Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Paul, everything IP is a part of the Intel licensing agreement.
Where you say:
because licensing for Redwood is not a part of the Intel-Rambus agreement.
This is incorrect.
Cheers
Cor
ON: You jumped the gun. I'm sure Cor realized he was wrong.
No way! As I was typing a response I saw Elixe was doing a great job of it. Who cares about the Mhz? Even Intel saw the light and has given that up:) (try to understand their model numbers)
And another thing, just because $ 53 or whatever more is charged does not mean it costs a penny more. It is called marketing. You get the $$ you can get for the better product.
So the extra price was not for the pins (which I said were cheap these days) but for the market opportunity and for the extra memory channel.
Cheers
Cor
Calbiker you should have been a politician, as in how to lie with statistics:
S754 A64 3000, 2 GHz, 512 MB L2: $141
S939 A64 3200, 2 GHz, 512 MB L2: $194
Comparing a 3000 with a 32000 tsk tsk....
In comparing equal speeds on pricewatch 3000 or 3200, I see $ 10 to $ 14 difference. That's only 4 cents per extra pin and we throw in the second memory controller for nothing.
(reference: a northbirdge costs about $ 25)
read my lips: pins are cheap!
Cheers
Cor
On "Turbo PCIe"
(posted similar message on the Fool, where the only interest is now stock price and not the technology this company stands for.
There was an announcement yesterday of double speed PCIe coming up. A we know Rambus already announced this as "turbo PCI Express" some time ago, which led me to this:
I had been wondering when it would happen (given that the hypertransport.org already upped their specs twice). We knew of course that Rambus had defined a "turbo PCIe" physical specification at double the normal speed of 2.5 Gb/s, but I was wondering when will there be a use for it.
It seems to be a future spec, sort of, at the pci special interest group (pcisig.org), see for the announcement:
http://www.pcisig.com/news_room/news/press_releases/2004_12_15/
or
http://www.pcisig.com/news_room/news/press_releases/2004_12_15/2004_12_15.pdf
PCI-SIG Announces PCI Express Architecture Performance Extension
Planned Specification Doubles Current Data Rate to 5GT/s
[emphasis mine]
PORTLAND, Ore. – December 15, 2004 — The PCI-SIG®, the Special Interest Group responsible for PCI Express™ architecture, announced today that the data rate for the next planned revision of the PCI Express specification will be 5 Giga transfers per second. This new specification extends the performance capability of its flagship I/O interconnect architecture to meet the anticipated system requirements across computer and communication industry applications. The doubling of the data rate follows historical performance increases for I/O specifications in the industry.
The PCI-SIG board of directors considered market requirements and technical analysis of a range of data rates to obtain the most feasible, highest performance, backward compatible solution within the current PCI Express ecosystem. Platform implementation cost, high-volume manufacturability, system topologies, validation and interoperability were the key considerations studied by the PCI-SIG board members in the process to extend the PCI Express data rate to 5GT/s.
Many of the leading electrical experts who developed the PCI Express 1.0a specifications have been chartered to draft the new specification. The PCI-SIG expects to deliver the new specification in the second half of 2005 in time for product introductions beginning in 2007.
"The PCI-SIG continues to deliver on the promise of timely PCI Express enhancements to meet the ever-growing bandwidth requirements of the I/O industry," said Tony Pierce, PCI-SIG chairman. "The board of directors thoughtfully considered an array of technical and market data and arrived at the best decision for the next revision of the PCI Express architecture. We plan to deliver the new 5GT/s PCI Express specification consistent with our members’ expectations for 100% backward compatibility, low-cost manufacturability and world-class compliance and interoperability."
"The new 5GT/s PCI Express specification will enable the required performance boost for bandwidth-hungry applications such as cinema-quality graphics and multimedia, enterprise servers and storage, and multi-gigabit networking," said Ajay Bhatt, chairman of the PCI Express steering committee, responsible for managing the technical development and coordination of PCI Express specifications. "The electrical specification will introduce evolutionary methodologies to meet the technical challenges of running at 5 Giga transfers per second while maintaining compatibility with existing system topologies and silicon processes."
