Intel leaving low-end chipset business?

  • Thread starter Thread starter Yousuf Khan
  • Start date Start date
Del said:
Or offload the manufacturing to tsmc/chartered/etc

That's exactly where they are produced right now. Only Intel makes
their own chipsets, everybody else goes through one of these contract
manufacturers. When they're already being produced at
tsmc/chartered/etc. who have other customers they are committed to
already, who takes precedence? VIA & SiS because they can sell a few
thousand more cheap chipsets; or do they put off Broadcom, Altera,
Nvidia, or whoever else is also their customers?

Yousuf Khan
 
YKhan said:
That's exactly where they are produced right now. Only Intel makes
their own chipsets, everybody else goes through one of these contract
manufacturers. When they're already being produced at
tsmc/chartered/etc. who have other customers they are committed to
already, who takes precedence? VIA & SiS because they can sell a few
thousand more cheap chipsets; or do they put off Broadcom, Altera,
Nvidia, or whoever else is also their customers?

Yousuf Khan
You make the assumption that the independent foundries are running at
full capacity. I would guess that is not true.

Del
 
Del said:
You make the assumption that the independent foundries are running at
full capacity. I would guess that is not true.

It's not my guess, it's the premise that these articles are about.

Yousuf Khan
 
That's exactly where they are produced right now. Only Intel makes
their own chipsets, everybody else goes through one of these contract
manufacturers. When they're already being produced at
tsmc/chartered/etc. who have other customers they are committed to
already, who takes precedence? VIA & SiS because they can sell a few
thousand more cheap chipsets; or do they put off Broadcom, Altera,
Nvidia, or whoever else is also their customers?

While we may see some short-term supply problems, the laws of supply
and demand will tend to correct things fairly quickly. TSMC,
Chartered, UMC, et al. are in the process of building more capacity
all the time. SiS doesn't even have to worry about this because,
unless something has changed recently, they still have their own fab.

Simply put, there is a LOT of semiconductor fab space out there. At
any given time there are dozens of companies going through upswings or
downswings in demand. It may take a few months for companies like VIA
to get a bit of extra fab space, but not much more. Fortunately it's
not like Intel will immediately run out of chipsets on one given day,
there will be LOTS of inventory floating through the channels for a
few months.

As I said, we might see a short-term supply issue that will result in
some odd pricing (ie low-end stuff going up to be only a few dollars
cheaper than higher-end stuff), but it will be correctly quickly
enough.
 
Tony said:
While we may see some short-term supply problems, the laws of supply
and demand will tend to correct things fairly quickly. TSMC,
Chartered, UMC, et al. are in the process of building more capacity
all the time. SiS doesn't even have to worry about this because,
unless something has changed recently, they still have their own fab.

Yeah, actually something did change, and not so recently, UMC now owns
SIS. According to SIS they have obtained guarantees from their UMC
parent for additional fab space if necessary. Not sure how much
flexibility UMC has in moving aside their other existing customers --
can't see UMC playing too much favourites with their subsidiary vs.
their customers, otherwise it would become a public relations nightmare
with their paying customers.

Here was the announcement of the merger between SIS and UMC:
http://www.xbitlabs.com/news/other/display/20040226125122.html
Simply put, there is a LOT of semiconductor fab space out there. At
any given time there are dozens of companies going through upswings or
downswings in demand. It may take a few months for companies like VIA
to get a bit of extra fab space, but not much more. Fortunately it's
not like Intel will immediately run out of chipsets on one given day,
there will be LOTS of inventory floating through the channels for a
few months.

Another issue is that neither SiS nor VIA seem to have their
competitive integrated graphics chipsets ready to go yet. For SiS, it
is the SIS662 chipset, which it only expects will be ready to sample in
Q1 2006. Meanwhile, VIA's P4M890 integrated graphics chipset may be
ready for Q4 2005. They've both been concentrating on AMD chipsets for
the last little while, so it seems they had deemphasized their Intel
products since they were expecting to be just niche players in Intel
territory before they found out about these Intel plans.

http://www.digitimes.com/news/a20050809A6027.html

Yousuf Khan
 
Yeah, actually something did change, and not so recently, UMC now owns
SIS. According to SIS they have obtained guarantees from their UMC

Hmm.. I must have missed that one..
parent for additional fab space if necessary. Not sure how much
flexibility UMC has in moving aside their other existing customers --
can't see UMC playing too much favourites with their subsidiary vs.
their customers, otherwise it would become a public relations nightmare
with their paying customers.

