Quote:

From the core yes. But there are pretty many other things within the package for real high end networking (several 10 Gb controllers, packet inspector, buffers, security stuff). Not that MorphOS gets hindred by these compounds, but I guess they are rather bad for the wallet. On teh other hand the other QorIQs have also some fancy stuff on board and are not too expensive.
And - unfortunately the 5500 has not Altivec. Still if someone would try to build new hardware for MorphOS now I would still rather recommend MPC 8610 or 8640.

The P5010 could be a good alternative to the ibm 476 (which I also still consider interesting)

> several 10 Gb controllers

No, just one.

> Not that MorphOS gets hindred by these compounds, but I guess they are
> rather bad for the wallet.

Yes, probably. We'll see what the price tag for the P5010 is going to be.

> On teh other hand the other QorIQs have also some fancy stuff on board
> and are not too expensive.

I wonder how the price of the P5010 will compare to the one of the eight-core P4080, which Freescale sell currently for as much as 380 USD.

> unfortunately the 5500 has not Altivec.

Yes, that's sad. If it would implement Power ISA 2.06's VSX at least...

> if someone would try to build new hardware for MorphOS now I would
> still rather recommend MPC 8610 or 8640.

Yes, reasonable choice. But that train left the station, I fear.

> The P5010 could be a good alternative to the ibm 476

Do you mean the LSI Axxia? After all, the IBM PPC476 ist just a core while the P5010 is a complete SoC CPU. If yes, I think the dual-core P5020 to be a better alternative to the dual or quad-core Axxia because the P5010 is only single-core.

Quote:

Easy: $45.5 (that's 380/8, cheapest joke ever - me ducks).

Quote:

Sure but... COuld there be a way to turn this around, and make that train enter the station back? See, perhaps, these chips are about to get veryu cheap soon. Oh, well, that depends if freescale really sells looots of them. Ouch...
Alright, price is not the only hurdle to build a computer around them, but it can only help. See, it's listed at two hundred dollars...

> $45.5 (that's 380/8 [...]).

Not quite. But nice joke :-)

> COuld there be a way to turn this around, and make that train enter the station back?
> See, perhaps, these chips are about to get veryu cheap soon. Oh, well, that depends if
> freescale really sells looots of them. Ouch... [...] it's listed at two hundred dollars...

"Interestingly, feanor revealed to have received a direct offer from Freescale for massively reduced MPC8610 pricing." (quoting myself)
https://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=6684&start=99

That project is still listed btw:
http://www.powerdeveloper.org/program/mpc8610
http://www.power2people.org/bounty_047.html (82% to go)

With MorphOS being available for G4 Mac mini I'm personally not too much interested in this train turning around because I'm not confident about the extent of real world performance advantage a 1.33 GHz MPC8610 or 1.25 GHz MPC8640 offers over the 1.5 GHz MPC7447A/B (+ Apple northbridge). It would be new hardware, though.

> I normally wouldn't trust anyone claiming in 2010 that there'll be an e700 coming,
> especially if that person held the position of Director of Strategic Marketing for
> Freescale's Networking Systems Division from March 2001 to April 2009 and thus
> simply *must* know about the cancellation of Freescale's e700, which took place
> long before he left the company.

It may be that I did him injustice by mocking his February 2010 claim of the e700 being a future product. See:

https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7001&forum=3&post_id=74210#74210

Update:

> the most recent Power Architecture roadmap revision, which is not yet published
> on power.org website

It is now:

http://www.power.org/resources/devcorner/roadmap/Power.org_PA_Roadmap_Aug_2010.pdf

Interesting bits (flaws?):
- QorIQ P50xx a 32 bit processor?
- QorIQ P55xx? What's that?


Edit: Adjusted PDF file URL because it was changed.

[ Edited by Andreas_Wolf on 2010/9/28 1:19 ]

Update:

> he presents the QorIQ P1 and P2 families to be based on
> e500mc core (while in reality based on e500v2).

