Quote:

Yes, few years late when he was under heavy x5000 development or similar. This gap to task wait seems to be crucial since there seems to be no one else assigned to task.

Quote:

Let's face it, the situation is pathetic. The more I compare the X1000 to the X5000/20, the greater my appreciation for the former is.
Not a bad little system.
Not as powerful as a Quad G5, but still pretty cool.

However, since we aren't likely to see MorphOS on an X1000 or a Quad G5...and things seem pretty terminal for big endian PPCs...it is what it is.

Quote:

PA Semi has awful MIPS per Mhz ratio degrading its 1.8Ghz nominal rating, but when used dual core, with FPU instructions, it has fast mem and SATA and board is quite expandable. Its bi-endian arhitecture, 64-bit, G5 class overall in all but integer performance. Too bad its not that easy to add support for another mobo with no financial stimulation (MorphOS). Its quite enjoyable under Linux, but sigh, change to Southern Island cards created no 3D problem under Linux, rendering many installations unusable incl. X1000 Ubuntu Remix sold by AEON. In the end there is no system that completely use it, and CFE is ... built for routers and should stay there.

saga of wasted computer potential. So contrary to Amiga spirit of squeezing maximimum out of it.

Quote:

Best way out: Don't look back and welcome spectre/meltdown x64

Quote:

Yeah, but the system has a nice set of expansion slots backed up by more PCI-e channels.
And the processor is slightly better than a G4 (as well as being full 64 bit).

The more I'm learning about the X5000's expansion , the more troubled I am about it's limitations. Then again there's the difference in memory bandwidth.

> The more I'm learning about the X5000's expansion , the more troubled I am about
> it's limitations. Then again there's the difference in memory bandwidth.

"Then again" or rather "furthermore"? ;-)

Grammatically, you are right to point that out.
But actually my final comment was a bit recursive.

I am beginning to think the e5500 and e6500 cores may not be reaching anything close to their supposed memory bandwidth.

> I am beginning to think the e5500 and e6500 cores may not be reaching
> anything close to their supposed memory bandwidth.

If the cause of the memory performance issue is in the SoC, I think it's in the SoC's 'uncore' (memory controller) rather than its core.

Quote:

I'd like to see some bandwidth figures for the T10xx CPUs when interfaced with DDR4.
The DDR3 controller used with other e5500 and e6500 cored CPUs does seem to be the issue.

Quote:

I thought the memory bandwidth and speed for the X1000's PA6T, was one of its few advantages, when compared to the older Apple G4 & G5 computers? Or is the PA6T an exception for e6500 cored CPU's, when it comes to memory bandwidth?

It's probably that I'm not paying attention to this discussion, so I have made an incorrect assumption (again).

Edit: I probably should have changed "memory bandwidth" above, to read/write speed to RAM.

[ Edited by amigadave 27.01.2018 - 16:04 ]

Quote:

Yep, I think Andreas mentioned this before in response to a similar comment.
The PA6T in the X1000 was produced by PA Semi and does not use a Freescale core.
And yes, it's got better memory bandwidth then a G4 (not sure about the G5 though) as well as the e5500 and e6500 based Socs.

So, even with older DDR2 memory, the X1000's memory controller performs much better than the X5000.

>> is the PA6T an exception for e6500 cored CPU's [...]?

> I think Andreas mentioned this before in response to a similar comment.

Indeed, I did:
http://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=12137&start=94

> yes, it's got better memory bandwidth then a G4 (not sure about the G5 though)

Yes, also better than the G5.

http://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=11137&start=624

Addendum:

>> drivers for [...] onboard Ethernet are still not completed.

