Message 14682530
From: tinkershaw
"Those same engineers are now working with DDR.
I have to take issue with this conclusion first it is totally untrue in regard to my friend as LSI. Second, AMD, for example, is hiring droves (well a
material amount anyways) of engineers to work with RDRAM. RDRAM is on their product road map and speculation is now, if the P IV works as
many expect, that RDRAM for AMD will be here much sooner than expected.
No, RDRAM is not a failure. It is just one of the most emotional technology debates I have ever seen. RDRAM is currently (1) dominating in
workstations, (2) 100% Playstation II, (3) shipping in digital set-top boxes, and HDTVs as well as 40 other products. It will also be the exclusive
memory chip for Intel's P IV set to launch next month (at least exclusive until the middle of next year). And even at this point INTC has restated on
several occasions that RDRAM is their memory solution of choice and that other memory options with the P IV are only being made in case
RDRAM pricing and production has not sufficiently matured.*
*QUICK NOTE: Intel has been frustrated with the slower than expected RDRAM ramp-up by the DRAMs. Intel is now directly subsidizing the
difference between the SDRAM and RDRAM prices for DRAMs to expedite this ramp.
RDRAM is not a failure. It is just in the typical battle to overthrow the status quo; a battle in which Ericsson admitted to having made blatant false
propoganda as a weapon in its battle with QCOM and CDMA; the DRAM industry is not any different. In addition, and this is the big point, what
substitute is there? DDR does not scale and it is near the end of its life cycle even as it has yet to see any market other than a few cards. It will
shortly be a bottleneck to higher processor speeds, at the high-end within less than a year; much less its inferior abilities in regard to multi-tasking,
graphics and streaming media.
To my mind I'm willing to except evidence that RDRAM is dead, RDRAM is a failure, etc. However, the problem with all this evidence is the pure
venomous glee that most seem to have when they present this evidence, as well as failing to present an alternative to RDRAM. What DDR-II? Or
shall the industry allow DDR to be the peak bandwidth capacity of computers going forward until 2004 or so when this fabled DDR-II can be
produced. Intel depends on ever increasing processor speeds for its business model Intel cannot allow this. AMD, on the other hand, as the
purveyor of low-end chips (even as it tries to once again take on INTC on the high-end) can afford to let this happen to the detriment of INTC's
high-end ambitions.
A lot of the truth in this matter is found in the politics and strategic positioning of the companies. IT is a fascinating analysis and will be my topic in
strategic analysis this quarter. My group will be applying game theory to the Rambus strategy.
Again, like many on this thread, I am not an engineer. I am an attorney now at business school. I can understand and apply technical knowledge but
I am no expert in regard. What I am very good at, and what the Gorilla Game teaches us, is that one does not need to be a technical expert to
invest in high-tech. Following the politics, value chains, product adoption curve, etc. is the more valuable tool.
Tinker
P.S. As for the product adoption curve RDRAM is clearly across the chasm and at a minimum in the bowling alley. As described above it is taking
out the workstation market, moving on to take out the set-top box market, HDTV market and game player markets like the PS II. In addition,
RDRAM's greater "granularity" compared to DDR makes RDRAM not only better performing but much more economical in applications like
set-top boxes and Playstation II, as well as in things like cellular phones. RDRAM is not yet to the price point, nor is its performance benefit
sufficient enough at this stage of processor development, to be a major player in the low end computer market, but it should be moving its way
down the field over the next year. You can get an RDRAM equipped system from DELL for under $1500 today.
In addition RMBS has agreements with close to 25% of the DRAM industry for royalties on SDRAM, DDR and RDRAM. Rambus' revenues are
clearly in the tornado. Should they have a beneficial outcome in the ligitations (al lah QCOM/Ericcson) the full-blown tornado stage will be
complete without a competitive threat for years to come. The twister should only intensify as Rambus moves beyond just DRAM into collecting
royalties on the controllers from the likes of INTC and AMD, and if successful in penetrating its designs into the wireless and communications chip
world. Two opportunities that at least Rambus management has claimed to be even larger than the DRAM opportunity.
So my conclusion: RDRAM in the bowling alley; Rambus is in its second quarter of a company specific tornado. Last quarter royalty revenue
increased 55% sequentially, this quarter royalty revenue was nothing less than a tsunami. The potential is for more tsunami like quarters for what
could be an extended period of time. Rambus has a revenue model even more compelling than Gemstar's.
As always use your own judgments. I've looked at Rambus and the industry battles long and hard. I am not so confident to be 100% Rambus
(although I know a few people who are - and being 100% anything is kind of boring anyways). However, it is my feeling that when you look at the
totality of the situation, and the current Rambus valuation, it is a very compelling place to be for the less risk adverse and something to watch very
closely for the traditional more risk adverse gorilla investor.
I'm taking the position of buying in before the litigation resolutions. Perhaps a position of speculative money in now and more conservative money in
later may be the best compromise.
P.P.S. in regard to the latency issues: this is somewhat disputed. RDRAM, SDRAM and DDR seem to be relatively equal in this regard. However,
Samsung's White Paper on the topic concludes that RDRAM is superior to other memory alternatives in latency.
I have heard some statements to the effect that the 820 chip is the principal problem in regard to latency issue and that it is not inherent in RDRAM.
The latency problem seems to disappear when you mount the RDRAM chip directly on the processor.
So the best answer I have for latency is that worse case all the memory chips have very similar latency characteristics; Samsung has claimed
superior RDRAM latency characteristics; the 820 was the cause of any latency concerns and not an inherent RDRAM problem; and finally the
issue goes away when you mount the RDRAM directly onto the processor.
As one can see understanding the technical issues is not always that helpful. Far more helpful is understanding the gorilla elements and understanding
the politics and strategies involved in this titanic industry struggle. It is fascinating if nothing else." |
