I sent the following email to Disk/Trend:
Sorry Bob,
Please, try this link: 164.195.100.11
and TYPE OR PASTE: 5,787,445 then click on the description. You may want to go to the bottom of the document to save you some time.
AGAIN:
In your response you stated the following:
"One of the problems with flash is that the need to have an erase cycle when rewriting will slow it down. Also, the higher capacity chips that use multibit/cell technology are inherently slower than single bit cell technology."
My questions: does invention 5,787,445 resolve the flash memory problems mentioned above? If no, why not?
Do you foresee a significant market for this invention? How would you expect it to affect other markets?
Thank you.
I RECEIVED THIS RESPONSE:
From: C=US/A=INTERNET/DDA=ID/bkatzive(a)disktrend.com, on 12/3/98 4:33 PM:
"OK, I looked at the patent. Remembering that I am a market research analyst and not a system architect, software or systems engineer, and unqualified to be an expert witness, it seems to me that the scheme of using block linking to locate unused blocks for writing, while avoiding the need for an immediate erase operation, may yet incur significant overhead and ultimately slow down writing as file fragmentation increases. The benefit obtained probably will be greatly influenced by the speed and architecture of the associated processor. I am reminded of a somewhat similar linking notion employed by M-Systems in their flash file system, which I believe exhibits such overhead effects. I am also reminded of the log structured transaction schemes employed in certain disk drive arrays, which write incoming data to the first available block, restructuring the data for more efficient storage and retrieval as available background time permits. Certain flash card producers (Lexar, for one) claim significant write performance improvments based on architectural improvements within the card. I have no design
specifics, but I wonder if they are using the approach referenced in your inquiry.
However, the method of locating available blocks for writing does not address the issue of the write response time of physical cells, which is slower for the multi-bit cells used for high capacity flash memory chips than for bit-per-cell implementations. I suggest that in any side by side performance comparisons of this approach vs. other recording devices, the selection of an optimum device may be determined by the nature and quantity of the data to be recorded and the transaction rate that must be supported.
In any event, it is my impression that flash memory users would want any new products to look like currently existing products from an interface and control standpoint, unless there was some overwhelming economic or performance advantage obtained from doing otherwise.
I hope this helps.
Bob Katzive"
DISK/TREND, Inc.
1925 Landings Drive
Mountain View, CA 94043
Tel 650-961-6209 Fax 650-969-2560
E-mail: bkatzive@disktrend.com
Web: disktrend.com |