> The long story as told by Hyperion director Costel 'Cyborg' Mincea five days ago:
> http://translate.google.com/translate?sl=de&tl=en&u=http://www.os4welt.de/viewtopic.php?p=32421%23p32421
> (relevant part starts at 5th paragraph)

A-Eon's view:

"A-EON Technology Ltd contracted Costel Mincea of Hyperion Entertainment CVBA to [...] complete the X1000 driver project. The contract was signed at the Amiga 30th event in Peterborough, UK in August 2015. Payment was made for the contract in full within the following month and as I understand work has been ongoing steadfastly since [...] with breaks in the development [...]."
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=42440&forum=32#809150

>"...The contract was signed at the Amiga 30th event in Peterborough, UK in August 2015..."

Two and a half years and they have not produced a working network driver, and along with De Groot, Mincea is now running Hyperion?

Guess they are just to 'busy' to complete a contract that has already been paid for.

Doesn't bode well for future support of older Aeon products.

For that matter, do any of the Aeon AmigaOne boards have functioning on-board network drivers (if not under OS4 then Linux at least)?

And we are complaining about the delay in the release of MorphOS 3.10 to WHAT end again?
We don't pay for updates, and when our developers committed to projects in the past, they completed them.

[ Edited by Jim 01.03.2018 - 11:38 ]

> do any of the Aeon AmigaOne boards have functioning on-board network drivers
> (if not under OS4 then Linux at least)?

OS4: no
Linux: X1000/Nemo - yes, A1222/Tabor - probably yes, X5000/Cyrus - currently no

IMHO.
The PA6T (made by IBM) is under utilized on the X1000. The SB600 does most of the work and moves data to/fro the CPU through a single high speed pipeline. That includes the PCI network card.

Pretty much the X1k was a 'just-get-it-working' and kick it out the door type of project. I don't blame Trevor for the shortcomings. He is an AmigaNut, but not a hardware guy. If you think about it the X1k is not that much different than the A1SE/XE, which was supposed be a system only to get OS4 working on PPC. A development system, and it was sold as such. Unfortunately the A1XE somehow became the standard base design for all that followed. The XMOS chip has never been supported at a base level. Sad really, looks like the X5000 will follow suit, no innovation, nothing for the hardware enthusiasts, going to do everything in software. Pfffffft! Microsoft.

Of course you can't blame A-EON for everything when 90% of the squeaky wheels want a dirt cheap Amiga that most of them would not purchase anyway. And if they did they would want the OS for free.

The X1000 is a very good PPC Linux machine however, and CFE allows you to probe around the hardware and find some interesting undocumented features.

As my meaningless rants would only fall on deaf ears my X1k now sits in the closet along with other offerings to the Gods-of-technology.

Perhaps someday............. well, one can dream.

Quote:

Of additional interest:
from Jamie Krueger

If I read this correctly, even IF the driver being worked on currently was completed/released it would be (how to say this politely...) a handicapped driver?

#6

Quote:

Well the Tabor driver HAS to function (since there are no extra expansion slots for a wireless networking card). I suppose you could use a USB adapter, but that is sub-optimal.
The tabor driver may point the way to fixing the X5000 driver.

But the X1000 driver is going to be a bit different.

> The PA6T (made by IBM)

The PA6T was designed by P.A. Semi (bought by Apple a decade ago) and manufactured by Texas Instruments.

> is under utilized on the X1000.

Hardware-wise, I consider the Nemo board a good design by Varisys, despite the obscure xCORE/Xena stuff. All SerDes lanes of the PA6T have been put to good use and the GbE controller is routed to a PHY and jack.

> The SB600 does most of the work

You mean the allegedly under-utilized PA6T should do more of what the SB600 does? If so, what tasks would that be?

> and moves data to/fro the CPU through a single high speed pipeline.

These 4 lanes (1 GB/s total) should provide sufficient bandwidth for the SB600 and its controllers (SATA/PATA, USB, audio, PCI). The SB600 doesn‘t support a wider connection anyway.

> That includes the PCI network card.

If the OS can‘t use the PA6T‘s GbE, a supported PCIe NIC can be used, so that the network traffic doesn‘t have to go through the SB600.

> Pretty much the X1k was a 'just-get-it-working' and kick it out the door type of project.

That‘s certainly not my impression.

