> which leaves us at the same point - not knowing how
> many PCIe lanes the processors support.

Have a look at what I just found:

http://lists.denx.de/pipermail/u-boot/2011-January/085252.html

The information therein bring me back to my idea of configurable SerDes lanes. Near the end you find a config table listing no less than 29 different pre-definitions of how to assign the 18 SerDes lanes. Definition '0x14' is the one with the highest amount of lanes dedicated to PCIe: three PCIe controllers used with 4 lanes for each of them, so 12 lanes overall for PCIe. Definition '0x10' assigns only 8 lanes overall to PCIe, but all of them to one single controller.
This makes me conclude that my idea was correct that the SerDes lanes on QorIQ are fully configurable like with PA6T, i.e. no hard-wiring involved on a processor level. And I also think that my example configuration of 15 lanes dedicated to PCIe alone is indeed possible, even if it's not part of the pre-definitions.
So, to answer the question: it's possible to assign up to 18 SerDes lanes to the PCIe controllers, which makes the processor support no less than 18 PCIe 2.0 lanes (but which would leave you with no Ethernet and no SATA).

That is a GREAT reference.
I haven't been able to get this kind of info out of Freescale directly.
Noted, bookmarked, and I will be returning to review this after this post.

Absolutely fantastic stuff. Thanks again Andreas.
If you find further info like this, please post a link.

> I wonder what problems have kept it from being released to the beta testers?

It seems Varisys had problems obtaining some of the parts:

http://www.amiga-news.de/en/news/AN-2011-03-00017-EN.html *

That's an interesting read overall as it seems to indicate that sourcing of the parts (including the PA6T) is done by Varisys, not by A-Eon, which effectively makes Varisys A-Eon's CPU supplier. On the other hand, that would mean that Varisys lied to you, and that Dickinson lied to us several times.

* Edit: now also to be found at http://www.a-eon.com/news.html

[ Edited by Andreas_Wolf on 2011/3/11 12:35 ]

Two pieces of info on the P50XX:

Available PCI Express controllers configurations are:

PCIe1 PCIe2 PCIe3 PCIe4
x4 x4 x1 x1
x1 x1
x8
x4 x4
x4 x4 x4
x4
x2 x4 x2
x2 x2
x2 x2 x2

Eval Board Status:

The P5020DS development system will be similar to P4080DS.
Uses 2.0GHz versions of the P5020 (early versions may use slower speeds).
Available Q2-2011
Cost: Approximately $4000.

Please contact Freescale sales office or our distributors to order the P5020DS.

> Two pieces of info on the P50XX:

Thanks. Appreciated.

Quote:

I think in that list they spared the configs x4 x1 x1 and x4 x2, which are part of the config table linked to before. But anyway, seems my 'dream config' of 15 lanes for PCIe (x8 x4 x2 x1) is not possible on this processor, unfortunately. So as suited configurations we have config x4 x4 x4 with 12 lanes for three ports (i.e. most lanes overall), config x8 with 8 lanes for one port (i.e. most lanes per port) and config x4 x4 x1 x1 with 10 lanes for four ports (i.e. most ports). Which of those would be your favourite on a board for our market?

Quote:

That's interesting as the announcement in June 2010 said 2.0 GHz initially and 2.2 GHz later.

Quote:

Okay, as you said.

Quote:

As expected.

x4 x4 x4 makes more sense than a single X8. An X4 PCIe slot has about the same bandwidth as an 8X AGP slot.
I've sent a second message inquiring if additional configurations are possible by re-assigning SerDes lanes.
Looks like Freescale is right on schedule, but you're right they did originally state "starting at 2.0". I too am surprised that they may release early evaluation boards at a slower speed.

Update: Freescale support just responded to the above mentioned inquiry.

>Only configurations from my previous response are implemented.

So X8 limits PCIe to one controller. 3 controllers at X4 looks like the best option. Now how do I scrape up four grand...

[ Edited by Jim on 2011/3/10 19:54 ]

Andreas_Wolf wrote:
which effectively makes Varisys A-Eon's CPU supplier. On the other hand, that would mean that Varisys lied to you, and that Dickinson lied to us several times.

Lied? A-EON paid for the PA6T CPU's which Varisys hold on our behalf.

TrevorD(ickinson)

Take it easy Trevor. We've always supported your project here and personally I think you picked a great partner for the design of the Nemo.
We just have been getting conflicting information as to the source of the PA6Ts that Varisys is holding for you.
This is not an easy to obtain component and even before you announced your association with Varisys, Paul Gentle had discouraged me from considering this same CPU (pointing out Freescale's QorIQ line as an alternate).