If they were a North American company they might, but things tend to
work a little differently in China/Taiwan. They seem to have ways to
just sort of make bad public relations kind of go away and if anyone
asks their questions seem to get lost in the translation somewhere.
We all put up with it though because of the cost advantages.
Another issue is that neither SiS nor VIA seem to have their
competitive integrated graphics chipsets ready to go yet. For SiS, it
is the SIS662 chipset, which it only expects will be ready to sample in
Q1 2006. Meanwhile, VIA's P4M890 integrated graphics chipset may be
ready for Q4 2005.

Even if Intel were to stop producing low-end chipsets tomorrow it
would still give them lots of time to ramp up production. Besides
they would probably start by replacing the old i845GV and i865GV
chipsets that are still being sold in pretty large quantities. They
don't necessarily need the latest and greatest. For example, VIA's
P4M800 Pro should be more than sufficient to fit the bill for any
low-end chipsets that HPaq or Dell might want.
They've both been concentrating on AMD chipsets for
the last little while, so it seems they had deemphasized their Intel
products since they were expecting to be just niche players in Intel
territory before they found out about these Intel plans.

http://www.digitimes.com/news/a20050809A6027.html

Hang on a second... This article says absolutely nothing about low-end
chipsets, it says that Intel is getting rid of two of their
"mainstream" (ie middle of the road) chipsets, the i915GL and i915PL.
They will still have their i865GV, i910GL and i915GV, all of which
come BELLOW the pecking order of the two chipsets that they are
discontinuing! Hell, the i915PL doesn't even come with integrated
graphics!

What Intel is doing is getting rid of most of their non-DDR2 chipsets.
The only difference between the i915G and the i915GL is that the 'GL'
model lacks DDR2 support. Given that these two chipsets almost
certainly use the exact same piece of silicon (with DDR2 support
purposely disabled in the 'GL' model), it probably only makes sense to
discontinue the lower-end model and just sell the i915G for the same
price now that these chipsets have been upstaged by the i945 series.
Same story goes for the i915P vs. i915PL.

This is QUITE a different story than what you first posted.
 
Tony said:
Even if Intel were to stop producing low-end chipsets tomorrow it
would still give them lots of time to ramp up production. Besides
they would probably start by replacing the old i845GV and i865GV
chipsets that are still being sold in pretty large quantities. They
don't necessarily need the latest and greatest. For example, VIA's
P4M800 Pro should be more than sufficient to fit the bill for any
low-end chipsets that HPaq or Dell might want.

I think all of the old 800-series chipsets were announced to be
mothballed a long time ago.

BTW, Dell just introduced its first system based on an Nvidia Nforce
chipset.
Hang on a second... This article says absolutely nothing about low-end
chipsets, it says that Intel is getting rid of two of their
"mainstream" (ie middle of the road) chipsets, the i915GL and i915PL.
They will still have their i865GV, i910GL and i915GV, all of which
come BELLOW the pecking order of the two chipsets that they are
discontinuing! Hell, the i915PL doesn't even come with integrated
graphics!

What Intel is doing is getting rid of most of their non-DDR2 chipsets.
The only difference between the i915G and the i915GL is that the 'GL'
model lacks DDR2 support. Given that these two chipsets almost
certainly use the exact same piece of silicon (with DDR2 support
purposely disabled in the 'GL' model), it probably only makes sense to
discontinue the lower-end model and just sell the i915G for the same
price now that these chipsets have been upstaged by the i945 series.
Same story goes for the i915P vs. i915PL.

This is QUITE a different story than what you first posted.

Maybe, but it seems to be a constantly evolving story too. There was a
story on Dow Jones Newswires today saying that Intel has now asked SiS
to help it out by building chipsets for it:
08/09/2005
Dow Jones News Services
(Copyright © 2005 Dow Jones & Company, Inc.)



TAIPEI (Dow Jones)--Intel Corp. (INTC) has placed an order with Taiwan's Silicon Integrated Systems Corp. (2363.TW) to help ease a chip shortage, the Commercial Times reports, without citing sources.

Silicon Integrated may sell chips to Intel as early as October, the report says. The chips would be used in Intel's personal computer motherboards.

United Microelectronics Corp. (UMC) will make the chips that Intel has ordered, the report says. UMC owns a controlling stake in Silicon Integrated.

