy about this, it will be interesting to see what happens now (even if it'd have been nice for the semiconductor start-up market if they had fetched a higher price).

Exactly. But what you're missing (along with most other people) is that you can build a kit even if you're two separate companies. This acquisition won't change much on that front - it's all about making more money via bundling (short-term) and roadmap synergies (long-term) along with making sure Icera doesn't fall into enemy hands (ala Intel buying Infineon).

CSR and Broadcom are both going to continue providing very strong solutions on the open market for a very long time. And anyway, the process tech lacks behind too much so there's little opportunity for integration unlike basebands (unless you'd integrate with the cellular RF which might be interesting). And there are quite a few target applications where Tegra wouldn't fit in, so the bundling synergies aren't optimal either.

Developing 60GHz (WiGig) in-house would be interesting and potentially a good fit for Icera's architecture, but I'm skeptical it's worth the time/resources/risk given the synergies and market potential aren't quite as good as with cellular. So at this point, I'd say they should focus on what they know, and try to expand into that market instead of wasting time and effort elsewhere.

"Vell, [strike]Zaphod's[/strike] Arun's just zis guy, you know?"

 

Now all of a Sudden Nvidia has the missing pieces of its Puzzle. GPU that works great with GPGPU, Desktop Games, and they are working their way onto Mobile GPU, Licensee of ARM and working on 64Bit ARM, and finally a very good baseband processors.

I wonder why Apple didn't buy it, 350M is pocket money to them.

 

Probably multiple reasons:
- integrated baseband is very useful for low or mid end, but less so for high end.
- margins on baseband chips are typically smaller than socs. So there is less benefit it making it yourself.
- baseband integration can seriously complicate time to market because it's much harder to qualify silicon. This is bad if you want to have the best time to market for new features.
- a baseband is a baseband: not much more than a generic data pipe and interchangeable with other components with serious impact on the overall system that's user visible. SOC feature are much more of a determining factor for the final product.

 

Icera does have excellent base band processor techniques. It's a pity that nVidia did not get any success design wins after bought ICera, especially considering that Icera play very well in era of 3G.
nVidia has just claimed not focus on main-stream phones. That's bad news for Tegra 4i, integrated with Icera soft-modem. Tegra 4i could be a good platform for cost-sensitive main-stream phones in 2013. But until now, no product has coming out. The quad-core A9 is too old for main-stream market for 2014.

 

I actually think that Tegra 4i is a pretty well balanced SoC for the mainstream market. I'm just wondering what's taking them so long. There should have been smartphones on the market with Tegra 4i three months ago. I guess NVIDIA is having trouble delivering, again.

 

Well, any device based on Tegra 4i was not expected to come to market until Q1 2014 at the earliest. So I would expect to see a product launch at MWC 2014 later this month (Feb 24-Feb 27). And technically, it is the OEM that has to deliver the finished product. Anyway, T4i is NVIDIA's first voice modem product, so they went through very rigorous certification testing through AT&T. This resulted in a nearly one year lag in time between when the SoC is first sampling and when the SoC is incorporated in a finished product for sale. As for more widespread adoption of an Icera modem, the issue here in the USA is that some major wireless providers (including Verizon and Sprint) still rely on CDMA technology for their 3G network, which means that use of Qualcomm's modem is required.

 

CDMA is technically inferior (last I heard anyway) as it couldn't mix voice data and "data" data at the same time, which is crap when you have a smartphone. This arbitrary limitation is silly, because it's all digital data to the system anyway regardless if it's voice or not being transcieved...

 

Well it seems we would need a new title like Qualcomm to pay for Icera

 

I wonder how hard it would be to move baseband processing onto the GPU cores.

 

Basebands have real-time constraints, not sure that would work well on GPU.