I don't think Andreas intended to insult you, A-eon, or Varsys. We just have had conflicting comments in an information vacuum.
Personally, I can't wait to see the version 2 boards released.

@Jim,

Andreas may, or may not have intended to insult Trevor, but I think 99.9% of the people that read his post would have been insulted if it had accused them of lying, instead of Trevor, so he has every reason to be pissed off.

As for your quote:

"If Apple has ceased production then the product will only be available till supplies run out. This is not a good situation for A-eon."

I don't see how this could impact A-EON much. How many X1000 systems do you think will ever be produced and for how long will they be produced before a different design using a newer CPU would replace it? I am quite sure that there are enough PA6T CPUs already produced and stocked for A-EON to complete their limited production run(s). The one problem of using a discontinued CPU is that as the supply runs low, the price of each CPU will go up, so if there is a steady demand for this chip in military applications to the point of depleting the available supply, it could make it very expensive to obtain, or possibly impossible.

At the estimated selling price of the X1000, I don't think anyone (even A-EON & Hyperion) thinks that it will ever sell more than a few hundred boards. It is possible that A-EON already has that many PA6T's being held for them by Varisys, as they were probably required to purchase a minimum number of CPUs, just to get any of them.

[ Edited by amigadave on 2011/3/10 13:50 ]

> Lied?

*Would* mean that you lied if Varisys is your CPU supplier. Let me explain: At VCF you told that you found a supplier for the CPU and this supplier then suggested Varisys as a design company to you. This story wouldn't make sense if Varisys was also the supplier, now would it?

> A-EON paid for the PA6T CPU's which Varisys hold on our behalf.

Thanks for clarification. So Varisys is in fact *not* A-Eon's CPU supplier, which means you did not lie.

> I think 99.9% of the people that read his post would
> have been insulted if it had accused them of lying,
> instead of Trevor, so he has every reason to be pissed off.

My statement was meant in a conditional way, with the premise that Varisys is A-Eon's CPU supplier. The discussion was about whether this premise was true or not, and I outlined the conclusion that would have to be drawn if that premise was taken as truth. So no, I didn't intend to accuse him of lying.

Quote:

Actually, the interesting thing is, we still don't know where the PA6Ts came from.
It was never my intention to suggest that A-eon didn't have a supply, or that they hadn't already been paid for. It's just that further supplies seemed unlikely and as of a year ago Varisys was recommending other processors.
They even turned down offers to purchase their own PA6T based board the VM31.
So if Varisys help procure these CPUs, how long ago did that occur (considering they're just getting ready to produce the beta testing boards).
Andreas statement is directly linked to contradictory statements that have been made about A-eon's source for their CPUs.
It wasn't personal and what he stated is true.

Someone has provided false/incorrect information. If the term lying was offensive, it wasn't meant to be.

Quote:

A company that supplies CPUs "CAN" suggest to the buyer that they also be used as the design company. It makes perfect "sense" and does not make Trevor a liar, unless you have other proof that he specifically stated that Varisys was "NOT" the supplier of the CPUs.

I hope that is not the only example you are using to infer that Trevor might be lying. Perhaps it is not a big deal to you who you infer might be lying, or not, but to some people, it is a very big deal.

> A company that supplies CPUs "CAN" suggest to the
> buyer that they also be used as the design company.

Go listen to his VCF speech, it's on Youtube. He said it were different companies, not that the CPU supplier suggested itself as design company.

> I hope that is not the only example you are
> using to infer that Trevor might be lying.

You're too late. I already stated that with the recent information given by him I conclude that he did not lie. It is in my posting you just quoted. But to answer your remark: Let's pretend for a moment that Varisys is also A-Eon's CPU supplier (which we now know it is not). Do you think the whole NDA story that went until December would have made any sense from June on, when Varisys was officially revealed as A-Eon's partner?

> Perhaps it is not a big deal to you who you infer might be
> lying, or not, but to some people, it is a very big deal.

I say the key words again: condition, premise, conclusion.

This does bring up a valid point Dave. Who claimed that Varisys was A-eon's CPU supplier?
It doesn't appear in Trevor's statements.He only claims to have paid for them and that Varisys is "holding' them for A-eon's production run.
So, Andreas would be correct in claiming that he has to conclude that Trevor has been truthful.

Even if we never find out where they came from, we do know it wasn't Varisys.

