Paul, do you think PC133 really has the headroom to scale up with the faster Pentuim III's which will be coming out with the introduction of Coppermine? According to Intel they will be running at 700 Mhz across the board by this time next year. Will PC133 really be sufficient with these CPU's?
My viewpoint is that Intel is driving the industry toward DRDRAM sooner rather than later so the volumes get up to sufficient levels to achieve cost parity with current SDRAM. From what I've been reading, I have come to the conclusion that the first 600mhz Coppermines will do ok with SDRAM. However, since Intel is charging ahead so fast to .18 micron we will be well above 700 mhz before we know it. It is at these clock rates that RDRAM will really be necessary - clock rates that approach 1 Ghz. Do you agree with this statement?
As for Camino, I listened to the Rambus conference call two weeks ago and Geoff Tate (the CEO) said that there wasn't a problem with the 820 chipset (Camino) per se. He said Intel decided to hold back on releasing the chipset until there were sufficient volumes of RDRAMS available. At least I believe this is what he said. There was no recording so I couldn't double check. On the other hand, Edelstone from Morgan Stanley says the production worthiness of the latest version of Camino is critical to the timetable of the Rambus ramp up. Edelstone says we should know this any week now. What do you think? Something doesn't jive with what Tate is saying and what Edelstone is saying!
As far as supplying the market with RDRAMS it appears like the Koreans (especially Samsung) will be able to supply the whole pc market in the latter part of 99 assuming Camino is on schedule. Samsung will be able to produce 5 million units a month starting in three months. My hope is that the rest of the industry Micron, NEC etc. pick up the slack in 2000. Although I believe NEC will have the capability to produce around 3 million units/month toward the fall.
As far as Sony is concerned, I believe it is a great design win but it will never make up a significant portion of Rambus' revenues. The Playstation II will sell something like 30 million units per year with 16 megabytes of RDRAM per unit. Contrast that with a minimum of 128 megabytes for a pc and PC's obviously sell at a much higher clip than 30 million per year. If we have a Nintendo II type of machine we may get up to some serious volumes for Rambus in the game console market. The current Nintendo only uses about 4 megabytes of RDRAM.
As a result, if Rambus doesn't corner the pc market the company's stock price will be in trouble. I believe that there is a slim chance that they won't get a majority of the pc market over the next couple of years. I'm betting that they will! What do you think about these thoughts - especially the technical questions I posed?
If I remember correctly, you were selling Rambus puts. If so, I know you have a vested interest as well. I have a break even of 66 for my latest batch of May Rambus puts and an average stock entry point in the low 70's.
Thanks in advance,
Doug
PS: a good article from the Samsung web site follows.
The Resurgence of Rambus DRAM
Electronic News, 4/5/99
by Nick Schwartzman
Sr. Product Marketing Manager, DRAM Marketing
Samsung Semiconductor, Inc.
An expert, a perceptive man once observed, is anyone a mile away from home. The recent delay in Rambus technology has created a brand new group of expert doomsayers ready to cast a pall over Rambus' coffin and put a couple of nails into it. Well, the hasty funeral arrangements turn out to be a farce of sizeable proportions (a rare event in our field, inundated with sub-micron chores) - the coffin is empty. Rambus is alive, Rambus is healthy, and its future looks bright to say the least.
So, is there any merit to the theory that Rambus may never take over the DRAM Olympus, as some opportunists profess? Although support for PC133 SDRAM is growing, there is a widespread belief that it
is an interim solution before RDRAM becomes available in volume. It should be remembered that PC133 requires a new chipset to support the 133 MHz front-side bus on the Intel PIII.
However, the 133 MHz front-side bus is not expected until September, coincident with the Camino chipset and RDRAM rollout. Aside from the technical advantages of RDRAM over PC133 SDRAM, it does not
make sense from a marketing perspective to split the industry into two camps. Inasmuch as PC OEMs are not openly prepared to pay higher premiums for RDRAM, the lower premiums for PC133 are not
compelling enough to stop technological progress. From the pure premium rationale, Luddites should rule
the world. Fortunately, they do not.
Is PC133 really here? Yes, PC133 SDRAM has been around for some time successfully carving its niche in the server market that needs Gigabytes of memory space, and is content with incremental improvements in memory performance. DRAM manufacturers can definitely supply PC133 in volume now, but what about chipset manufacturers? Can they really supply PC133 desktop chipsets, which would rival Camino's
performance, in volume? Is the business risk worth it when they know that PC133 may be a short-timer in the desktop environment? The notion that PC133 will be a mainstream main memory for generations of ultra-fast CPUs coming to market seems absurd. There is no doubt that PC133 will take some portion of the PC market for a short period of time, but by no means will it enjoy a long life. Let us briefly discuss the ubiquitous performance benchmarks. By employing simple rules of multiplication, we can see that 133MHz
devices on a 64-bit bus cannot squeeze more than 1GB/s of sustained bandwidth. This is 600MB/s less than 800MHz RDRAM. 600MB/s less! Moreover, Rambus can offer something that no other technology
can - it can double its effective bandwidth with minimal effort by introducing new channels. Today, RDRAM uses two channels with 1.6GB/s bandwidth. Tomorrow, it will use four channels and push the
bandwidth envelope to 3.2GB/s. Is there any other technology that can do the same thing so effortlessly?
The Rambus delay that received so much publicity and negative comments is surprisingly beneficial for the DRAM industry. First, DRAM manufacturers used the unpredicted respite to polish their technology
offerings and scrap plans for low-density RDRAM. The industry consolidating and focusing around 128/144M RDRAM, leaving no quarters for 64/72M. Second, there is a fledgling spirit of cooperation
among Rambus suppliers to deliver fine-tuned solutions. The numerous setbacks are motivating manufacturers to seek guidance and provide assistance to achieve a common goal of industry-wide standardization.
Despite the firm belief in the Rambus technology, we have to admit that initial plans were overly aggressive. Samsung made a bold presumption that other manufacturers were as advanced and dedicated
to RDRAM as we are. Unfortunately for the industry, they are not. There was a typical breech (a well-documented case in the history of modern warfare) between fast-advancing tank regiments and
slow-moving infantry brigades. The advance stalled, and Intel's generals prudently decided to stop and regroup. Regretfully, they did this so abruptly that the troops became confused and disoriented. However, Samsung stayed focused, and we still hold the commanding heights.
|
