cube

physical design @ nvidia | prev ugrad ece @ cornell | opinions are my own

[cube @cube75703632]

@bubbleboi i think the idea that design cycle times can be reduced with simply mask changes, and not redoing the whole PD, is quite interesting..

[cube @cube75703632]

@beaversteever lemme in

https://t.co/m50ci3xgWm

[cube @cube75703632]

straight outta taiwan

[cube @cube75703632]

@bubbleboi @InvertedBe tcl scripts ready whenever the RTL is

[cube @cube75703632]

@Willschips cool!

what’s the application?

[cube @cube75703632]

@backslashvarphi how big are these pads?

[cube @cube75703632]

@humphilomath @jwt0625 aimed to be fully open source

[cube @cube75703632]

straight outta taiwan

[cube @cube75703632]

@humphilomath maybe in the US. i know georgia tech started around the same time. now a lot do it, and more will do it. money from CEMiD and chips act funding a lot of programs across the country.

this will be my research profs new class. i’m helping make a lot of this.

[cube @cube75703632]

@humphilomath first university to have undergraduates tapeout.

i’ll be graduating and moving back to the bay soon.

[cube @cube75703632]

@humphilomath for my university project team Cornell Custom Silicon Systems.

there’s an Analog-Mixed Signal implementing an 8bit SAR ADC.

Digital chip implementing an async FIFO, FFT and simple classifier, with some BIST.

RF chip implementing a 915 MHz OOK transceiver.

[cube @cube75703632]

@bubbleboi show us the PDK. even intel struggles to create a customer-friendly PDK.

you can make a process, but if your PDK is unusable, tough luck

[cube @cube75703632]

@Willschips agent skin investment strategy remains unpatched

[cube @cube75703632]

@bubbleboi every post a banger can’t lie

[bubble boi @bubbleboi]

We need an EE chat summer school.

June to August we teach you how to build a chip from scratch. No bullshit. From the transistor and up.

[bubble boi @bubbleboi]

I didn’t know this a few years ago but the gap between like an “ok” physical design and a spectacular one is wiiiiiide as hell.

It’s the difference between hitting F_MAX at 600 MHz and 1.5 GHz on an advanced node like 65 nm. And now with so many chip startups all relying on a few design houses the gap is getting wider and wider.

Broadcom isn’t going to optimize the living shit out of your chip because the engineering hours explode to do so. If only there was a way to get good fast before the ladder is pulled up.

[cube @cube75703632]

@beaversteever lemme get that PDK

[cube @cube75703632]

@ptrschmdtnlsn @ico_TC MUSE (MPW University Service)

[i2cjak @i2cjak]

Have you heard of "Tiny Tapeout" but couldn't afford a submission?

Together with Avi Brown and Adam Taylor, American Embedded is PROUD to announce the FIRST EVER "tiny tpot," including 8 FULL sponsorships (analog or digital) for November's shuttle! What's Tiny Tapeout...? 1/x

[cube @cube75703632]

@fishPointer @i2cjak sign me up

[cube @cube75703632]

@beaversteever unfortunate retweet