Intel, which is phasing out some of its older chips, has ordered chips from Silicon Integrated to stem the shortfall, the report says.


Newspaper Web site: http://www.chinatimes.com


-By Alan Patterson, Dow Jones Newswires; 8862-2502-2557; (e-mail address removed)

-Edited by Sharon Buan


(END) Dow Jones Newswires



Yousuf Khan
 
It's only these builders that'll care.

I have to agree that builders are not likely to use cheap parts, so
this isn't likely to affect them.
No, as I said, it's not baseless, Intel is definitely saying they are
having trouble producing enough chipsets themselves. That's completely
established. The only rumour here is what Intel is going to do to
alleviate the situation, and in this rumour they're going to alleviate
the situation by cutting their less profitable chipsets. HOw's that for
a "grounded" rumour?
There is such a demand they are having a hard time keeping up, so they
are going to leave the market? That doesn't sound like sound business
practice to me. Logically, they should stay in the business because
there is high demand.
 
I have to agree that builders are not likely to use cheap parts, so
this isn't likely to affect them.



There is such a demand they are having a hard time keeping up, so they
are going to leave the market? That doesn't sound like sound business
practice to me.

You are missing the point: Intel has the fab capacity to fill
the demand for high-profit chipsets *OR* low-profit chipsets.
They do NOT have the capacity to fill BOTH demands. Hence they
have chosen to use their limited fab capacity to make the
chipsets that have the higher profits. It would be absolutely
idiotic for them to do anything else.

Logically, they should stay in the business because
there is high demand.

No. Staying in the low-profit chipset business means diverting
some of their limited fab capacity away from the manufacture of
more profitable chipsets. It would be extremely stupid for them
to do that. If they think demand will stay high for a long time,
they can build more fab capacity but that takes several years:
until then they have no choice but to use their limited fab
capacity in the most profitable way they can.
 
No. Staying in the low-profit chipset business means diverting
some of their limited fab capacity away from the manufacture of
more profitable chipsets. It would be extremely stupid for them
to do that.
If they think demand will stay high for a long time,
they can build more fab capacity but that takes several years:
until then they have no choice but to use their limited fab
capacity in the most profitable way they can.

Rob, I'm under the impression that chipsets are built on fabs that are
no longer usable for front-line CPU manufacture. Such as .13u or .18u
now. Intel doesn't build chipset fabs, they wait for existing CPU
fabs to become obsolescent. As you say, making cheap chipsets
apparently is not their best use for their existing .13 -.18u fabs.
No way is Intel (or anybody else) going to start construction on
additional .13u fabs at this time.
 
There is such a demand they are having a hard time keeping up, so they
are going to leave the market? That doesn't sound like sound business
practice to me. Logically, they should stay in the business because
there is high demand.

These are mainly support chips for their CPUs. There is high demand for
their CPUs (obviously), and therefore there is high demand for their
chipsets. But chipsets are low profit. Also if they try to maintain
their stranglehold control over their chipset market, it's going to mean
there's less chipsets available and it will reduce the demand of their
CPUs (nobody buys a CPU without having some kind of a chipset).

Anyways, the deed is already done. Intel has already contacted SIS and
UMC to produce more of their chipsets for Intel's processors. Intel is
likely going to be out of the low-end chipset business soon.

Yousuf Khan
 
Rob, I'm under the impression that chipsets are built on fabs that are
no longer usable for front-line CPU manufacture. Such as .13u or .18u
now. Intel doesn't build chipset fabs, they wait for existing CPU
fabs to become obsolescent. As you say, making cheap chipsets
apparently is not their best use for their existing .13 -.18u fabs.
No way is Intel (or anybody else) going to start construction on
additional .13u fabs at this time.

The problem is that chipsets are having higher demands put on them. A
1GHz FSB isn't an easy thing to do in .18u. Chipsets ain't what they used
to be.
 
Felger said:
Rob, I'm under the impression that chipsets are built on fabs that are
no longer usable for front-line CPU manufacture. Such as .13u or .18u
now. Intel doesn't build chipset fabs, they wait for existing CPU
fabs to become obsolescent. As you say, making cheap chipsets
apparently is not their best use for their existing .13 -.18u fabs.
No way is Intel (or anybody else) going to start construction on
additional .13u fabs at this time.

