Why did Intel kill off their modem program?
semiaccurate.com
semiaccurate.com
I'm also an unfortunate owner of an Arris box. If only the person who showed me this website could have done it a few years ago...
This guy said it added about 8ms of latency: https://forums.timewarnercable.com/t5/Connectivity/arris-tm1...
So maybe not a huge issue, but personally I’d just spent the $40 on a SB6183 that doesn’t have the issue.
That's because the software is essentially lowest bidder and untested.
(Note: running in a lab is not the same as adversarial testing.)
Mine is a hardware version 3 though.
It's most noticeable in online games, for regular usage, you probably won't see much of a problem.
[1] My most recent purchase had the Intel® Dual Band Wireless-AC 8265. This is a list of all of Intel's wireless PC chips: https://ark.intel.com/content/www/us/en/ark/products/series/...
Is it technical, IP-related, or both?
Living in the information age is great! I just wish I had more information.
Once they're out of the business, they have no incentive to say why, they just focus on other things.
Since Intel has not always played fair against AMD, they don't really want to talk about anticompetitive business practices in a market they're exiting anyhow.
1. Intel's own Culture and Incompetence. There is a very long lead time in Semiconductor industry. What you are seeing today are likely work started 3 to 4 years ago. But even having that in mind, the sale of Infineon to Intel were completed in early 2011. It was announced and preparation ( assuming they did prepare ) in 2010. And the first Intel self Fabbed Modem? The Intel 7560 Modem, on 2018 iPhone.
2. What ever politics they had at Intel, Pat Gelsinger was out in 2009. Like I said there are 4 years of his work release after he left. That is 2013. And you can now look at what Intel has achieved in 2014 - 2019.
3. Intel Custom Foundry is a complete mess. And anything non-CPU are done on the custom foundry node. Such as Modem and Altera's FPGA. And we all know their 10nm is 2 years late, or 3 years depending on how you count them.
4. It is not IP, or technically hard to design a working LTE modem. It is extremely hard to design a modem that works well in all edge cases around the world. And if I remember correctly from the first day of Qualcomm/ Apple trial. The intel Modem were priced closer to $20, not $10 as SemiAccruate Report. Qualcomm spends lots of money in Field Testing as well as working with all Mobile Phones vendor / MNO in testing. When Qualcom has 100s of clients and design wins they get much more testing done and data than all other players on the market. There is a reason why Qualcomm has the best modem on the planet and you are only paying a slight premium for it. Its because they throw money into testing. While Intel, well I don't know what they have done in those 8 years, at least they have Apple as a client ( For a brief moment ).
While you may question how much Qualcomm charges for their Patents, ( In my opinion, not a ridiculous amount as media likes to paint that picture ). I surely would pay $20 - $30 more RSP for a quality Qualcomm Modem, than saves the money on a worse Intel Modem.
5. I have always felt there is a strong Intel Culture that makes any acquisition fail to have any synergy.
I think SemiAccurate reported that Intel's price was $-10, even.
Until the company is at an existential crisis level (a la the DRAM changeover--and even that almost killed Intel), this simply will not change.
There are a few objective reasons making it difficult to compete in the cellular business and a few Intel specific reasons:
- Intel's process and FABs are designed for high performance x86 chips. Not low power cellular chips. Given that the margins on x86 chips are also way higher, there was no incentive to change anything. So the cellular modems were produced using a power hungry process and were less competitive.
- At the time, Qualcomm was holding most essential patents required for implementing 3G (and probably now 4G). Licensing fees are enormous burden on anyone entering the market.
- Cellular technology undergoes tons of tests by operators, handset manufacturers, base station equipment manufacturers. It takes a lot of effort (and manpower) to execute all these tests and make the required modifications. It doesn't help that the reference platform is Qualcommm. So if you had made different a different tradeoff than Qualcomm, you may do better in some tests but will have to make fixes for other tests.
- Some problems are only seen in the field and only in certain places (due to quirky base station equipment, non standard deployment, etc.) Again, it takes a lot of effort to detect and fix those issues.
- Qualcomm (The big incumbent in this business) is huge: over 30,000 employees. No way you can compete if all you have is about a 1000-engineers in your cellular division.
By the way, the only successful cellular maker I'm aware of besides Qualcomm is MediaTek (Taiwan) which:
- Concentrates on the low end market.
- Ships complete solutions. You can find lots of different phones which are essentially a single MTK design with very few modifications.
From what I know, you don't need any licensing to design and implement a 3G/4G/ or even 5G Modem. ( Unless you want CDMA Support ) It is the ODM or Phone Manufactures that has to take care of it. Much like you don't need a license to include H.264 decoder, not even to ship it. But you will need to pay if you decide to enable it.
>only successful cellular maker I'm aware of besides Qualcomm is MediaTek
Mediatek is awesome. By concentrating on low end, something Qualcomm doesn't want to do, they have an economy of scale for both Fabs Capacity, Design and Modem testing.
And just a note, Huawei, assuming they will Settle with Qualcomm, and Samsung both have their own Modem Implementation. These two represent close to 40% of market. And if you include Apple down the line, that is roughly 55%. Adding Mediatek, that is close to 70%. And that is excluding BBK and Xiamoi plans to make their own modem, together they represent 15% to 20%. Qualcomm doesn't have a monopoly as media likes to paint it.
3G is based on CDMA technology (wideband CDMA) and does require licensing fees. It doesn't really matter whether the chip manufacturer or the phone maker pays the fees. The end result is that non-Qualcomm chips make the product more expensive (and Qualcomm gets stronger anyway due to licensing fees).
That was before the settlement, and Samsung needed Qualcomm to access market with Network that was built on CDMA2000 ( Verizon ) going forward Samsung will likely be expanding their SoC across their product line.
>3G is based on CDMA technology (wideband CDMA)
The CDMA in reference were CDMA owned by Qualcomm or CDMA 2000, WCDMA and the China version of TDS-CDMA has cross patents agreement with other W-CDMA vendors.
I'm seriously unconvinced it's serving a wider public interest. LTE is a standard. It should be RAND to implement good FPGA for it.
> Update April 18, 2019 @ 8:18am: My math is off, it is closer to $10K/phone, $8750 to be exact. I will still take the cash option.
yikes!!
Why not? Presumably they would be opening themselves up to some liability?
If you exclude Unisoc and Intel, the other could represent a total of up to 70% Smartphone Market. That is assuming Mediatek continues to dominate the lower end, and the top 3 vendor, Samsung, Huawei, Apple's market share are stable.
There are quite a few other players focusing on M2M LTE solution.
Got a citation for that?
They design motherboards, and share their designs with their OEM's- much like reference GPU's.
They (via acquisition) have some of the better wired NIC's on the market (their features for sure beat Realtek's, and are more or less on par with Mellanox's.)
[Intel doesn't] have a clue about how to design circuits or software.