What’s running at the tower? Router, wireless mgt etc. I see a lot of “we need a server because 5G” in the press. I was wondering if it was “solution in search of a problem” like most of the 5G/Edge talk is in industrials (my domain).

After spending 10+ years in the cellular wireless industry in both engineering & planning roles…I don’t see much value in stuffing a GPU into these chassis.

Contrary to popular belief, the amount of actual compute power in the typical cell site is not what you think…it’s probably lower than what you think. Stuffing a GPU into the chassis increases the thermal (it generates heat) & power (it needs power) demands of the install while adding another “point of failure” (does the site still work without the card? what redundancy exists for that card?). So anyone pushing cellular service providers (CSP) to do all these fancy things out at their cell sites seems to forget that those CSP still have a business to run and service calls out to remotely located cell sites are NOT a desirable business cost; human beings in trucks driving long distances from a service center takes time & costs $$$. CSP want solid, reliable, and preferably redundant hardware that does not require much human attention…even if that hardware seems antiquated, low-tech, and-or backwards to the rest of us.

@Patrick

This is what “edge computing” properly looks like, not some aenemic A53 powered toy-box with the “closed source” ImaginationTech GPU that you mentioned in a previous article. Another paid-for vendor insert STH article maybe?

Intel demoed their “project endgame” in a recent presentation … sharing gpu processing via a wireless connection. I wonder if they have in mind adding their GPUs in the expansion slots on these 5G edge servers…