400+ Mhz DDR...system ram.

we all know that the current top line DDR memory sticks are rated for 133(266) or 150(300) Mhz for PC2100 & PC2400 respectivly.

i was overclocking my Geforce2pro last nite and i was thinking:

"i can overclock the DDR on this sucker over 480, maybe even 500!...
where the HELL is my 250Mhz FSB and 512mb of 500Mhz (PC4000) DDR SDRAM???

is it just the excessive cost of making ram rated less than 5ns?
Surely all DDR uses a common architecture, thus it would be relatively easy to transpose chips from a card onto a ram stick?
or are the individual chips too large? too hot? have timing issues? etcetcetc

another invention i have thought of would be to have DUAL head mechanisms on opposite sides of a hard drive platter. thinking about the logistics i guess implementation would be considerably difficult. but i still think its a good way of reducing the bottlenecks inherent with hard drives.
1. theoretical doubling of transfer rates (both heads work independtly and do striping?)
2. doubling the maximum number of seeks possible in a given time frame

tell me what u think... of both ideas.

am i a genius... or just as loopy as my Hamster?


"i love the smell of Overclocking in the morning!" Says my Hamster.

There are several reasons because there isn't (yet) 250Mhz DDR DIMM modules:
-As you mentioned, the cost of making ram rated less than 5ns is high. And you'd need quite more of it than in a video card - you really don't want to have a 64MB system - how about paying $500 for 256MB of 250Mhz DDR stick?
-The typical margins of DIMM modules are way lower than Graphic boards.
-The RAM modules used in video cards can have a more custom physical format than the allowed by DIMM sticks, thus being easier to produce high frequency modules.
-Market size would be low, due to high price tag of the memory modules.
-Even if it was profitable to produce such RAM modules for a small market (it would be quite possible), you'd need:
.support from the chipset for such an high memory bus frequency
.to balance performance they would also want to up the FSB frequency - this would require the CPU maker support
.the higher frequency would probably mean extra PCB layers - higher motherboard cost.
.Also higher frequency also means extra heat - the northbridge would probably need to be produced in a smaller process.

The natural industry evolution will lead us to higher frequency memory modules, although it can take a while, and there are other ways of increasing bandwith - dual channels (128 bit DDR) and QDR.

Regarding the HDD question, what you are proposing is a on drive RAID system. That has been thought before (and dismissed), but it has some problems:
- there are several heads in each HDD - duplicating their number and creating a control system that allowed them to work in tandem isn't easy
- Heat dissipation - HDD are hot as they are - extra heads...
- You'd have to include one on-HDD RAID like controller
- The total cost compared to a traditional RAID system would, most likely, be higher (the HDD itself would probably cost more than twice as much, plus the added controler cost) and performance wouldn't be superior.

I'm not saying it can't be done, I just doubt the financial viability of the venture.


How terrible is wisdom when it brings no profit to the wise