I find interesting that has PCI Express (not 16x and just 8x but well, it's for embedded) and PCI.
The bad things is the lack of ethernet, USB and SATA. I think that will use all the available interfaces if wanting to make a full desktop computer, so no expansion possibilities.
This is clearly designed for the embedded market. What about something more powerful?
MPC8641D seems more interesting as it have four 10/100/1000 ethernet ports (with TCP and UPD checksum acceleration, QoS support, and packet header manipulation), 64 bit (72b with ECC) DDR2 memory controllers, Dual 8-lane PCI Express ports, 4-lane serial RapidIO port and two a600 (G4-like) cores.
Both aren't really chips designed for desktop but the later one could be used for different systems and be an interesting substitute of Pegasos II for MorphOS users.
The only problem can be the lack of stuff like USB and SATA, I'm not sure what stuff can be connected to RapidIO for example.
What is perfectly clear is that PowerPC abandoned the desktop market and who knows about the server one if IBM and the Power group don't help to make it more interesting compared to competitors (x64 and SPARC). X86 is still there for stay for a long time and learn to like it, ARM and PowerPC will be really scared when X86 will be able to compete with them in terms of power efficiency (that will can be possible with new manufacturing process and future optimized SoC designs by AMD and nokia).
Please keep in mind about an interesting workstation machine for development and MorphOS, find a clear market for MorphOS 2.0+ and pay the developers if they release a product with the requirements for your business needs. MorphOS has a lot potential, now it needs to be guided in a right and clear way into the market and providing very strong killeraps into other OSes in the embedded world like Linux.
1) the performance difference on most graphics cards between 8x and 16x lanes is absolutely minimal - in the single percentage points. So, you may lose 4% or 6% of your framerate. Plenty of x86 boards come with only an 8x slot for graphics especially if they are SLI or CrossFire, and nobody notices the minor performance difference.
2) none of those things are bad. It just means, you need support chipsets. A southbridge perhaps. This is no different to the Pegasos. It keeps the MPC8610 cost-effective. It doesn't have USB, SATA or Ethernet because it is meant as a processor (think, more in line with AMD building in memory controllers to their processors, rather than the northbridges).
3) The MPC8641D is not as interesting because it is far more expensive and also.. MorphOS doesn't support dual core processors anyway. What would you do with 4 gigabit ethernet ports in MorphOS? As you said, what would you even do with RapidIO even in another OS such as Linux? It's hard to justify a board costing hundreds of dollars for a 4-port gigabit router :)
All Freescale processors are designed for the embedded market. The MPC8641D is meant for different embedded market (high performance communications processor) than the MPC8610 (high performance imaging systems and kiosk). Right now, the MPC8610 makes sense as it DOES fit with MorphOS, it makes sense from the lead of the Efika (keep it simple) and it makes sense for cost.
There is probably scope for an MPC8641D board, *after* we have gone through the motions with the MPC8610 and created a successful product. All the same AltiVec work we did on the G4 and hope to do on the MPC8610 will work, with the advantage of a dual core processor and slightly higher integration.
But, the expense, compared to an MPC8610/Southbridge combination, may not justify the solution. This is the key point where most products fail in any market.
No ethernet, no SATA and no USB? Then developing this board is a total waste of money. It is absolutely superfluous. No serious person would ever buy a board without only one of these 3 connectors.
@Anonymous: it's the chip (the 8610) that does not have support for USB, Ethernet and SATA ONBOARD. You can add it with additional components (Just like in a normal PC).
How does this proccessor compares to an equivalent 1Ghz x86 (lets say an Athlon or Pentium 3)?
I have bought an Efika which I use as a mini-server. If only I could load it with more memory it could be a much more interesting and useful machine. Having said that I think Genesi is aiming the wrong target with their latest offering.
Okay, only says smp for the kernel so who knows how well MorphOS works with multiple cpus & cores.
The 8641D is too expensive and would only attract a few, hard core users. Not worth it. Better to stick with low to mid end markets instead. That is where Genesi can attract the new users.
What was actually needed for Efika system was a processor with 550-800 Mhz speed at a good price. Maybe e500 core? But that means missing out on Altivec because that is only available on e600 cores.
@angus The 8610 has top speed of 1.33 Ghz w/Altivec. I believe these are the chips they would use. I think it'd be comparable to Athlon 2.0 Ghz CPU but can't say for certain.
Efika form factor would be a mistake. For upper end systems like the 8610, it is better to go with Mini ITX or Micro ATX form factors. Efika size is only good for lower processor speed, fully integrated systems ( sub entry market ).