Today Freescale announced a new QorIQ P2 processor, which is not based on the e500v2 core like the earlier P2 processors but on the e500mc core like both the P3 and P4 families:

http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=P2040
http://www.freescale.com/files/32bit/doc/fact_sheet/QP2040FS.pdf
http://media.freescale.com/phoenix.zhtml?c=196520&p=irol-newsArticle&ID=1462516

A 1.2 Ghz e500 core processor? That's a worse solution than the 86XX processors which were suggested before, and which you are right seem pointless with Mac G4 support.

A quick question here. How is it that the development board for this processor uses the same outdated ULi Southbridge that the 8640 development system uses?

Does someone have a stash of M1575s sitting somewhere? Freescale refers to this as an Nvidia component, but as far as I know it was discontinued when Nvidia bought out ULi.

[ Edited by Jim on 2010/8/25 1:02 ]

> That's a worse solution than the 86XX processors

I only mentioned the P2040 announcement for sake of completeness. As you can see in my self-quote, two months ago I claimed QorIQ P2 to be only e500v2 core based (which was true back then), while Mr. Behmann presented it as being e500mc core based (which was false back then). Now with the new P2040 the QorIQ P2 family has e500v2 core based members as well as (upcoming) e500mc core based members.

> How is it that the development board for this processor uses the same
> outdated ULi Southbridge that the 8640 development system uses?

You mean the P2020/P2010 development board (as I can't find any specs for a P2040 development board yet, and P2040 is not pin compatible to P2020/P2010)?

> Does someone have a stash of M1575s sitting somewhere?

It may even be Freescale themselves having hoarded some ;-)

> Freescale refers to this as an Nvidia component

In fact, Freescale's reference to this part is rather inconsistent: "ULi/nVidia", "nVidia/ULi", "nVidia", "ULi" (and also both names in different combinations of uppercase and lowercase letters).

Good points. So far I haven't seen any other new introduction as interesting as the e5500 you pointed out.

As far as those processors having too much additional I/O capability, since Freescale is unlikely to demand a significantly higher price than an already overpriced e600 cored product (after all, consider the market for the e5500 cored devices) why would the extra capabilities be a problem?

> As far as those processors having too much additional I/O capability, since
> Freescale is unlikely to demand a significantly higher price than an already
> overpriced e600 cored product (after all, consider the market for the e5500
> cored devices) why would the extra capabilities be a problem?

You better don't ask me but the ones who regard that as a problem ;-) I said two months ago that we should wait and see what the price tag for the P5010 is going to be. Then we'll be able to compare to the prices of e600 core based processors.

Yes, I don't get it. Why would a $300 P5010 be a problem if they want $556.28 for an MPC8641D 1.5 Ghz processor (quoted price from Newark)?

> Why would a $300 P5010 be a problem if they want $556.28 for an
> MPC8641D 1.5 Ghz processor (quoted price from Newark)?

I think for the sake of consistency we should rather compare the P5010 to the 1.5 GHz MPC8641 (which is 518.48 USD at Newark in single quantity), and the 1.5 GHz MPC8641D to the P5020. We'll see if the P5010 will be no more than about 300 USD in single quantity.
The 500+ USD price tag for the MPC8641(D) in single quantity being indeed highly problematic from our view doesn't mean that 300 USD for the P5010 in single quantity couldn't be problematic as well for somebody. It depends on the personal view and on the purpose and the calculation of the product. If somebody finds 300 USD for the P5010 problematic then I can accept his opinion (which doesn't mean I'd share it of course).

OK, I should have taken the # of cores into account.
I think you've also made a really important point about having to wait to see what the prices are going to be.
And, those quoted prices (from distributors) doesn't reflect the often much lower pricing obtainable directly from the manufacturer.

> those quoted prices (from distributors) doesn't reflect the often
> much lower pricing obtainable directly from the manufacturer.

Yes, correct. While Freescale's price for the MPC8641(D) is 27% lower than Newark's single quantity price, Freescale only sells quantities of multiples of 100. On the other hand, Newark also sells quantities of 50+ where the price for one MPC8641(D) is 19% lower compared to their single quantity price. Overall, that makes Freescale's price 10% lower than Newark's for large quantities.

