4 ms·
No, and only 64 GB max of non-ECC at that. I typically don’t run simulation of more than 1 hour though, and I’ll know when the results are wrong. My use case
by madengr 6y ago
No, and only 64 GB max of non-ECC at that. I typically don’t run simulation of more than 1 hour though, and I’ll know when the results are wrong. My use case is CAD/real-time design for RF and EM; multiple smaller simulations of parametric sweeps, Monte Carlo, and optimization.
Ryzen does support ECC, and > 64 GB, so that is a plus.
- jrockway 6y agoI don't care too much about the results of my own programs, but I do worry very much about OS-level data structures. If some bit gets flipped in memory while you're extracting an update, that program will just randomly be weird forever. If some bit gets flipped in your filesystem, your data randomly disappears. Given how frequent both auto-updates and bit flips are in practice, no doubt something on my computer is corrupted somewhere. Or maybe only my backup is corrupted. Maybe it will never cause a problem, or maybe my computer will just one day become unusable. I have spent so much time debugging random failures that I see ECC memory as at least avoiding one set of possibilities, reducing the inevitable system integration time that building your own computer requires.