This actually makes sense to me; 1) Efika for low end ( 400 - 600 Mhz ) 2) Mini ITX for mid end ( 600 - 850 Mhz ) 3) Micro ATX for high end ( 850+ Mhz )
PowerPC is very limiting. For instance, in regards to Freescale.
They only provide Altivec on e600 cores. SMP is only possible with e600 cores. When you start going over 400 Mhz, costs start increasing rapidly for every 200 Mhz., No L2 cache on lower end CPUs, Very few new CPU designs coming out, ( very ) high prices for mid and upper end CPUs, etc., etc. Unfortunately, MorphOS is made for PowerPCs so Genesi will have to stick with them and do the best they can in this situation.
What would have been very cool is making a motherboard with dual 5200b CPUs ( 2 x 400 Mhz for Linux or MorphOS; MOS kernel is SMP ). Maybe 5200b are SMP compatible because of the MPC603e core they're based on? ie: Could be possible to do but maybe hard?
Tundra allowed SMP of MPC603e cores. http://www.tundra.com/product_detail.aspx?id=1309
PS It sucks that SMP is only available for higher end CPUS on PowerPC. Intel started offering SMP back with Pentium 1 MMX ( or non MMX? ) CPUs ( Intel HX chipset motherboard ). And I've owned Intel SMP Pentium II & III systems ( dual CPUs ). Very sad to see that you can only do this with PowerPC on their upper end, pricey CPUs.
It has no USB/SATA/Ethernet and Sound (I suspect).
It is not a problem for me if it is cheap. I can buy whatever I like in the form of Expansion cards. I like this option.
On the graphics side it seems OK. Hopefully the gfx support is open source.
Actually this is the motherboard I have been dreaming for. Haiku and Syllable can have a viable platform. Even Gnu-Darwin with some tweaking/drivers can boot on this machine.
Great Job/Choice/Product. I am waiting for price quotes.
Better. It's more in line with something like Intel's A100 and A110 processors (for reference they are sub-GHz ~6W chips with ~400MHz FSB and a companion northbridge, based on the Pentium M design). The MPC8610 should be available at up to 1.3GHz, though.
An aside;
If you take a 1GHz G4 made this year and put it against a 1GHz Pentium III, you'll notice the Pentium taking up 4 times the power and operating far slower.
This is because there is an 8 year gap between the technologies, 130nm Intel chip with 256k L2 cache and 133MHz front-side bus vs. a 90nm G4 with 2MB L2 cache and 200MHz front-side bus, who will win? :D
So, it's never very good to compare old and new processors. There are lots of ~1.3GHz processors on the market right now from Intel, Via, AMD and everyone else, so you can compare with those. But, Intel A100 and A110 compare best with featureset, except the built-in memory controller and PCI Express root complex :)
Back to what you said;
The MPC8610 is very power efficient, probably a little more so than even Freescale's website suggests (they are being very conservative for once). Performance should be good; far, far better than Via's CoreFusion platform at the same clock speeds, perhaps with not quite as good power consumption, but still in the levels you could cool passively, and effectivfely.
angus: I can't see the full URL for what you posted here, but I am assuming it points to a Mini-ITX board. Yes, something like that is what we would go for. There are many advantages to ITX now that simply did not exist when we had the Pegasos - the market was immature and Via have a propensity to ditch form factors and build in proprietary Via-only features (like the Nano ITX form factor, which has been thrown away after a distinct lack of boards or cases, in favor of Pico ITX, and has a custom power connector, and won't fit in ANY standard cases). AMD DTX form factor builds in ITX compatibility, and Intel have embraced it for their reference boards. Dell are even shipping ITX PCs. And, importantly, Freescale concentrates on ITX for all it's multifunction reference designs now. This has all happened *since* Pegasos was discontinued, it all makes it easier to justify as a product design decision, where there was none but "geek chic" before.
To allay any of the fears here, the BOARD will have USB, SATA and Ethernet, you could not live without these on a desktop board and we know this of course. They are simply not in the SoC package like the MPC5200B - however this means we can drop in new chips as technology advances, making it a far more futureproof design. Which chips are chosen is still being pondered over, but we have a remit to choose the chipsets which offer the best and most mature Linux drivers and importantly open documentation, if that is at all possible.