You are basically right. However, even stipulating that, it
still takes years for Intel or AMD or anyone else to build a
modern .09 (or .065 ?) CPU fab so that they have surplus .13 fabs
that they can retire from CPU production and recycle for chipset
production.

And, as Keith said (even if not in exactly these words) chipsets
these days are getting to the point where they need fabs that are
not that far behind the CPU fabs.
 
Yousuf Khan said:
These are mainly support chips for their CPUs. There is high demand for
their CPUs (obviously), and therefore there is high demand for their
chipsets. But chipsets are low profit. Also if they try to maintain their
stranglehold control over their chipset market, it's going to mean there's
less chipsets available and it will reduce the demand of their CPUs
(nobody buys a CPU without having some kind of a chipset).

Anyways, the deed is already done. Intel has already contacted SIS and UMC
to produce more of their chipsets for Intel's processors. Intel is likely
going to be out of the low-end chipset business soon.


Thus increasing their profits and my retirement income.
Good move!

--

... Hank

http://home.earthlink.net/~horedson
http://home.earthlink.net/~w0rli
 
Rob said:
And, as Keith said (even if not in exactly these words) chipsets these
days are getting to the point where they need fabs that are not that far
behind the CPU fabs.

Which sort of makes the decision by AMD to move the memory controller
into the CPU more prescient. Now the chipset can putter around
controlling hard disks, serving USB, passing PCI-E around, etc. without
having to worry about keeping up with the memory demands of the CPU.

Yousuf Khan
 
Which sort of makes the decision by AMD to move the memory controller
into the CPU more prescient.

Prescient? No, it is the obvious thing to do, for many reasons including
getting the traffic off the FSB. Why hasn't everyone gone there?
....search me. Memory interfaces on processors are a hassle, but what
isn't?
Now the chipset can putter around
controlling hard disks, serving USB, passing PCI-E around, etc. without
having to worry about keeping up with the memory demands of the CPU.

....which they cannot. It's like real estate; Latency, latency, latency.
Why piss away the gold on a trailer-park?
 
Which sort of makes the decision by AMD to move the memory controller
into the CPU more prescient. Now the chipset can putter around
controlling hard disks, serving USB, passing PCI-E around, etc. without
having to worry about keeping up with the memory demands of the CPU.

Err, chipsets for AMD processors do still have a 2.0GT/s
hypertransport connection or two on them, as well as a 2.5GT/s
PCI-Express connector. Ok, these tend to be somewhat easier to handle
than, for example, the P4's 800MT/s (or 1.06GT/s) bus, but they still
aren't trivial things.

I think one of the real tricks that has changed the chipset business
though is the widespread adoption of integrated graphics. At a guess
I would say that integrated graphics chipsets probably outsell
chipsets without built-in graphics by a relatively small margin (maybe
60% with graphics, 40% without?). This is quite a change from just a
few years ago where on-board graphics was limited to a fairly small
low-end niche. If you look at a modern chipset with a built-in
graphics controller you will probably find that more than half of the
die space is actually taken up by the graphics controller portion (in
the case of ATi's chipsets I hear it's up around 90% of the die).
This transition to integrated graphics as being fairly standard is,
IMO, a major factor in wanting to move chipsets to the latest and
greatest fab technology as quickly as possible.
 
Here's an update to this story. It looks like according this story,
Intel may be trying to prevent the chipset makers from becoming too
dependent on business from AMD.

By exiting the entry level segment Intel is reducing itself as a
threat to Taiwan chipset makers, and providing them more opportunities
to expand their market presence on the Intel platform. To prevent AMD
from gaining more share, Intel is opening the door for chipset vendors
to increase their chipset production, as well as their market focus, on
the Pentium 4 platform.

http://www.digitimes.com/news/a20050812A9052.html

Yousuf Khan
 
Yousuf Khan said:
Here's an update to this story. It looks like according this story,
Intel may be trying to prevent the chipset makers from becoming too
dependent on business from AMD.

By exiting the entry level segment Intel is reducing itself as a
threat to Taiwan chipset makers, and providing them more opportunities
to expand their market presence on the Intel platform. To prevent AMD
from gaining more share, Intel is opening the door for chipset vendors
to increase their chipset production, as well as their market focus, on
the Pentium 4 platform.

Oh my! Are these the same vendors who could not
possibly increase their production (grin).



--

... Hank

http://home.earthlink.net/~horedson
http://home.earthlink.net/~w0rli
 
Back
Top