SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Rambus (RMBS) - Eagle or Penguin -- Ignore unavailable to you. Want to Upgrade?


To: Jon Khymn who wrote (76377)7/30/2001 8:49:58 PM
From: Don Green  Respond to of 93625
 
Intel Unleashes Speedy Notebook CPU

Exclusive PC World tests show impressive performance boost in first Tualatin-based notebook.

Tom Mainelli, PCWorld.com
Monday, July 30, 2001

SAN FRANCISCO -- Intel went after speed-hungry notebook users Monday, with the launch of its 1.13-GHz Mobile Pentium III Processor "M." Early PC World tests indicate the chip could lead to noticeable notebook performance gains, and scores of vendors are lining up to offer products based on the new chip.

Calling the awkwardly titled chip "a generational leap forward," Frank Spindler, vice president and general manager of Intel's mobile platform group, told analysts here today the new processor is the company's first produced using its new 0.13-micron process technology.

The new process lets the company shrink the size of the chip while speeding up performance and adding extra features.

Intel is offering the CPU, formerly codenamed Tualatin, at two new top speeds: 1.13GHz and 1.06GHz. It will also backfill its lineup with chips at 1 GHz, 933 MHz, and 866 MHz.

PCWorld tests of the 1.13-GHz chip in a pre-production Dell Inspiron 8100 running Windows 2000 racked up the highest PC WorldBench 2000 score ever for a notebook, earning a 210 on our tests of office, communications, and graphics applications and multitasking. Its 15-percent advantage over the average for five previously tested non-Tualatin 1-GHz PIII notebooks (also running Windows 2000) is impressive--a bit higher even than the chip's 13-percent clock-speed increase.

Ready for Purchase
Numerous vendors will support the new chip, but only a few are selling products today. Among them is Compaq, which is offering its Presario 1700 through its Web site, says Lorena Kubera, director of portable product marketing.

The site offers various numerous configuration possibilities for sale now and shipping within the week. A sample configuration with the 1.13-GHz PIII-M chip, 128MB of PC-133 memory, a 10GB hard drive, 24X CD-ROM drive, 8MB of video memory, a 14.1-inch active-matrix display, Microsoft Windows 2000, and Microsoft Office 2000 sells for $1997.

Dell is now selling both Latitude and Inspiron notebooks using the new processor, although it could take up to 25 days for products purchased today to ship, says Steve Hollington, product marketing manager.

A sample configuration of an Inspiron 8000 priced at $2607 nets you the 1.13-GHz chip, 128MB of PC-133 memory, a 30GB hard drive, 16MB of DDR video memory, an 8X DVD drive, a 15-inch active-matrix display, Microsoft Windows ME, and Microsoft Office XP.

Better Process, Better Performance
Intel's switch from its 0.18-micron process to the new 0.13-micron process also lets the company squeeze 512MB of performance-improving secondary cache on the new chip. That's twice the L2 cache of existing mobile PIII chips. The larger, better L2 cache should mean better performance across the board, meaning a new 866-MHz PIII-M should outperform an existing 866-PIII, Spindler says.

Another side benefit of the new process is that the chip uses less power (and produces less heat) to do the same job. Intel claims the new chips run 20 percent faster than existing chips based on the older 0.18-micron process, while using 40 percent less power.

To further enhance the chip's power-saving capabilities Intel improved its SpeedStep technology for the PIII-M, Spindler says. The enhanced technology lets users do more than choose between maximum-performance or battery-optimized modes by choosing an automatic setting. The new setting revs the chip between low and high frequency speeds based upon the needs of the application running at the time.

Slow to Ship
While Dell and Compaq are selling products now, other vendors will take longer to ship their first products. Other vendors on hand to show support at the chip launch included Acer, Asus, Clevo, Fujitsu, Gateway, Hewlett-Packard, IBM, Sony, and Toshiba.

Representatives from HP, Gateway, Acer, and Toshiba all discussed plans for future PIII-M products, but say the first products won't ship until next month or later. HP's first PIII-M notebook will be the Omnibook 6100, says Lara Kahler, worldwide product marketing manager for HP's mobile computing division. Due to ship in the coming weeks, the new notebook also will soon include integrated wireless networking.

A slight lag between Intel's launch and the introduction of new products isn't unusual in the mobile market, says Kevin Krewell, senior analyst with MicroDesign Resources. While most desktop vendors can launch a new product at the same time as Intel announces a new desktop processor, notebooks can be a bit trickier.

