3 ms·
The architectural innovation is relatively cheap, since it's all hdl that can be synthesized and then simulated. The problem is the costs of masks and silicon
by outsomnia 5y ago
The architectural innovation is relatively cheap, since it's all hdl that can be synthesized and then simulated.
The problem is the costs of masks and silicon production. The US has been asleep at the wheel about the implications of just letting all production happen elsewhere and Europe has not even got in the vehicle.
> Trying to make more interfaces available, commodity
This has been happening for simpler IPs up to 10G Ethernet and USB2 at opencores. The problems are complexity, patent licenses, and the phy parts being tied to a specific process.
- brucehoult 5y agoPatent expiries moves along. Anything associated with Pentium Pro or PowerPC G3/G4 has expired, and we're getting to Pentium M ("Centrino") and Athlon64. CPU design is only tweaking after that, nothing fundamental. GPUs is a bit more of a headache. Patents for things such as Rage 128 and TNT2 should be free now, but we've still got five years to go until patents for unified shader model expire. That's again, basically the last fundamental advance in GPUs. Good open-source DRAM controllers are desperately needed. There are examples for DDR3. A problem is that some of the standards themselves for critical things such as USB, WIFI, DRAM are not free.