New Hyperion Entertainment Website http://a-eon.com/ - The Mystery Continues

[MskoDestny]

My current bet is that the CPU is something in the Xilinx Virtex 5 family. The two PowerPC 440 cores explains the 8 x 25% hint and the ability to host soft cores in the FPGA would explain the more recent comment about cores "No, it doesn't have the number of cores you are thinking of. Whatever number you are thinking of. At least, not necessarily." The dual core Virtex 5 support 3 or 4 PCIe endpoints (depending on the model) and from what I understand each endpoint supports 8 PCIe lanes, so it would seem to match the slots (assuming two endpoints can be ganged together for a single x16 card).

I don't think the Xorro slot is an HTX slot like someone suggested. If the larger connector was a reversed PCIe x16 slot (physically, electrically the slot is HyperTransport in HTX) it would stick out further than the two x16 slots on the board. The notch is also on the wrong end. My assumption is that it's some sort of custom expansion slot. Not sure what it's for though.

The placement of the PCIe and PCI slots is rather strange. Much further from the edge of the PCB than on a standard motherboard. I wonder why.

[MskoDestny]

The hard reality Hyperion are going to face here is that current patterns effect future trends. The trend is for $300 netbooks, $150 consoles, and low cost machines. Big Desktops sales are down and all my buddies would only spend $1200 or more on a KICK ASS laptop or Desktop for a highly specialzed business purpose...or a MAC

But Hyperion know all this and will price the machine for Amiga Enthusiasts who are prepared to pay for this. I hope that this X1000 will spawn a X500 with expansion capabilities. I think Hyperion will realize this just like Commodore did when they released the A1000.

These days it's almost impossible to compete with x86 on the desktop. There just isn't enough volume for PowerPC to drive the R&D necessary to create competitive chips for the desktop (and laptop) market. Apple realized this and switched to x86. If Hyperion wanted OS4 to be anything more than a niche OS for nostalgic hobbyists, they would head in the same direction.

There is some chance that ARM could gain a foothold in the low end market. They're starting to approach netbook class processing power, they're dirt cheap (considerably less than PowerPC chips with similar oomph aimed at the embedded market ) and they tend to have reasonably decent video hardware onboard (it seems most of the PowerPC based SoCs tend to be oriented at network hardware and thus don't have integrated video hardware). Even then, it's hard to say whether ARM will make it in the personal computer market (though it's doing just fine in the mobile device market obviously).

So anyway, I wouldn't expect an X500, or if there is one I wouldn't expect awesome price/performance. PowerPC desktop hardware is going to be a niche for the forseeable future and as a result it's going to be expensive for what you get.

Sorry for the negativity. It's cool that Hyperion is trying to keep the Amiga platform alive, but I think it's good to have realistic expectations.

[MskoDestny]

IIRC Apple switched to Intel because Motorola were unable to make a G5 processor run cool enough to use in a laptop. This was a serious issue for Apple who were stuck with the increasingly obsolete G4 chips for laptops, they had to do something. Whether Motorola's failure was due to the volume issue you describe is debatable, but I would suggest it was more of an engineering/architecture problem.

The problem was that IBM didn't want to foot the bill to do further work on the PowerPC 970 because their only major customer for the chip was Apple and Apple didn't deal in high enough volume for it to be worth IBM's time unless Apple paid for the R&D.

And the only reason they were getting chips from IBM rather than Motorola/Freescale is that Motorola gave up on making chips for the desktop market for essentially the same reason.

Keeping up with Intel is expensive. Even AMD, who pushes a lot more desktop oriented CPUs than Motorola or IBM ever did for Apple, has a hard time of it. If Intel hadn't fumbled with the P4 they probably still would have been a minor player.

Hopefully the new PPC chip spoken of relating to the X1000 will run cool enough for any potential use in a future laptop configuration.

I'm sure it will. It's almost certainly a chip aimed at the embedded market and those tend to have rather modest TDPs. According to Wikipedia, the AMCC Titan (the most likely contender for the CPU) draws only 2.5W per core at 2GHz (though it's unclear whether that's TDP or average draw). An article from EETimes, suggests the TDP for the whole chip might be around 15W or so. For comparison, the Atom 330 has a TDP of 8W, but that's just a CPU not a SoC.

[MskoDestny]

My bet's still on a 8640D (e600) CPU. The Titan might be too fresh to be included in a finished(?) design. The 8640D has the right age and it's spec'ed at 21W on full load (remember the '8 threads to get it up to 20W'?). And it's got dual memory controllers.

The e600 is only compliant to the PowerPC spec, not Power Architecture 2.04. The e500mc is Power Architecture 2.06 so that would technically meet their description.

It's almost certainly not the PowerPC 970. They claim the CPU they're using has a "very low" TDP and that "you've probably never seen one in the wild". Neither applies to the 970.

The PA6T from PA Semi also fits the description, but who knows how they would get their hands on it now.

[MskoDestny]

certainly at two full PCIe x 16 slots (well, at least they tried on that... they have two full lenght PCIe slots, but they degrade to x8 if you use them both... so there's really only one x16 slot).

