Good article in EE Times on disk drive IC integration. Interestingly, the article contains an important factual error. To wit, Lucent is quoted as having the first integrated controller-channel chip in development. This is stupid because CRUS has had such a chip already available for a few months. Other than that key oversight, it is a worthwhile article. It is significant that WD's COO is cited as saying a single-chip drive isn't necessary for the desktop market, since WD is the most important initial target for this CRUS chip. It doesn't seem that he is that strong against it based on his other comments though and appears he could be convinced by the performance boost expected from the CRUS second-generation integrated controller-channel.
G.P.
eet.com
Diskcon abuzz over single-chip drives, home markets
By Craig Matsumoto
EE Times
(09/17/99, 12:11 p.m. EDT)
SAN JOSE, Calif. — At the Diskcon trade show next week, drive makers will
continue aiming their sights at the single-chip drive, with LSI Logic Corp.
(Milpitas, Calif.) becoming the latest entrant in that field. Meanwhile, disk-drive
companies continue to face pricing pressures but have high hopes of a
storage-heavy home market driven by demand for video and music.
Among semiconductor makers, the single-chip drive will continue to be a
buzzword at Diskcon. LSI Logic already has announced it will be unveiling its
single-chip drive at the show. The device is based on the previously announced
Merlin 2 Fibre Channel core. LSI officials were not available to comment by
press time.
Lucent Technologies Inc.'s storage division will be touting the prospects for
single-chip integration of drive controllers and read channels at Diskcon, said
product manager Joe O'Hare.
Already a leader in producing read-channel components, Lucent is hoping to
leverage its ASIC expertise for the storage market by becoming the first
company to integrate the read channel on to the same chip as the drive
controller. Such a part is still in development, O'Hare said, and is particularly
tricky because the read channel is a mixed-signal part that is being called upon
for increasingly high levels of precision.
But it's not certain that integration will take off in all sectors of storage.
Single-chip drives aren't necessary for traditional markets like desktop
computers, said Russell Stern, co-chief operating officer for Western Digital
Corp. (Irvine, Calif.).
"Silicon gets so cheap, the density of gates and Moore's Law has made it so now
the only penalty" is package cost, he said.
"You have to make careful trade-offs between channel architecture and
controller architecture," Stern said. Western Digital develops its own controller
architecture and uses an external channel architecture.
In the case of most disk drives, getting a low-noise read channel supercedes the
benefit of a single-chip drive. "A couple decibels in a channel is worth more than
a packaging cost," Stern said.
Meanwhile, the consumer market is particularly promising to disk-drive vendors
because the storage needs are greater than for office desktops.
"The great thing about the A/V [audio/video] market is that it needs storage,"
said Russell Krapf, general manager of A/V solutions for Western Digital. Krapf
cited the oft-quoted figure of 4.7 Gbytes of space needed to store a movie.
"There's a lot of talk about home servers," said Bill Moon, vice president of
advanced technology at Quantum Corp. (Milpitas, Calif.) and chair of a Diskcon
technical session about future disk-drive designs. Given enough storage
capability, the home server "virtually could be the entertainment center for the
home," storing movies, music and PC applications together, he said.
In addition to needing more space, multimedia drives have to eliminate latency,
often caused by the error-correction algorithms typically run during data
retrieval. "We have developed new commands to the disk drive to tell it when it's
receiving A/V data," Krapf said.
Home drives also need to be quiet, a feature Western Digital touts in its new
WD Performer line of drives for consumer devices. The advent of home offices
is making that requirement look appealing for some business desktops as well.
"We're learning things in the A/V space that we're finding applicable to the
desktop space," Stern said.
Capacity in desktops has gotten cheap, so that PC OEMs can avoid using
multiple heads or multiple platters and still get plenty of storage.
But while the drives themselves continue to carry more data, they may be
approaching a point where the extra storage isn't worth the cost.
"Disk drives will become a commodity, and they'll manage to have a fixed
commodity point, [a certain number of Gbytes] that's going to be the 'brick' that's
used in the future," Moon said. "The issue is, people just don't need the storage."
Moon noted that a similar application lies in Internet caching, wherein Web
pages are saved on a local disk drive, letting users view them without
reconnecting to the Web. Popular with corporate IT departments, the idea could
begin to catch on with home networks to boost performance, he said.
Similarly, Stern sees the creation of "very specific appliance devices that serve
up a very vertical function." In particular, these will be easy to install and
maintain, and they won't be controlled from a host CPU but rather will stand on
their own in order to deliver higher performance. "There's no reason for the
number-crunching processor to know or have to worry or care about the address
where data is stored," he said.
Also tied to the consumer market is the 1-inch drive, an area being tackled by
only IBM Corp. and startup Halo Data Devices. Most other vendors aren't
tackling 1-inch drives yet, as the technology looks promising for digital cameras
but doesn't have the huge-volume potential of standard drives.
"I'd much rather not have the moving parts, but it's just so darned inexpensive
compared with solid-state technology," Moon said. "They'll sell a lot of them."
But 1-inch drives — and other exotic technologies — often escape the radar
screen of companies like Quantum and Seagate Technology Inc., Moon said.
"All of those emerging market start with such small volumes that it's hard to get
the attention of a large company," he said.
Elsewhere, the transition to GMR heads is all but complete and expected to last
the industry for at least five years.
The shift to GMR from MR heads came as disk-drive densities rose. With more
data crammed onto the same size of platter, the signals that the head needs to
read get weaker, and GMR heads are better able to pick up a weaker signal. In
addition, the read channel — which passes data to the disk controller — has to
get better at filtering out noise, including residual data from neighboring locations
on the disk platter. Various "partial response" algorithms are available for the
read channel to interpret those signals.
A change "is a very long way off, because of the advancements the [disk-drive]
companies expect in the capabilities of spin valves, which is the basic technology
behind GMR," Stern said. So head transducers are going to be a stable
technology for some time.
However, other technologies will have to be tweaked if the industry is to
continue ramping areal densities — the number of bytes per square millimeter of
disk platter.
"The physical number of tracks that you're trying to put on a drive and the
electronics capable of delivering the data rates that we need will cause us to be
creative with other technologies. We're not going to get there with just head
transducer technology. We're going to hit a wall," Stern said.
The merging of NGIO and Future I/O could bring some promising results for
disk drives, Stern said.
"The creation of a high-speed switching fabric is critical in storage technology,"
he said. "As a result, we've seen Fibre Channel make an extremely strong play
when in fact it's probably not the right technology."
Fibre Channel isn't scalable enough to serve as a storage switching fabric, Stern
said, but IT groups have been adopting it because it's the only technology
available. Stern's hope is that the upcoming "system I/O" specs can deliver a
more suitable protocol. |
