@Shadowolf

Oh, come on. You're splitting hairs here. What would you say is missing for the PASemi 1682M to be used as the basis for a "desktop system"? Just about the only thing I can come up with is "it won't run Windows"

@CodeSmith

Quote:

won't run Windows

That's pretty much it. There is no desktop OS out there for that cpu, and thus they won't target it as such. Markteing bubble. That's all. Hardly the performance. Well, I'd like to see some NVidea drivers, though. Cg running on a 16x PCIe 7900GT or even 88xx would rock! ($1500 board plus $800 gfx board ;) But maybe that will be supported thru linux? (otho, nvidea linux drivers are all x86, or are there any ppc ones ?)

_________________
AmigaOne XE G3/933/VIA/FM801/R200 (fixed),
G3SE/600/Voodoo3/Sil680/RTL8139/SBLive! (noiseless!)
µA1-MK2/G3/933/R200/CMI8738
XE/G4 (broken 7450/800)

@CodeSmith

Quote:

Oh, come on. You're splitting hairs here. What would you say is missing for the PASemi 1682M to be used as the basis for a "desktop system"? Just about the only thing I can come up with is "it won't run Windows"

There's nothing 'missing' as such, just that the 1682 as a desktop processor looks really poor next to basically any current x86. It's slower, more expensive, and less flexible. Non x86 processors cannot succeed on the desktop to any real degree. PA Semi are not stupid, they're well aware of that fact.

The 1682 is quite clearly designed for embedded use. There are several aspects of the design that are very desirable for embedded use, but are negatives for desktop systems. That's not to say it can't used, just that it isn't a perfect fit.

That said, it's miles better than any other PPC part right now and the obvious choice for a PPC desktop board.

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater

Quote:

the 1682 as a desktop processor looks really poor next to basically any current x86

Really? a superscalar CPU, with two cores that can each issue 3 instructions per cycle (out of order), running at 2GHz and with a 1066MHz DDR2 memory controller with 2MB of onboard L2 cache doesn't sound too shabby compared to my Core 2 Duo. OK, so the intel chip may end up being better, but I wouldn't say the PPC chip looks "really poor" next to it.

Last edited by CodeSmith on 06-May-2007 at 12:43 AM.
Last edited by CodeSmith on 06-May-2007 at 12:33 AM.

@CodeSmith

Nothing is going to make DrBombcrater or ShadowWolf happy, unless they want OS4 on x86. So it's really not worth discussing anything with them.

Adam

@ackcontrls

Fair enough. I wouldn't mind seeing some benchmarks for that chip though. It looks awesome on paper and I'm sure it's a lot faster than even a 7448, but it would be nice to compare it to a recent Intel chip.

@ackcontrls

I'd still like to see a cell based board, though. But the current cell with the XDR is well, not really suitable and the new one with DDR2 (is it 2 or 3) is probably to expensive and not even available. But spu power would be easier to utilize thru a secondary kernel (or kernal extension) then multicore with the current AmigaOS. (and well, north/southbridge would be needed for the cell as well, unless one hooks up a RSX directly to it).
Anyway, PA Sema looks the natural choice beside it.

_________________
AmigaOne XE G3/933/VIA/FM801/R200 (fixed),
G3SE/600/Voodoo3/Sil680/RTL8139/SBLive! (noiseless!)
µA1-MK2/G3/933/R200/CMI8738
XE/G4 (broken 7450/800)

@CodeSmith

Quote:

Really? a superscalar CPU, that can issue 3 instructions per cycle, running at 2GHz and with a 1066MHz DDR2 memory controller (all times 2) doesn't sound too shabby compared to my Core 2 Duo. OK, so the intel chip may end up being better, but I wouldn't say the PPC chip looks "really poor" next to it.

It's more than just performance, although that's an issue. The most recent information I can find has PA Semi claiming the 1682 scores about 1000 in SpecInt2000 @ 2GHz. AMD's and Intel's best desktop parts score roughly two and three times that, respectively. (in x86 terms, 1000 SpecInt2000 means a Sempron 3200 or similar - about £25 worth)

Configuration flexibility is a problem, too. With the 1682 you have to accept the functions built into the chip. If it doesn't provide what you need (not enough PCIe lanes, for example) then you're stuck. It's not like say, and Athlon 64, where you can just pick a suitable hypertransport system controller chip from the wide range available.

The cold truth is the PA Semi design costs more than normal X86 desktop parts, performs worse, and is less flexible. It can't run Windows and even with Linux it takes a hit because of the many x86 binary only drivers out there (ie, NVidia's Linux drivers, among many). So it's never, ever going to get used for a desktop board unless PPC compatibility it essential.

But inside the reality distortion bubble that is the PowerPC world, the 1682 actually looks pretty good. It certainly shows how utterly rubbish IBM's and Freescale's designs are.

_________________
Who do you serve, and who do you trust? - Galen

@Jorge

Agree completely. I guess we'll have to wait until Monday to see the complete board spec, but I'm very pleased with the choice of CPU. This is way more than I was expecting, and as I said above, it will at last put amigas back on an even footing with PCs and Macs.

I did a bit more digging into the company, and it looks like its founder was the head designer for the DEC Alpha, widely recognized as one of the best CPUs ever made. This is good technology; all we need now is that their sales people are better than MAI Logic's.

@ackcontrls

Quote:

Nothing is going to make DrBombcrater or ShadowWolf happy, unless they want OS4 on x86. So it's really not worth discussing anything with them.

Being able to read people's minds is such an interesting talent...

Meh. I hate it when people run away from discussions. It takes all the fun out of them.

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater

Well a discussion should at least stay on a common ground. x86 is not an option. That said, it is pointless to dicuss this. (IMO).

