> Well someone else obviously thinks PowerPC is still useful:
> http://www.eetimes.com/electronics-news/4399118/Startup-plugs-Power-chips-into-energy-efficient-servers

From that article:

"Along with IBM, Servergy is the only company currently making Power-based servers."

There has always been Hitachi's SR series using IBM POWER chips:

- SR16000 with POWER6 and POWER7
- SR11000 with POWER4 and POWER5

I've been under the impression that these machines are built by Hitachi around the POWER chips they obtain from IBM. Or maybe these are just relabelled IBM machines?

"It’s not clear whether Servergy designed its own Power processor or is using an off-the-shelf chip. [...] analysts said they are unaware of any company making merchant Power chips for servers. One possibility is the company could be using a chip derived from an existing IBM chip design."

This is interesting. I found an Oracle blog article from a month ago stating:

"The first system I'd like to tell you about is a really cool 8 core Power Architecture Hyper-Efficient Enterprise Server from a company in Dallas called Servergy."

And from Servergy's Principal Software Engineer Ben Collins' blog, also one month ago:

"The company I've been gainfully employed with for the past 1.5 years seems to be using something quite different than your grandmother's Power chip. Not quite the behemoth of the IBM Power7 iron (in size nor noise), but not the wussy of your old PowerMac neither. We're talking multiway SoCs with full floating-point running at a fraction of the wattage of just about anything else on the market. Add with it full hardware virtualization (via KVM), and you begin to see where in the market this is headed."

No more specific details than "8 core" and that it's not IBM POWER given unfortunately. I have a feeling that it could be a QorIQ chip, judging from Ben Collins' Freescale-focused comments here and there, as well as from his personal development repository.
Digging somewhat deeper and searching for his development mailing list postings from the last few months answers the question I think:

https://lkml.org/lkml/2012/2/27/237
http://lists.gnu.org/archive/html/qemu-ppc/2012-05/msg00243.html
https://lists.ozlabs.org/pipermail/linuxppc-dev/2012-June/098255.html
https://lists.ubuntu.com/archives/kernel-team/2012-June/020738.html
https://lists.ubuntu.com/archives/kernel-team/2012-June/020741.html
https://lists.ubuntu.com/archives/kernel-team/2012-June/020655.html
https://lists.ubuntu.com/archives/kernel-team/2012-June/020693.html
https://lists.ubuntu.com/archives/kernel-team/2012-June/020735.html
https://lists.ubuntu.com/archives/kernel-team/2012-November/022652.html
https://lists.launchpad.net/ubuntu-powerpc/msg00066.html
https://lists.launchpad.net/ubuntu-powerpc/msg00073.html

And the most revealing one regarding adding e500mc support to Ubuntu's Linux kernel:

"Our company is in talks to become a full partner with Canonical, so I suspect this makes it more of an official offering than anything else. Also, I am being paid to do community work on this kernel and maintain it fully. [...] Also, this line of processors has a large community base. Most PowerPC developers have similar machines (benh)."
https://lists.ubuntu.com/archives/kernel-team/2012-June/020733.html

In conclusion, my guess is: It's Freescale's 32-bit QorIQ P4080 chip (8 e500mc cores), so nothing to write home about and in my opinion a rather questionable choice for a server platform.

[ Edited by Andreas_Wolf 24.12.2012 - 11:57 ]

> if his investment so far has been $100,000 to $200,000 US dollars,
> I doubt he is charging enough to recoup all of that investment already

"Trevor said that it cost over US$400,000 to develop the X1000."
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=36690&forum=2&start=80#685861

More info from Trevor's AmiWest 2012 presentation (direct link, Youtube):

"It's all a matter of scale. If I could build 10,000 units, the NRE (non-recurring engineering) costs would be 5 GBP (8 USD) a board, 8 USD a system. That's nothing. If you build 1000 units that's now 10 times. If you build 100 units that's now 10 times again. So, someone's got to eat that non-recurring engineering costs. On the X1000 I'm eating it, because I wanted to produce the system. On the next systems [...], we're building that into the price, assuming that we sell a certain number of units. And that will govern the price."

I personally don`t understand the fuss over SMP. The ability to spawn an asynchronous process on the other core with it`s own memory space would almost as useful, although my knowledge is limited to scientific computing.

AFAIU SMP benfits from the fact that the task scheduler just balances the tasks between available cores itself and actual programs don't need to be SMP aware. The scheduler schedules taks and threads to the cores. On AMP only those applications benefit from additional cores that explicitly are written for multi core processing. Probably no big deal for new multithreaded programs, but old programs will utilize the master core only.