And those prices still don't take into account the opportunity to negotiate an even lower price on a product Freescale has a significant quantity of on hand.
I remember years ago approaching TI about a VDG that supported sprites and overlay functions (for a video overlay/sprite board for AppleIIs).
They price I finally was quoted (for 100) was less than their stated price per 1000.

> those prices still don't take into account the opportunity to negotiate an even lower
> price on a product Freescale has a significant quantity of on hand.

Yes, that's what feanor was offered with the MPC8610.

This would be the only real reason to consider the 86XX processors. I'm sure Freescale has a fair number of these stockpiled.
Still, maxing out at 1.5Ghz, its hard to argue for the creation of new motherboards that would be quite expensive when there are so many G4 Macs available.
Even the P5010 with its higher performance potential may not make sense when faced with G5 Macs.
Then again, if there's a market for the relatively weak performing Acube motherboards...

Quote:

Well, there are some pros for the 86xx. Very fast memory bus. No 74xx solution comes even close to that. Could increase the speed for data intesive processes significantly.
High integration, low wattage. The Mac mini is pretty tiny, lw wattage and a real hardware gem, but with the 86xx tiny boards with low wattage are easy to design, too (I think t he NEC redtail board looked nice).
It's new. I personally don't care about that too much, but others have a strong preference for new devices.

Quote:

While relatively cheap and powerful the Apple G5 computers have also some significant draw backs: They are big boxes (but looking goooood ). And they draw a lot of electrical juice. And they are to get used only.

Update:

> the eight-core P4080, which Freescale sell currently for as much as 380 USD.

New prices:

http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=P4080&tab=Buy_Parametric_Tab

530 USD for the 1.5 GHz version, 442 USD for the 1.3 GHz version and 368 USD for the 1.2 GHz version. All versions in quantity of 100 pcs.

Just for comparison the P4040 (quad-core version of the P4080) prices:

http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=P4040&tab=Buy_Parametric_Tab

397 USD for the 1.5 GHz version, 331 USD for the 1.3 GHz version and 276 USD for the 1.2 GHz version. All versions in quantity of 100 pcs.

Update:

> Freescale announced a new QorIQ P2 processor, which is not
> based on the e500v2 core like the earlier P2 processors but on
> the e500mc core like both the P3 and P4 families

P2041 announced:

http://media.freescale.com/phoenix.zhtml?c=196520&p=irol-newsArticle&ID=1561029
http://www.youtube.com/watch?v=kUkoQ4kzNW4

It's like P2040 + 25% higher clocked + L2 cache + 10GbE.

Addendum:

> http://videos.cache.magnify.net/HYS8JG0X56WGX7G3-PowerOrg_DAC_483_336_384x216.mp4
> (Design Automation Conference 2010)
> Quite disturbingly, he claims the XBox360 and the Wii to be powered
> by the Cell CPU.

Seems he still didn't learn:

"Then there's a number of cluster or hybrid type of approaches such as mini ...micro core or mini cores actually that's being used in the game consoles. And we have evidence of that in the whole area in terms of game consoles for XBox, of Playstations and Wii type of game consoles."
http://videos.cache.magnify.net/3D16ZQ1TW91KG78S-Power_483_336_384x0.mp4 (at 1:24)

(And at 4:37 he seems to indicate that Freescale's e6500 core is 24-way threaded. Funny guy.)

Addendum:

> Funny guy.

He's been uttering interesting statements again:


"Power.org came onto the scene back in 2005, with IBM and Freescale as founding members."

Freescale, while not actually being a founding member, got founding member status when it joined Power.org only in 2006, not in 2005.

https://www.power.org/news/articles/freescale/
https://www.power.org/news/pr/view?item_key=f9079dfb4b41fd262943334cf0cdfd41adb2c878
http://media.freescale.com/phoenix.zhtml?c=196520&p=irol-newsArticle&ID=831602


"we recently celebrated Power.org’s 20th anniversary."