One interesting point in the "benefits" of being a PCI SIG member is the following:
Benefits of PCI-SIG Membership
PCI-SIG members can participate in the review of all PCI specifications before they are released to the industry. If you are interested in becoming a member, please visit the PCI-SIG Web site at www.pcisig.com/membership. PCI-SIG members develop and maintain PCI Express specifications and are actively involved in defining compliance criteria and checklists, as well as other technical enabling collateral.
As an additional and extremely valuable benefit of PCI-SIG membership, members are given the right to receive patent licenses from any other member of our organization with necessary claims of patent embodied within our specifications. These licenses may be limited in scope to an implementation of a particular specification, but must be granted to all members on reasonable and non-discriminatory terms. ...
This sounds similar but a little different from the JEDEC patent licensing stuff. (imo)
Afaik Rambus is not a member of the pci-sig and one wonders what barriers to selling their solution for the PHY that gives.
Cheers
Cor
Ha Calbiker, I knew you were an oldie:)
Ha; calbiker used it in the seventies.
The (dma) memory transactions don't go through the cpu in the AMD64, the corssbar switch ahs a similar function to the northbridge in this case (with the advantage that caching can take place at the same time).
Yes I agree it makes the die bigger, but then it has worked apparently, Intel's die are getting bigger because they need all the chache to keep up, LOL:)
Why should video streaming data have to travel through the processor? The added complexity also presents pin limitation problems. Socket 754 processors have only one memory channel. Why?
Pins are cheap apparently, 1000 pins no problem these days. Anyway most applications are perfectly happy with one memory channel.
On the rigidity, yes , to some extent, but the advantage is simpler chipsets (and the IO chipsets go over from one cpu to the other, because it is all HT). I believe a next version will have a FB-DIMM type IO for the memory so that solves that. (multiple ones)
Cheers
Cor
OK enough AMD stuff, my next post is about rambus/PCIe
Guess you haven't heard of DMA?
LOL, I worked with dma back in the early eighties.
I agree that much of the disk IO is through the dma channel so in the Intel case through the northbridge and in the amd case through the crossbar; let's say that cancels out. All the other IO makes the intel fsb choke (including the IO commands to the hd) and not reach full bandwidth for memory. Which is not the case with the amd hypertransport io system.
Cheers
Cor
That's clearly wrong. Dual cores are shipping next year.
Last I saw they were saying 2005H2, but if Intel makes it, that's another matter. They were not all that good in coming up to their promises recently
Cheers
Cor
All I/O traffic and graphics must first go to the processor (an unnecessary path).
Not really, there is a crossbar switch inside the processor die to handle that much as that is handled in equivalent cases by a crossbar inside the Northbridge in the Intel case.
And you are not suggesting that most IO is directly from/to memory to/from IO are you? A lot of IO does involve the processor and in the Intel case has to cross that (as you pointed out yourself) narrow bandwidth pentium fsb.
Cheers
Cor
Speaking of lousy, what about the FSB? It's a bottleneck. A 800 MHz FSB has a bandwidth of 6.4 GB/s while 1066 FSB has 8.5 GB/s. Dual channel DDR3-800 has a peak bandwidth of 12.8 GB/s. The memory has a bandwidth that's 50% larger than the FSB. That's lousy!
Plus that all the IO traffic has to go through the same channel... (well I will not bandy about names but you know where that is not necessary, LOL :)
Cheers
Corroller
(bought in-line rollerskates, man o man that was 35 years ago me on the ice...., don't laugh, practice a little daily will get there before I am 65, 2.5 years to go)
on:Comm. Leibowitz: talks about the 3rd cause? of the case???? “Unfair Method of Competition” Assuming you haven't proven your case; can we still find a SECTION 5 Violation?