Quote:

Looks like I'll be happily proved wrong: 3 months! (potentially less than a year from that post)

[ Edited by boot_wb 02.11.2012 - 02:29 ]

@boot_wb,

Wow! That is some quote you dug up on yourself regarding how long it would take to port to the G5.

3 months is a lot different than 4 years, but then the bounty is only for porting to one specific model of the G5 PowerMac, which might work on more models of the G5 PowerMac, or might not. Since most of the work probably was already done when the "Proof of Concept" was done a few years ago, there is less work to do now, to get the port completed.

I am not going to go searching through all of my old posts to see what my estimate(s) were of how long I thought it would take to port to the G5, because I am sure any estimate I gave probably would have included more than a port to just one model.

I am just happy that this one port is happening, and there is a chance that it will work on my dual 2.7GHz G5 PowerMac tower, even though I would keep my G5 just to run MacOSX Leopard & Final Cut Studio even if it never runs any version of MorphOS3.x. At least now there is a chance that I will be able to run MorphOS3.x on my dual 2.7GHz G5 PowerMac.

> 3 months is a lot different than 4 years

Since the 3-months figure is what it is supposed to take not from now but from when the porting starts, I think it's more fair to compare it to boot_wb's 2-year estimate. Still a big difference, of course.

> Since most of the work probably was already done when the "Proof of
> Concept" was done a few years ago, there is less work to do now, to
> get the port completed.

I think this has already been taken into account in his original estimate.

Update:

>> Well someone else obviously thinks PowerPC is still useful:
>> http://www.eetimes.com/electronics-news/4399118/Startup-plugs-Power-chips-into-energy-efficient-servers

> I found an Oracle blog article from a month ago stating:
> "The first system I'd like to tell you about is a really cool 8 core Power Architecture
> Hyper-Efficient Enterprise Server from a company in Dallas called Servergy."

Seems that Servergy has another platform in the works called P-Cubed with a yet unnamed "Dual Core Power System on Chip (SOC)":

http://www.servergy.com/wp-content/uploads/2012/12/P-Cubed.jpg
http://www.linux.com/images/stories/714/P-Cubed.jpg
http://www.linux.com/news/enterprise/high-performance/147-high-performance/659106-servergy-announces-linux-on-power-enterprise-development-platform
http://www.servergy.com/news.html

What can be seen from the board layout is that at least GbE, USB and HDMI come from the SoC itself.

Addendum:

>> It's simple really.

> Indeed it is:
> 1. The X1000 exists.
> 2. An unknown to us number of X1000 units was available for sale.
> 3. All of those X1000 units available for sale so far have been purchased.

In his AmiWest 2012 presentation (direct link, Youtube), Trevor kind of reveals that there may currently be 400 to 500 X1000 owners. Whether this is meant to include those buyers from the current batch who still have to get their machine (supposed to be delivered until January 2013 at the latest) is unknown to me.

Update:

> Seems that Servergy has another platform in the works called P-Cubed
> with a yet unnamed "Dual Core Power System on Chip (SOC)":
> http://www.servergy.com/wp-content/uploads/2012/12/P-Cubed.jpg
> [...]
> What can be seen from the board layout is that at least GbE, USB
> and HDMI come from the SoC itself.

More information:

http://ben-collins.blogspot.com/2012/11/servergy-announces-new-powerpc.html

From there:

"The platform is geared toward making modern Power systems available to developers for a fraction of the cost of full fledged server systems (Servergy's primary market). While the board is aimed at increasing the ecosystem and community around Linux-on-Power, the pricing is sure to attract hobbyists and students as well. While Servergy did not say the exact price, they are aiming at a sub-$200 system."

Update:

> Seems that Servergy has another platform in the works called P-Cubed
> with a yet unnamed "Dual Core Power System on Chip (SOC)":
> http://www.servergy.com/wp-content/uploads/2012/12/P-Cubed.jpg
> [...]
> What can be seen from the board layout is that at least GbE, USB and
> HDMI come from the SoC itself.

More information:

"The platform is geared toward making modern Power systems available to developers for a fraction of the cost of full fledged server systems (Servergy's primary market). While the board is aimed at increasing the ecosystem and community around Linux-on-Power, the pricing is sure to attract hobbyists and students as well. While Servergy did not say the exact price, they are aiming at a sub-$200 system."
http://ben-collins.blogspot.com/2012/11/servergy-announces-new-powerpc.html

Ben Collins says in a comment on the P-Cubed:

"We are definitely aiming this to be a low-end version of our CTS product line."

As I said I believe the CTS-1000 to feature Freescale's e500mc-cored QorIQ P4080 SoC. However, I'm not aware of any dual-e500mc chip. My feeling is that at the announced low price it should be some e500v2-cored chip (QorIQ P1 or P2, probably P1022 with DIU due to HDMI traces coming from the chip).

