Log In   |  Register Free Newsletter Subscription
Skip navigation
Zibb
Subscribe to Semiconductor International

Mystery Wirebonds in Marvell Sheeva..

July 6, 2009

A few weeks ago we found this:

Sheeva CPU

It’s the CPU chip from a Marvell SheevaPlugTM plug computer.  This is a gadget that plugs into a wall socket, designed to run network-based services that normally need a dedicated PC, but at a cost of about $100.  It incorporates a 65-nm 1.2 GHz Marvell CPU with 512 MB each of flash and DRAM memory, and connects via Gigabit Ethernet as well as USB.

The question is, what are those wire bonds doing on the top of the die?  (In the images they have collapsed onto the die surface, but before de-packaging they would have looped away from the die.)  It’s the first time we’ve ever seen such a thing.

An optical image shows that they link pads at the die edge to pads in the chip interior, so they have obviously been put there for a reason, but what is it?  Our functional analysis of the die indicates that they loop over the CPU core of the chip.  Our speculation is that the initial chip layout simply couldn’t get enough power into the CPU block, so it couldn’t run fast enough to meet the 1.2 GHz specification.  This was likely the cheapest way of solving the problem, even though it meant a rework of at least the top metal and pad masks.

But that’s our speculation; otherwise, it’s a mystery.  Any other ideas?

Posted by Dick James on July 6, 2009 | Comments (4)

9/4/2009 11:49:23 AM CDT
In response to: Mystery Wirebonds in Marvell Sheeva..
Colorado33 commented:

Similar bonding was done on GaAs logic die for the the Cray3 during early development. It was faster than creating new masks and used only to test ideas and certain functions. I was new to the industry when I worked there, so I don’t know the full details on why. fyi, we used super-glue to insulate the wire.


9/4/2009 11:48:32 AM CDT
In response to: Mystery Wirebonds in Marvell Sheeva..
Shin commented:

100$ to see a bunch of wires …


9/4/2009 11:48:32 AM CDT
In response to: Mystery Wirebonds in Marvell Sheeva..
Etna commented:

The chip version of jump wires.


9/4/2009 11:48:31 AM CDT
In response to: Mystery Wirebonds in Marvell Sheeva..
Steve commented:

Dick, odds are that your assumption is correct. Normally, in the power business, these pads would have gone directly to the package posts but they probably had a height limitation in the package. To go all the way to the package from the middle would require a very high loop.

I would also assume that this is probably temporary and that at this moment, they are working on a revised die layout for the long term.

POST A COMMENT
Display Name
captcha

Before submitting this form, please type the characters displayed above. Note the letters are case sensitive:

Advertisement

Advertisement
NEWSLETTERS
SI NewsBreak and Special Reports
Photovoltaics Report
Wafer Processing Report
Litho & Metrology Report
Packaging Report



Please read our Privacy Policy

OTHER NEWS FROM RBI
About Us   |   Advertising Info   |   Site Map   |   Contact Us   |   FREE Subscription   |   RSS
© 2010 Reed Business Information, a division of Reed Elsevier Inc. All rights reserved.
Use of this Web site is subject to its Terms of Use | Privacy Policy