Hi, does anyone know how TXN's DSP market is diveded between
different products, customers??
I've been using DSPs since late 70s, and used the first
generation TMS-DSPs up to appr 1990 for modems, also looked into GSM phones at that time.
However, after that I've concentrated on custom designed DSPs, to
be able to integrate pheripherals to lower chipcount, cost and power consumption.
Anyway, the reason for this question is that it seems as there are two main (high volmume) markets for TI DSPs, modems (USR) and
GSM phones ( Nokia, Ericsson??).
In both markets there is a huge pressure to integrate more on one chip, so (to me) it is a question of Texas providing solutions where one can add custom designed blocks to a TI DSP, or doing the whole thing from scratch, but maybe with a DSP license from TI??
The outcome of 56k battle between Rockwell and 3com (USR) might help or hurt TI, same if pressure on Cellular phones takes them to more custom design DSPs. ( H,ADSL market might also go with or without TI DSPs??)
Any input or comments??
Lucent's DSP share for 1996 was 28%. This has slipped to 26% in 1997 and is expected to be about 25% in 1998. The major reason is that their DSPs are not compatible with TI DSPs.
The company with a growing DSP share is NEC. SOME NEC DSPs are compatible with TI DSPs. Look for NEC to be somewhat of what AMD is to Intel.
You failed to comment on my major point, that becasue of the huge markets, the extreme pressure on cost, that the actual product manufacterers will have to integrate a DSp + a lot of peripherals on the same chip.
This is not only to save costs directly, but also to save power, save cost through saving battery costs.
ANother factor is that cost of manyfactering goes down when the level of intergration is higher.
However, there is a threshold for TI's customers to do this, the volume has to be large enough.
As a result I would guess that TI must try to avoid this by selling their DSPs at a very low price, this keeps the market theirs, but doesn't pay high dividends<g>
( and buying time like this they must scramble to build a licensing or chipset designing system)
<You're saying 35%. Source??>
Wasn't me, I quoted somebody else.
From what I know Lucent or AT&T had some share in third party phones some years ago, and probably lost most of it.
They had an architectural edge at that time, but has lost it when all DSPs have gotten faster. (plus some other factors)
I actually was very surprised at somebody quoting 35% for LU, and also wonder how or where that figure comes from.
PS On the other side, if true, that would be a good addition of 35% of the marketshare for somebody else <g>..
TXN practically owns the high end of the dsp market, which is growing at a 40%/year clip. TXN currently hold 45% of the dsp market of $800 million. At the current growth rate TXN would hold 45% of a $32.8 billion industry in ten years. TXN thinks it will be a $50 billion industry. DSP's are used in disk drives, DVDs, LANs, Cellular apps, Modems, Telecom, etc.
But most importantly DSPs are crucial to xDSL modems. xDSL may not be the real deal in high bandwidth data transfer, but it is the next big deal. The big telcos with tons of copper pairs are out there right now are pitching xDSL as the answer. This is due to the pressure the real deal (cable modems) is putting on the big telcos in the data trnsmission game. Thats another story but suffice it to say telcos will sell xDSL connections by the
millions before the real deal is availible to everyone and the telcos are on the penny stock page.
TXN will sell a bazillion DSPs to the xDSL industry before xDSL vaporizes. Add to that the previosly mentioned growth industries that eat DSPs and TXN's purchase of Amati (Amati's DMT has been selected by ANSI as the STANDARD for ADSL equipment) and we can't make enough DSPs.
IMHO DSP's are going to produce huge revenue for TXN well beyond the horizon of the average investor. That said can TXN extricate itself from the hopless DRAM market without a signifigant per share loss? What other potential pitfalls lurk in TXN's future?
The figures I've seen for Lucent's share of the DSP market are around the 26% level. Bear Stearns was one source I think. You're saying 35%. Source?? I'm curious which is correct because Lucent is TI's major competitor so the size of their share of this market is important
As I understand this is a $5B market presently, growing to $50B in about 10 years, (TI's numbers, I believe). TI
has about a 45% share of the current market. Lucent has 35%, and I believe Analog Devices has about 8%. >
This is the information which makes me wonder who is tweaking the numbers. From what I know about two major DSP markets:
- mobile phones: (digital only, forget the analog)
Nokia, Ericsson and Motoroal are the big ones, Motorola have lost markershare lately. The first two probably uses mostly TI DSps, but Motorola obviously uses their own.
- Modems (high quantities, although last year was low waiting for the 56k standard, and is going to explode and of this year)
Clear market leader is Rockwell, who uses their own DSP, USR (3COM) uses TI DSP, but is well behind Rockwell (who has historical etc connections with Motorola)
In both markets Lucent lost probably most of it's market share some years ago.