Today we have fedora 9-beta PPC running on both the Efika and Pegasos II. Some of us probably want 256-512MB RAM embedded platforms like the EFIKA and maybe the MPC8610 and ATI 2400 HD Pro. Maybe we'll get that before Easter?? ~K
SMP is not just for the low end. Take a look at the MPC8572E (dual core, 1.5GHz e500 - with SPE SIMD, so not entirely lacking against AltiVec, just not as powerful). You may also want to look at the MPC5510 which is a dual-core e200 (one of them is used as a VLE coprocessor).
The MPC5121E has the AXE coprocessor which is effectively an on-core 200MHz RISC DSP - with a surprisingly comprehensive instruction set. There are lots of definitions of multi-core, SMP or otherwise, and it's not restricted to simply "two of the same processor". You might well consider the announced Intel and AMD "CPU with GPU" chips multi-core as they will include a graphics processor capable of outstripping the main processor cores (all 4 of them, potentially!) in floating point operations.
We are definitely looking more towards ITX than custom form factors, so don't worry about that. There is very little reason with the high integration to produce a board any bigger.
To clarify a misunderstanding, the MorphOS kernel does NOT support SMP. What you run is what Ralph calls "ABox" which has absolutely no fundamentally required support for multiprocessing. Since it is based on the exec.library concept from the old 1980's AmigaOS, the entire OS relies on writing into memory for message passing - something which is quite safe on a single processor, but incurs a *full cache flush* when scheduling the next processor. This design has already been attempted; PowerUP does it every time you switch from m68k to PPC, but remember m68k didn't have 2MB of L2 cache to manage and neither did the 604e!
You would throw all the performance of both processors in the toilet undertaking this multiprocessing model inside ABox.
The underlying "Quark" microkernel may support SMP, but until this functionality can be exposed (perhaps through abstracted DSP libraries or so) you will not be able to use it, and certainly not be able to use it as a dual-core e600 running seperate MorphOS apps on seperate processors, or multithreading across two cores.
Lack of USB/Ethernet/Sata/Sound will accelerate time to market, bring down the cost and make Linux compatibility better. I do not think that they must be included if enough slots are present. I also have no problem with PCI. The only requirements I think are Serial, PS/2 and Floppy. Another improvement from the previous Efika would be expandable RAM in the form of slots, not soldered on board. DDR2 or DDR400 makes no difference to me. Maybe I am a bit conservative.
Genesis has the last word. Looking forward to the new Efika.
I want to first make some corrections to your statements.
You said, "The MPC8610 should be available at up to 1.3GHz, though."
Freescale site says, "Up to 1333 MHz performance and 3060 MIPS"
1333 Mhz = 1.33 Ghz like I previously said :-) & NOT 1.3 Ghz; Okay, I admit you were pretty close.
You said, "There are lots of definitions of multi-core, SMP or otherwise, and it's not restricted to simply 'two of the same processor'."
Sorry, but that is not entirely true; partly right. That is correct for multi-core but not for SMP. SMP stands for Symmetric Multi Processors. Symmetric meaning the same type & speed ( identical ).
From Wikipedia, "Symmetric multiprocessing, or SMP, is a multiprocessor computer architecture where two or more identical processors are connected to a single shared main memory."
Unfortunately multi-core can mean something different like you said. For instance, the 5121e is a triple core ( CPU + AXE + PowerVR ) - instead of just refering to # of CPU cores which can cause confusion.
You said, "To clarify a misunderstanding, the MorphOS kernel does NOT support SMP."
From MorphOS Developer Connection site: "Basic kernel information MorphOS/Quark has the following basic design goals: ..... Symmetrical multi processing (SMP) ....."
Seems like a contradiction to what you said. From my understanding, MorphOS has the basic goal of being SMP capable according to MorphOS site ( or at the very least the kernel is designed for SMP ). I wonder why MorphOS developers would make that statement if it wasn't true? I can't comment further on this issue without using/testing MorphOS on an SMP machine myself. All I can say is that either the developers are misleading the public or you're incorrect.
I still believe that Micro ATX is the way to go for higher / upper end products but if you believe Mini ITX will work out then go for it & hopefully things work out for the best. Mini ITX is good for mid range products and Micro ATX for upper end. Just my thoughts. On upper end systems, I like to have 2 to 3 PCI slots, plus AGP or PCI-E slot, USB 2.0, SATA, PATA, 2 DIMM slots, Ethernet and Audio; ie: I like the works on upper end systems as I'm sure others do too. On low & mid range systems I like them to be more on the compact size.
I actually believe doing dual processor or dual CPU core system is not worth it on PowerPC because of the cost and little benefit gained. A dual CPU 5200 system would have been cool though but I don't believe it is technically possible. So, it seems that the better course is to go for the fastest speed single CPU core instead.
