Well, at 1.3GHz you can expect 10-15W power consumption. This is about the same as the MPC7447 at 1GHz on the Pegasos - but not including the Discovery II, VT8231 or your Radeon which obviously bump up the requirements!

At 800MHz, the specs say roughly 6W (this is about the same as the 600MHz G3 on the Pegasos).

At 600MHz you can expect somewhat proportionally less, we could reduce the core voltage a little, and the internal bus speed to a still reasonable level.
Negotiation is key here. We buy our chips from Freescale based on prices we negotiate from Freescale, not "retail" bundles. We no more buy at those prices at Spoerle than Dell buys retail Core 2 Duos.
Yes it definitely needs ethernet for networking, USB for peripheral support, and perhaps some controller for disk, and a few other minor elements (power management) - or a single discrete southbridge encompassing all of those, but there are some good solutions here which have a great benefit in software support (the Freescale reference designs use a ULi1575 which is already fully supported in Linux)

We are still waiting and will still be waiting past this Christmas for certain peripherals in the MPC5121e to have BSP drivers available for it. Alternatively we could write them ourselves.. it depends if you want low cost in terms of BOM, or low cost in terms of development.

The MPC8610 is an absolutely perfect little device, it's not too complex, nor is it lacking in features at the same time. It's an extremely effective update from the discrete G4 even at the same clock speed - one of the reasons we never released 7447A or 7448 processor cards is we could not justify the extra cost to consumers for the relatively minor performance boost they gave.

For "three times the price" you get probably 6x the performance (of the 5121e) at barely 3x the power consumption. You still gain the benefit of increased integration. It is about twice as fast as the 7447 at 1.3GHz - more if you are doing memory-intensive operations - and the same or lower power consumption than the discrete processor (far lower than the 7447+DiscoveryII).
You're right, but the difference would pay for the extra performance you get out of it.

_________________
Matt Sealey

Thanks for your explanation, I appreciate it.

The mpc5121e is not hardware coherent, because of the multi-port D-RAM controller it uses. The multi-port D-RAM controller was designed in, to increase memory bandwidth. The multi-port controller means that the D-RAM controller itself listens to incoming requests on multiple busses: the display controller, the graphics engine, the PowerPc core at the same time, and always accepts this request that has highest priority and can be serviced at the given time by the D-RAM. As the D-RAM has different banks, 8 of them most of the time, and each of these banks may be in a different state, the requests pending on the different busses are many times for different banks. Some of these requests hit open banks, and can advance, others hit closed banks, and must wait. The multi-port will give priority to these requests it can advance, while opening the banks for the requests it cannot service now. A single port would arbitrate between the requests before entering the D-RAM controller, and then the requests that hit closed banks would require the D-RAM controller to wait, without the possibility to do any useful work in that time.

Moving to a hardware cache coherent mechanism means that a multi-port D-RAM controller cannot be used, because all traffic must pass first on the snoop bus, this is the PowerPc coherent bus. Making the 5121e hardware coherent, would make a bottleneck of the PowerPc bus, and on top would mean a normal D-RAM controller needs to be used, not a multi-port any more, leading to considerable loss of bandwidth.

As the 5121e needs lots of bandwidth, mainly for its display controller and its graphics engine, the decision was taken to give priority to designing for high bandwidth, at the cost of requiring maintenance of coherency in software. For the 5200, the tradeoff was different. The 5200 does not have the bandwidth requirement of the 5121e, as there is no display controller, nor a graphics engine.

niceee :)

@henri234

Henri, as *the designer* of the memory controller in the 5121e we truly appreciate your post here in this thread. You and others at Freescale made the design decisions for the 5121e. That is your prerogative. The chip was designed as you intended for the purposes you sought to address -- that is all good. In these boundaries, the 5121e is a very interesting chip.

The caveat *in late 2006-early 2007* should have been that the 5121e was not designed to provide compatibility for software developed for the 5200B. As you know, if you are trying to do many things at the same time and you don't have the ability to keep track of everything with a central memory bus you will have issues. In this environment, you lose more together than you gain from the better performance of the subunits. The 5121e was not designed with a desktop or mobile consumer device in mind (though you could build a acceptable thin client out assuming it was to be supported by enough bandwidth and you were willing to invest in customizing the required software). Unfortunately, hundreds, if not thousands of developers have suffered because of this lack of disclosure. That is our grievance. This is much more of a 'people' issue vs. a technical one.

It seems we all agree now where the 5121e is best directed and where it is not. It is too bad we could not have worked closer together and much more complementary when it would have made a bigger difference.

Again, thank you for posting here.