?????????????????
Great work from Pat (Scruffy).
Here is what that means:
An Antitrust Primer
The antitrust laws describe unlawful practices in general terms, leaving it to the courts to decide what specific practices are illegal based on the facts and circumstances of each case.
* Section 1 of the Sherman Act outlaws "every contract, combination . . . , or conspiracy, in restraint of trade," but long ago, the Supreme Court decided that the Sherman Act prohibits only those contracts or agreements that restrain trade unreasonably. What kinds of agreements are unreasonable is up to the courts.
* Section 2 of the Sherman Act makes it unlawful for a company to "monopolize, or attempt to monopolize," trade or commerce. As that law has been interpreted, it is not necessarily illegal for a company to have a monopoly or to try to achieve a monopoly position. The law is violated only if the company tries to maintain or acquire a monopoly position through unreasonable methods. For the courts, a key factor in determining what is unreasonable is whether the practice has a legitimate business justification.
* Section 5 of the Federal Trade Commission Act outlaws "unfair methods of competition" but does not define unfair. The Supreme Court has ruled that violations of the Sherman Act also are violations of Section 5, but Section 5 covers some practices that are beyond the scope of the Sherman Act. It is the FTC’s job to enforce Section 5.
* Section 7 of the Clayton Act prohibits mergers and acquisitions where the effect "may be substantially to lessen competition, or to tend to create a monopoly." Determining whether a merger will have that effect requires a thorough economic evaluation or market study.
* Section 7A of the Clayton Act, called the Hart-Scott-Rodino Act, requires the prior notification of large mergers to both the FTC and the Justice Department.
Some cases are easier than others. The courts decided many years ago that certain practices, such as price fixing, are so inherently harmful to consumers that a detailed examination isn’t necessary to determine whether they are reasonable. The law presumes that they are violations (antitrust lawyers call these per se violations) and condemns them almost automatically.
Other practices demand closer scrutiny based on principles that the courts and antitrust agencies have developed. These cases are examined under a "rule of reason" analysis. A practice is illegal if it restricts competition in some significant way and has no overriding business justification. Practices that meet both characteristics are likely to harm consumers -- by increasing prices, reducing availability of goods or services, lowering quality or service, or significantly stifling innovation.
The antitrust laws are further complicated by the fact that many business practices can have a reasonable business justification even if they limit competition in some way. Consider an agreement among manufacturers to adopt specifications that require fire-resistant materials for certain products. The set of specifications may be called a standard. The agreement to adopt the standard is restrictive: the manufacturers have limited their own ability to use other materials, and they have limited consumer choice. But the agreement to adopt the standard may benefit consumers in that it provides assurances of safety.
What if manufacturers did not use a uniform standard for electrical outlets and plugs? The likely result would be incompatibilities between parts produced by different manufacturers. But because of the standard, parts manufactured by different companies become interchangeable; competition for the parts increases, and prices go down.
http://www.ftc.gov/bc/compguide/antitrst.htm
Cheers
Cor
Any idea when these commissioners who are in a huddle now will lay an egg?
re:
The five-member commission made no decision Thursday.
Cheers
Cor
Elixe I agree the numbers don't gell, they admitted in the sound bite that they were a bit lost here and were looking for feedback from the analysts (sic!).
So I am only the messenger here with no opinion on it.
Cheers
Cor
Threejack, you listened to the call yesterday. Apart from mentioning the inside cell stuff, did they say anything about using/or not using XDR with Cell?
Cheers
Cor
LOL, true :))
Cheers
Cor
The person who goes farthest is generally the one who is willing to do and dare. The sure-thing boat never gets far from shore.
Hey Stowboat, ever ran out of fuel? I did, but then mine is a sailboat, just matter of waiting for some wind.
