|||

[ Edited by Velcro_SP 20.04.2011 - 07:22 ]

Quote:
The prob is when a discussion starts to turn into an argument or gets too childish it isn't always clear to decide who actually started and whose fault it was. It just happend. It is wise then to just cut it, leave out any accusations and just move on to the topic again. It's not always easy, but most of the times the best approach.

>> May you enlighten me what's so hard in telling computer chips from computer
>> boards? I really don't get it.

> You're the only one who doesn't.

That's rather unlikely. I even dare to claim that here on MZ there is no one besides you that cannot tell computer chips from computer boards.

>> A PPC460 based core like the APM821xx has? A PPC405 based core?
>> Or something else?

> 800 MHZ are expected to deliver 1216 MIPS i.e. only 1.5 MIPS/MHz.

Thanks for the pointer. I overlooked this important piece of information in the press release. A value of 1.52 DMIPS/MHz clearly points to a PPC405 core (compare: PPC405EX, PPC405EXr). So I think we have it. Thanks again :-)

> IIRC the 460 core should do about 2 MIPS/MHz (like the e300).

Yes, PPC460 (as well as PPC440) delivers 2.0 DMIPS/MHz. But e300 delivers slightly less, namely 1.9 DMIPS/MHz.

> I'd love to discuss the substantive comment I made about
> energy consumption and X86 v. PPC.

Nobody's stopping you from discussing your own comment.

> I can't entirely ignore when someone blows off the substantive talk

As far as I'm concerned, you are free to discuss whatever you feel is substantive (e.g. "energy consumption and X86 v. PPC"). Don't act as if I was somehow hindering you from doing that. I'm definitely not.

|||

[ Edited by Velcro_SP 20.04.2011 - 07:31 ]

If you both were able, somehow, to end up the discussion in a friendly way (take your time, gentlemen), it can be very good both for you and this place.

It has happened to me a couple of times, a technical fight became a great friendship some months later.

You both now that this bitterness can't last forever.

I'm dropping this and leaving it alone, we're nowhere near a switch but when we do, I will fight against x86 and suggest ARM or G5

> we're nowhere near a switch but when we do, I will fight against x86 and
> suggest ARM or G5

I wouldn't call porting to G5 a "switch". PPC970/G5 offers an optional 32 bit mode which is binary compatible to "G2"/G3/G4.

PP970 is still ppc, yes but in one thing i read back when 2.4 just came out was that the devs didn't want to touch it because of "incompatibilities" with current software may happen. I have faith in the team, and I will be patient and stop whining about Power Mac G4 support. If I were worried about incompatibilities, maybe a 64-bit kernel, 32-bit userland setup would be better than a full 64-bit OS?

> i read back when 2.4 just came out was that the devs didn't want to touch it because
> of "incompatibilities" with current software may happen.

Quote from the Bigfoot interview about the chances of G5 Mac support:

"These machines are a bit more troublesome: They use 64 bit CPUs, which would require quite some modifications to Quark to support. Other than that, they don't have much in common with earlier Macs when it comes to hardware, so the driver circus starts from new again."

So there seem to be two obstacles pointed out by Bigfoot : (1) 64 bit CPU and (2) need for new drivers for the remaining onboard components.

Regarding (1), my naive understanding of the matter is that the kernel would be able to run in 32 bit mode but somehow would have to tell the CPU to switch to that mode upon system start. Given that Hyperion seem to have managed to overcome the obstacle of switching a 64 bit CPU into 32 bit mode and running a 32 bit OS on it, I can't imagine it would be that hard for the MorphOS Team's kernel wizard to achieve the same ;-)

> If I were worried about incompatibilities, maybe a 64-bit kernel, 32-bit userland
> setup would be better than a full 64-bit OS?

Yes, probably. Or even both kernel (Quark) and userland (ABox) in 32 bit?

Addendum:

> A value of 1.52 DMIPS/MHz clearly points to a PPC405 core

Confirmation:
http://i.cmpnet.com/dspdesignline/2010/APM801_presentation.pdf (page 2)

Answer to myself:

>>>> Just yesterday I contact a UK company that builds PPC boards (mainly
>>>> because the head of the company had designed an XMOS based PCIe
>>>> sound card). He basically discouraged me from looking into the PA Semi
>>>> based board he had designed because he was having trouble obtaining the
>>>> processors.

>>> It's not Varisys you're speaking of, is it?

>> Yep, you obviously know this company.

> Sounds as if my assumption that Varisys would be the still unnamed Nemo
> designers won't turn out true.

Actually, my assumption just turned out true:

http://www.a-eon.com/news.html :-)

> Actually, I should have known this, because it was said
> the Nemo designing company was 15 years old, whereas
> Varisys has been founded by Barnes only 10 years ago

Hmm, seems Hermans's calendar is 5 years fast ;-)

> probably as some kind of Transtech spin-off

"The partnership developed at Transtech Parallel Systems, where they worked for a decade or more before that on the company's Inmos T800 based Transputer products. [...] While at Transtech, Adam designed embedded processors boards based on Inmos Transputer, Intel i860, and both TI and Analog DSP Devices. During the mid to late 1990s he led the development of a series of PowerPC products based on G2 (603), G3 (750) and G4 (7400) CPUs. [...] Having known Transputer inventor and XMOS head David May (then head of Inmos) since their Transtech days, Varisys was eager to work with him on his latest project, and helped XMOS get up and running with validation boards and development systems."
http://www.amigans.net/modules/xforum/viewtopic.php?post_id=51602#forumpost51602
http://dev.amigans.net/a-eon/NewsRelease-20100618.pdf

Hey Andreas,
I see you quoting things you pointed out to me a few weeks ago.

Do you think these news release might have something to do with Varisys discouraging me from looking further into PA Semi PPC processors?

Also, its hard to believe that everyone on this site isn't aware of how weak Acube's next board is going to be.

Come on guys, we don't need to worry about Acube's products, they make the Efika look good (and that's about all). We had better boards than this years ago.

Quote:

Couldn't agree more

|||


[ Edited by Velcro_SP 20.04.2011 - 07:28 ]

> Do you think these news release might have something to do with Varisys
> discouraging me from looking further into PA Semi PPC processors?

I really don't know. I'm fairly puzzled that while developing a PA6T based board (for which the CPU parts have been sourced by A-Eon according to the extended press release PDF file) Varisys discouraged you from looking into the PA6T. Maybe you could ask them for clarification?

> its hard to believe that everyone on this site isn't aware of how weak Acube's
> next board is going to be.

I think I'm quite aware of the weakness of a 1 GHz PPC440 core + smallish L2 cache based CPU in comparison to e.g. a G4 CPU of the same clock rate ;-)

> we don't need to worry about Acube's products

What you take for worrying is probably rather some's consideration of the Sam460EX as a possible candidate for a future MorphOS hardware platform ;-)

>>> Sounds as if my assumption that Varisys would be the still unnamed Nemo
>>> designers won't turn out true.

>> Actually, my assumption just turned out true

> I'd like to express my support for ACube.

Thanks for your insightful contribution to the discussion of Varisys.

> It'd also be nice if ACube came up with a 5121e-based system

ACube already evaluated the MPC5121e:

------------------------------
(03) Could you please tell us a bit about how and why the ADS5121/Hellrosa/Efika2 was on display by ACube Systems at Pianeta Amiga?

We, as hardware manufacturer, evaluate many different kind of processors. Since this is a reference board we are studying and evaluating it (as we are doing with other processors).
------------------------------
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=27164&forum=42#459651

The fact that now there's still no MPC5121e based hardware or announcement thereof by ACube seems to indicate that they rejected the utilization of the MPC5121e.

Update:

> it's still not official that they use the PA6T. Up to now it's just my and other's
> educated guess, based on the specs and hints revealed so far.

It's quasi-official: the AmigaOne X1000 sports the PA6T :-D

http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=31890&forum=33

Quote:

If they read the according threads at www.powerdeveloper.org I can imagine why they refused the 5121.
And since Acube made good experience with Aplied Micro (former AMCC) I would suggest tehm to try to build a sub 100 EUR board around the APM 801xx. That chip is dirtiest cheap and delivers at 800 MHz a little more power (about 1.5 times MIPS wise) than the e300 @ 400MHz. From the specs a nice chip for ultra low price / lowest wattage / low end applications.

> I would suggest tehm to try to build a sub 100 EUR board around
> the APM 801xx. That chip is dirtiest cheap

While the chip is cheap I don't think that ACube would be economically able to build a sub 100 EUR board with it. The 440EP and 460EX are just not expensive enough to support your reasoning.
Maybe the 821xx at about 10 USD extra per chip would be more suited. That wouldn't lead to a sub 100 EUR board either but would have much more bang for only very little more buck than a 801xx based board ;-)

> and delivers at 800 MHz a little more power (about 1.5 times MIPS wise) than the
> e300 @ 400MHz.

Yes, 1.6 times the DMIPS value to be precise. And the 821xx @ 1.0 GHz delivers 2.6 times the DMIPS value of the e300 @ 400 MHz (and 1.6 times the DMIPS value of the 801xx @ 800 MHz).

> From the specs a nice chip for ultra low price / lowest wattage / low end applications.

In an AmigaOS/MorphOS context I think the presence of an FPU is mandatory. Does it have one?


Btw, would you please take a look at this?

Quote:
Yes, Acube are not yet known for sane prices. But with some courage and a sub 100 design would be possible with the APM 801xx.
Quote:
At least the 821xx could result in a smarter solution than the coming 460 board (which is insanely expensive and has some severe limitations).
Quote:
Ouch - it seems to be absent (405 core in the schematics, no dedicated fpu ploted). But haven't read too much about the 801xx yet. The tech documentation is not published yet (announced for September). But given teh low price I guess they discarded everything that is not necessary for their target market and a fpu is not needed everywhere, so it seems likely they left it out.
Quote:

Well, I think I read it in a Freescale newsletter, but I am not too sure any more. I think there is something like the e5500 a 64 bit core derived from e500 for QorIQ in the pipeline, but no e700 (which would AFAIK a 64 bit derivate from the e600) . But I cannot provide valid links now.

>> The 440EP and 460EX are just not expensive enough to support your reasoning.

> Yes, Acube are not yet known for sane prices.

I'm not sure you got my point.

> I think I read it in a Freescale newsletter, but I am not too sure any more. I think there is
> something like the e5500 a 64 bit core derived from e500 for QorIQ in the pipeline

You're spot on. Thanks for info :-)

Introduction of the 64 bit dual e5500 core based QorIQ P5020 processor (also to be available as single core P5010):

http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0712.pdf

Seems that "e5500" became the official designation of what was called "e500mc64" before:

http://www.google.com/search?q=e500mc64

Other FTF presentations with references to P5020/P5010/e5500/e500mc64:

http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0425.pdf (pages 11 to 14, 16 and 26)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0444_PDF.pdf (page 20)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0676.pdf (pages 13/14 and 22)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0691.pdf (page 3)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0704.pdf (pages 9 and 11)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0874.pdf (pages 7/8)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_ENT_F0748.pdf (page 42)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0598_PDF.pdf (pages 4, 16 and 18)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0707.pdf (pages 17 and 19/20)
http://www.freescale.com/files/ftf_2010/Americas/WBNR_FTF10_NET_F0421.pdf (pages 10 and 15/16, with later P5020 release dates than officially announced ones)
http://www.freescale.com/files/ftf_2010/Americas/FTF10_ENT_F0453.pdf (pages 6 and 80)
http://www.freescale.com/files/ftf_2010/Americas/FTF10_ENT_F0273.pdf (pages 5/6 and 13, with confused families)
http://www.freescale.com/files/ftf_2010/Americas/FTF10_IND_F0427.pdf (pages 6/7, with confused families and confused cores)


Edit: Added some more FTF presentation links.

[ Edited by Andreas_Wolf on 2010/7/18 23:46 ]

|||

[ Edited by Velcro_SP 25.04.2011 - 07:13 ]

Quote:

Nah, I know the price of the currently used chips are not necessarily the reason why Acube sells their boards for a premium. But with the 821xx and 801xx the design could be even more simplified and Acube could try to change their business model from a few exclusive boards to affordable bulk production. It is easier to produce an ultra low cost board with *ultra* low cost chips than with low cost chips. But that needs courage and funds. And of course that attempt is risky and I am not sure whether I would invest my money into such an approach or not.