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.
Technology Stocks : Rambus (RMBS) - Eagle or Penguin -- Ignore unavailable to you. Want to Upgrade?


To: dumbmoney who wrote (30086)9/21/1999 2:41:00 PM
From: Dave B  Respond to of 93625
 
dumbmoney,

That's odd, the latest controllers use a single SDRAM, which provides adaquate bandwidth.

They must not think so, as they're planning to go one way or the other (RDRAM, as per engineering, or DDR, as per procurement). I'll ask.

The problem that hard drive's have is they need less than a single chip's worth of memory, so they will eventually go to embedded memory - not for performance, but as a cost reduction.

Ditto about asking re: embedded memory.

Dave



To: dumbmoney who wrote (30086)9/23/1999 5:23:00 PM
From: Dave B  Read Replies (2) | Respond to of 93625
 
dumbmoney,

I just had a very quick conversation with the friend at the HDD company (he was late for a meeting) and asked about the SDRAM/RDRAM/embedded DRAM issue for hard drives.

I'll tell you what he said with the caveat that we didn't get to explore it fully and he's going to call back later. So I may have to adjust what I say below. Here's what he says:

Right now they use a single 4MB PC-100 SDRAM chip in their drives for caching. The manufacturing process for ASICs and SDRAM is very different and SDRAM can't be placed on ASICs. You can embed SRAM on ASICs, but it takes up much more room and makes the chip prohibitively expensive if you put on any amount of memory at all. They've looked at embedding 128K of SRAM on an ASIC for a very, very low cost drive (like, below PC usage, or maybe e-machines level performance), but with that small of a cache, performance is really poor. When they upped it to 256K, the single-chip solution was more expensive than the ASIC with an external memory chip.

They are planning to move to PC-133 short term. [Now we're getting into the area where he was talking really, really fast]. The bandwidth requirements for the cache for them, however, is such that they will need about 400MBs next year out of the cache (still from a single chip; sustained rate, I believe), which SDRAM will not be able to supply. So he says that they'll move to a single-chip RDRAM solution to meet this need -- it supplies the bandwidth in a single chip and has a low pin count. He didn't say it, but it'll probably be 8MB of cache by then, or a 64Mb RDRAM chip.

If I've misquoted anything, or get additional information, I'll post it.

Dave