...
Tapeout is on Wednesday. We had additional delay because we had to instigate the option to use other process corners (the foundry provides also the option to produce fast transistors (FF model)). We found little advantage of using the FF corner at or above 0.6V, but after some additional simulations we found that the chips would work very nice in the FF corner at 0.50V - 0.55V. To be able to operate at this voltage not only the hashing cores but also the IO transistors must be "tuned". We made sure the chips will work at that voltage.
Provisional estimations are below:
vdd m gh/s watt gh/J
0.60 tt 3.34 0.96 3.49
0.60 ff 8.02 3.67 2.18
0.55 ff 6.84 2.23 3.07
0.50 ff 5.51 1.38 3.99
[m: model, tt: standard, ff: fast transistors]
=> at 0.5V we get faster chips that have higher gh/joule than tt chips at 0.6V
Of course these are just simulations.
...
We understand the the most important is timing of deployment. We loos a lot of money with every day of delay. We still think start on July 1st is feasible. We selected a packaging facility that is in the same city as the foundry so we hope to have the packaged chips in house end of May. The rest is fast :-)
whoa..does that mean we're getting a 65% speed boost?!