4 ms·

Most chip design tool startups are designed to be bought by a big player. The big guys (cadence, mentor graphics) don't really do much feature development. They

Most chip design tool startups are designed to be bought by a big player. The big guys (cadence, mentor graphics) don't really do much feature development. They just buy startups that innovate around one feature. In many cases, these startups make a plugin for one of the major tools. The plugins are able to integrate pretty well because the founders of these startups are often former employees of the big guys. Revolving door... Your company might have had trouble brcause it is hard to convince engineers to risk a million dollar tapeout on untested tools.

kev009 11y ago
You hit the nail on the head for my experience. In a domain I am much more at home at, data storage reminds me of the same. Most storage vendors want to be gobbled up by EMC, IBM, Hitachi, Oracle StorageTek.. it takes a lot of guts to go it like PureStorage. But even something as critical as SAN and Network storage is magnitudes less conservative than circuit design software.
eli_gottlieb 11y ago
>Your company might have had trouble brcause it is hard to convince engineers to risk a million dollar tapeout on untested tools. As opposed to existing, battle-tested tools that are known to be pretty crappy?
buserror 11y ago
Funny bit is that they'll know buy pretty much any crap. I've worked lately with a NAND controller IP that ticks /every single box/, has every single feature and absolutely all the acronyms you can think of, and is the most useless piece of crp ever* to use. It's so bad that I wish there was a 'passthru' mode so let me talk to the NAND flash on GPIOs, because it's likely I'd have better performance anyway. But, that's a brand new IP bought by cadence to add to their list, and therefore it was also sub-licenced by some hardware designer somewhere (who has no idea what 'software' is) from the feature sheet alone. Then it lands on the poor unsuspecting person who needs to write the driver (me). So yes, that startup/feature/IP cycle is a very weird economy indeed.

kev009 11y ago

You hit the nail on the head for my experience. In a domain I am much more at home at, data storage reminds me of the same. Most storage vendors want to be gobbled up by EMC, IBM, Hitachi, Oracle StorageTek.. it takes a lot of guts to go it like PureStorage. But even something as critical as SAN and Network storage is magnitudes less conservative than circuit design software.

eli_gottlieb 11y ago

>Your company might have had trouble brcause it is hard to convince engineers to risk a million dollar tapeout on untested tools. As opposed to existing, battle-tested tools that are known to be pretty crappy?

buserror 11y ago

Funny bit is that they'll know buy pretty much any crap. I've worked lately with a NAND controller IP that ticks /every single box/, has every single feature and absolutely all the acronyms you can think of, and is the most useless piece of crp ever* to use. It's so bad that I wish there was a 'passthru' mode so let me talk to the NAND flash on GPIOs, because it's likely I'd have better performance anyway. But, that's a brand new IP bought by cadence to add to their list, and therefore it was also sub-licenced by some hardware designer somewhere (who has no idea what 'software' is) from the feature sheet alone. Then it lands on the poor unsuspecting person who needs to write the driver (me). So yes, that startup/feature/IP cycle is a very weird economy indeed.