PS Using the Efika 8610 naming made me ( and others ) think Genesi was going for the custom Efika1 size. People associate Efika name with the custom form factor that Genesi came out with ( for Efika 5200 ). Also, I think of a lower end system whenever I hear Efika. Since 8610 is going after the high end, it should either use Pegasos or new name. ie: Pegasos name means upper end motherboards, Efika name for lower end boards. Just another one of my suggestions.
* Up to 1333MHz G4 core with Altivec * DDR/DDR2 SDRAM Memory Controller with ECC (up to 533 MHz/1066 DDR) * Integrated graphics, SXGA 1280 x 1024 resolution @ 24bpp * Two PCI Express® Interfaces - one with 1x/2x/4x/8x lanes for optionally connecting a GFX card (8 lanes routed to a standard 16 lane slot) for those not happy with the integrated graphics - the other one routed to the ULI southbridge * "Traditional" PCI 2.2 Interface at 32-bits and 66 MHz * etc, etc...
From the ULI M1575:
* 4x SATA-II, 3Gbps (RAID) * 2x PATA (also RAID) * 8x USB ports (2x USB2 and 6x USB1.1 if I read things correctly) * 1x 10/100 Mb/s Fast Ethernet * HD Audio * etc, etc...
I would put it on a Micro-ATX form factor motherboard and call it "Pegasos 2.5" or something similary catchy! It would essentially be a Pegasos 2 but with improved performance and specs (tightly integrated on-chip controllers, DDR1066, USB2, SATA-II, better audio, the AGP slot replaced with a x16 PCI-Express slot, etc), and a complete ODW system (without external gfx card) could probably be sold at a price similar to or even lower than the price tags that we saw at the end of the Pegasos 2 line.
There will be SATA etc. on the board it's just not integrated into the chip
@jones:
You're wrong. SMP support in "QBox" means you can run QBox tasks on seperate processors. You are not running apps in QBox, you are running them in ABox which is a task of QBox.
@anonymous #2:
There are lots of reasons not to use ULi chips, it's a good idea, and a southbridge would be better and cheaper than several discrete chipsets, but it's down to part availability and nothing more.
I just ordered a Via Epia SN :D I hope I will be able to add a Pegasos 3 next to it soon!
Also I stumbled upon this this preview of an upcoming CPU. I certainly looks interesting for the kind of boards Genesi is making. Just an idea you could produce x86 in addition to PowerPC boards! :o
This isn't about the current (er..February 14th) post, but: hope all is well. I see you haven't posted in awhile. What's cooking? What *evolutionary* changes have occurred lately? :)
(I should have named my blog "Revolution of The Species"...haha)
Of course ARM is an interesting architecture, too. But it doesn't scale up as the PPC does, nor does MorphOS run on it (well, I still see MorphOS as the "gem"). Also while I like and endorse to be openminded it also holds some danger to lose yourself in too many approaches and alternatives.
So, if you have capacities left it is fine and worthwhile to evaluate ARM, but if all your capacities are taken already, keep away from additional challanges. It is important to keep a clear focus.
My ARM comment wasn't taken out of the air. Genesi's latest version of the firmware, "AURA", supports PPC but also ARM and x86 (AFAIK). At AmigaWorld.net BBRV wrote in a post the title of their next blog entry (it was deleted though by a site moderator with a Genesi allergy). The new blog has never appeared but it was about this, which I believe is closely related to this issue, which in turn I believe is closely related to this subject.
30 comments:
Interesting thing.
Where to buy?
A new SoC prozessor ?
Put it on your own board and sell it as a computer ....
What about MorphOS?
http://www.freescale.com/imageprocessor
I find interesting that has PCI Express (not 16x and just 8x but well, it's for embedded) and PCI.
The bad things is the lack of ethernet, USB and SATA. I think that will use all the available interfaces if wanting to make a full desktop computer, so no expansion possibilities.
This is clearly designed for the embedded market. What about something more powerful?
MPC8641D seems more interesting as it have four 10/100/1000 ethernet ports (with TCP and UPD checksum acceleration, QoS support, and packet header manipulation), 64 bit (72b with ECC) DDR2 memory controllers, Dual 8-lane PCI Express ports, 4-lane serial RapidIO port and two a600 (G4-like) cores.
Both aren't really chips designed for desktop but the later one could be used for different systems and be an interesting substitute of Pegasos II for MorphOS users.