_________________
AmigaOne XE G3/933/VIA/FM801/R200 (fixed),
G3SE/600/Voodoo3/Sil680/RTL8139/SBLive! (noiseless!)
µA1-MK2/G3/933/R200/CMI8738
XE/G4 (broken 7450/800)

SMP etc: PA Semi has a single core BTW.

Adam, this CPU captures the imagination! How PPC-ish is it? Will it be theoretically difficult to port OS4 to it?

If OS4 continues to develop, the high end system could support multicore & 64 bit!

What will we run with all this P O W E R !?

No, wait... I'm serious. What WILL we run with all this power?

_________________
If you're going through hell, keep going. -Winston Churchill

@DrBombcrater

Quote:

It's more than just performance, although that's an issue. The most recent information I can find has PA Semi claiming the 1682 scores about 1000 in SpecInt2000 @ 2GHz. AMD's and Intel's best desktop parts score roughly two and three times that, respectively. (in x86 terms, 1000 SpecInt2000 means a Sempron 3200 or similar - about £25 worth)

You're falling into the familiar trap of comparing artificial benchmarks across cpu architectures. PPC and x86 instruction sets are so different that the only benchmarks that really count are wall-clock ones. I'd like to see comparisons of say the time it takes to do a ray trace, or to encode a certain piece of music into MP3. Then we can do meaningful comparisons.

Quote:

Configuration flexibility is a problem, too. With the 1682 you have to accept the functions built into the chip. If it doesn't provide what you need (not enough PCIe lanes, for example) then you're stuck. It's not like say, and Athlon 64, where you can just pick a suitable hypertransport system controller chip from the wide range available.

Wide range available? There are just a handful of NB chips than can talk Hypertransport, and in any case you're constrained by the 2000MT/s bandwidth so you can't just keep adding peripherals. The description of PCI express support is a bit confusing, but it sounds like you get what most x86 NBs give you: 1x16 and a few more lanes split between x8, x4, x2 and x1. Sure, you don't get 2x16 Crossfire support, but I think only one NB from NVidia supports that.

Quote:

It can't run Windows and even with Linux it takes a hit because of the many x86 binary only drivers out there (ie, NVidia's Linux drivers, among many). So it's never, ever going to get used for a desktop board unless PPC compatibility it essential.

OK, but this is not exactly affecting the particular use of the chip that we're discussing here. The claim was made that this chip is not suitable for desktop use, not that it's unsuitable to run Windows or Linux.

Quote:

But inside the reality distortion bubble that is the PowerPC world, the 1682 actually looks pretty good. It certainly shows how utterly rubbish IBM's and Freescale's designs are.

I think Steve Jobs took the reality distortion bubble with him when he left I agree with you about the other manufacturers dropping the ball. If you look at how things were in 2001 (500MHz 7410), you'll see that according to Moore's law the 1682 is roughly where the PPC world should be now (we should be at 4GHz, we're at 2x2GHz).

Last edited by CodeSmith on 06-May-2007 at 02:03 AM.
Last edited by CodeSmith on 06-May-2007 at 01:42 AM.

@debrun

The chip implements the PowerPC instruction set, same as the Gx chips from Freescale and 74xx chips from IBM. It also implements Altivec, although PASemi gives it a different name probably for trademark reasons. It is more advanced than a G4, but I don't know enough about the PPC architecture to know if it could be called a G5. I'm fairly certain that OS4 can be tweaked to run on this CPU, in fact I suspect (since OS4 has a HAL) that the biggest changes needed will be for the board that the CPU is mounted on, rather than the CPU itself.

@ackcontrls

Agree TOTALLY. How much stupidity around here (sadly, as usual)



@all

How a processor can be considered a "desktop" system???

Just making a f*cking desktop system!!!!!

End of the story.


@ackcontrls

The only thing that makes me worry about is the real date for the real availability of the product. You talk about it as it is a bit far from having everthing well finished or even secured. And I think you know better than anyone what happens with estimations and hopes...

Last edited by vision on 06-May-2007 at 02:23 AM.

@vision

Actually, if a CPU only supports industrial interfaces, it won't make a good desktop CPU. For example, many PPC northbridges so far support only PCI-X, which is common in industrial applications but almost unheard of for home or business use.

@CodeSmith

Quote:

Really? a superscalar CPU, with two cores that can each issue 3 instructions per cycle (out of order), running at 2GHz and with a 1066MHz DDR2 memory controller with 2MB of onboard L2 cache doesn't sound too shabby compared to my Core 2 Duo

Core 2 core issues 4 instructions per cycle. AMD64 issue 3 instructions per cycle.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Jorge
Quote:

x86 is not an option. That said, it is pointless to dicuss this. (IMO).

It's only not an option because it was decided that it is not an option, not because it is not an option, technically speaking. If Hyperion is out of the picture (are they?), opinions and options might change. OS4 compiled for a virtual 68k CPU, minus some PPC and address space magic, would truly be "Amiga Anywhere". Isn't that Amiga Inc's vision? Right now they seem to jump from "very unavailable" to "maybe available", within the PPC camp. But maybe that's just because the people at Amiga Inc. don't have a technical background and not set-in-stone strategy?

@Hammer

Well, I had an amd64 before my core 2 duo, so there

@dietmar

We are talking about the immediate (and possible?) future, with a pending announcement on Monday (that is in two days). In that context x86 is not an option, no.

Last edited by Jorge on 06-May-2007 at 03:23 AM.

_________________
AmigaOne XE G3/933/VIA/FM801/R200 (fixed),
G3SE/600/Voodoo3/Sil680/RTL8139/SBLive! (noiseless!)
µA1-MK2/G3/933/R200/CMI8738
XE/G4 (broken 7450/800)