[dba-Tech] Software on a USB Drive

Lembit Soobik lembit.dbamail at t-online.de
Fri Nov 17 04:49:17 CST 2006


not the bond wires, they are gold, and they are thick enough to not become 
victim of electromigration.
I meant the leads on the chip. there are (similar to PCboards) several 
levels of interconnect. one is usually aluminum.
When the power supply lines are designed with right angles (without 45 
degree (bevel?) on the inside, there is current crowding at the inside 
corner when the current takes the shortcut, and that is where in many cases 
the aluminum is then eroding (transported away by the current). this effect 
is strongly supported by high temperature. So cool your CPU as good as you 
can, if you want a long life for it).
Some technologies today use copper for this reason. Copper is much better 
conductor and gets less electromigration.

btw, not knowing what a honcho is (thanks for the compliment, John :-)), but 
I was development manager for Integrated Circuits and during that time 
developed a patent how to design output transistors which are much less 
susceptible to ESD.

Lembit

----- Original Message ----- 
From: "JWColby" <jwcolby at colbyconsulting.com>
To: "'Discussion of Hardware and Software issues'" 
<dba-tech at databaseadvisors.com>
Sent: Thursday, November 16, 2006 9:45 PM
Subject: Re: [dba-Tech] Software on a USB Drive


> And since Lembit was a honcho for a TI manufacturing plant, that is info
> "from the source"!
>
> Thanks Lembit.
>
> I didn't realize they used aluminum for the wires though.  Back in the 
> day,
> when I would break chips open, they were all gold.
>
>
> John W. Colby
> Colby Consulting
> www.ColbyConsulting.com
>
> -----Original Message-----
> From: dba-tech-bounces at databaseadvisors.com
> [mailto:dba-tech-bounces at databaseadvisors.com] On Behalf Of Lembit Soobik
> Sent: Thursday, November 16, 2006 2:57 PM
> To: Discussion of Hardware and Software issues
> Subject: Re: [dba-Tech] Software on a USB Drive
>
> the real problems with non volatile semiconductor memory are things like
> this:
> the information is stored in a totally isolated electrode, the electrode 
> is
> inside silicon dioxide (SiO2 - glass), thus it is floating.
> when you drive a junction under this electrode far enough into breakdown,
> electrons will tunnel thrugh the SiO2 and be trapped there.
> the electrode is also a gate for a transistor and will make the transistor
> conductive.
> when such process is repeated often enough, there are also electrons 
> trapped
> in the SiO2, and this prohibits erasing or writing this cell.
> I remember it was a great achievement when some company could produce NV
> memories with 100 000 read/write cycles.
>
> reliability of bond wires is pretty good today, and not specific to NV
> memory.
> however, there are other problems (also not specific to NV), which cause
> long term failures, like Electromigration:
> the metal connections in semiconductors are pretty narrow. when you have
> current flowing through these leads (Aluminum) the aluminum atoms are
> carried to one end. this is among others temperature deendent, i.e., with
> higher temperature the electromigration increases, and the thinner the 
> lead
> becomes on one point, the more electromigration increases, since the
> temperature of the lead increases at that point, until the lead breaks.
>
> Lembit
>
> ----- Original Message ----- 
> From: "JWColby" <jwcolby at colbyconsulting.com>
> To: "'Discussion of Hardware and Software issues'"
> <dba-tech at databaseadvisors.com>
> Sent: Thursday, November 16, 2006 5:24 PM
> Subject: Re: [dba-Tech] Software on a USB Drive
>
>
>> Rocky,
>>
>> Flash have faster access times than hard drives, but slower read/write
>> times.  IOW, the drive has to rotate to get the data under the head, 
>> which
>> can be milliseconds, but once the data starts to stream on/off the disk,
>> it
>> does so faster than flash.  The other issue with Flash is that they can
>> only
>> WRITE any given memory location a certain number of times before the
>> location "wears out".  Associated with this is the problem that flash is
>> written in pages, not individual bytes.  Which means that if only a 
>> single
>> byte of a page changes, the entire page has to be written anyway.
>>
>> The question is, is this an issue?  It really depends on how the drive is
>> used.  Many files are written once and then stay there for a LONG time.
>> Word documents, excel spreadsheets etc.  OTOH, things like the area of a
>> disk used to contain data for a SQL Server or Access database might 
>> change
>> many times per second, just depending on usage of the database.
>>
>> OTOH, flash is purely electronic vs. mechanical for hard disks.
>> Electronic
>> circuit failures group into two types, electrostatic and mechanical (yes
>> mechanical).  We all know the issue with ESD.  The mechanical failures
>> with
>> chips come into play with the bonding wires that link the chips to the
>> external pins leading off the container.  These wires are made of gold,
>> and
>> are finer than a human hair.  They will fail from flexing caused by
>> heating
>> up and down as the chip is turned on / off.  You might have noticed that
>> incandescent light bulbs never seem to just fail once they are on, but
>> rather they fail (typically) when you turn the light bulb on.
>> Incandescent
>> lights are a hot wire.  The wire flexes as it heats up and the metal wire
>> expands.  Eventually that flexing causes a stress fracture.  The same
>> phenomenon occurs in electronic chips, though much slower simply because
>> the
>> heat is usually less so the amount of flexing is less.  Unfortunately
>> neither of these failure types is the issue with FLASH "max write 
>> cycles".
>>
>> Anyway... If electronic chips were never turned off, and never subjected
>> to
>> ESD, they would likely last for centuries, or perhaps "forever".
>>
>> The reality is that rotating and solid state memories both have their 
>> uses
>> and the boundaries of where each is most useful shift back and forth but
>> both will be around for awhile.
>>
>> There are other "flash" (solid state) memories on the horizon though,
>> based
>> on magnetics and the likes.  We'll have to wait and see how they effect
>> the
>> equation.
>>
>> You have heard that Vista can use FLASH for an intermediate level cache.
>>
>> John W. Colby
>> Colby Consulting
>> www.ColbyConsulting.com
>>
>> -----Original Message-----
>> From: dba-tech-bounces at databaseadvisors.com
>> [mailto:dba-tech-bounces at databaseadvisors.com] On Behalf Of Beach Access
>> Software
>> Sent: Thursday, November 16, 2006 10:48 AM
>> To: 'Discussion of Hardware and Software issues'
>> Subject: Re: [dba-Tech] Software on a USB Drive
>>
>> True.  Can a 100GB flash drive be far off?  And at a price competitive
>> with
>> a hard drive?
>>
>> Do you know anything about their reliability?  Ever hear of a flash drive
>> failing? And are the speeds of reading and writing comparable?
>>
>> The 8GB one I used seemed to generate a lot of heat.  I wonder if they
>> will
>> degrade quickly from their own heat.
>>
>> Rocky
>>




More information about the dba-Tech mailing list