The only problem can be the lack of stuff like USB and SATA, I'm not sure what stuff can be connected to RapidIO for example.
What is perfectly clear is that PowerPC abandoned the desktop market and who knows about the server one if IBM and the Power group don't help to make it more interesting compared to competitors (x64 and SPARC). X86 is still there for stay for a long time and learn to like it, ARM and PowerPC will be really scared when X86 will be able to compete with them in terms of power efficiency (that will can be possible with new manufacturing process and future optimized SoC designs by AMD and nokia).
Please keep in mind about an interesting workstation machine for development and MorphOS, find a clear market for MorphOS 2.0+ and pay the developers if they release a product with the requirements for your business needs. MorphOS has a lot potential, now it needs to be guided in a right and clear way into the market and providing very strong killeraps into other OSes in the embedded world like Linux.
What about MOS for 8610? Ten more years? :)
What about MOS for 8610? Ten more years? :)
Fedora is nice, but I would be happy to install openSUSE on an mpc8610 :-)
anonymous, to answer some things you posted
1) the performance difference on most graphics cards between 8x and 16x lanes is absolutely minimal - in the single percentage points. So, you may lose 4% or 6% of your framerate. Plenty of x86 boards come with only an 8x slot for graphics especially if they are SLI or CrossFire, and nobody notices the minor performance difference.
2) none of those things are bad. It just means, you need support chipsets. A southbridge perhaps. This is no different to the Pegasos. It keeps the MPC8610 cost-effective. It doesn't have USB, SATA or Ethernet because it is meant as a processor (think, more in line with AMD building in memory controllers to their processors, rather than the northbridges).
3) The MPC8641D is not as interesting because it is far more expensive and also.. MorphOS doesn't support dual core processors anyway. What would you do with 4 gigabit ethernet ports in MorphOS? As you said, what would you even do with RapidIO even in another OS such as Linux? It's hard to justify a board costing hundreds of dollars for a 4-port gigabit router :)
All Freescale processors are designed for the embedded market. The MPC8641D is meant for different embedded market (high performance communications processor) than the MPC8610 (high performance imaging systems and kiosk). Right now, the MPC8610 makes sense as it DOES fit with MorphOS, it makes sense from the lead of the Efika (keep it simple) and it makes sense for cost.
There is probably scope for an MPC8641D board, *after* we have gone through the motions with the MPC8610 and created a successful product. All the same AltiVec work we did on the G4 and hope to do on the MPC8610 will work, with the advantage of a dual core processor and slightly higher integration.
But, the expense, compared to an MPC8610/Southbridge combination, may not justify the solution. This is the key point where most products fail in any market.
No ethernet, no SATA and no USB? Then developing this board is a total waste of money. It is absolutely superfluous. No serious person would ever buy a board without only one of these 3 connectors.
Nice video !
I hope the Efika 8610 will come out of the box soon... :-D
@Anonymous: it's the chip (the 8610) that does not have support for USB, Ethernet and SATA ONBOARD. You can add it with additional components (Just like in a normal PC).
How does this proccessor compares to an equivalent 1Ghz x86 (lets say an Athlon or Pentium 3)?
I have bought an Efika which I use as a mini-server. If only I could load it with more memory it could be a much more interesting and useful machine. Having said that I think Genesi is aiming the wrong target with their latest offering.
Efika is:
1) power, heat & size efficient
2) performance & resource limited
For desktop use #1 seems mostly irrelevant and #2 is a disadvantage but for service usage (think home-server) #1 turns into a definite advantage.
Now if Genesi could take 8610 and produce a board like this:
http://via.com.tw/en/products/mainboards/motherboards.jsp?motherboard_id=450
@matt sealey
What no smp in MorphOS. You sure about that? :-)
http://developer.morphosppc.com/index.php?mode=library&submode=view&artarticle_id=5&artchapter_id=9
Okay, only says smp for the kernel so who knows how well MorphOS works with multiple cpus & cores.
The 8641D is too expensive and would only attract a few, hard core users. Not worth it. Better to stick with low to mid end markets instead. That is where Genesi can attract the new users.
What was actually needed for Efika system was a processor with 550-800 Mhz speed at a good price. Maybe e500 core? But that means missing out on Altivec because that is only available on e600 cores.
@angus
The 8610 has top speed of 1.33 Ghz w/Altivec. I believe these are the chips they would use. I think it'd be comparable to Athlon 2.0 Ghz CPU but can't say for certain.
Efika form factor would be a mistake. For upper end systems like the 8610, it is better to go with Mini ITX or Micro ATX form factors. Efika size is only good for lower processor speed, fully integrated systems ( sub entry market ).