Cheers
Cor
Hi Elixe,
I understand you don't want to listen to all 25 minutes. But that's the ease of my MP3, you can download it (it is 22.75 minutes) and just listen to relevant pieces. The HDTV question starts at minute 21.
Btw I use the total recorder program, which allows me to cut and paste in the audio file.
The cellphone piece is minute 13 and foll. They are talking about volatile memory (excl. flash): pseudo sram, cellular ram, mobile ram, low power dram, ....
For these categories together 1.5 MB per phone produced in 2004 is expected to go to 15 MB per phone 05H2. (then representing 15% of capacity rather than the 2% it takes now)
To show off, I cut the piece and posted it at
http://chipstocks.net/mu/csfb-cellular.mp3
(283 Kilobytes, 2:21 minutes)
Cheers
Cor
So putting Elixe's news and Paul's and Threejack's comments into the great mixer we get, that Rambus admits GDDR3 is the fastest memory on the market (so XDR is not officially on the market yet or the speed grade not available yet?).
Furthermore Tate expects that to change soon, so there must be a fster XDR coming out. We will have to be patient, but I am seeing some shimmers of the old DDR/RDRAM stuff coming again...
Its here ... its not here ... is works ... it's faster .. it does not work ... etc etc.
Cheers
Cor
Calbiker, those garden houses look terrible, good thing you are doing your own design. In those things you show from the catalogue the first squatter round the corner will settle down in it.
Maybe you can make something sinewave shaped, nice Fourier transform stuff
Cheers
Cor
MU conference presentation:
posted this at TMF, also link to audio on my site.
http://boards.fool.com/Message.asp?mid=21711867&sort=whole
the audio link is:
http://www.chipstocks.net/mu/csfb041202.mp3
(2.7 MB, about 25 minutes)
Elixe, you may be interested in some mkt numbers in that, in my transcript, snip:
The average memory per cellphone is about 1.5 Mbyte this year.
Is supposed [according to industry analysts] to go up to 15 Mbytes per phone in the 2nd half of 2005; this amounts to about 15% of total dram consumed in 05H2, up from about 2% today.
(the 15% corresponds to about 30-35% penetration of phones which have some kind of memory in them; memory per phone ranges from 32 MB to 92 MB)
Most aggressve estimates say 25% of dram will be consumed in cellphones.
...
Q. about penetration of memory in HDTV. Huge range of estimates. They see a memory capacity NOW of about 58 MB per HDTV
External studies show 4M to 15M sets sold. [this is not a lot of memory, as yet]
the [ ] brackets are my comments; this is a rough transcript for my own use so I can find audio portions quicker and typed as I listen, so bear with me.
Cheers
Cor
Hi calbiker, you can rest assured there are more fanatic AMDers than I
I have not studied either the patent or eachus' post sufficiently to answer in detail. So I won't.
It interests me that Paul thinks this infringes the Rambus patent and yet this is before the uspto. Do you think it is the same stuff?
On your point 8:
It might be possible, but why slow down the system with slower memory?
There could be situations where part of the memory is less important. There is a hierarchy already of level1 and level2 cache and then system memory, why not have different levels of importance (urgency) of system memory?
There is some memory coming up which is a lot cheaper than dram (polymer types), but may have some access disadvantages. It would still be a lot faster than a disk drive, but slower than dram. It could hang at the end of the chain and give you a faster total system. Heck you could even leave out the hard disk altogether...
Cheers
Cor
Paul, I put the pictures on my site for easier access, they are tifs, can be read directly in your browser with alternatiff.
see:
http://chipstocks.net/amd/patentapp0230718/img1.tif .... img18.tif
Cheers
Cor
Paul, about the AMD patent
Yesterday I have put the question on the AMD board at about the same time as putting it here.