"Notebooks take a lot longer to design," he says. That's especially true when you're talking about a new type of processor. And in addition to the new chip, Intel is launching a new chip set to match, so that takes time to integrate, he says.

A Chip Set to Match
The Intel 830 MP chip set incorporates a handful of improvements to spur better performance, says Intel's Spindler. Among those improvements is a 133-MHz processor system bus (up from 100MHz); support for the faster PC-133 RAM (up from PC-100); and the ability to support up to 1GB of memory.

Today the company is offering the 830-M with support for external graphics; in the near future it will offer two separate products with different levels of integrated graphics, he says.

In the meantime, MicroDesign Resources' Krewell says it it's very important that Intel's introduction of the PIII-M and 830-MP chip set go smoothly. The chip in particular represents the company's future, since all of its processors will one day move to the 0.13-micron process, he says.

"This is the first one on this process," he says. "Launching this successfully is key."



To: Jon Khymn who wrote (76377)7/30/2001 10:22:51 PM
From: dumbmoney  Respond to of 93625
 
What do you think the chance of RMBS go bankrupt in next 3 years? (10%? 50%?,,,,)
As I read more of your posts, I am getting gloomier and feel this might be another GSTRF: great idea, great technology, but not able to execute economically to attract the customers. (If you keep at it, you'll might be able to convert me from long to shortie ;o)


Rambus isn't that sort of company, financially speaking. They won't go bankrupt.



To: Jon Khymn who wrote (76377)7/30/2001 11:02:24 PM
From: Bilow  Read Replies (1) | Respond to of 93625
 
Hi Web Myst; Ah, so you're just looking for the DCB. That's what Zeev is doing, he knows nothing about memory, and he has a pretty good record of making money on RMBS to the long side so I wouldn't worry too much. But Zeev does keep very tight stops on things. I'm guessing that if RMBS gets to the point where you can see the whites of the eyes of the $6 trades, Zeev will take his (small) loss.

As far as the Rambus story goes for long term investors, there are two different stories, which is what I was asking you about.

There's the growth of RDRAM story, and that's my specialty. Until early 2000, Wall Street's story for Rambus was the support of Intel forcing RDRAM onto the market. Those of us who are more familiar with the memory industry could smell blood in mid 1999, and by late 1999 it was obvious that RDRAM was going to be stuck to niche status. That wasn't obvious to Wall Street until mid 2000, and by that time the other story had taken over, that Rambus would collect high royalties on all memory types.

To the memory designers, it was obvious that RDRAM wasn't going to make it as soon as the Camino launch started being delayed. But for Wall Street, which doesn't have the intuitive connection to memory, the clue was the total absence of any other design wins other than Intel (and Sony). In late 1999, Nvidia started ripping market share out of the rest of the graphics market with their DDR solutions. Up till that point, RDRAM had been commonly used in graphics cards. But when every graphics house began announcing DDR boards and dumped their RDRAM based boards, it started becoming clearer to Wall Street. The other clue was the absence of chipset support for RDRAM from AMD and the many chipset manufacturers like VIA, ALi, ServerWorks, SiS, etc. And the last nail in the coffin for RDRAM as a mainstream memory was Intel's announcement of eventual DDR support for the P4, and also Intel's exec saying that Rambus was a "mistake". By that time, the hope that RDRAM would become pervasive was gone from Wall Street, but the Rambus had gotten several memory makers to sign up for royalties so the stock stayed high. Since mid 2000, the news for RDRAM continues to be very poor. The P4 was late, and volumes were well under what Intel had predicted. Everyone was waiting for the SDRAM and DDR versions of the P4. Intel had to drop the prices on P4s to historically low levels, (for the amount of silicon and freshness of the design), and even subsidize RIMM prices. Meanwhile, DDR boards started showing up from AMD and others, and these didn't prove to have the unreliability (at least in the memory interface) that Rambus had predicted. When Compaq and Hp started selling DDR based computers, all taint of DDR as an unreliable memory technology had to dissolve (except with the Rambus faithful).

So Edelstone and others predicted that Rambus would be highly profitable due to their patent portfolio covering SDRAM and DDR.

Then, on March 15th, Judge Payne in the Rambus v. Infineon case gave a Markman ruling that essentially decided that SDRAM and DDR did not infringe on RDRAM. This decision was widely described in the trade press in exactly that way, but Rambus put out press releases saying that they would prove their case even with the brutal Markman decision. But Wall Street wasn't so sure, and the stock dropped.