This actually makes sense to me;
1) Efika for low end ( 400 - 600 Mhz )
2) Mini ITX for mid end ( 600 - 850 Mhz )
3) Micro ATX for high end ( 850+ Mhz )
PowerPC is very limiting. For instance, in regards to Freescale.
They only provide Altivec on e600 cores. SMP is only possible with e600 cores. When you start going over 400 Mhz, costs start increasing rapidly for every 200 Mhz., No L2 cache on lower end CPUs, Very few new CPU designs coming out, ( very ) high prices for mid and upper end CPUs, etc., etc. Unfortunately, MorphOS is made for PowerPCs so Genesi will have to stick with them and do the best they can in this situation.
What would have been very cool is making a motherboard with dual 5200b CPUs ( 2 x 400 Mhz for Linux or MorphOS; MOS kernel is SMP ). Maybe 5200b are SMP compatible because of the MPC603e core they're based on? ie: Could be possible to do but maybe hard?
Tundra allowed SMP of MPC603e cores.
http://www.tundra.com/product_detail.aspx?id=1309
PS It sucks that SMP is only available for higher end CPUS on PowerPC. Intel started offering SMP back with Pentium 1 MMX ( or non MMX? ) CPUs ( Intel HX chipset motherboard ). And I've owned Intel SMP Pentium II & III systems ( dual CPUs ). Very sad to see that you can only do this with PowerPC on their upper end, pricey CPUs.
It has no USB/SATA/Ethernet and Sound (I suspect).
It is not a problem for me if it is cheap. I can buy whatever I like in the form of Expansion cards. I like this option.
On the graphics side it seems OK. Hopefully the gfx support is open source.
Actually this is the motherboard I have been dreaming for. Haiku and Syllable can have a viable platform. Even Gnu-Darwin with some tweaking/drivers can boot on this machine.
Great Job/Choice/Product.
I am waiting for price quotes.
Vasileios Anagnostopoulos
@agnus;
Better. It's more in line with something like Intel's A100 and A110 processors (for reference they are sub-GHz ~6W chips with ~400MHz FSB and a companion northbridge, based on the Pentium M design). The MPC8610 should be available at up to 1.3GHz, though.
An aside;
If you take a 1GHz G4 made this year and put it against a 1GHz Pentium III, you'll notice the Pentium taking up 4 times the power and operating far slower.
This is because there is an 8 year gap between the technologies, 130nm Intel chip with 256k L2 cache and 133MHz front-side bus vs. a 90nm G4 with 2MB L2 cache and 200MHz front-side bus, who will win? :D
So, it's never very good to compare old and new processors. There are lots of ~1.3GHz processors on the market right now from Intel, Via, AMD and everyone else, so you can compare with those. But, Intel A100 and A110 compare best with featureset, except the built-in memory controller and PCI Express root complex :)
Back to what you said;
The MPC8610 is very power efficient, probably a little more so than even Freescale's website suggests (they are being very conservative for once). Performance should be good; far, far better than Via's CoreFusion platform at the same clock speeds, perhaps with not quite as good power consumption, but still in the levels you could cool passively, and effectivfely.
angus: I can't see the full URL for what you posted here, but I am assuming it points to a Mini-ITX board. Yes, something like that is what we would go for. There are many advantages to ITX now that simply did not exist when we had the Pegasos - the market was immature and Via have a propensity to ditch form factors and build in proprietary Via-only features (like the Nano ITX form factor, which has been thrown away after a distinct lack of boards or cases, in favor of Pico ITX, and has a custom power connector, and won't fit in ANY standard cases). AMD DTX form factor builds in ITX compatibility, and Intel have embraced it for their reference boards. Dell are even shipping ITX PCs. And, importantly, Freescale concentrates on ITX for all it's multifunction reference designs now. This has all happened *since* Pegasos was discontinued, it all makes it easier to justify as a product design decision, where there was none but "geek chic" before.
To allay any of the fears here, the BOARD will have USB, SATA and Ethernet, you could not live without these on a desktop board and we know this of course. They are simply not in the SoC package like the MPC5200B - however this means we can drop in new chips as technology advances, making it a far more futureproof design. Which chips are chosen is still being pondered over, but we have a remit to choose the chipsets which offer the best and most mature Linux drivers and importantly open documentation, if that is at all possible.