@DrOctavius

The 5121e could be a great chip for your application. Please don't let us confuse you because of our bad experience with the some of the former Freescale folks that managed the marketing of the chip.

@all concerned

Now you know why the EFIKA2 never saw the light of day. Nevertheless, we still support this effort: openSUSE MPC5121eADS and will continue to do so for the foreseeable future.

R&B :-)

_________________
http://bbrv.blogspot.com

I stand corrected, but I wonder why the MPC8610 manages to do without it, I assume the OCeaN switch fabric and the much faster bus alleviates the need to squeeze every last drop out of the SDRAM bandwidth?

We appreciate your post, thanks for joining PowerDeveloper :)

_________________
Matt Sealey

And nobody told you early enough.
So you could actually prepare your software for non cache coherency, but it would be too slow? By the way, which software is that? Every driver?
Then, for what was it designed? Because it was born with fancy audio and video assistance.
So consumer products are completely ruled out... Perhaps it's only a matter of making different software for it, instead of using PPC Linux.
Genesi used to "own" a special operating system, way back. But that's not a guarantee of anything.

I suppose that "special" software can be made for this chip, in order to make it useful, to have sense.
But, if it takes too much time, it might be wasted effort. Competition is fierce now in the embedded market, a very quiet place only a couple of years ago. Waste some months developing something, of fighting a particular problem, and someone else announces a completely different solution that rules you out.
Big, big pity. Sounds like an official cancellation announcement. Thankyou again for your transparency.
Sure. In the end, the MPC5121e is a clever chip, worth success.

No OS alone will sell devices like that. Relevant applications is required. I can fully understand how the focus has been to make the OS itself reach a certain standard, a certain level of usability and usefulness. But for the OS to get some real commercial future, I think the focus must shift now when MorphOS 2.1 has got us there. Now the relevant applications (whatever they are) needs to be developed before the OS can be used as a central part of a business like this.

Completely agree. In my irrelevant (and perhaps wrong) opinion, MorphOS development is slow because of:

1.- Closed source: There's no choice about this, either you accept it, or go somewhere else.

2.- Lack of developers: Mostly due to above cause, also to cause below, but also because general lack of amigans, and the fact that amigans are so split and confronted.

3.- Lack of easily available hardware: This is where something like the LimePC could be a blessing.

4.- Lack of development tools: Actually, tools exist, but are not very user friendly. That leads to scare away casual, stupid developers (like me!), which can be considered a benefit from the sense of humour of the Team.

5.- Lack of documentation, or developer forums. The sight of "developer.morphosppc.com" is devastating. It was gently provided by Genesi, in all good faith, but its state of abandonness tells a lot about the lost relation between MorphOS Team and Genesi. Which were almost the same thing in the early, gloriuos Thendic days, when there were "big bucks" for giving away exotic computers to developers. What an exciting time was that.

6.- And perhaps the worst: Lack of users (?!). Perhaps due to cause three, but also because MorphOS is shockingly expensive (while it was free before!). Maybe cost is only relative to its sheer quality. But how can you compete with other solutions in the same category that are completely free?

All in all, dreaming is free. Amigans never lived in the real world.

jcmarcos, I strongly agree with those points. I have never understood what the MorphOS Team gained from keeping so much of their work to themselves. Surely their actions (inaction?) hurts them just as it hurts would-be developers, users, etc.

I won't pay €150 to be a mere spectator.

+1
pity, because the potential is great.

Hi Folks, please check the date. That is the cover of a 48 page document that Genesi developed for MTC/THTF. With the exception of MorphOS, they have done pretty much as we outlined in the marketing requirements document for 'Lime.' And, now MTC/THTF will try to create their own version of PowerDeveloper. As far as we can tell, they have never had an original idea of their own. Their business practices are shameful. The management has absolutely no sense of ethics. The manner in which they have treated a number of the developers involved is unforgivable. We have nothing but contempt for the CEO. The Executive management of Freescale has been fully appraised of the situation.

<break>

We have worked with Freescale on the 5121e extensively. After MTC/THTF disregarded our agreement, they went on to basically screw everyone else involved. That includes Freescale, because they have over-promoted the capabilities of the chip and continue to hurt its reputation in the marketplace with their exaggerated claims.

We have LimePC/CherryPal and LimeBook units -- they are not commercially viable. The LimePC is the CherryPal with a different logo. CherryPal does not have a cloud and they don't have money. MTC/THTF is trying to sell the LimeBook to users in China that don't know any better. It is a mess.

<break>

We will have a new product in the market in 2009 if things keep moving along as they are. This product will meet or exceed the expectations MTC/THTF and CherryPal have created. We will succeed where they have failed.