> Who claimed that Varisys was A-eon's CPU supplier?

We recently discussed three statements that could be interpreted in that way:

1. https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7183&forum=3&post_id=82084#82084
2. https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7183&forum=3&post_id=82208#82208
3. https://morph.zone/modules/newbb_plus/viewtopic.php?topic_id=7183&forum=3&post_id=82293#82293

But then, all of these statements were made before Trevor Dickinson gave his clarification right here in this thread.

Very Concise.
We may not know where they came from, but we know where they didn't come from.
Again, I hope Trevor was not offended as I'm fairly certain that both of us (actually almost everyone on MorphZone) have been quite positive about his project.

Mathias 'Corto' Parnaudeau's statement "the Amiga X1000, designed by the company Varisys, former PA Semi client and holder of a stock of processors to be resurrected" suddenly makes more sense. This is not a rough translation. It literary means holder of a stock not supplier.

So statements made to me by Varisys' staff and statements made by Trevor Dickinson both here today and prior are true.

Well, we're still in the dark about the supplier, but I feel much better about this apparent ambiguity as there have been no false statements.

By summer we should have our first objective reviews of the X1000 and the P5020 evaluation board. Good news all around. Thanks Andreas.

[ Edited by Jim on 2011/3/11 3:16 ]

[ Edited by Jim on 2011/3/11 3:16 ]

My point is that before anyone writes in a public forum a conclusion that states that someone else has lied several times, if condition "X" is true, should be more careful, or verify the condition before writing such a statement, or perhaps choose to write the statement in a different way without the words liar, lying, etc.

Being labeled as a liar can be very damaging to future marketing of a commercial product, as well as being very offensive personally. I understand and agree with Trevor's angry response to the original statement by Andreas. That is all.

> My point is that before anyone writes in a public forum a conclusion
> that states that someone else has lied several times, if condition
> "X" is true, should be more careful, or verify the condition before
> writing such a statement, or perhaps choose to write the statement
> in a different way without the words liar, lying, etc.

I don't think your point makes sense. Maybe you should look up the concept of conditions and what they're used for. And I stand by what I said: If Varisys was A-Eon's CPU supplier, Trevor Dickinson would have lied to us. You see how the conclusion to become true requires the premise to be true beforehand? For this statement to make I don't need to verify if the premise is true or not.

> Being labeled as a liar

Nobody labelled Dickinson a liar here.

> I understand and agree with Trevor's angry
> response to the original statement by Andreas.

I don't see anything angry in his response.

> Mathias 'Corto' Parnaudeau's statement "the Amiga X1000, designed
> by the company Varisys, former PA Semi client and holder of a stock
> of processors to be resurrected" suddenly makes more sense.

Indeed. I must say though that another statement ("we were able to secure a supply of CPUs through Varisys") is still a bit of a mystery to me. Do you have an idea what that could mean specifically and how it would fit the story as we know it by now?

> By summer we should have our first objective reviews
> of [...] the P5020 evaluation board.

Who's going to do that review? :-)

Odd news on the P5020 evaluation board. I inquired about the default SerDes configuration and got this info.

Factory SerDes configuration for the P5020DS board is 0x36:
PCIe1 x2
PCIe3 x2
SGMII EC1
SGMII EC2
SGMII EC3
SGMII EC4
Debug x2
XAUI 10GEC
SATA1
SATA2

Of course it is possible to modify the RCW to use another SerDes configuration.

So as shipped, the PCIe lanes are relatively limited. Luckily we already know that we can reconfigure for one X8 PCIe or three X4 PCIe configurations.

As usual Andreas, you don't see anything you don't agree with.

> As usual Andreas, you don't see anything you don't agree with.

And you continue failing to grasp some simple concept. Don't now if that's usual with you, though.

> Factory SerDes configuration for the P5020DS board is 0x36

Alright, so overall only 16 of 18 available lanes in use per factory default.

> Of course it is possible to modify the RCW to use another SerDes configuration.

Am I reading this right that the SerDes lanes are not fixed on the P5020DS board and can be re-assigned? So far I thought that a board designer had to decide for one of the available configurations and hardwire that one. Reconfigurability on a fabricated board would be even nicer, of course.

Quote:

That part surprised me as well, but as this board is not a commercial product but rather a form that allows evaluation of the P5020, maybe they felt the need to include this option.
I was more surprised by the lack of an X4 configured PCIe slot in the default setup.
Running a PCIe video card at X2 would be possible but painful.

[ Edited by Jim on 2011/3/12 1:15 ]