Today we have fedora 9-beta PPC running on both the Efika and Pegasos II. Some of us probably want 256-512MB RAM embedded platforms like the EFIKA and maybe the MPC8610 and ATI 2400 HD Pro. Maybe we'll get that before Easter?? ~K
@jones:
SMP is not just for the low end. Take a look at the MPC8572E (dual core, 1.5GHz e500 - with SPE SIMD, so not entirely lacking against AltiVec, just not as powerful). You may also want to look at the MPC5510 which is a dual-core e200 (one of them is used as a VLE coprocessor).
The MPC5121E has the AXE coprocessor which is effectively an on-core 200MHz RISC DSP - with a surprisingly comprehensive instruction set. There are lots of definitions of multi-core, SMP or otherwise, and it's not restricted to simply "two of the same processor". You might well consider the announced Intel and AMD "CPU with GPU" chips multi-core as they will include a graphics processor capable of outstripping the main processor cores (all 4 of them, potentially!) in floating point operations.
We are definitely looking more towards ITX than custom form factors, so don't worry about that. There is very little reason with the high integration to produce a board any bigger.
To clarify a misunderstanding, the MorphOS kernel does NOT support SMP. What you run is what Ralph calls "ABox" which has absolutely no fundamentally required support for multiprocessing. Since it is based on the exec.library concept from the old 1980's AmigaOS, the entire OS relies on writing into memory for message passing - something which is quite safe on a single processor, but incurs a *full cache flush* when scheduling the next processor. This design has already been attempted; PowerUP does it every time you switch from m68k to PPC, but remember m68k didn't have 2MB of L2 cache to manage and neither did the 604e!
You would throw all the performance of both processors in the toilet undertaking this multiprocessing model inside ABox.
The underlying "Quark" microkernel may support SMP, but until this functionality can be exposed (perhaps through abstracted DSP libraries or so) you will not be able to use it, and certainly not be able to use it as a dual-core e600 running seperate MorphOS apps on seperate processors, or multithreading across two cores.
Some more thoughts.
Lack of USB/Ethernet/Sata/Sound will accelerate time to market, bring down the cost and make Linux compatibility better. I do not think that they must be included if enough slots are present. I also have no problem with PCI. The only requirements I think are Serial, PS/2 and Floppy. Another improvement from the previous Efika would be expandable RAM in the form of slots, not soldered on board. DDR2 or DDR400 makes no difference to me. Maybe I am a bit conservative.
Genesis has the last word. Looking forward to the new Efika.
Hey Matt,
I want to first make some corrections to your statements.
You said, "The MPC8610 should be available at up to 1.3GHz, though."
Freescale site says, "Up to 1333 MHz performance and 3060 MIPS"
1333 Mhz = 1.33 Ghz like I previously said :-) & NOT 1.3 Ghz; Okay, I admit you were pretty close.
You said, "There are lots of definitions of multi-core, SMP or otherwise, and it's not restricted to simply 'two of the same processor'."
Sorry, but that is not entirely true; partly right. That is correct for multi-core but not for SMP. SMP stands for Symmetric Multi Processors. Symmetric meaning the same type & speed ( identical ).
From Wikipedia, "Symmetric multiprocessing, or SMP, is a multiprocessor computer architecture where two or more identical processors are connected to a single shared main memory."
Unfortunately multi-core can mean something different like you said. For instance, the 5121e is a triple core ( CPU + AXE + PowerVR ) - instead of just refering to # of CPU cores which can cause confusion.
You said, "To clarify a misunderstanding, the MorphOS kernel does NOT support SMP."
From MorphOS Developer Connection site:
"Basic kernel information
MorphOS/Quark has the following basic design goals:
.....
Symmetrical multi processing (SMP)
....."
Seems like a contradiction to what you said. From my understanding, MorphOS has the basic goal of being SMP capable according to MorphOS site ( or at the very least the kernel is designed for SMP ). I wonder why MorphOS developers would make that statement if it wasn't true? I can't comment further on this issue without using/testing MorphOS on an SMP machine myself. All I can say is that either the developers are misleading the public or you're incorrect.
I still believe that Micro ATX is the way to go for higher / upper end products but if you believe Mini ITX will work out then go for it & hopefully things work out for the best. Mini ITX is good for mid range products and Micro ATX for upper end. Just my thoughts. On upper end systems, I like to have 2 to 3 PCI slots, plus AGP or PCI-E slot, USB 2.0, SATA, PATA, 2 DIMM slots, Ethernet and Audio; ie: I like the works on upper end systems as I'm sure others do too. On low & mid range systems I like them to be more on the compact size.