Eventually Rambus lost the case entirely, and even got convicted of fraud. You have to remember that this case was chosen by Rambus as the best way they had of proving that SDRAM and DDR used their IP. In history, the number of companies that get convicted of fraud after suing another company for royalties is somewhere very close to one, and that one is Rambus. With that ruling, there was no way that Wall Street could possibly get involved with this stock.

Let me put it this way. Rambus was not an earnings stock, there never were high earnings (compared to the stock price) in RMBS. Instead, Rambus was always a story stock, and when their first story failed, they migrated to a new story. There is no way that Wall Street is going to pony up to see what the third story that Rambus is going to come up with.

Suppose you've got a trader for a mutual fund. He buys RMBS. If he makes money, great, but if he loses money, what will management do to him? The company's just been convicted of fraud, fraud in the collection of what is now the majority of their royalties, and this after their management repeatedly, publicly stated that they had the patents. The end result, there are plenty of stocks to buy out there where no one will second guess your loss, but Rambus isn't one of them. That's what's keeping institutional investors out of Rambus.

Rambus' main attraction is their great story. But the "great technology" story was broken by the combination of the Camino fiasco, the continuing high prices of RDRAM, and Intel's backing away. Similarly, the "great IP" story was broken by the courtroom losses. At this point, I wouldn't be surprised to see RMBS trade below $2 per share, and that's before the expected RMBS market share losses and the expected further courtroom losses.

One thing about Rambus investors is that they are very vocal, and a bit of a collection of nut cases. You've seen the posts by Rambus longs here and on the other Rambus threads. Imagine what the mail of that guy who wrote the Fortune article must be like. That bad attitude towards Rambus now present in the mainstream financial press was present long before in the (electronics) trade press, and I'll bet the quality of the e-mails that Rambus losers are sending in to the Fortune author are just like the garbage and half truths they post on this thread. Here's a good link for what the trade press thinks about Rambus investors:
e-insite.net

As far as comparing RMBS to another stock, my choice would AENG. Advanced Engine Technology was a story IP stock. The story was that they had an improved internal combustion engine (ICE) design that would make the investors wealthy. As usual, the DA investors didn't think to wonder why they were being given the opportunity to buy in rather than the firm simply remaining private until they had gotten the royalty stream on line. This is identical the to the Rambus story, why did Rambus management bring the company public, they didn't need the money.

AENG had some big names signed on as early investors, names that every race car fanantic in the country would recognize. (I.e. Carroll Shelby, of the Shelby Cobra.) That's similar to the Rambus story, or maybe a bit better, as all Rambus had were a couple of Stanford professors.

AENG gave a simple to understand reason why their engines were better: "The OX2 Engine having less than thirty parts with only three major moving parts was hailed by those present as the first real breakthrough in internal combustion engine design since the introduction of the Otto Four Cycle Engine some One Hundred and Thirty years ago."

AENG gave demonstrations where their engines produces more horsepower per pound than any modern car engine. This advantage was supposed to be so attractive that the car makers would be forced to sign up for high royalties.

The amazing thing is that while everything the AENG people said about their technology was true, they never got the thing into production. The problem was a subtle engineering problem, one understood only by automotive engine designers. This is similar to the problem with Rambus, that the weaknesses of its technology are not well understood except by memory design engineers. So in both cases, mom and pop were sucked in by the press release.

Unfortunately, AENG never got GM to sign up to make nothing but OX2 engines for their 2004 production, LOL! The AENG thread on SI is now dead, as is the stock, more or less. I don't know jack about engine technology, but I followed the AENG story, and collected up a set of links to the story and to posts from the engineers who explained why AENG was a bad idea here: #reply-12453797 If you like researching stocks, go to the above link and click through the dozen or so links, the AENG story is fascinating.

Of course both cases largely consisted of the company insiders dumping shares to the public. At this time, AENG is trading around $.80 per share, and it sure looks like Shelby is buying shares. (Check it out on Yahoo, I bet it's more likely to bounce than RMBS.)

-- Carl



To: Jon Khymn who wrote (76377)7/31/2001 12:18:18 AM
From: Bilow  Respond to of 93625
 
Hi all; Why apeNEXT chose DDR over RDRAM. 44 Physicsists decided to use DDR for their next big Lattice Quantum ChromoDynamics super computer project. Perhaps reading this will give a hint of the decisions faced by a memory designer:

apeNEXT: A Multi-TFlops LQCD Computing Project
[44 authors from Italy, France, Germany and Switzerland.], Feb 25, 2001
...
5.2 Memory Technology
We limit ourselves to memory systems used in future high-end PC's or low-end workstations. This choice (the same as APE100 and APEmille) should be the most effective to provide the highest level of integration, reduce costs and guarantee part availability.

