Are you sure that a single core is clocked at 1.25 GHZ? I thought that the frontsizebus is clocked at this speed, but not the cores, which are still at 2.5 GHZ each.

Compared with the Peg2 and the G4, the Peg2 has 133 MHZ FSB and 1 GHZ CPU, while the OSW would have 1.25 GHZ FSB and 2.5 GHZ CPU (wich is because of the two cores kind of 2x 2.5 GHZ per CPU), so MOS would run on a single core still at 2.5 GHZ.

At least I understand in this way the information about the G5 which can be found on the Apple website:

http://www.apple.com/g5processor/architecture.html

Maybe users have to connect more devices than one CDROM and one HD via USB. Would be an option, if it is USB2.0... and if MOS will have USB2.0 support until then

After what Neko wrote it seems that the PPC used in OSW really is a 1,25 GHz. Apple usually uses 1/2 CPU clock for FSB. If Genesi takes the same approach this would mean FSB 625 MHz for the OSW.
A Serial ata connection is way faster and much cheaper. Also with at least two HD drives you could have the option of doing a "light" raid setup.

But we are going Off Topic maybe we take this thread at Morphzone for further discussion: http://www.morphzone.org/modules/newbb_ ... 8&forum=11 or open one here...

C'ya

The southbridge has 4 SATA-II ports. The OSW is a design for a "complete system" more than it is solely for the motherboard. Like the ODW it will have an HD and DVD but the rest is not up to us..
Some clarifications

* the diagram shows REQUIRED connections for system design, not ALL connections. It also shows possibility for legacy PCI which still may or MAY NOT be in the final design. The chipset supports 4 SATA ports, we will use two of them at least for the complete system. Do not use the block diagram previously posted as a hard-and-fast end specification, where if you only see two lines then that means two lines. Usually it means two lines because it is more complicated and redundant to draw 10 of them on a block diagram

* CPU speed is not chosen, but IBM offer all speeds from 1.0GHz up to 2.5GHz as you well know. There needs to be a balance between processor speed and the prospect of a $350-cost liquid cooling solution (as on the PowerMac).

Please remember we are targetting this device at $1500 sales point, which is $500 less than the cheapest PowerMac G5 with 2 cores. OSW will be faster than the cheapest PowerMac by some degree because it has 4 cores. It will also be cheaper than the NEXT cheapest PowerMac and is guaranteed to have greater aggregate performance. The device it won't be competing with is the $3500 baseline 2.5GHz quad G5 PowerMac which to be honest you get what you pay for - even if we think that is a little too much to pay.

Again, it is all about balance, and remember this is a multicore, multithreaded world, and this system is being designed with IBM in collaboration for a lot more than "running MorphOS".

We at Genesi do not stoop low for the "numbers game" and have always valued low temperature and low noise operation over some meaningless display of raw, unadulterated power (and heat output), for our entire operation.

The goal is to produce a USEFUL workstation and server platform (in 1U-capable format which poses some restrictions), and not one which simply complies to a mathematical subtraction game to see who wins.

This is computing, not Top Trumps.

_________________
Matt Sealey

OK, thanks for taking time for this extensive reply. I really appreciate that. Going to waiting mode for final OSW specs....

>The chipset supports 4 SATA ports, we will use two of them at least for the complete system.

So, the third one is reserved for another HD and the forth could be used for external connection or a 3 disk RAID5?

>Usually it means two lines because it is more complicated and redundant to draw 10 of them on a block diagram

But on the other hand USB has 4 in that diagram.

>There needs to be a balance between processor speed and the prospect of a $350-cost liquid cooling solution (as on the PowerMac).

You chose the BTX formfactor, so please be aware of the recent beneficial developments in the x86 world.

>OSW will be faster than the cheapest PowerMac by some degree because it has 4 cores.

While I am not against using a power-efficient low voltage chip, single threaded performance is always important (that's why IBM goes to 4+ GHz with their Power6) and the G5 was really designed for high clockspeeds with advanced power management features to keep the average power consumption down. Scaling for two CPUs is easy, scaling to four cores is not that trivial anymore in a workstation environment as can be seen with the Quad PowerMac.

>and have always valued low temperature and low noise operation over some meaningless display of raw, unadulterated power (and heat output), for our entire operation.

...

>The goal is to produce a USEFUL workstation and server platform (in 1U-capable format which poses some restrictions)

Sun puts two Dual-Core 2.6 GHz Opterons into 1U X4100. Clever engineering!

>not one which simply complies to a mathematical subtraction game to see who wins.

>This is computing

And computing is pure mathematics.

It's 4 SATA-II ports, you can do what you like with it.
Different strokes for different folks; let's define this as *at least* 4 ports. The block diagram is a MINIMUM specification.


