Semiwiki 400x100 1 final
WP_Term Object
(
    [term_id] => 151
    [name] => General
    [slug] => general
    [term_group] => 0
    [term_taxonomy_id] => 151
    [taxonomy] => category
    [description] => 
    [parent] => 0
    [count] => 442
    [filter] => raw
    [cat_ID] => 151
    [category_count] => 442
    [category_description] => 
    [cat_name] => General
    [category_nicename] => general
    [category_parent] => 0
)
General
Posted on by Don Dingee

How many consortia does POWER need to succeed?

How many consortia does POWER need to succeed?
by Don Dingee on 08-06-2013 at 1:02 pm
Categories: EDA, General

Sometimes press releases just make me scratch my head. Today’s example comes from IBM: after tying PowerPC and Power.org in knots for almost 20 years with rules and restrictive licensing, IBM breaks ranks and sets up ANOTHER consortium with different players.

The OpenPOWER Consortium ditches those boring embedded guys – pretty much down to Freescale and LSI – and retargets POWER for servers. This time, IBM aligns with a much more pop culture crowd, with Google and NVIDIA as the headliners. I’m now having a flashback to the PowerPC 620, the ill-fated 64-bit attempt from the last generation …

This time, it’ll be different, I’m sure. IBM has opened the kimono to POWER8, including open-sourcing the firmware. By bringing in NVIDIA, the idea is to integrate CUDA into the flow – remember, GPUs aren’t just for graphics, they can be used as scalable computing engines nicely. By bringing in Google … well, I’m sure that does something, but it is hard to tell exactly what right now. Other than attaching their name to the press release without a quote, Google is very quiet about this. One of the B list players in this is Mellanox, of InfiniBand fame, which is fascinating because Power.org had gravitated to RapidIO as the interconnect.

From IBM’s chair, if they hope to compete with Intel and the ARM ecosystem in server hardware space down the road, they need some new friends. But I’m very curious if IBM is the only one fabbing silicon, just how open is this? I doubt Freescale or other big processor players will jump on huge, complex, hard-to-fab chips to sell only a few thousand parts. Tyan isn’t exactly a huge name in servers. There may be something to the Google thing: Google, Facebook, Amazon et al are taking to a model where they design their own servers and networking boxes for internal consumption. That may prove to be IBM’s real motive in this.

I’m sure Intel and ARM aren’t losing much sleep over this yet.

Bookmark this, it might get more interesting (one would hope): open-power.org

lang: en_US

Share this post via:

Comments

0 Replies to “How many consortia does POWER need to succeed?”

You must register or log in to view/post comments.