I actually believe doing dual processor or dual CPU core system is not worth it on PowerPC because of the cost and little benefit gained. A dual CPU 5200 system would have been cool though but I don't believe it is technically possible. So, it seems that the better course is to go for the fastest speed single CPU core instead.
PS Using the Efika 8610 naming made me ( and others ) think Genesi was going for the custom Efika1 size. People associate Efika name with the custom form factor that Genesi came out with ( for Efika 5200 ). Also, I think of a lower end system whenever I hear Efika. Since 8610 is going after the high end, it should either use Pegasos or new name. ie: Pegasos name means upper end motherboards, Efika name for lower end boards. Just another one of my suggestions.
I would combine a Freescale MPC8610 with a ULI M1575:
From the CPU:
* Up to 1333MHz G4 core with Altivec
* DDR/DDR2 SDRAM Memory Controller with ECC (up to 533 MHz/1066 DDR)
* Integrated graphics, SXGA 1280 x 1024 resolution @ 24bpp
* Two PCI Express® Interfaces
- one with 1x/2x/4x/8x lanes for optionally connecting a GFX card (8 lanes routed to a standard 16 lane slot) for those not happy with the integrated graphics
- the other one routed to the ULI southbridge
* "Traditional" PCI 2.2 Interface at 32-bits and 66 MHz
* etc, etc...
From the ULI M1575:
* 4x SATA-II, 3Gbps (RAID)
* 2x PATA (also RAID)
* 8x USB ports (2x USB2 and 6x USB1.1 if I read things correctly)
* 1x 10/100 Mb/s Fast Ethernet
* HD Audio
* etc, etc...
I would put it on a Micro-ATX form factor motherboard and call it "Pegasos 2.5" or something similary catchy! It would essentially be a Pegasos 2 but with improved performance and specs (tightly integrated on-chip controllers, DDR1066, USB2, SATA-II, better audio, the AGP slot replaced with a x16 PCI-Express slot, etc), and a complete ODW system (without external gfx card) could probably be sold at a price similar to or even lower than the price tags that we saw at the end of the Pegasos 2 line.
@anonymous #1:
There will be SATA etc. on the board it's just not integrated into the chip
@jones:
You're wrong. SMP support in "QBox" means you can run QBox tasks on seperate processors. You are not running apps in QBox, you are running them in ABox which is a task of QBox.
@anonymous #2:
There are lots of reasons not to use ULi chips, it's a good idea, and a southbridge would be better and cheaper than several discrete chipsets, but it's down to part availability and nothing more.
I just ordered a Via Epia SN :D I hope I will be able to add a Pegasos 3 next to it soon!
Also I stumbled upon this this preview of an upcoming CPU. I certainly looks interesting for the kind of boards Genesi is making. Just an idea you could produce x86 in addition to PowerPC boards! :o
Hey Raquel & Bill Buck...
This isn't about the current (er..February 14th) post, but: hope all is well. I see you haven't posted in awhile. What's cooking? What *evolutionary* changes have occurred lately? :)
(I should have named my blog "Revolution of The Species"...haha)
Anyway, just checking on you.
Don't worry! We are fine. We will post a blog today or tomorrow that will surprise a few people...
Thanks to all of you for the feedback (via email or posted here) on the 8610 effort. We are still contemplating the best form factor, I/O, etc.
R&B :)
Me and my amiga frends in poland wait from good news from genesi :)
Jerzy "Drako" Guc
LimePC handhelds and an 8610 dev-station?
We can compete with Apple.
Good tip Luca.
We can all be happy.
Where to buy?
...
Or ARM?
Of course ARM is an interesting architecture, too. But it doesn't scale up as the PPC does, nor does MorphOS run on it (well, I still see MorphOS as the "gem"). Also while I like and endorse to be openminded it also holds some danger to lose yourself in too many approaches and alternatives.
So, if you have capacities left it is fine and worthwhile to evaluate ARM, but if all your capacities are taken already, keep away from additional challanges. It is important to keep a clear focus.
@ulrich
My ARM comment wasn't taken out of the air. Genesi's latest version of the firmware, "AURA", supports PPC but also ARM and x86 (AFAIK). At AmigaWorld.net BBRV wrote in a post the title of their next blog entry (it was deleted though by a site moderator with a Genesi allergy). The new blog has never appeared but it was about this, which I believe is closely related to this issue, which in turn I believe is closely related to this subject.
"Genesi's latest version of the firmware, AURA, supports PPC but also ARM and x86"
Whaaaat?
Post a Comment