In the near future, planned memory systems are either RAMBUS DRAM's or DDR SDRAM's.

The DDR SDRAM (Double Data Rate Synchrous DRAM), is the evolution of the mature SDRAM (Synchronous DRAM) technology (widely used in the APEmille machine). The SDRAM is a low latency burst oriented device made of multiple (2 to 4) banks of asynchronous DRAM controlled by a synchronous controller which allows pipelining of the I/O interface (one word is accessed for every clock cycle). The Double Data Rate architecture realizes two data transfers per clock cycle using both edges of the clock and one special reference signal to fetch corresponding data.

The RAMBUS is a more advance memory architecture which works as a chip-to-chip system-level interface rather than a conventional memory device. The RAMBUS RDRAM (which stands for RAMBUS Direct Dram) shares the same architectural idea of the SDRAM one, a core asynchronous plus a synchronous controller. It makes use of a large degree of parallelism (32 interleaved memory banks) on a narrow internal bus. The RAMBUS RDRAM is based on the Direct RAMBUS Channel, a high speed 16-bit bus at a clock rate of 400 MHz, which thanks to the adoption of a dedicated signalling technology (RAMBUS Signalling Level) allows 600 MHz to 800 MHz data transfers.

In table 7 we summarize the main features of the two technologies, for currently available and next generation (less than 2 years from now) chips.

                 DDR          RDRAM          DDR          RDRAM
Data rate 200 MHz 800 MHz 400 MHz 800 MHz
Memory size 256 Mbit 128/144 Mbit 1 Gbit 256 Mbit
Organization x4,x8,x16 x16,x18 x16,x32 x16,x18
Peak bandwidth 0.4GB/s(x16) 1.6 GB/s 1.6GB/s(x32) 1.6 GB/s
Package TSOP(66) BGA TSOP(80) BGA
Power (VCC) 2.5 V 2.5 V 1.8/2.5 V 1,8 V
I/O type SSTL2 RSL SSTL (?) RSL (?)
Power cons. 80mA 330mA ? ?
Cost (norm). 1.0 1.8 ? ?
Sample/Prod. Now/Now Now/Now 3Q99/4Q00 ?

Table 7: A summary of several important figures for two options of dynamic RAM's. The second and third columns refer to presently available DDR and RAMBUS devices. The fourth and fifth columns refer to the expected evolution of these devices in the next two years.

Some comments are in order:

* The simple architecture of the DDR SDRAM allows larger memory size per device. For a given amount of memory, this reduces the number of used components.

* Since power consumption is proportional to the interface clock (a factor 4 between RAMBUS e DDR), aggregated memory systems using the DDR SDRAM reduce the global consumption.

* On the other hand the extremely high peak bandwidth of the RAMBUS allows to build a very fast memory system with minimum impact on board space occupancy (compact BGA packaging).

* The logic complexity of a RAMBUS interface is much larger than for a DDRAM controller (the latter could be easily designed on the basis of the experience done in the realization of the APEmille memory controller). On the other hand, several silicon foundries make a RAMBUS controller available as a core cell.

We conclude this section by presenting in table 8 two possible DDRAM-based memory systems for apeNEXT. The performance target is set by our basic performance figure, discussed in the previous subsection of 1.6Gflops and R=4, leading to a bandwidth requirements of at least 3.2 GBytes/sec (assuming double precision data words throughout).
...

pcape2.pi.infn.it (page 17-18)

I should note that the above paper was probably written well before the date shown, that may be the date it was translated into English, or the date it was published. In addition, there are numerous errors in the paper, (i.e. their power calculations, their board space assumptions for RDRAM, number of banks in SDRAM, etc.) but what do you expect, these guys are not going to be the kind of experts on memory design that work in industry. The point is that they are choosing DDR SDRAM because it's cheaper, you can get denser parts, and it's compatible with the previous memory types. These are factors that everyone faces.

ApeNext decided to go with DDR SDRAM, and expects to use 1Gbit DDR SDRAM chips, as I noted in this post: #reply-16077134

More on the apeNEXT project, which is using Synopsis compilers, is here:
synopsis.com

-- Carl