POWER Updates Submitted For Linux 4.13

[phoronix]

Phoronix: POWER Updates Submitted For Linux 4.13

A variety of POWER architecture updates have been submitted for the Linux 4.13 kernel merge window...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-4.13-POWER

[torsionbar28]

I know there have been 3rd party efforts like the Talos workstation, but if IBM wants to get serious about making POWER an open architecture that people actually adopt, they ought to produce their own line of ATX motherboards and sell them at or below cost to enthusiasts. It would be an investment in their ecosystem. The geeks that buy these and play with them at home, and get familiar with them, are the same folks who will recommend to their bosses to buy an IBM POWER server at work. I'd buy one today if I could get a decent ATX board with a POWER chip for around $699, i.e. roughly the cost of a Supermicro ATX board and basic Xeon chip.

[starshipeleven]

if IBM wants to get serious about making POWER an open architecture that people actually adopt,

FYI: they don't.
POWER is for server stuff, and there they are priced aggressively vs xeons (and can actually pwn them with workloads optimized for them).

About "adoption by the people" topic, I'm pretty sure that there isn't enough money to be made by selling a high-power consumer/enthusiast system that can't run Windows.

[Dopefish]

Personally, I'd wait and see if Tyan puts out a revision to their OpenPOWER lineup. They did offer a quad core POWER8 server system with an ATX format-factor motherboard. If IBM won't provide this, a third party likely will.

Sadly, with how much screwing around IBM tends to do these days, I am thinking a future POWER workstation purchase will be supplanted by a RISC-V based one, assuming they get a fab partner that can crank out high performance chips.

[torsionbar28]

FYI: they don't.
POWER is for server stuff, and there they are priced aggressively vs xeons (and can actually pwn them with workloads optimized for them).

I'm very familiar with POWER servers and AIX, I'm a fan of them and recognize their competitiveness.

About "adoption by the people" topic, I'm pretty sure that there isn't enough money to be made by selling a high-power consumer/enthusiast system that can't run Windows.

To be clear, my intent was not to say "adoption by the people" as if grandma would go to wal-mart and choose a POWER machine over an x86 pc. Lol. Not talking about consumer mass market. What I was getting at was something more in the spirit of CentOS, MIcrosoft MSDN, etc. I.e. a low cost way to get the technology into the hands of enthusiasts and IT geeks at home. Not something grandma buys at the corner store. You play with CentOS at home and get familiar with it, you're more likely to recommend RHEL to your boss at work. If the only way to learn RHEL was to pay $799 for a server license, Red Hat would have no where near the success they do today.

Back to your point, the POWER servers being competitively priced vs xeon, and faster than xeon in certain workloads, I think these qualities are unknown to a lot of IT folks, only because they have no familiarity with POWER. What I was suggesting was not low cost POWER boards as a profitable venture for IBM, but rather as a way to increase awareness of the POWER brand and its capabilities, by selling those boards at/near cost. Label them as a "development" board with 90 days of hardware-only support or something minimal like that. Geeks get to play with them at home, and the geeks will start recommending POWER servers to their bosses at work.

As for Windows OS support being an indicator of success, I think the Raspberry Pi, Nvidia Jetson, etc. would beg to differ!

[starshipeleven]

What I was getting at was something more in the spirit of CentOS, MIcrosoft MSDN, etc. I.e. a low cost way to get the technology into the hands of enthusiasts and IT geeks at home.

Hm, it would make sense, although it would probably be a large risk, and afaik it would be far far far too modern thinking for average IBM bean counter. The main thing holding them back on doing this is their pre-80s mentality (in this sector anyway, IBM is a multi-headed monster, not all heads are stupid)

As for Windows OS support being an indicator of success, I think the Raspberry Pi, Nvidia Jetson, etc. would beg to differ!

I said high-power consumer/enthusiast system. Raspi or Jetson is embedded stuff and embedded isn't really high-power in an absolute sense, we all know Linux dominates in embedded due to much higher flexibility because it is opensource.

[edwaleni]

The OpenPOWER conference for 2017 completed back in May.

Rackspace has released their reference design for OpenPOWER platforms
NVidia showed their CUDA/OpenCL integration
IBM showed their deep learning tools
Ubuntu showed off their Open Stack / containers solution
IBM showed OpenStack and DBaaS solutions
Red Hat did a cloud compute demo

Probably the eye opener was the demo of the NVidia/POWER platform being used for cryptocurrency processing.

Curious what kind of Etherium mining this platform could perform and what the power requirements would be (as compared to a typical Linux/Radeon desktop)

Up to now, the only company that has attempted to crowdsource an Open POWER ATX board is Talos, and they have only raised $517k out of the $3M they say they need.

If IBM or a partner could fab a scaled down POWER SoC with 4 threads, then you could build an ecosystem, similar to what ARM has done with Broadcom (Pi) and Rockchip (Asus/Pine).

Until then, OpenPOWER is making the most headway in the large scale hosting providers. That is where the money is.