Hacker News new | past | comments | ask | show | jobs | submit login

From 7:40~ "The signal integrity doesn't work out."

I don't understand, but maybe someone else could explain.




See https://electronics.stackexchange.com/questions/29141/interc...

At some point (if you go to high enough frequencies), the capacitance of the copper traces will become high enough (i.e. without any capacitor component connecting to a trace between a CPU and a RAM module) to prevent further frequency increases. One way to deal with that is to have shorter traces. This is exactly what CAMM memory modules do - they have shorter (total) traces than DIMM. Even shorter traces are possible if you get rid of modules completely (i.e. solder the RAM chips to the motherboard). Better yet is to place RAM and CPU cores on one chip, skipping even the motherboard traces between the CPU and RAM chips.


I think there's like electromagnetic interference if signals across various buses in computers are too close together, making it more likely that the signals get corrupted or noisy, which could increase latency for trying to clean the signal or make it impossible to get any data of value.

Not an engineer though so please correct me. I only have a vague understanding of this.


> which could increase latency for trying to clean the signal

There's none of that here. That's a concept for uncontrolled interfaces. This is a memory interface, where you either have a good signal or a flawed design. Things like ECC do exist, but those are to detect bit corruption in the memory, but a flawless communication between is still required.




Consider applying for YC's Summer 2025 batch! Applications are open till May 13

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: