X1000 CPU is PWRficient PA6T

[Jose]

@Piru

So, you'll be lining up to buy one as it comes out ?

[haywirepc]

"8,800 MIPS at 2.0 GHz is about where the leading Pentium processors were back in 2003."

So its as fast as a computer I picked out of the trash recently, a dell pentium 2ghz machine...

Amd has new 64 bit 3ghz chips with 6 and 8 cores that will cost 300$ for 6 cores or 400$ for 8 cores, and remember thats brand new bleading edge tech price. 18 months after release, they will be half that price.

I sure don't see an x1000 in my future, I'd rather for 900$ buy a new linux box with 8 processing cores,12 gigs of ram and 2 terrabytes of hard drives.

For the same 900$ I can buy a sam board and run amiga os on hardware thats slower than 10 year old pc hardware. I just don't see the whole lets stick with ppc, its just retarded at this point, in my view anyhow.

Even people like me who are huge amiga fans can not justify the expense based on how outdated the hardware is. They are killing any chances it had by offering substandard slow ass hugely inflated priced hardware.

[Jose]

"
"8,800 MIPS at 2.0 GHz is about where the leading Pentium processors were back in 2003."
So its as fast as a computer I picked out of the trash recently, a dell pentium 2ghz machine...
"

At what wattage, 95 watts ? Even the lower power ones are about 45 Watts, no comparison.

[KimmoK]

Anyway the issue is A-EON are marketing it as the second coming,

I think that's because they have now the control of Amiga & OS developments.
First time since 199x AOS devs have control of AOS future.

And the HW is a big jump forward in terms of performance (IMO, it's 5...10 fears forward from SAM) and up-to-date expansion slots (15 years forward from SAM) + GPGPU etc.

(xena is there mainly/just to bring some I/O flexibility over standard HW, like the video slot of original Amigas, waiting for people to utilize it, or like the geekport of bebox, just like they seem to say now, after the tease is over, since january)

about CPU performance:
It should be like p3@2Ghz. A lot better than P4@2Ghz. + it has integrated RAM controllers + integrated PCIe bridge + more cache , unlike those 2ghz pentiums, unlike any other PPC on desktop. It should significantly outperform systems like original xbox360. And whole system (HW) is more modern than the last G5 mac. Only time will tell how many years it takes for the SW to support it fully.

@everybody

A-Eon does not seem to plan to sell many hundreds of x1000. Surely they "just" pave the way for next gen multicore HW & ASMP AOS4.

[persia]

And yet a far more apt description somehow.

"monkey bazooka" sounded more awkward

[Piru]

GPGPU

Isn't that a bit premature considering there isn't even 3D support for the included Radeon X700 card yet?

As far as I can tell both ATI and nVidia are keeping their GPU stream processing drivers closed source (and binaries are only available for x86 and amd64). So where is this GPGPU stuff coming form then?

[lsmart]

Per design, Amiga can't do what outsiders would call *real* SMP, not even "later".

By the wikipedia definition of SMP, even classic AmigaOS can be extendet to do real SMP. Are you perhaps thinking about NUMA?

[the_leander]

By the wikipedia definition of SMP, even classic AmigaOS can be extendet to do real SMP. Are you perhaps thinking about NUMA?

Oh dear gods no...

AmigaOS as it is today, either in classic or NG cannot and likely will never support SMP. The APIs simply do not allow for SMP and would need major reworking before SMP is possible.

[SamuraiCrow]

Using a gang scheduler for SMP would allow Forbid/Permit-style disabling of the multitasking to be implemented.  Nasty code that uses Forbid to speed things up will have the dubious distinction of being slower though because it will still run on one core only.

Personally, I think the FPGA approach to expanding the Amiga (eg. making slave processors to the master single-threaded CPU) is the best way but SMP is certainly possible.

[the_leander]

Right up until something running in one core wanted to talk to something running in another.

[Trev]

Using a gang scheduler for SMP would allow Forbid/Permit-style disabling of the multitasking to be implemented.  Nasty code that uses Forbid to speed things up will have the dubious distinction of being slower though because it will still run on one core only.

Personally, I think the FPGA approach to expanding the Amiga (eg. making slave processors to the master single-threaded CPU) is the best way but SMP is certainly possible.

There's no reason why AmigaOS couldn't do something entirely different. More than a few folks are working on system designs that don't rely on cache coherency and related paradigms to do local parallel processing. And I'm not talking about co-processors or other types of dedicated out-of-band hardware.

[the_leander]

Using a gang scheduler for SMP would allow Forbid/Permit-style disabling of the multitasking to be implemented.  Nasty code that uses Forbid to speed things up will have the dubious distinction of being slower though because it will still run on one core only.

The problem is that there is a large proportion of the Amiga's software catalog falls into the "nasty code" category - mostly for optimization reasons. If you implement SMP, you have to break API compatibility. To attempt to have both would at best leave you with a system with (application and probably system) stability somewhat south of Charles Manson.

At which point you may as well just go over to Haiku or some other OS that has these features built into it from the beginning and run UAE.

--edit--

Tbh I think the Dragonfly's experimental internal clustering model would probably be easier to apply to th

[lsmart]

Using a gang scheduler for SMP would allow Forbid/Permit-style disabling of the multitasking to be implemented.  Nasty code that uses Forbid to speed things up will have the dubious distinction of being slower though because it will still run on one core only.

Personally, I think the FPGA approach to expanding the Amiga (eg. making slave processors to the master single-threaded CPU) is the best way but SMP is certainly possible.

That´s what I was thinking. Note though, that even the nasty code will have the benefit of the fastest single core Amiga ever. So quit your legacy apps and you have full multicore speed or run them and expect twice the performance impact they would normally have. This is certainly faster than any emulation.

[the_leander]

So quit your legacy apps and you have full multicore speed or run them and expect twice the performance impact they would normally have. This is certainly faster than any emulation.

Except that your legacy apps will always need emulation to some degree. They're 68k remember.

And if you're going down that road then I'm sorry to say that it's been done before and better courtesy of Amithlon. Which had the added benefit of not running on a 10 times overpriced under performing hardware platform.

[Fats]

Using a gang scheduler for SMP would allow Forbid/Permit-style disabling of the multitasking to be implemented.  Nasty code that uses Forbid to speed things up will have the dubious distinction of being slower though because it will still run on one core only.

I think gang scheduler would work if Forbid()/Permit() would only be used by user programs. Problems is that Forbid()/Permit() pairs are used throughout drivers, system libs and user programs. Therefor I think a gang scheduler will either severely limit heavy IO task by CPU intensive tasks or the other way around.
In  the end you will need to replace all the Forbid()/Permit() around IOs by specific locking mechanism in order to combine heavy IO with efficient SMP.

greets,
Staf.

[warpdesign]

What if you had enough *native* apps to get rid of legacy ones ?
You will always be able to run them with UAE anyway...

Why are we still talking about *legacy* in 2010 ? We need to move forward...