Ah, so it's year 2025 already. Nice to know.


"Android + MPC5521e, and Android + MPC5525"

I guess that should read MPC5121e and MPC5125 (MobileGT), as MPC55xx is really another kind of thing (Qorivva).


"If you have watched the news lately, you probably know that in terms of high-performance computing, two of the Top 10 and five of the Top 20 most powerful computers are based on Power Architecture."

Actually, in the Top 10 this is just one, namely Roadrunner on rank #10.

http://www.top500.org/list/2011/06/


"Included in these are the BlueGene/Q and also the BlueWater, which will be installed next year."

Actually, BlueGene/Q is not in the Top 20 but on rank #109. And how can BlueWaters be in the current Top 20 when it's not even installed yet as he says?

http://www.top500.org/system/ranking/11074
http://www.theregister.co.uk/2011/07/15/power_775_super_pricing/


"The Blue Gene/Q [supercomputer] is characterized by a 16-core chip and [...] uses Power7"

Actually, it uses A2 core based Power BQC chips, not POWER7 (which has only up to 8 cores anyway).

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7675&start=69


"Editor’s note: Blue Gene/Q was also ranked No. 1 on the list of the world’s fastest supercomputers for six years in a row (see www.top500.org)"

That's complete nonsense. BlueGene/Q was a new entry to the list in 2010 and hasn't even reached Top 100 once. The most successful BlueGene to date was BlueGene/L, which was #1 from 2004 to 2008, i.e. 3.5 years, not 6. It was superseded by Roadrunner as the new #1 in 2008.

http://www.top500.org/system/ranking/11073
http://www.top500.org/system/ranking/11074
http://www.top500.org/system/ranking/7101
http://www.top500.org/system/ranking/7605
http://www.top500.org/system/ranking/7747
http://www.top500.org/system/ranking/8968
http://www.top500.org/system/ranking/9485


"We have gone from 596 teraflops in 2004 [with Blue Gene/L]"

Actually, BlueGene/L was #1 in 2004 with only 92 teraflops. 596 teraflops had not been reached before 2007 (with BlueGene/L on #1).

http://www.top500.org/list/2004/11/
http://www.top500.org/list/2007/11/


"to 1 petaflop in 2008 with Blue Gene/P."

Actually, 1 petaflops was reached in 2008 by Roadrunner, not by BlueGene/P. BlueGene/P reached 1 petaflops only in 2009.

http://www.top500.org/list/2008/06/
http://www.top500.org/list/2009/06/


"And now with Blue Gene/Q, as I mentioned, we are reaching 20 petaflops."

Actually, 20 petaflops for BlueGene/Q has been postponed to at least 2012. It's currently at 0.1 petaflops, which is only 0.5% of the performance they're aiming at for 2012 (or later).

http://www.top500.org/system/performance/11074


Source for this pile of crap statements: http://www.mil-embedded.com/articles/id/?5269

I dont know wtf this subject title means but from reading the thread thought i'd share this:

When we (Genesi) were exhibiting at the SoCal Linux Convention (think it was called SCALE) in Los Angeles, CA we had about 6 full Pegasos 2 (pre odw) workstations deployed with matching LCD monitors. We had a small booth and the event electricians went to all the booths to check how much current we were drawing. (Dont forget IBM and other vendors are there with server racks) Apparently if you have too much wattage you have to pay additional fees or something.. well much to the suprise of the techs, who thought there gear needed recalibration, we were running extremely low wattage. So much so that they stated they have never seen full desktop systems deployed running that low wattage!

Those were hopeful times as we had Dale Rahn (OpenBSD porter for peg2) Sven Luther, bbrv, Paul Adams, and myself showing off the systems. We had great interest and meetings with some top companies like Novell, IBM, Suse, Real Networks and others..

ah so sad now ..

> I dont know wtf this subject title means

I can tell you: It's Velcro_SP's way of blurring his traces ;-) Original title was: 5121E v. Atom in Wattage Sweepstakes.