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.
Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?


To: tejek who wrote (121005)7/26/2000 10:49:26 PM
From: richard surckla  Read Replies (2) | Respond to of 1572512
 
tejek... >>I would not be surprised if Intel has been unloading RMBS shares...have you checked their SEC reports lately?<<

Intel does not own shares in Rambus.



To: tejek who wrote (121005)7/27/2000 12:19:49 AM
From: richard surckla  Read Replies (2) | Respond to of 1572512
 
<font color=red>tejek, found the following on The Fool... Rambus VS SDRAM (DDR)...

Author: mazepa
Number: of 11005
Subject: Rambus VS SDRAM (DDR)
Date: 7/26/00 7:46 PM

Email this to a Friend
Format for Printing
Post New • Post Reply • Reply Later • Create Poll
Problem Post • Recommend it!
Recommendations: 9

Here is a little write up I put together from various sources for the benefit of those who
would like to understand the technology. I have tried to keep it simple and hope you
enjoy it. Please comment and correct me if I have inaccurately presented something.

In today's systems you will require around 3.7GB/s of memory bandwidth. The AGP
4X video card requires 1.06GB/s and a 64 Bit PCI bus running at 66MHz requires 528
MB/s.

RDRAM is basically the same as SDRAM. RDRAM doesn't change the way memory
is addressed or any of the fundamentals of memory storage/retrieval. The way it works
is it manages to speed up the actual process of transferring the data from one end
(RDRAM) to the other (your chipset). This acceleration is achieved using a small
interface present on the RDRAM chips themselves as well as the memory controller,
which for a PC is located on the North Bridge or Memory Controller Hub of the
chipset. Through these interfaces RDRAM is capable of transferring data at 400 MHz
DDR (double pumped) which means that the transfer rate is equal to that of an 800
MHz channel. RDRAM features a 16 bit wide RAMBUS channel which has a peak
transfer rate of 1.6GB/s.

Currently PC133 SDRAM provides 1.06GB/s of memory bandwidth (I am not going to
even consider 100MHz SDRAM in any comparisons). When the migration to PC133
DDR occurs, it will provide 2.1GB/s of memory bandwidth. That still falls short of what
is needed in today's high-speed systems. That's why Intel and AMD did not include it
beyond their current 2000 roadmap.

As you can see PC133 DDR SDRAM (running at 266MHz) and RDRAM 800 MHz
will not satisfy these needs. So both RDRAM and PC133 DDR (which translates into
PC266MHz) won't provide the necessary bandwidth of 3.7 GB/s.

One solution to this problem is to develop Quad Pumping technology (will be a while).
Another solution is to increase clock speed (which isn't possible until the manufacturing
process improves). The final option is to increase the width of the memory bus, which
brings us to the reason DDR SDRAM cannot grow much further than its current
implementation. Here is why.

Data can be transferred in to ways: serially (one bit at a time). Parallel (multiple bits at a
time). When multiple bits of data are transferred simultaneously, you require a larger pin
count and signal integrity becomes an issue as the transfer rate increases. Pin counts are
a huge factor and play into how motherboards are designed. The higher the pin count,
the more difficult it becomes routing traces on the motherboard to connect everything
together. At a certain point you run out of room to run traces. This in turn forces you to
add layers to the motherboard printed circuit board. This becomes very, very, very
expensive. The norm for motherboards is 4 layers.

Here is the advantage RDRAM has over SDRAM.

RDRAM transfers data serially where SDRAM transfers data in parallel. A RDRAM
channel is only 16-bits wide where a SDRAM bus is 64 bits wide. To add an extra
RDRAM channel is quite simple since the pin count is less than SDRAM. This would
increase your memory bandwidth to 3.2GB/s with one extra channel. Real close to what
is necessary in today's systems. SDRAM on the other hand will require Dual SDRAM
banks (for a total of 4.2GB/s). The problem with DUAL SDRAM banks is it will
require twice the amount of traces and will require an 8-layer motherboard. Well if you
thought RDRAM was expensive go buy an 8-layer motherboard. Some motherboard
manufactures will not even consider an 8-layer design. So unless they can reduce the pin
count on SDRAM (I don't think so), it will not be feasible solution.

RDRAM having a lower pin count is very attractive for companies outside the PC
world. As you may know the Sony Playstation II uses RDRAM (32 MB of PC800
RDRAM) since it saves lots of room in their PS2.This will prove attractive for other
devices that will be small in nature (Internet Appliances). Another factor where
RDRAM proves to be attractive is power. SDRAM requires 900mW of power per
chip for every chip on the module. Every chip on the SDRAM module draws the same
amount of power all the time. With RDRAM each chip on the module isn't active unless
required. So one chip could be using 1165mW (active) of power with other chips using
10mW (nap) or 250mW (standby). This will extend battery life in portable solutions
(laptops). It will also cut down on heat generated inside the device. A definite advantage
over SDRAM.

I hope the above explanation brings to light why SDRAM DRR is a limited solution that
will not scale with tomorrow's devices.

My take on Intels announcement.

Why Intel is supporting SDRAM with the release of the Pentium 4? It's in Intel's best
interest. My take on the announcement is that they want to make their products as
compatible as possible with all memory. This way they can capitalize on the folks who
will upgrade (purchase their processors and new motherboards and reuse your
SDRAM) or buy low-end systems (using SDRAM). If you choose to reuse your
SDRAM then be prepared to cripple your new Pentium 4. This may be acceptable for
some folks who could care less how their PC's or Internet devices perform but in the
business world this is not an option (high-end workstations, Servers) since performance
does matter.

Eventually manufactures will have to use Rambus technology to produce memory if they
haven't already been doing so.

Let me list some of the companies that have a license to use the Rambus interface in
their chipset designs: Intel, Compaq, HP, National Semiconductor, Acer Labs, AMD,
and Sony. I know that I have missed a few.

I'll let the numbers people run the numbers and give us their estimates on what Rambus
should be/will be worth.

Disclosure: Long RMBS as of today…..

Maz