Addendum:

> My feeling is that at the announced low price it should be some e500v2-cored chip
> (QorIQ P1 or P2, probably P1022 with DIU due to HDMI traces coming from the chip).

On another site someone says he "saw that they use a P1025 chip", i.e. an e500v2-based chip without DIU. Where he saw that I've no idea though.

Update:

>> there's an interesting tweet by IBM from June 2011 which seems to confirm that
>> the Wii U actually uses the very same POWER7 *chip* (not just the same core) as
>> IBM's Watson computer, which would mean an eight-core POWER7 chip. I find that
>> very hard to believe.

> If the recent information about Wii U's CPU being a three-core chip is true, this
> would mean that IBM purposely spread false information in June 2011. There's not
> even a POWER7 chip with 3 cores (4 cores being the smallest variant). And from
> last month two other tweets from the same IBM twitter account that, citing an
> old Engadget article, suddenly read way more vague and contradict the older tweet
> (same chip as in Watson vs. custom chip):
> http://twitter.com/IBMWatson/status/231902074890248192
> http://twitter.com/IBMWatson/status/240241146213842944

2 months ago, IBM reinforced the claim that the Wii U processor was based on POWER7:

http://twitter.com/IBMWatson/status/247460688283463680

Only few days later, IBM finally admitted that their claim that the Wii U processor was a POWER7 chip of some sort or somehow derived from POWER7 was a misinformation:

http://twitter.com/IBMWatson/status/248820933618442240
http://twitter.com/IBMWatson/status/248929547842641920

Love being made to look foolish repeating dumb rumors only to find out that the Wii U's processor is a warmed over three core version of the last processor.

Makes the rumors about Power 8 look less likely.

> Love being made to look foolish repeating dumb rumors

Always depends on who started the rumour in the first place. In this case this was no less a figure than IBM itself. So if IBM spreads misinformation about their own product, taking it for truth is not a foolish act in my opinion.

> Makes the rumors about Power 8 look less likely.

You mean the rumour about future POWER chips with Cell SPE-like units?

>You mean the rumour about future POWER chips with Cell SPE-like units?

Yes. it calls into play how trustworthy these guys can be before a products release.

I looked stupid repeating this.

> I looked stupid repeating this.

Regarding your praise of the Wii U processor only for the last 2 months, after IBM had admitted their misinformation ;-)

Andreas_Wolf,
Quote:

Thanks you Andreas for that boost to my ego.
Better go have a hearty breakfast.
Have a feeling its going to be a "kicked in the ass" day. ;)

@Jim,

We all make dumb mistakes from time to time. I think I do so more often than almost anyone else on this forum site. I have thought several times about just not posting so often, so I could cut down on the number of dumb mistakes I make, but then what would be the fun in that?

If you can't laugh at yourself regularly, then you are probably going to end up with an ulcer, or some other nervous disorder eventually.

Jim,
Quote:

IBM do have a habit of being a little vague about this sort of thing.

e.g. they said POWER6 and Cell were related.
This is actually true - but this referred to the circuit technology *not* the architecture.

Probably the same thing here - and there are certainly similarities such as the use of eDRAM.

Would that be kind of like my toaster is Power7 related because they both run on electricity?

> IBM do have a habit of being a little vague about this sort of thing.
> [...] Probably the same thing here

IBM tweed from September 16th:
"The Wii U is a custom 45nm #power7 chip. Same SOI design in #ibmwatson"

IBM tweed from September 20th:
"Pardon the error. It's a custom chip built on Power Architecture base"

I wouldn't call this "a little vague", but rather spreading false information and then posting a correction and apology later. Sincerely, I hope that such is *not* IBM's habit (spreading false information in the first place that is). To me that's a quite remarkable incident, as they started this POWER7 nonsense talk right when the Wii U was announced in mid-2011 and were subsequently questioned ad nauseam about further details on the chip and only 1.5 years after the announcement bothered to come forward with a correction.

Andreas has pin pointed the truly disturbing thing, this looks like they were deliberately making misleading statement. Certainly not "vague", and not only incorrect, but false.

So, can you trust "Big Blue"?

> this looks like they were deliberately making misleading statement.

That's also my suspicion. To me it seems like the purpose of IBM making the misleading connection between the Wii U and their Watson supercomputer was to keep Watson in the news, riding on the coattails of the Wii U's news coverage.

I wonder what, if any, involvement IBM has had with the development of the Xbox 720 or PS4 cpus?
WiiU's cpu looks weak even when compared to current generation of game consoles.
Does not bode well for Nintendo.

[ Edited by Jim 20.11.2012 - 17:46 ]