There is a long answer from eachus there:
http://boards.fool.com/Message.asp?mid=21706035
We are working on getting the images more accessible.(probably on my site, I've got more bandwidth than I need anyway:)
Here is a piece of the post for those who do not have TMF access:
As near as I can tell the idea here is to replace DIMMs with something similar to FB-DIMMs, but with two significant differences. First, the idea uses the same sort of protocol, and perhaps the same protocol as Hypertransport, with a 16-bit uplink and 16-bit downlink. We could guess at speeds, but it would be just that, a guess. (Like 6.4 to 25.2 GBytes/sec.
The idea would be that AMD manufactures CPUs that interface to this new (and probably slightly more expensive) memory module, and motherboard manufacturers put the appropriate wiring and sockets on a motherboard. I can imagine it having a slow start in the desktop market, but for servers? It might knock FB-DIMMs right out of the park.
Why is this so much better than FB-DIMMs? No guarantee that it is, but if I was designing a system, I'd much prefer to live with this design. It should have about as many pins per memory module as current DDR, but:
1) All high-speed signals are unidirectional, like HT.
2) All high-speed signals are point to point, like HT.
3) No active or passive termination required, like HT.
4)The bandwidth per channel should be about three or four times that of FB-DIMMs,
5) The latency for the first memory-module should be significantly less than for FB-DIMMs.
6) The latency for subsequent modules should be less than with FB-DIMMs.
7) Mix DDR, DDR2, and DDR3 modules in one system. (In practice, I suspect that we are talking all DDR3, but why not? This would be a new memory layout, and there is no need for a non-AMD standard off the memory module.)
8) Mix different speed and latency memory with no problem. My read is that, requests go out, tagged data comes back in a burst.
9) Collisions between memory module replies are handled by buffering up to one line of data. (No sending the next burst without an ACK. Technically, no bus, all paths one way, there are no collisions. All that happens is that a memory module has incoming data from another module to pass through while it is already transmitting to the CPU.)
10) Potentially, NUMA aware OSes can use the first memory module for time-critical data, and put other data further out.
Seem like a lot of wins and no losses? Well to win, there will need to be a food chain. If this goes into next generation Opterons, I assume that Sun is on board in a big way. Add a memory manufacturer (Infineon? Micron?) to build the memory modules, and probably Newisys to build the motherboards, and you have a minimal food-chain. Of course, HP, IBM, Tyan, Samsung, etc., will all want to participate if there is a critical mass to begin with, and Sun can provide that.
Oh, and once the infrastructure is in place, it should migrate to the desktop as soon as AMD wants it to. Um, let me speculate a bit in that area. If you could have memory equivalent to dual (DDR2) CL3 PC2-6400, would you be willing to throw away your existing memory? As I see it, that is about what it will take to make this the next desktop standard. (That's where the range comes from above. This is interesting at 6.4 GBytes/sec, compelling for servers at 12.8 GB/sec, and compelling for everybody if the protocol can handle 25.2 GB/sec.)
What does AMD get out of this? A simpler on-CPU memory controller, with potentially lower latency than DDR2 or FB-DIMM, and a protocol that should make 12.8 GB/sec/channel practical. (Oh, and probably lower on CPU latency, as the data is in the right form for the crossbar.
Cheers
Cor
It really is a pity the Ihub does not have a rec system (with all it's shortcomings, I now have to write a whole post to recommend this.
And especially to exhort the lurkers here to speak up.
One great feature here is that if you make a mistake, you have 20 minutes to correct it:)
Even then, if you see a horrendous mistake later, you can ask Threejack by private mail to axe the thing, as he more or less has the power of the supreme being on this board (and in good hands, may I add)
Cheers
Cor
Hi Paul, if that is so (Rambus' patent prior art to this one), it will not be granted, simple as that. We will see from the ensueing procedure what happens.
The interesting point is that you seem to think it is a similar memory patent. Did you see differences? for example upstream and downstream traffic organized differently?
I am thinking of the history where in 2000 everybody was caught off-guard (and I know from inside Intek they were blindsided) with the slew of Hypertransport (then called Lightning Data Transport) patents, which have been applied for in 1996. Shortly after there was suddenly a Hypertransport consortium with many members (Cisco, Apple, Xilinx, .... ) who are still now applying this bus/method.
