Testing out staf's SRAM generator, the results are really impressive (warning: these srams have not been silicon proven and the tool is still very experimental)
128x8, 256x8, 2x 256x16
Is the tinytapeout run with the test cells on run 2? I'm super excited to see if this and Tholin's custom cell library work to see if my club can speed up our project
28+11
Is the tinytapeout run with the test cells on run 2? I'm super excited to see if this and Tholin's custom cell library work to see if my club can speed up our project
I think some projects on run #2 use it, but the wafers are barely out of the fab, it will take some time before they get finalized / bonded / sent out ... and then the people with such design still need to test them and report the resuts.
I have another related question, how much confidence is there in the 3.3 V SRAM macros? I think it’s at least been proven out once in silicon 2 but wanted to know how many other projects were using it
I might have missed something, but I don’t think there’s been more than a basic check of functionality on a couple of run 1 dies. A number of run 2 projects use it though so there should be significantly more confidence before the run 3 deadline.
Until run 2 testing starts, I think this is still accurate:
https://discord.com/channels/1361349522684510449/1532458465379750028/1532694419164233878
1
glaxus
I have another related question, how much confidence is there in the 3.3 V SRAM macros? I think it’s at least been proven out once in silicon 2 but wanted to know how many other projects were using it
I'm looking forward to getting a bond tester unit with a "programmable precision voltage supply" add-on "dongle" (the dongle made and ordered; the tester with the dongle programmed) to do retention (and with it, power-up-bit-pattern) tests on some of the dies (especially if I can get the dongle to someone who has a bond tester unit and access to a bunch of the dies and time to cycle a few dozen through it (technically should only take like 5 minutes or thereabouts at a single temperature)).
That's for FlexMem, Stef's SRAM generator.
Technically that covers 5V and two DRC variants of 3.3V SRAM.
And it's pretty much the densest flexible/varied SRAM we have access to without signing any NDAs.
@proppy we use XLS for datapath blocks on gf180mcuD and would like to upstream a GF180 delay model (run_op_characterization.py on the foundry 7-track cells) once our library qualification wraps up mid-November. Would a PR to google/xls be the way to go, and is anyone already on it?(edited)
4
1
1
