Breaking News
News & Analysis

Google, Intel Prep 48V Servers

12V rails not going away soon
1/21/2016 00:01 AM EST
4 comments
NO RATINGS
1 saves
More Related Links
View Comments: Newest First | Oldest First | Threaded View
NeilO940
User Rank
Rookie
Re: 48V is nothing new to servers
NeilO940   1/26/2016 7:41:03 PM
NO RATINGS
The focus of the panel discussion was on direct (single step) 48v to PoL conversion as distinct from dual or multi-step conversion. The latter is likely to be less efficient and more expensive particularly for higher power rails.

PowerRox
User Rank
Blogger
48V is nothing new to servers
PowerRox   1/22/2016 8:34:10 PM
NO RATINGS
Rick:


48V input to server boards (or front-end power supply output, if you prefer to think about it that way) is nothing new and has been around for a long time even outside of the Telco sector (typically supporting a -36-72VDC input).  Particularly in the days of the "big iron" servers, a 48V midbus was required to pump enough current around large boxes without sacrificing too much voltage drop and/or requiring comically oversized busbars to handle the 12V current.

You are correct in that most boards use 12V as the primary rail as input to most of the downstream, non-isolated converters.  So any system that outputs 48V from the front-end simply has an extra, isolated bus converter to do the 48-->12V conversion.  It can be more efficient (in terms of power & cost) to co-locate this conversion step directly into the front-end, which is why you may see a direct 12V out.  So really, we are just talking about moving it down the line from one place to another.

A SIDE NOTE: PoE-enabled network switches (and any other enterprise HW that integrates PoE) may still utilize a direct "48V-ish" (actually more like 52-58V range) to support the growing power needs of multiple PoE ports.

Thanks for the article!

Brian Zahnstecher

PowerRox

PowerRox
User Rank
Blogger
Re: Google comment
PowerRox   1/22/2016 8:24:32 PM
NO RATINGS
Rick:

The migration to 3.3V disk drives should be inevitable with the increasing use of solid-state drives (SSDs) and integration of other solid-state, non-volatile technologies.  If anything, then I think the 5V is merely a carryover from the legacy infrastructures of 12V & 5V for spinning drives.

Hot-swap and surge protection are really the only things I can think of that dictate a higher voltage at all.  NOTE: an exception is if you are bussing drive power across a backplane and need to minimize the IR drop and thermal loss due to higher current handling of lower voltage rails.

rick merritt
User Rank
Author
Google comment
rick merritt   1/21/2016 6:14:36 PM
NO RATINGS
In an email, Mobashar Yazdani, a semiconductor manager at Google Inc. who moderated the panel, noted that hard disk drives currently use 5V and wondered if they would be interested in migrating to 3.3V to enable a more efficient 48V/3.3V server.

Most Recent Comments
AKH0
 
Andy_I
 
Rick_Hille
 
sranje
 
chipmonk0
 
DrewTech
 
dt_hayden
 
sixscrews
 
resistion
Most Recent Messages
8/29/2016
1:51:22 PM

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed