> Freescale insists that e6500 delivers 6.0DMIPS/MHz.

They seem to have changed their tune. According to https://www.nxp.com/files-static/training/doc/ftf/2014/FTF-NET-F0032.pdf (page 7/8), it's still 3.3 DMIPS/MHz per thread, but now only 5.4 DMIPS/MHz (i.e. 10% less than before) per core.
On the other hand, https://www.nxp.com/files-static/training/doc/ftf/2014/FTF-NET-F0140.pdf (page 8/9) and https://www.nxp.com/docs/en/brochure/BRT1T2FAM.pdf (pages 2 and 7) still show 6 DMIPS/MHz for the e6500 core.


Edit: The following documents show 7 DMIPS/MHz
https://www.nxp.com/docs/en/product-brief/T4240PB.pdf (page 2, 10/2014)
https://www.nxp.com/docs/en/fact-sheet/T4240T4160FS.pdf (page 2, 07/2016)

[ Edited by Andreas_Wolf 17.05.2023 - 14:23 ]

Quote:

I agree. I can't remember if I have already said this:

But I would love continued support of PowerPC device, even while supporting a second architecture. Support all available PowerPC, while selecting 2 - 4 x86 64 items: a desktop computer and laptop computer, each with a "low end" (affordable one) and a "media center" one.

I do not believe there needs to be supporting more than four X86 64 items. It might increase User Base significantly, but it might not. According to Wikipedia, there are other Amiga "inspired" OSes, almost all on X86, yet none are more popular than MorphOS. (Or perhaps it is because these OSes are inferior compared to MorphOS.) However, if X86 64 support increases User Base by thousands of thousands, then perhaps they should support dozens or hundreds of X86 64.

I do have one respectful disagreement. I do not think ABox should be supported. It should be moved to an emulator. Another idea: perhaps if MorphOS attempts SMP, then one core can be ABox and another core can be QBox. ?

It would be nice for MorphOS to support the X5000, but also it would be nice if MorphOS to support the SAM 460.

> Support all available PowerPC

This would dash against missing support for GPUs in machines where the GPU is onboard and you can't put in another one.

> I do not think ABox should be supported. It should be moved to an emulator.

On little-endian non-PPC, ABox couldn't be supported anyway. And on PPC, a box or virtual environment for ABox wouldn't be something I'd call an emulator.

> The T2080 looks like a good possible CPU for any PPC Amiga/Morph/AROS system,
> if any kind of multi-processor support can be completed. The fact that it includes
> Altivec is a plus.

Prices for 1000 pcs. qty. revealed: 127 100...111 142 USD
Reference design board T2080RDB is 1499 USD.

http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=T2080&tab=Buy_Parametric_Tab
http://www.freescale.com/files/32bit/doc/brochure/PWRARCHQIQSG.pdf (page 7)
http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=T2080RDB
http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=T2080RDB&tab=Buy_Parametric_Tab
http://www.freescale.com/files/32bit/doc/fact_sheet/T2080RDBFS.pdf


Edit: new chip prices

[ Edited by Andreas_Wolf 06.11.2014 - 15:54 ]

Quote:

Am I reading the pdf correctly that the Eval board has 4 PCIe slots that are 2.0/3.0 Generation compatible?

If the MorphOS Dev. Team would port to this Eval board, couldn't this board be used as the basis for a brand new MorphOS compatible computer system at a price of not much more than $2,000 US Dollars, or less if you use existing case and PSU and other peripherals that you already own?

At 1.2GHz to 1.8GHz with Altivec, 8 cores, and 4 PCIe slots, as well as USB2.0/3.0 and Ethernet on board, I don't see why this wouldn't work. The 7 other cores would go to waste, but could be utilized if the owner also wanted to run Linux on the same box part of the time.

Edit: We could use PCIe to PCI and/or AGP adapters for video and sound cards, until we could get drivers for
PCIe cards, couldn't we? Just wondering if this could be the answer for users who want new equipment, instead of used Mac gear. That is not me, as I am satisfied with what we already have, but some are not.


[ Edited by WB_Coder 04.06.2014 - 15:03 ]

a) 2000$ is still far far to expensive to make any sense for what is essentially G4 class performance

b) even if we could make good use of all cores (which we can't) it would still be a bad bang/back ratio

c) getting PCIe GFX cards with allready supported GPUs is not a problem

I would fear supporting that proposed computer could overwhelm MorphOS developers. Instead of jumping at multiple computers simultaneously, it should be taken one step (or computer) at a time. Example SAM 460, which unless something changed I am not aware of, it is a computer still in production, not an Apple, and is not expensive. If angry mobs demand X1000 etc. then I would certainly join the request to port MorphOS to the X1000s.

Quote:
Well, I wouldn't call it cheap or even reasonably priced ;)

Quote:

I would rather see a 32 bit Amiga compatible PPC Mac plus a NG MorphOS 64 bit non-PPC machine worked on simultaneously. With the G5 you could even have both on the same machine as a stop gap measure.

But hey, I'm no programmer and I'm not a developer. I trust the MorphOS Team that for whatever reason they don't port it to PPC machine X or x64/ARM machine Y today, it's because they have a good reason not to. And my bet it that it's mostly a question of TIME. If they could live by coding for MorphOS, then I bet more cool things would happen. But since they can't, we can't be too demanding.

Maybe we could start a foundation, like they have for the Haiku camp? Then we could maybe pay salary for a developer to work full or half time for a while. Or use the money to reward people who make big contribution for MorphOS in general? I know I can spare 25 euros every month for such a thing. Just my 2 cents.

[ Edited by Yasu 05.06.2014 - 10:28 ]

Perhaps the person most desperately wanting MOS on HW X should put up a bounty for it.

(I'm personally very convinced that current low end ARM or off the self x64 is out of the question, so until custom x86 becomes cheaper than custom PPC, let's stay with the PPC.)

>> http://www.freescale.com/files/32bit/doc/brochure/PWRARCHQIQSG.pdf (page 7)

> Am I reading the pdf correctly that the Eval board has 4 PCIe slots that
> are 2.0/3.0 Generation compatible?

No. While the T2080 chip has 4 PCIe controllers, the T2080RDB board offers only one PCIe x4 slot (not sure whether PCIe version 2 or 3). You can see this "PEX x 4 Slot" in the block diagram there:

http://www.freescale.com/files/graphic/block_diagram/products/processors/31877-T2080RDB-BD.jpg

...called "PCIE x4 Slot" on the photograph there (page 4):

http://www.freescale.com/files/32bit/doc/quick_start_guide/T2080RDBQS.pdf

One PCIe slot could run short when we consider the need for graphics and audio. So what I think may be better suited than the T2080RDB is the T4240RDB:

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=7780&start=34

It has the following main advantages relevant for general-purpose computing:

- PCIe slots: x8 + x4 (vs. x4)
- CPU: 1.67 GHz (vs. 1.53 GHz)
- RAM: 6 GiB in 3 DIMM slots (vs. 4 GiB in 1 SODIMM slot)
- price: 1445 USD (vs. 1499 USD)
- allegedly available in production volumes from Nexcom

However, it also has one main disadvantage relevant for general-purpose computing:

- SATA ports: 1 (vs. 2)

Using a SATA card in the PCIe x4 slot of the T4240RDB would more than make up for its SATA port shortage compared to the T2080RDB (but then no slot left for an audio card).

> If the MorphOS Dev. Team would port to this Eval board, couldn't this board
> be used as the basis for a brand new MorphOS compatible computer system
> at a price of not much more than $2,000 US Dollars, or less if you use existing
> case and PSU and other peripherals that you already own?

I think yes, it could. Both boards come with case and PSU already. The board form factors would complicate using a standard case anyway.

> 1.8GHz

1.67 and 1.53 GHz respectively.

> 8 cores

4 and 12 cores respectively. They are dual-threaded, though, so amount to 8 and 24 hardware threads respectively.

> 4 PCIe slots

2 and 1 PCIe slot(s) respectively.

> USB2.0/3.0

2.0 only.

> G4 class performance

Can you give any meaningful benchmark comparison between e6500 and e600 cores? At least Dhrystone benchmark results in 43% better per-clock performance of e6500 (single-thread) vs. e600. This would need a 2.4 (2.2) GHz e600 to match a 1.67 (1.53) GHz e6500 (single-thread).

> SAM 460, which unless something changed I am not aware of, it is a computer
> still in production

It should have re-entered production some weeks ago, according to information from ACube:

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=8931&start=143

> I would rather see a 32 bit Amiga compatible PPC Mac plus a NG MorphOS 64 bit
> non-PPC machine worked on simultaneously. With the G5 you could even have
> both on the same machine as a stop gap measure.

A non-PPC OS on the G5? ;-)

> we could maybe [...] use the money to reward people who make big contribution
> for MorphOS in general?

This sounds like the bounties that happen once in a while.

> off the self x64 is out of the question, so until custom x86 becomes cheaper
> than custom PPC, let's stay with the PPC

Why must it be custom x86(-64) instead of off-the-shelf x86(-64)? And if we're talking about custom boards, then why shouldn't x86(-64) be cheaper than PPC considering that x86(-64) chips usually have better price-performance ratio and all other costs are similar?