Well, I would hope that is not needed (consider that those solutions are $150 per processor and require considerable chassis area to install)
You think about single applications scaling to all 4 processors, but you forgot about 4 applications spreading over 4 cores. You generally run more than one task at once on a multitasking operating system.

_________________
Matt Sealey

>>You chose the BTX formfactor, so please be aware of the recent beneficial developments in the x86 world.

>Well, I would hope that is not needed (consider that those solutions are $150 per processor and require considerable chassis area to install)

Rumour has it these units might be offered for about $50.
IBM also uses a 12 cm system fan with airduct for their 970MP tower systems.

So does it mean MorphOS will run on the Pegasos III?
If so i'm in!

Yes, I know that, and I also know that there are maybe a lot more Pegasos / ODW owners who use only Linux on their machines and never run MorphOS.

But at least _I_ will not buy such a OSW if MorphOS will not run on it with a considerably higher speed than on a G4 today. Because I dont see a reason in spending $1500 or $799 if I only get about the same speed as now with my Peg2 and MOS.

This would mean that MOS has to run on the OSW _and_ either one G5 core of the OSW must be clocked at at least 1.7 GHZ or MOS gets multicore ability. Only then I would buy a OSW, and I would buy it definitely on this conditions.

Thats right. But if I run 4 apps which all together do not use 100% of one single core, why should it be faster to run these four apps on four cores? There will only be a speed gain if all apps I run at the same time would use more than 100% of a single core to run smoothly. But if I e.g. convert a a DVD to DivX I need either an application which is multicore-able or higher single core speed.

Exept I consider to convert 4 DVDs at the same time... but then I would need 4 DVD drives.

Only examples.

It depends on the ratio of the CPU they would use on a single core.

If you have 4 apps which use 10% each of a single core they will not run faster. They will just use 10% of a seperate core. But then they also do not need a single 2.5GHz core to run on anyway if they will not fully utilise it.

If you have 4 apps which require 33% each of a single core, they will run faster as they can all run at their full potential on a core each, instead of being "truncated" to 25% each. Therefore 4 cores will be faster.

Either way there is no disadvantage to a multicore setup. Apps should multithread; this has been the design philosophy for computing for a very long time, SMP has been around for a very long time, it not only increases performance on SMP but it increases interactivity on single-processor machines.

On AmigaOS for years it has been the recommendation to put your user input thread seperate so that you can collect it while your main task is busy.

Take for example a networked app - a good one is most web browsers, which spawn a process for networking and especially DNS lookups because gethostbyname() is a blocking call. You don't want your entire browser to freeze when it looks up a hostname. Also, Networking should not block layout, and the same way, layout should not cause the HTTP stream to pause being received. Image decoding should not freeze HTML parsing. And most of all NONE of these tasks should stop you from clicking the "STOP" button.

On MorphOS, look at Voyager and IBrowse for examples.

Basically there are two kinds of task in the world, the compute-bound and the io-bound. The vast, vast majority of applications that people run are predominantly io-bound, in that they access disk, wait for mice, handle network, hand data to graphics cards, etc. and your performance is limited by that - a DVD player is limited by how fast it gets data from the DVD drive and then by how much data it gets (up to 15MBit/s usually, it only has to decode 30 frames per second), the rest of the time it is idle.

_________________
Matt Sealey

I think that maybe you missed this due to the page addition right after the post; is there any place where I can find this information? Any help in this regard would be highly appreciated.

Even if MorphOS is not multi-core aware, thanks to Xen you should be able to run at least 4 MorphOS instances in parallel on an OSW with no slowdowns at all. You'd switch between them like you switch users on Mac OS X. One MorphOS could be dedicated to Digital Entertainment, the Other could be your work environment, another one could run a game or run a webserver, ... They all run at the same time. You could even have Linux and MorphOS running at the same time. To your network it would like like there were two different machines connected to the network.

The possibilities are endless: you can make distributed apps using LAM/MPI or PVM. You could use TCP/IP sockets to implement communication between client and server module of an app you're writing: server could run in ppc linux, while frontend runs in MorphOS, both on the same machine!

Imaging being able to ssh from MorphOS to Linux running on the same machine! Power of MorphOS combined with power of Linux. What more can you want?

It gets really cool when OpenSolaris/PPC gets to beta stage. Solaris, MorphOS and Linux all running on the same machine. Remember, Genesi has access to the source code and full register-level documentation for state of the art 3D hardware. MorphOS, being very lightweight, could be turned into a perfect game platform. Client stuff runs on MorphOS, server could run under Linux or Solaris on the same machine. Sun has already ported Java to the platform and there's IBM java as well. Have a look here:

http://www.gridtoday.com/04/0329/102930.html

Lots of opportunities for the MorphOS people to make MorphOS even better.

@Pieter : that looks really cool !
Hope we'll have all that stuff soon
We'll sure make nice apps with that

Are the first Pegasos III/OSW for developers still planned for June 2006 ?