Cheers
Cor
Disclosure: I have no shares or options AMD, long Rambus and calls, no puts. (for the time being
AMD cooking up a new memory system?
(especially for Paul, the resident FB-DIMM sleuth)
http://tinyurl.com/3szch
This is a patent application, published in november this year, so a few weeks ago. i have not yet ascertained the full pedigree, filing date etc.
System including a host connected to a plurality of memory modules via a serial memory interconnet
Abstract
A system including a host coupled to a serially connected chain of memory modules. In one embodiment, each of the memory modules includes a memory control hub for controlling access to a plurality of memory chips on the memory module. The memory modules are coupled serially in a chain to the host via a plurality of memory links. Each memory link may include an uplink for conveying transactions toward the host and a downlink for conveying transactions originating at the host to a next memory module in the chain. The uplink and the downlink may convey transactions using packets that include control and configuration packets and memory access packets. The memory control hub may convey a transaction received on a first downlink of a first memory link on a second downlink of a second memory link independent of decoding the transaction.
I wonder how this compares with FB-DIMM stuff?
There is also something cooking at AMD, but could be just the usual rumor mill of the inquirer (as if we don't listen to rumor mills here
http://www.theinquirer.net/?article=20049
Cheers
Cor
Hi Paul, thks for that link.
I find the description by nAo (nov 30) very interesting. Almost as if Elixe is speaking
Cheers
Cor
Hey Elixe, did you see the latest from CRAY?
Reminded of that as you mention the BlueGene again
http://www.cray.com/downloads/sc2004/SC2004-CAE_nov4_pp.pdf
not into the pretaflops, but 50 teraflops.
The interesting one is where they explain how they made a co-processor from a Xlilinx virtex FPGA, let's see if I can find it; (application acceleration system)
http://www.cray.com/downloads/Cray_XD1_Datasheet.pdf
There is a more detailed pdf about that fpga somewhere on the site, if I find it I will post it. Got it on my hd, I wish they would automatically save the original url as well:(
Cheers
Cor
On GORX also check out that the two top guys take out $ 1 M together a years whilst the comapny has minimal sales and negative profits. Stinks imo.
Cheers
Cor
Thks Elixe,
used to be in Holland that there was a detectable signal at the beginning and end of commercials and somebody made a box to switch off the sound. Don't know if that still holds. (the advertisers obviously didn't like it)
On car raios we have such a signal for traffic news comign through which will switch the volume up; do you have such a system? (I believe Philips initiated it)
Where do you believe "cell" wil have it's earliest penetration? I would expect in hdtv type devices.
Cheers
Cor
Hey Elixe, quite geeky that setup:)
Here is an aricle from today about the media stuff:
http://www.reed-electronics.com/electronicnews/article/CA484385
Home Media Servers Not Ready for Mass Market Until 2006
By Jessica Davis -- Electronic News, 12/2/2004
For a market that hasn't seen a lot of sales as of yet, the home media server/hub space is bustling with activity.
These next generation devices are designed to be the center of the digital home, organizing and networking audio and video and providing intelligent functions, such as the time shifting and automatic recording available with a personal video recorder.
And companies are coming at the market from many different angles – from next generation set top boxes to "entertainment" PCs to enhanced personal video recorders to entirely new devices– all looking to get a piece of the action, even though most observers agree that the action won't really arrive in earnest until 2006.
It's a market at the Wild West stage of development with lots of players and no rules.
"The end result is consumers are very confused," said Connie Wong, a senior analyst at Semico. "There isn't a standard name for this. There are all sorts of different names -- media hub, media server…"
"And the price points are significantly different because all have different capabilities," she said. "Some handle some media formats, some handle a lot of media formats. They have different sized hard disks. And the price points are all over the place from low end to high end depending on capability."