Anyway, the problem is that both markets are very tough in terms of price, the modem chips have dropped form maybe $100 to $10 in a couple of years, and same is about to happen in mobile phones.
(target end price for a phone is $100, doesn;t leave much for TI)
It seems like the only way to get the price that low is to integrate everything in one or two chips, so if TI wants to be in the business, they have to provide the phone-modem manufacterer with this capabilty, and not "insist" on selling standalone DSPs.
Otherwise these companies are forced to do the design themselves, leaving TI outside the business.
Another factor is when the models and the technology stabilizes enough, so that the same product can be sold for 2-3 years.
This would probably be when the Japanese step in, and use their consumer production skill to take the prices down to another level (and the quality up another level)
Anyway, I do not suggest that TI should go down this path, as it might be a path better taken by somebody else, but when it happens, there is going to be a lot of "pain" in TI and other companies involved??
Hard disks and multimedia are other DSP markets, but they will probably go the same path, become standard, very-very-low margin markets, and that's a tough place to be?? (hard disks are already??)
On the positive side TI already bought Amati (ADSL) which seems to mean they are really getting into the DSP-chipset business, not just stand-alone DSPs.
However, they cannot get into all DSP markets with chipsets, so they have to find a suitable compromise??
The HiFi Japanese companies + Philips are also getting deeper and deeper into DSPs through audio,etc, so at some point whatever will happen??
Intel would of course like to grab the PC-part of the DSP market, but have to do some tricks because of monopoly problems ( like coming up with a useless MMX to follow up with a better MMX-2)
In a way I feel very unsure about an Intel-like "dominance of DSP markets" on a long range, because DSPs are not the same as DOS and Windows, as they are embedded super-low cost systems which doesn;t have to be compatible with a lot of third party products. (although Microsoft would like every refrigator to run ontop of Windows.. SPOX, NSP and all that BS.. but maybe WIndows CE??)
Even if the cost of DSP would keep a stand-alone DSP market alive, the power consumption demand will kill it for mobile applications ( includes phones + modems + harddisks for Laptops)
Here I am also thinking of some early microcontrollers which dropped from $10 to $1.50 when they got into the refrigator market..
PS btw. Cars are also a hugh market for DSPs, soon there will be 10-50 DSPs in every car... and the plastic around them will be more expensive than the chip itself.
Hi, Young TI
Thanks for the post and welcome to DSPs.
I wouldn't say DSP is so difficult, if one takes care to go down to the real "basics". However, one needs to have some basic "skills" in the math used, that is, Laplace (poles and zeros) and Fourier transformation (FFT).
They might seem like "magic" when wieved only from a high level, but if one takes a short trip down to a couple of simple examples, they all become very "understandable".
Same applies for bit accuracies, etc, there is a lot of BS in the academic papers, where academics "decline" to go down to the real bit-level, where things are simple.. which kind of suits them..
To make efficient programs one should have the blockdiagram of the CPU on the wall, and try to think of how the data comes out of the RAM, goes to here and there, registers are changed, etc, and then write the code to do it.
If one just thinks of the code as "text" on a paper, it ones more becomes kind of "magic".
I guess I was lucky because I started from building CPUS almost from scratch and programming in assembler, I found it difficult to take many high level C-guys through the needed learning process, as they "decline" to leave the high level world where they feel secure.. and same for many mathematicians who happily spend time dividing 1 by 0..
Feel free to post me any questions Gunnar
Integrating circuits onto DSPs is very doable and TI is
in a great position to do it (if they're not already) with
outstanding mixed-signal and analog departments.
However, the question remains: when does it become cost
effective? For high-volume customers it makes a lot of
since, but there are still technical issues that can hurt
yield and time-to-market.
I just pray that TI will maintain its DSP dominance, because
we all know that DSPs will be the single largest growing
electoric device and the company with a big chuck of this
market is sure to do extremely well.
ADI has, and will continue to give ti a run for it's money ( and market share ). for all the reasons stated as assets for ti ( i.e. good mixed signal position ), plus they have good offerings, at lower prices in some respects. their software tools don't match the quality of ti's, but they are improving. but for now, ti still hold the lead in market share.
but i'm in on both, so let them both thrive! so far, ti has been the more lucrative investment.
Reason for the run up today were comments by Charles Boucher of UBS Securities. He said, "TXN's focus on high-speed communiction markets and it's broad tech base provides the potential for very strong growth.
Young Tier, I don't give advice other than Don't put your eggs in one basket, buy wise, hold. If I told you my status in life you would think me a blow-heart and I don't want to do that.