> the X1k is not that much different than the A1SE/XE

I think it‘s worlds apart, both design-wise and component-wise. The Eyetech AmigaOne (Mai Logic Teron) boards are bug-ridden evaluation boards with questionable components (ArticiaS northbridge and 82C686B southbridge). You can see that Varisys is a professional board design company whereas Mai Logic was not.

> which was supposed be a system only to get OS4 working on PPC.

The Eyetech AmigaOne boards were developed by Mai Logic under the Teron moniker as evaluation boards for their northbridge chipsets and for running Linux.

> A development system, and it was sold as such.

The AmigaOne SE (Teron CX), yes. The AmigaOne XE (Teron PX) was sold by Eyetech as end-user system, especially after OS4 became available.

> the A1XE somehow became the standard base design for all that followed.

There‘s no way that anything that followed, except the Micro-A1, was based on the AmigaOne XE.

> looks like the X5000 will follow suit

In my opinion, the biggest problem with the Cyrus Plus board‘s hardware design is its compatibility with both the P5020 and the P5040, which leaves the P5020 halfway under-utilized in terms of SerDes lanes and PCIe controllers.

http://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=12137&start=97

Moreover, the X5000 seems to suffer from a severe memory bandwidth problem, caused by either the SoC‘s memory controller or a firmware bug or mis-configuration.

http://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=11137&start=623

> Pfffffft! Microsoft.

Huh? :-)

> If I read this correctly, even IF the driver being worked on currently was completed/released
> it would be (how to say this politely...) a handicapped driver?

Only if it was ported from U-Boot instead of Linux, which I don‘t think Hyperion is doing or will do.

Quote:

Quote:

Source

I suppose I got the wrong idea from that posting then.

#6

> The tabor driver may point the way to fixing the X5000 driver.

I doubt it as the P1022's eTSEC/VeTSEC is less complex than the P5020's DPAA with dTSEC.

> I suppose I got the wrong idea from that posting then.

...or I‘m reading the word 'suggestion' wrong ;-)

IBM is printed on my PA6T, but maybe TI did manufacture it(?).

I'm not saying it was poorly designed, just that the design was unimaginative.
Why not another southbridge type chip with some different functions for example.

When I say the SE/XE are not that much different from the X1k I'm speaking of functionality not 10 years of technological advancement.

Does it have a PPC CPU?
Does it have card slots?
Does it have a southbridge?
Etc.

Well, anyway that's why I prefaced my post with IMHO, and that's worth about nothing.

Too bad the PA6T disappeared. Rotten luck.

> IBM is printed on my PA6T

That‘s interesting. I have seen many photographs of PA6T-1682M chips and all looked like this (with P.A. Semi logo and name):

http://www.amigaos.net/sites/default/files/styles/large/public/pasemi.jpg

Can you make a photo of yours and upload somewhere?

> Why not another southbridge type chip with some different functions for example.

Which other southbridge chip with what different functions that was available a decade ago are you imagining?

> When I say the SE/XE are not that much different from the X1k I'm speaking of functionality
> not 10 years of technological advancement.

Me neither. I was speaking about the quality of the board design and the quality of certain components. Nothing to do with technological advancement.

> Does it have a PPC CPU?

Yes, of course it does. Using a non-PPC CPU would have made it unsupported by OS4, which wouldn‘t have made much sense for a board specifically designed with running OS4 in mind, would it?

> Does it have card slots?

Yes, of course it does. The users are supposed to be able to expand the functionality of the computer, after all.

> Does it have a southbridge?

Yes, because the PA6T is just a minimal SoC and doesn’t provide the required I/O functionality except Ethernet, opposed to Cyrus‘ P5 or Tabor‘s P1. Another possibility would have been to use a dedicated chip for each I/O function, which was the original plan, but Varisys ultimately decided for an all-in-one package as the cheaper and technologically superior solution.

As I see it, there were no alternatives to incorporating a PPC CPU and card slots. And it made good sense to incorporate a southbridge chip like the SB600, given the restrictions of the PA6T.