The Linux Kernel Continues Prepping For POWER9 Support

[phoronix]

Phoronix: The Linux Kernel Continues Prepping For POWER9 Support

IBM continues prepping the Linux kernel for supporting the upcoming POWER9 processors...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-4.8-More-POWER9

[milkylainen]

How about power envelope? Or efficiency? While fast, Power CPU's are notorious power hogs.
There are not many applications left that do not care about compute/power density in HPC solutions.
Cooling is still a major issue. As is consumption.

The industry does care about ISA's any more. Simply put, if a platform yields better compute/power density, it is usually chosen. If economical constraints allow it.
The rest is fixable code. Especially in HPC solutions.

[microcode]

How about power envelope? Or efficiency? While fast, Power CPU's are notorious power hogs.
There are not many applications left that do not care about compute/power density in HPC solutions.
Cooling is still a major issue. As is consumption.

The industry does care about ISA's any more. Simply put, if a platform yields better compute/power density, it is usually chosen. If economical constraints allow it.
The rest is fixable code. Especially in HPC solutions.

As I understand it, that's exactly what they've been working on. I think the POWER9 ISA is probably there to make sure that it remains a patent minefield so that IBM can keep control of it while they push their "Open" POWER initiative.

I think they will lose and lose painfully to RISC-V; even research RISC-V machines with source available (under commercial-friendly licenses) are approaching the performance of today's Intels with vastly simpler designs. The open licensing and the blossoming collaborative community have made immense progress in just the last two years; and there are defense industry products already shipping RISC-V.

[CrystalGamma]

RISC-V machines […] approaching the performance of today's Intels

Do you have a link for this?

[starshipeleven]

I think they will lose and lose painfully to RISC-V;

IBM is a platinum founding member of RISC-V foundation (among pretty much everyone apart Intel and crap chinese IC manufacturers). Not gonna "lose" against themselves.

even research RISC-V machines with source available (under commercial-friendly licenses) are approaching the performance of today's Intels with vastly simpler designs.

Pics or GTFO, as they say.

[microcode]

IBM is a platinum founding member of RISC-V foundation (among pretty much everyone apart Intel and crap chinese IC manufacturers). Not gonna "lose" against themselves.

Pics or GTFO, as they say.

Nobody says "Pics or GTFO". They say "Tits or GTFO" or "Pics or it didn't happen.".

In either case, nobody is stopping you from watching some talks on youtube, reading a whitepaper, or downloading and running the designs yourself on an FPGA. If you're not even willing to do that, then why is the burden on me to prove it?

The Berkeley out-of-order machine (BOOM) is the most promising open source design of the Out of Order Superscalar variety. It's also likely that there are (unpublished) faster RISC-V OoO superscalars to be taped out in the coming years. NVIDIA is also choosing to integrate a RISC-V core as the realtime controller in their upcoming GeForce, Quadro, and Tegra chipsets to replace their proprietary Falcon cores.

[starshipeleven]

Nobody says "Pics or GTFO". They say "Tits or GTFO" or "Pics or it didn't happen.".

http://www.urbandictionary.com/define.php?term=pics%20or%20gtfo

http://www.thecrimsonflame.com/t2607p45-pics-or-gtfo

http://iwastesomuchtime.com/on/?i=16735

https://twitter.com/picorgtfo

If you're not even willing to do that, then why is the burden on me to prove it?

Because otherwise I would be busy proving myself all inane shit I see someone claim around, while the man making the claim should have some kind of thing on-hand because it's the thing that convinced him and should be easy for him to show me too.

The Berkeley out-of-order machine (BOOM) is the most promising open source design of the Out of Order Superscalar variety.

These are not the pics we were looking for.

You said "RISC-V machines with source available (under commercial-friendly licenses) are approaching the performance of today's Intels".

Please show some evidence of this.

[Zetbo]

http://www.urbandictionary.com/define.php?term=pics%20or%20gtfo

http://www.thecrimsonflame.com/t2607p45-pics-or-gtfo

http://iwastesomuchtime.com/on/?i=16735

https://twitter.com/picorgtfo

Because otherwise I would be busy proving myself all inane shit I see someone claim around, while the man making the claim should have some kind of thing on-hand because it's the thing that convinced him and should be easy for him to show me too.

These are not the pics we were looking for.

You said "RISC-V machines with source available (under commercial-friendly licenses) are approaching the performance of today's Intels".

Please show some evidence of this.

There is no evidence, because that claim is pure bullshit! RISC-V fanboys are getting ridiculous...

[starshipeleven]

There is no evidence, because that claim is pure bullshit!

Yeah, I'm somewhat expecting it is.
I'm following RISC-V relatively closely since a while, I would have noticed any such major news.

[bridgman]

Nobody says "Pics or GTFO". They say "Tits or GTFO" or "Pics or it didn't happen.".