Lots of PC chipsets do the same thing. I'm not sure what the point of having two x16 slots is though. I doubt OS 4 is going to support CrossFire/SLI anytime soon and given the relatively slow CPU (compared to a PC) you're likely to be CPU bound even with a single GPU anyway. I suppose it could be useful for multi-head setups (though the Radeon 5700 and 5800 cards support three displays per card these days). Does OS 4 support that?

[MskoDestny]

My comment  about military spec came from a young lady I know who is a project manager on the Hubble Space telescope. She told me that one of the Computers in the thing is a 1990's 486... not a chip from another supplier with equivalent CPU power... it's a 486..they had to go through so much engineering BS to get it approved and fabbed for aerospace use.  It's built to withstand the rigors of liftoff and whatever else happens up there.

Stuff that's going into space needs to be radiation hardened. Our atmosphere shields us from a lot of the radiation that comes our way. Apparently this is a lot easier to do with older slower designs.

Reagardless of all that am I crazy to believe that a FAST Amiga legacy compatible machine can be made using FPGA technology and come in under $300?'

Depends on what you mean by fast. The Terasic DE1 board goes for $150 and a modified Minimig with the 68K in the FPGA fits on the device on that board. There are other FPGA dev boards with bigger FPGAs that are under $300 so you could certainly do something faster than an Amiga 500 at such a price; how much faster I can't say.

Putting an Amiga clone chipset in the FPGA of the hardware the Atari Coldfire Project is producing could be interesting, but I doubt that will run under $300 when they're done.

I mean go to deal extreme and you see Super Nintendo, and Sega Genesis TV game units for like $30 bucks. Those consoles were more or less Amiga 500 like tech...

It's all about volume. You won't see an FPGA in those, but a custom IC. A custom chip is much cheaper per unit than an FPGA big enough to fit an equivalent design, but the upfront costs are much higher so it only makes sense if you know you can deal in a certain amount of volume.

[MskoDestny]

My G4 is 1.42 times faster than the Intel Atom.
My G4 is only 3 times slower than fastest overclocked i7.
My G4 is only 1.43 times slower my 6600 .

Comparing only the performance of a synthetic integer benchmark is rather misleading especially since the benchmark you used provided floating point and memory performance numbers. Your G4 is 1.42 times faster than an Atom at doing integer operations, but has no advantage at memory intensive tasks and only a small advantage at floating point. Your Core 2 Duo may only be 1.43 times faster at integer operations but it's 2.39 times as fast at memory intensive tasks and 3.2 times as fast at floating point. That overclocked Core i7 may only be 3 times faster at integer ops than your G4, but it's 4.64 times faster for memory intensive operations and 6.61 times faster at floating point. If you don't think that a 3x to 6+x speedup is not significant then I don't know what to say.

And all that is just for a single core. nbench is not a multi-threaded benchmark program.

x86 can be faster than the PowerPC, if
1.Problem can be carried out over multiple cores
2.Software developer knows how to write software for multiple cores.
3.Software developer has time to make the software on multiple cores.

Uh, no. It's faster even on single threaded code. IBM's POWER6 might be faster than an i7, but there aren't any PowerPC CPUs for the desktop market that are competitive with even midrange x86 CPUs like the Core2 and Phenom II. I imagine the chips for the embedded market (one of which is likely what's going to be in the X1000) fare even worse.

Further, things that really need a lot of CPU power have been moving to multithreaded code. The latest game engines and video encoders can use multiple cores.

[MskoDestny]

So You think that the software done on one core, will work by itself without a rewrite of several core faster.
You're not very smart.

You misunderstand. I'm saying that a single core on an i7 or even a Phenom II will outperform a G4 or even a PowerPC 970 core. nbench only measures the performance of a single core, not the whole chip.

The Cell in the PS3 has 8 cores clocked at 3,2 GHz, granted only 6 of those area available to the developer. Linux performes very well on the PS3. Its very cheap and I know people who use is as their desktop computer.

The Cell has a single PowerPC core and a bunch of SPEs. The PowerPC core may run at 3.2GHz, but since it's a relatively simple in-order design it's not all that fast (the Intel Atom has the same problem). The SPEs add a fair amount of oomph, but they're not really general purpose cores. They only have direct access to a very small amount of memory. Data from main memory has to be copied to this local memory via a DMA transfer in order for an SPE to work on it. They work in a manner more similar to a GPU than a CPU. They're a bit more general purpose than that, but GPUs have been getting more general purpose over time.

Also, you really need to look at performace vs powerconsumption. Which Intel x86 based CPU only uses less than the 440? and still gives the same performance.

It's not from Intel, but the VIA Eden ULV has a TDP of 1W at 500MHz (100mW at idle) and 3.5W at 1GHz (500mW idle). That's a little worse than the Titan (not sure about the chip in the SAM 440), but it's not bad. Not bad enough for x86 to have anything to worry about from PowerPC on the laptop/desktop anyway. ARM has more of a chance of putting a dent there. An ARM SoC intended for a smartphone has everything a cheap netbook needs whereas PowerPC SoC tend to be oriented towards communications equipment and thus lack things like video hardware.

I dont know eveyrthing there is to know about CPUs far from it, but my guess is that the CELL is more than powerfull enough and that the PPC does have a future.

PowerPC definitely has a future. Just not on the desktop.