Just what are these devices? Analyst firm InStat defines a media server as a device with a hard drive, a broadband connection and in-home networking capability. Some analysts choose a more liberal definition that does not include a broadband connection.
Still, people have started to buy the devices, in spite of the confusion. By the end of this holiday season, just over one million entertainment PCs will have shipped, a drop in the bucket compared to overall PC sales, said Mike Paxton, senior analyst for converging markets and technologies at InStat. And that million shipped as of this holiday season is nothing considering the devices have been available since 2002.
"The price difference between media server PCs and a more basic PC is an extra $300 to $400 for those additional capabilities," he said. The PCs are among some of the most expensive devices that fit the category.
Denon's media server won best of show at the Consumer Electronics Show in January 2004. The device did not function as either a cable or satellite box but rather simply was designed to manage video and audio throughout the home. The device was due for release in the September/October time frame but hit a delay.
"They are still trying to figure out the market for it," said Paxton.
Other media servers combine cable/satellite set top box functions with personal video recorder features such as time shifting. And some devices allow users to access IP-based video, adding all the content on the Internet to the regular channels that come into the home from satellite and cable carriers.
Texas Instruments is one of the companies that is looking to cash in on this emerging market by offering devices designed for streaming media. Such product would allow cable companies to offer an unlimited number of movies on demand without any constraints on start times. Or users could access "channels" on the Internet in addition to the traditional channels that are available today.
TI and others are also looking to tap into the vast emerging market for chips that will network the home media server/hub to all the other entertainment devices in the home, through both wireless technologies, existing coaxial cable already in the home, or enterprise networking technology such as Ethernet.
Ultra wideband (UWB) and a new Wi-Fi spec, 802.11n are two top contenders for the wireless digital home networking market, both offering enough bandwidth to send quality video signals from device to device throughout the home.
Cable operators are also catching on. Echostar offers consumers the option of buying a PVR-type box. Others are charging customers a fee of $5 to $10 a month for the time shifting capability available with PVRs.
And in an end run around all these CE players and other device makers, content company Disney contracted with Samsung to create its own set top box to offer movies on demand to consumers. Called MoonBeam, the service launched at the end of September in Jacksonville, Fla; Salt Lake City, Utah; and Spokane, Wash. The service charges consumers a per month equipment rental fee, plus a fee for each movie ordered.
"Disney is leaning the farthest on this experimentation of having a media product in the home," said Paxton.
"We are just at the leading edge of media server products getting out there," he said. "Operators are just dipping a toe in the pool, many consumers aren't even aware that these products are out there. And if they are aware they are asking, 'Is this something that is going to add value to our home experience?'"
Device makers in this market face a number of challenges before this market can take off, said Wong. Top among those are pricing, consumer education and an acceptance of the concept of this new device in the home.
Paxton believes prices need to go below $500, and preferably below $300 before the market is ready for the masses. And the PC devices will need to get down to the $800 price point at least, he said.
"Prices are huge," he said. "A lot of that price reduction will have to come with volume. But before we get to that point a lot of questions have to be answered in terms of features, networking technologies and partnerships."
Consumer education will also play a key role in the success of this market, said Wong.
"You have to make the consumer need and want this capability," she said. "On top of that, you have to make networking these products invisible to the consumer."
Consumers may also be confused by all the different media format technologies in the market today.
"We are still in the emerging stages of this market definitely," she said.
Paxton believes we are in the first year of a two-year long experimentation cycle. But, he says, the opportunity for semiconductor companies is very attractive.
"You are looking at a lot of processing power and a lot of memory," he said. "Devices will need processor modules, tuner modules, MPEG chip sets, wireless chipsets, networking chipsets…"
For me the atraction would be to record w/o the commercials. Does TiVo do that for you? I don't know how TiVo works, we don't have it here.
Cheers
Cor
Hi Elixe, I am not clear on that redwood stuff in ps3.
