RE: RMBS - Headache material for left brain types ;-)
From: tinkershaw on the G&K thread.....
So, as I see, you do not have anything to support your
claim. And from what you are saying, you have little
clue about DRAM manufacturing.
Typical engineering pessimism. Two weeks before I made my first ten bagger on Rambus an engineer friend of mine at LSI Logic stated that he hated RDRAM, and so did most of the engineers there, but they "were forced" to work with it anyways and didn't think much of it as an investment. I like that sort of coercive market power. It is a gorilla trait. The fact that the experts hated it but had to use it was very telling.
But in regard to the cost premium. Duh, it is more difficult to make than SDRAM. (1) its more complex, (2) it performs at higher speeds using far fewer pins and (3) RDRAM has only begun to hit mass production. One of the crowning achievements of the semiconductor industry has been the incredible ability to leverage experience into efficiency. How long has SDRAM been in mass production? How long RDRAM? Yet Samsung is saying that by 2002 RDRAM will be no more than a 10-15% premium to SDRAM; 20% by sometime in 2001. This after little more than a year, then two years in production. Would you pay a 10-20% premium for a 1 Ghz + chip that isn't bogged down by SDRAM? Well I don't know if you would but most of the rest of the world would.
In regard to RDRAM performance benefits, and I'm sick of repeating this (although I've never stated it here, but its been hashed and rehashed at all major Rambus forums over and over again) the 820 is not a great vehicle for RDRAM, neither are standard benchmark tests. RDRAM comes into its own the faster the chip speed and the more bandwidth required, particularly in multi-tasking, graphical and streaming media applications. By the end of 2001 I don't even think it'll be a contest. Heck, the P IV is out next month, check out those benchmarks.
In regard to DELL, currently 70% of workstations being shipped worldwide, from all vendors, are RDRAM equipped. That is 70% market share in less than 1 year. From 0 to 70% almost overnight. DELL, if memory serves is shipping workstations exclusively with RDRAM. DELL also, instead of keeping AMD SDRAM chips available as an alternative, used only INTC RDRAM equipped chips. It left itself no out in case RDRAM was a flop. Not the sort of risk characteristic of DELL (unless it really was not a risk).
Now, I'm going to MBA school at Duke currently. We did a case study on DELL last quarter. One thing DELL prides itself on is being the source where corporate America can go to to find out about practical technologies they need. That is PRACTICAL TECHNOLOGIES THEY NEED. Dell is not there to push the latest and greatest technology. It is there to make corporate America work better, more efficiently and cheaper. Its reputation is based on this. If RDRAM was such a bust, at an outrageous price, not only would DELL have lost several quartes of sales of workstations before it could make the switch to non-Rambus chips (as the 820 controllers to work with SDRAM were a bust) it would also, and much worse, lose its reputation.
This risk is nothing compared to what INTC did. They have based their future architecture on RDRAM. They have spent billions promoting it. They are currently subsidizing, directly, manufacturers to produce it, etc. Yeah I guess Intel doesn't think Rambus technology is something worth investing in.
BTW/ have you checked out how much radiation DDR will put out at 266 Mhz, if they can even get it to work at those levels without putting out so much interference as to make it practically useless. Fix one problem, and another problem is created with DDR. RDRAM, although it has very tight tolerances, solves these problems.
Rambus draws more hate, I mean just irrational hate, nothing to do with anything else, then any company I have ever followed (I guess I didn't follow QCOM during its really hated days as I was too busy practicing law at the time). Rambus may succeed brilliantly or it may go down in fire. Who knows. But one thing is certain, it is a much more elegant technology than DDR, works better, will be nearly as inexpensive as SDRAM within a year, and much less expensive relative to SDRAM for performance given, much less price performance compared to DDR, unless DDR will be cheaper than SDRAM.
IN regard to Samsung earning extraordinary profits. Ever here of sole-source contracts. Nearly 100% market share in their applications. Investing billions of dollars (yes billions) just to provide more production space for RDRAM. Lets see, the link I gave you before confirms that Samsung desires to move to the top of the semiconductor world. Yes, the best way to get there is to invest billions of dollars to produce a loser technology, that solves no problems, no one likes, is too expensive and difficult to mass produce, and after taking all these risks is not going to return extraordinary returns.
Please! The gorilla game does not require anyone to be a technical expert. If it were up to technical experts none of our retired thread members would be retired (ie, CDMA is impossible we know this because our engineers say so)and I'd never have bought into Rambus. What we follow is value chains, discontinuous innovations, switching costs, the whole shebang.
The fact is, Rambus has a value chain of the largest best abled players, not just in the DRAM industry but also in the communication chip industry as varied as TI to LSI Logic to many of the NGN companies if memory serves. It has a product which is discontinuous in its ability to handle bandwidth. Nothing that is commercially viable today, or even planned as far out as 2003+ (even DDR-II, the suppose next DDR step is much slower than RDRAM), and there is no other memory technology, even at todays high end speeds which will reach 2 Ghz by the middle of 2001, that can handle these speeds. No substitutes. Much less into 2002 and 2003 and beyond.
From a gorilla perspective Rambus has hair. DDR does not, it is continuous and at the end of its life cycle for high end applications - which as we know quickly become low end in the semiconductor industry. Maybe, maybe some DDR 133 will come out this year. I don't know, we are still waiting. The next step is DDR 266. I doubt, even if we do see DDR 133 (which will probably eventually happen) that we will ever see DDR 266. Ever wonder why DDR has been promised for well over a year and yet, time and time again its promised date of delivery keeps getting pushed back. Not one mass producable commercial board for it has yet come out? Maybe DDR is no sweetheart to produce either. Except with DDR you have no propoganda being flown about in the press to illustrate the extreme problems with all the pins and interference - no incentive for anyone to do so. And in the end who cares Rambus makes more money off of DDR under current agreements than it does off of RDRAM.
Sorry for the extensive length. But sometimes one needs to vent even as they get their point across.
Tinker
P.S. BTW would all non-engineers or experts in the technical fields please refrain from posting. I guess that might exclude carpetologists, and how about opera singers, lawyers and MBA students for certain, and probably most other scientists who may not be an expert in the particular fields they are posting in. Maybe I'll stick to Proctor and Gamble, I use consumer products.
Message 14671955
Ö¿Ö |