R&B :-)

_________________
http://bbrv.blogspot.com

It's not like MorphOS is comparable to Linux, where everything is centralized into the Kernel and you need access to that for whatever you would want to do. In MorphOS you can perfectly well develop any kind of application (and OS components for that matter, all the interfaces are there) without the need of having access to the Kernel sources.

MorphOS is mostly a closed source proprietary OS (with the exception of the desktop and some AROS components perhaps), and I don't see anything bad in that. Rather the opposite, it could bring economical opportunities, incentives and a future. People look at Linux which has a massive support from a huge community, and for some reason they translate this straight off to the Amiga world; "if only it was open source, the development pace would explode". But it doesn't work like that. It takes a momentum and size comparable to Linux, but few other open source OS's enjoys this. Look at AROS for instance. It has been i development for several years before MorphOS development started, but it's not even close when it comes to the MorphOS 2.1 standard (or OS4.1 standard for that matter), it's unusable in practice and it has about the same amount of users and developers (1-5 people or so), because those people are the same people. This despite being open source. And despite being x86 as well, for that matter. The closed source, proprietary alternatives got a lot longer, a lot quicker. Think about that for a minute.
Again, the MorphOS/Amiga model doesn't prevent anyone from developing.

If there are too few developers (I still think the Users/Developers ratio is better for the MorphOS platform than in many other OS's), I think it's because of lack of commercial opportunity. Which means that it's *in this area* (business development) that development should focus now.
"Every developer needs a desktop". Under any circumstances, 400MHz e300 based devices are no developer machines.

The Mac Mini however could turn out to be the blessing.
I think it's pretty much the same standard tools used everywhere.

But maybe it should become more easily accessible for beginners, I think so too. Guides, courses, a plug and play developer environment, etc, would be much appreciated. As well as a tolerant attitude towards beginner's stupid questions, so nobody is afraid of making a fool out of themselves.
All businesses stems from a demand from customers, and any product is an answer to those demands (or should be at least). It's not the other way around.

MorphOS is a product that lacks a demand from the masses, no-one needs it in its current shape and form (and with that I also mean the applications available, the hardware available, etc). Other than the couple of hundreds enthusiasts like you and me, practically no-one will buy into MorphOS since it has no purpose to them.

I think that had the price would been half of what they ask today, they would have made three times more sales to us enthusiasts (which would have meant more money as well). But even at a price of $0 it wouldn't make "outsiders" join in. Not without a purpose, not without the product filling a need and demand.
Indeed! :-)

This isn't the first time I hear stuff like this about Chinese companies, it almost seems like the Chinese way of doing business. Copy/steal good ideas and designs and make "their own" products out of it, and produce them at the incredible low cost that's only possible to do in China.

I think it's OK to assign production to China (like everyone does), but you shouldn't provide the entire context to them, only the components, and make sure you keep all essential IP and opportunities to yourself.
Will it be PPC based?

that's rubbish. It's impossible for someone to help, eg porting nvidia drivers (nouveau) to MorphOS, or adding eg. PS3/Cell support, or whatever related to the OS itself, unless signing an NDA, or being allowed in the "elite" MorphOS developer team. Or if not impossible, it's hugely more difficult.
And I somehow agree with them. It would. For one, it would mean that great developers might get interested and get involved in matters that for some reason are not in the MorphOS developers' agenda, like POSIX-compliance -which it isn't- or CPU/platform support, porting to another arch -even Intel/ARM- and who knows what else.
Look at Haiku also. It is not as old as MorphOS but it has far more developers than MorphOS ever did. And it has a brighter future, people are just drooling to install it. MorphOS could -if the developers wanted- have a similar impact. But they don't.
No they don't, but they don't really help either, do they?
I agree on the ratio, 50:1000 is better than 10k:1000M isn't it? and 10k is a rather pessimistic view of the number of total Linux developers...

A "blessing" a 4y old machine that the only way to get is via ebay?
I don't really know but is valgrind, kdevelop, cmake, eclipse (plus more) available for MorphOS? (Well, kdevelop mostl likely isn't because of lack of a Qt port, similar for eclipse I guess). It's not just the lack of tools that's the problem, it's the lack of incentive. Personally, I have little incentive to work on MorphOS if it means I have no say to the future of the OS. If MorphOS devs decide to dump it for some reason, or switch to ARM, or the porting stops at the MacMini, I will have no say in that, and basically, all my hard work will be for nothing, I'd rather stick with an OS that has a future (btw, future is not just 1-2 years from now. Linux has much more than that, even Haiku is greatly promising.
+1 on that
Most couldn't even if they wanted to, as it would mean buying old hardware from ebay AND software.