I see it primarily as an external -chip to chip- conenction. The ps3 will have one cell chip (which internally may consist of multiple cores), so does not need to use redwood per se.
Or it may use fredwood internally for the closely coupled cpu/coprocessor parts.
How do you see this?
Cheers
Cor
Hey threejack, did you notice the inconsistency to this "news" story?
Exactly how many CPU cores and how many synergistic cores will be integrated into a single Cell chip is unclear, but the rumor is that each Cell CPU will have eight synergistic cores plus a direct memory access (DMA) controller all residing on a set of Rambus serial links between the elements
Huh, serial?
Cheers
Cor
Here are the official Cell announcements I found.
I did not yet find the one from SCEI, difficult to navigate there.
IBM:
http://www-03.ibm.com/chips/news/2004/1129_cell1.html
Toshiba:
http://www.toshiba.co.jp/about/press/2004_11/pr2902.htm
Unofficial with some speculation:
EEtimes article:
http://www.eetimes.com/showArticle.jhtml?articleID=54200580
Note the IBM and Toshiba PRs don't explicitly mention redwood or rambus. The article does on page two.
Some more mentions, but remember they all ape each other
http://www.theregister.co.uk/2004/11/29/ibm_sony_cell_debut/
http://www.theregister.co.uk/2004/11/29/ps3_to_ship_2006/
http://www.technewsworld.com/story/news/38538.html
http://www.eweek.com/article2/0,1759,1733264,00.asp
In edit (I still looove that feature:), add the link to ISSCC:
http://www.isscc.org/isscc/2005/ap/ISSCC2005AdvanceProgramPreview.pdf
Cheers
Cor
Didn't Intel pay Micron several hundred million dollars to produce DDR2?
Hey Threejack,
They did (in the form of an investment as usual). A difference with the earlier deal is that MU actually has to pay Intel a large amount of money of they cannot produce a certain percentage of DDR2 by a certain date. So I think they will conform.
And, as elixe said: it's evolutionary
Cheers
Cor
Kate and others,
If you know any more sites (enthousiasts or whatever) to add there, I would be thankful. I have not found enough for my taste.
Cheers
Cor
Hey Kate, just to say hello.
I was visiting here to add this site to my "Intel Friendly Sites" on my website, Intel part:
http://chipstocks.net/intc/intc-sites.htm
Cheers
Cor
Hynix, I guess after this the EU duty is next, although I cannot remember now if it was for exactly the same reason, which might be important.
Cheers
Cor
PS have a look at the new look of my site http://chipstocks.net
I have worked on it for two weeks, for those in the know css,php and mysql.
It's friday if you click on the google ads you may buy me a beer:)
Nvidea stuff, Threejack, here is a knowledgeable thread on this on TMF:
http://boards.fool.com/Message.asp?mid=21630785&sort=whole
Cheers
Cor
Hey Will,
I am very aware of what those charts look like, also look at:
2 years,
http://finance.yahoo.com/q/bc?s=RMBS&t=2y&l=on&z=m&q=l&c=amd
1 year,
http://finance.yahoo.com/q/bc?s=RMBS&t=1y&l=on&z=m&q=l&c=amd
and including the recent runup, 3 months:
http://finance.yahoo.com/q/bc?s=RMBS&t=3m&l=on&z=m&q=l&c=amd
Looking at the chart since 1998 doesn't really do it for me
Cheers
Cor
OK Threejack, nothing "black" there then:)
It's where Rambus played the video clip of customer comments; obviously doesn't work after the fact with no feed, but left it there as a reminder of their intent.
Aren't you getting tired and impatient of it all sometimes?
On the one hand you kow Rambus had something there (rdram), which failed in the market for reasons which may or may not be completely uncovered. On the other hand you know the compay execs are spinning for their own bank accounts. It is not a good example of capitalism at work imo.
(maybe it's the season, I don't usually get this maudlin
Cheers
Cor