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.
Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?


To: Process Boy who wrote (54205)4/5/1999 5:20:00 AM
From: Tenchusatsu  Respond to of 1573921
 
Thanks, Process Boy. You seem to explain things pretty clearly compared to the rest of the crowd here.

Tenchusatsu



To: Process Boy who wrote (54205)4/5/1999 9:30:00 AM
From: Scumbria  Read Replies (1) | Respond to of 1573921
 
PB,

AMD/MOT process evidently is endeavoring to improve overall processor speed by reducing RC delay (go to Cu), which allows for smaller pitches (tighter design rules) and reduces die size.

I think you have hit on a key point- die size.

Like the 21264, K7 is architected to run at a high clock speed. It is much less dependent on process for MHz than PIII cores are. Instead, AMD may be targeting small die size and low manufacturing costs.

(The 21264 runs >600 MHz in a 0.35u technology.)

Scumbria



To: Process Boy who wrote (54205)4/5/1999 3:21:00 PM
From: Paul Engel  Read Replies (1) | Respond to of 1573921
 
PB = Re: " Cu's lower specific resistance allows thinner wires with less parasitic capacitance (reduces cross-talk), and also allows higher speed (less RC delay). "

You are ignoring SIGNIFICANT DETAILS.

Copper metallization lines in damascene processes need to be ENCAPSULTED by a thin film of barrier material to prevent diffusion into the surrounding oxides and silicon substrate.

With metallization linewidths approaching .25 to .35 microns, the 0.030 to 0.050 micron barrier film thicknesses ENCROACH ON BOTH SIDES of the copper, REDUCING the copper cross sectional area, thereby increasing the resistance of the copper traces.

For example, a 0.03 barrier film will reduce a 0.25 micron damascene "trench" to 0.19 microns wide for the copper to then fill.

Tradeoffs abound with this technology.

Paul



To: Process Boy who wrote (54205)4/5/1999 3:24:00 PM
From: RDM  Respond to of 1573921
 
PB Thank you for your Cu information.

The resistivity of CU is 1.6x favorable to Al. This is part of
the story. The layer thickness of the MOT process for Cu is thinner I believe that for typical AL. Do you have any data on this metal layer thickness. Half the the thickness would mean half the inter-metal capacitance, but reduce the transmission line resistivity to 30% advantage. My feeling from the talks is this is about where the processes are in relative layer thickness but I would like to know if you know any more precisely. I wonder if there are any published results of the performance of Cu processes in technical journals anyones has seen?

The 30% difference can mean the difference between 750Mhz and 1 Ghz operation of long traces.