>Andreas_Wolf wrote:

>> off the self x64 is out of the question, so until custom x86 becomes cheaper
>> than custom PPC, let's stay with the PPC

>Why must it be custom x86(-64) instead of off-the-shelf x86(-64)?

To build support for the HW.
Normal off the self x86 motherboards have production lifecycle of three months or so. (it takes months to reverse engineer the support for a HW componet when we can not get much documentation)

Industrial x86 boards have longer availability with higher price etc...

>And if we're talking about custom boards, then why shouldn't x86(-64) be cheaper than PPC

Only recently there has been x86 SoC chips with PCIex4 etc, so the board design would most likely require more components.
And from what I have found, usable x86 is not any cheaper than a usable PPC.

(but I'm not an expert, just learning etc...)

>considering that x86(-64) chips usually have better price-performance ratio and all other costs are similar?

Performance wise, yes, in that way price-performance ratio is better for x86.
(I believe a common 1.33Ghz x64 SoC has twice the performance of similarly clocked e5500 PPC, +SIMD +GPGPU)

For me, it is more like what we need than trying to compete with mainstream with CPU power.
Knowing how much better MOS runs on 1Ghz G4 vs Linux on younger 1.8Ghz AMD Sempron machine, some 1.4Ghz PPC would be ideal for LOW END NG miggy.
((T10xx consuming some 5W, I imagine it could be usable also in some (semi)mobile device))


(I personally prefer big endian + PowerPC as long as most of our SW is native on that etc...)

[ Edited by KimmoK 06.06.2014 - 11:45 ]

@Andreas_Wolf

G5 is 64 bit, right? So before a possible move we could base the Next Gen MorphOS on that. Or maybe PPC by then becomes a resonable candidate again. Then having a working 64 bit MorphOS with all the goodies could make the job easier. Maybe.

I just want a modern AmigaOS. MorphOS is by far the closest. It doesn't have to be the fastest or the cheapest, but it needs to look forward and not back. If I know there is a future (which has been confirmed, but no details revealed nor signs that it's being worked on atm) I don't mind waiting.

>>> off the self x64 is out of the question, so until custom x86 becomes cheaper
>>> than custom PPC, let's stay with the PPC

>> Why must it be custom x86(-64) instead of off-the-shelf x86(-64)?

> Normal off the self x86 motherboards have production lifecycle of three months or so.
> [...] Industrial x86 boards have longer availability with higher price etc...

It's still cheaper to use industrial off-the-shelf x86 boards than custom x86 boards.

>> And if we're talking about custom boards, then why shouldn't x86(-64) be cheaper
>> than PPC

> Only recently there has been x86 SoC chips with PCIex4 etc, so the board design
> would most likely require more components.

Why more components? Intel offers multicore x86(-64) SoCs with 8 PCIe lanes, USB, SATA and GbE for 60 USD. That's about the same feature set as PPC SoCs (ignoring PPC's DIU, which isn't suited for real desktop use anyway). And Intel also offers multicore x86(-64) SoCs with 4 PCIe lanes, USB, SATA and 3D GPU for mere 30 USD. Here you'd only need a cheap separate Ethernet chip. Audio would be needed separately with all those solutions, including PPC SoCs.
From AMD you can get multicore x86(-64) SoCs with 8 PCIe lanes, USB, SATA, GbE (using one PCIe lane), HD audio and 3D GPU for 70 USD. That's really all-in-one.

> from what I have found, usable x86 is not any cheaper than a usable PPC.

Well, 30 USD for an Intel x86 SoC with 3D GPU (only lacking Ethernet compared to PPC) or 70 USD for an AMD x86 SoC with really everything needed. I think this would come out cheaper than PPC.

>>> I would rather see a 32 bit Amiga compatible PPC Mac plus a NG MorphOS 64 bit
>>> non-PPC machine worked on simultaneously. With the G5 you could even have
>>> both on the same machine as a stop gap measure.

>> A non-PPC OS on the G5? ;-)

> G5 is 64 bit, right?

Yes, and PPC :-)

> So before a possible move we could base the Next Gen MorphOS on that.

Yes, a 64-bit SMP OS would be doable on the G5, of course.

*double post*

@Andreas_Wolf

Is G4 64 bit?

> Is G4 64 bit?

No, G4 is 32-bit (with a 36-bit address bus).

@Andreas_Wolf

I thought so. So what do you mean by:

Quote:

> what do you mean by:
> Quote:

I mean that the G5 is not only a 64-bit CPU but also a PPC CPU :-)