I gather people report the parallela board was bottlenecked by external memory,
Is the theoretical bandwidth of the 1024-core epiphany V significantly improved - how would it compare to similar GPUs for example. (or is this information still secret)
I gather the longterm plan with this architecture is to have huge amounts of onchip stacked memory directly on each core, but it seems in the present day off chip bandwidth is going to remain important
epiphany V external bandwidth
by dobkeratops » Fri Feb 24, 2017 1:11 pm
Last edited by dobkeratops on Tue Mar 07, 2017 10:40 pm, edited 1 time in total.
- dobkeratops
- Posts: 189
- Joined: Fri Jun 05, 2015 6:42 pm
- Location: uk