Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
New Chips Loaded With Dummy Parts (ieee.org)
61 points by naish on Sept 18, 2009 | hide | past | favorite | 16 comments


This is an interesting case of the relationship between abundant and scarce resources. It used to be that transistors were scarce, so using the silicon in order to pack them well was the optimization goal. Nowadays, transistors reached a point of zero cost, but yield and the process nodes to make them are incredibly expensive, therefore, you get "wasted" silicon.

In programming, this is akin to programming in assembly, when no cycle can be spared, versus programming in the highest available language, when the programmer's attention and productivity is scarce.


I don't see how transistors can be zero cost and at the same time be incredibly expensive.

Without the dummy devices you actually would not be able to pack them as close, so the dummy devices actually help achieve the goal of higher density.


What about compilers that stick in NOPs to keep pipelines in sync?


Less "to keep pipelines in sync" and more "to align jump targets to cache-friendly boundaries". But yes.


Yeah, didn't think about that one. Good one.


>Intel improved the resolution of its lithography with a technique called dipole illumination, a process that splits light into two beams, sends them through reduction lenses, and projects features from different angles.

That's so cool that someone had to figure out how to shrink LIGHT so that one can watch cats doing cute things on YouTube.


This technique has been used in PCBs for many decades. You fill in unused parts of the PCB layers with stripes, partly to improve the manufacturing yield, and partly (in ground planes) to "debounce" the circuitry.


Interesting. Kind of the junk DNA of chips.


Not really. The dummy features have a practical purpose to help the silicon deal with stress. Its similar to the ridges and lines that car manufacturers are putting on the panels of new cars. The metal is so thin on these panels that they have to put ridges to stop them from bending. The same thing has happened with processors. We have reached a point where the chips are so small that similar ridges have to incorporated to even out stress within the chip.


You might be dismissing the point too soon. To the lay-person, junk DNA looks like as useless a feature as a ridge in a door panel, or dummy pathways on a chip. To an expert auto-maker, the purpose of the panel ridges is obvious, structural rigidity. To an expert chip-maker, the purpose of dummy pathway is obvious. To Our Holy Creator Who Art in Heaven (or whomever you wish to credit our creation), it is possible that our junk DNA has a very obvious purpose that we have yet to divine.


Junk DNA does (or did, in many cases) have a purpose. Its just that any DNA which we don't understand the purpose for is labeled as junk. That definitely doesn't mean it doesn't do anything or that its useless.


Greg Bear's Darwin's Radio is an interesting scifi take on the idea that some junk DNA could be just one evolutionary step away from being activated.


Darwins Radio was a great read but I don't find the idea of ready made features for future use in our DNA very plausible.

He spins an amazing yarn around it though, and plenty of the social consequences are very believable.


Junk DNA could be protecting the useful DNA, so I can see an analogy.


Very good point!


Wooooosh ... sound over the head of a software guy reading this.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: