Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: asian1 on June 25, 2003, 07:17:06 PM
-
Hello
There are rumors about PowerPC 970 2 GHz accelerator boards / cards for old Mac.
Is it possible to create such accelerator for AmigaOne?
IIRC Cyberstorm is actually PowerPC accelerator for Amiga 68K. AmigaOS 4 will run on Cyberstorm.
If such card / module is available, Is it possible to port AmigaOS 4 to 64 bit PowerPC 970 accelerator for AmigaOne?
-
I think Amiga needs to move away from the idea of accelerator boards and concentrate on complete systems. The AmigaOne hasnt been released fully (i.e. with complete support from its OS, AmigaOS) and we're already talking about upgrading it.
That sounds like a major problem to me. Just how many people own an AmigaOne Earlybird system anyway?
-
IMO accelerators often leave you with various bottlenecks - for instance, a A1200 fitted with a Blizzard PPC would still have the same old I/O speeds. The 970 would need lots of technology such as DDR RAM, 8x AGP, PCI-X etc to take full advantage of the chip - in the end, what you'd get is a seperate motherboard on top of the A1; wouldn't it be just as easy to have an A2 or whatever?
I'd imagine that accelerators were so popular was as they allow far faster processors while keeping the custom chips and stuff resident on the mobo for apps that need them (ie the OS). With the A1 having no custom components that need to be kept, I personally don't think an accelerator would be much sense.
BTW, I know very little about PC architecture, PPC etc - I could well be wrong, but it's just my opinion.
-
I don't hink accelerators are allways bad. If you look at PC motherboard they 've had the same PCI bus and I/O for too much time. The major stuff that has changed was the processor, and memory controler. So why do people have to allways buy a new PCI bus, etc... with every new board? But maybe it all goes down to production costs.
-
I think that producing an entire motherboard is cheaper and easier than doing extrange accelerators.... for example look at our classic ppc accelerators, they require lots of layers to allow all that components to fit in such a small board... an entire motherboard would require less layers and would allow you to use standard northbridges, making an easier and cleaner design. In addition to that memory types change almost with each CPU generation so it makes more sense not to use accelerators. I like the idea of having the cpu in a slot or a simple cpu card like in the Pegasos, but if we are talking of such a big changes in memory types and cpu it's a far better idea to make a new motherboard.
-
I agree with Samuar. It's nice to have a good complete system rather than run "hacks" to provide more speed. It was ok on old Amigas when CPU horsepower really was too low to do anything worth mentioning, and it really worked. Some better some worse. Today, I think it should be avoided and stick with well designed complete systems.
-
by asian1 on 2003/6/25 14:17:06
If such card / module is available, Is it possible to port AmigaOS 4 to 64 bit PowerPC 970 accellerator for AmigaOne?
Question is just who is going to do that, MAI? I highly doubt they would bother. I mean all in all, why not spend the R&D on a new mobo design then spending all that time and money to only sell a few accelerators?
Dammy
-
jose: memory controller is in northbridge which is soldered on the mobo... (except with new cpus like opteron, ppc970, etc...)
-
asian1 wrote:
Hello
There are rumors about PowerPC 970 2 GHz accellerator boards / cards for old Mac.
Is it possible to create such accellerator for AmigaOne?
IIRC Cyberstorm is actually PowerPC accellerator for Amiga 68K. AmigaOS 4 will run on Cyberstorm.
If such card / module is available, Is it possible to port AmigaOS 4 to 64 bit PowerPC 970 accellerator for AmigaOne?
But what's the point with faster processors, if the CPU has to sit in it's ass and wait for the NB to shovel data through a narrow buses? IMO, there is too much emphasis on MHz and CPU performance, and too little on general system design. Perhaps we have all been so used to the x86 way of thinking, that we are simply stuck in it now?
The new Mac is a breath of fresh air in this sense!
-
takemehomegrandma wrote:
Perhaps we have all been so used to the x86 way of thinking, that we are simply stuck in it now?
The new Mac is a breath of fresh air in this sense!
I think given the board is basically pieced together from x86 tech with IBM 970's on it, its pretty funny that you think its a breath of fresh air. What exactly is innovative in the system to you?? Hypertransport, 400 Mhz DDR SDRam, PCI-X??
-Tig
-
@ Tigger
Yes I know that it has been put together mainly from standard (but state of the art) "x86 technology". But they have really gone as far as possible in their efforts to remove bottle necks, don't you think? OK, I don't really know, but how common is 2x 1GHz FSB? Memory would perhaps be a bottleneck (would RAMBUS/RIMM perhaps be better?), but "this high-performance frontside bus architecture enables each PowerPC G5 processor to discover and access data in the other processor’s L1 and L2 caches for ultrafast performance" will benefit from this, as well as data shoveling to/from the hypertransport connected I/O, the PCI and PCI-X, AGP, etc. And what about the "Point to Point controller" which shovels data around without even bothering the CPU's and the processor bus? Isn't that even a little "Amiga-like"? Perhaps I am wrong, but I really think this is the most innovative design this far, and I have not seen this on any desktop x86 motherboard, but perhaps I have only missed it? OK, perhaps "innovation" is a too strong word, we are basically talking about removing bottle necks, but at least this focus on removing bottle necks and not *only* pumping up the CPU MHz is a (at least tiny) breath of fresh air? ;-)
-
There are rumors about PowerPC 970 2 GHz accellerator boards / cards for old Mac.
Is it possible to create such accellerator for AmigaOne?
Bad idear.
New mobo pls.
If the PPC mobo market is ever going to catch up in the very long run, then hacking on cpu's that get chocked by the mobo has to stop & i saying with this new PPC cpu that is out now then the time is right for all ppc mobo makers to make new mobos for it.
-
-
The motherboards are already going to be here: Apple's G5 and IBM's PPC970 Blade.
The problem is licensing. You think the Amiga "market" is going to get a new motherboard when you just barely got an old motherboard with no OS? The only new motherboard on the horizon is the Pegasos II which still uses the MPX bus.
As for slapping on a PPC 970 on an A1, give me a freaking break. You may as well port AOS to MOL and run it on IBM's Blade running Linux. Seriously :roll:
-
Jose wrote:
I don't hink accelerators are allways bad. If you look at PC motherboard they 've had the same PCI bus and I/O for too much time. The major stuff that has changed was the processor, and memory controler. So why do people have to allways buy a new PCI bus, etc... with every new board?
.
Reasons for the new motherboard, i.e. the Northbridge and Southbridge cores gets updated. Certain benchmarks prove speed disparity between the chipsets (i.e. efficiencies issues) e.g. nVidia’s nForce2 vs VIA KT400.
With nForce2’s case, its 128bit bus****, 400Mhz DDR FSB, AGP 8X Pro, Hypertransport based Southbridge and Northbridge link, DASP (Dynamic Adaptive Speculative Pre-Processor)**, integrated DSP and 'etc'.
** "an intelligent agent that monitors CPU requests and looks for access patterns that it can
successfully predict. When it recognizes such access patterns, it exploits unused memory bandwidth to load its cache with data the CPU is expected to request later". This is built into the Northbridge chip.
****Useful for concurrent DSP, GPU and CPU main bus access.
-
what about the "Point to Point controller" which shovels data around without even bothering the CPU's and the processor bus? Isn't that even a little "Amiga-like"? Perhaps I am wrong, but I really think this is the most innovative design this far, and I have not seen this on any desktop x86 motherboard,
Have you heard of nVidia nForce series?
What you said is similar to AMD's K8 based 8xxx chipset, nVidia's K8 nForce3 Pro/ K7 nForce I/II and VIA's K8 chipsets. All of which employs AMD’s hypertransport technologies.
The speed of FSB is not quite important, it’s the throughput bandwidth i.e. measured in Mb/s or Gb/s.
With AMD’s Opteron and Athlon 64, the speed between Northbridge and the CPU core match CPU speed. This is partly due to on-die integration.
For nForce2 market share within AthlonX86 market refer to http://www.digitimes.com/NewsShow/Article.asp?datePublish=2003/06/23&pages=04&seq=22
("Nvidia’s nForce2 K7 chipset dominates market")...
Note that, Hypertransport technology is also being applied for AMD’s own 64bit MIPS CPU series.
-
>Who will create such board / module?
Perhaps Omnicluster or Totalimpact (if they can find suitable chipset for PowerPC 970).
By using slotserver / bewoulf approach (Omnicluster), it is possible to use the accelerator with PC, Classic 68K Amiga, PowerMac or other machines.
Omnicluster (http://www.omnicluster.com)
Perhaps the card /module will be connected to the main machine using PCI bus, Zorro, Ethernet or other methods.
If this card is finished and bug-free, is it possible to pay Hyperion to port the AmigaOS 4 to this product?
-
>The only new motherboard on the horizon is the Pegasos II >which still uses the MPX bus.
You should add "that I know of, based on publically available information".
Otherwise your statement is just plainly wrong.
-
Bad idear.
Perhaps not. How long do you think people are going to keep upgrading if they are paying the kind of money Eyetech's charging for an AmigaOne? At least it gives people a cheaper option.
If the PPC mobo market is ever going to catch up in the very long run, then hacking on cpu's that get chocked by the mobo has to stop & i saying with this new PPC cpu that is out now then the time is right for all ppc mobo makers to make new mobos for it.
The fewer options the market gives its customers, the quicker it will die.
-
It's nothing personal this time, and in no way connected to the thread or actual posts, just...
It's a custom by now, that comments/guesses made upon publicly available information are questioned by some with insider information. Why is it then that most comments, information and hinted dates/infos by those with insider information turn out to be less precise than the estimated guesses based on the public info?
I just speak in general of the phaenomenon itself. Maybe being an "insider" is a bias in itself to some degree?
As to
You should add "that I know of, based on publically available information".
Otherwise your statement is just plainly wrong.
Do you expect everyone to add to each statement the same? Publicly not available information is that: not publicly available. In other words: no one knows about that, and I don't see the point taking into account each time the fact that there may be some unknown factor unless it is known...
-
CPU slot is a nice to have.
One way to use it would be to have a modest & cheap CPU on the motherboard and a CPU slot for CPU upgrade (heterogenous) / second CPU (heterogenous / assymmetric multiprocessing).
I'm afraid that in the very low cost models CPUslot is not possible, but in high end models it would give an easy way to build a lot of different performance computer models per need.
And if the CPU slot is open for CPU card manufacturers to compete, it would accelerate CPU price drop & adoption of the latest & fastest CPU technologies....
(what is currently stopping people from instaling macintosh megarray CPU cards in A1XE???)
btw. Would CBM have survived better without competing CPU card manufacturers?
Would they have made more money by providing a new motherboard every few months instead of letting CPU card manufacturers to flourish...
-
HyperionMP wrote:
>The only new motherboard on the horizon is the Pegasos II >which still uses the MPX bus.
You should add "that I know of, based on publically available information".
Otherwise your statement is just plainly wrong.
:-o
Sawin these seeds of hype! :-)
(well, uh, we have these ... GREAT products coming up, it's secret and you now nothing about it now, but it will ROCK!)
Could it be that there exist an opposing phenomen to FUD? :-P
-
Madgun68 wrote:
Bad idear.
Perhaps not. How long do you think people are going to keep upgrading if they are paying the kind of money Eyetech's charging for an AmigaOne? At least it gives people a cheaper option.
If the PPC mobo market is ever going to catch up in the very long run, then hacking on cpu's that get chocked by the mobo has to stop & i saying with this new PPC cpu that is out now then the time is right for all ppc mobo makers to make new mobos for it.
The fewer options the market gives its customers, the quicker it will die.
I agree with MadGun 68.
Accelerators make much more sense instead of dishing out £500 every so often for a completely new system. Whether its Eyetech's fault or not, the prices for such mediocre hardware is a bit of a sick joke.
33,000 units? :-P
-
The main benefit of the PPC970/G5 is the improved CPU interface to the rest of the system. To cripple it in an "upgrade" by latching it to the same slow bus that the G4's currently use is like throwing money away.
In that case, the PPC970/G5 may perform WORSE than the G4's of the same or lower speeds. Why? The G4's are more optimized to get the most out of the limitations of their current interface, 'ell, check out their cache arrangements. The PPC970/G5 is designed from the ground up (originally as POWER4) to expect a certain minimum bandwidth feeding the CPU. This assumption is even higher than the POWER4 since IBM's going to likely be using the same processor interconnect technology it co-developed with Apple for their PowerMac G5.
The only thing you'll get with the PPC970/G5 over a current G4 is native crunching of 64-bit integers (oohhhh) and a 64-bit flat memory model. For the A1's 2 DIMM slots (it is 2, isn't it?). I can't imagine people putting in two 4 GB DIMMs to bump their G5-upgraded A1 to 8 GB. That would be on the order of insanity, far above stupidity--if the A1's motherboard/firmware/chipset even supported that.
So what do you end up with? A motherboard swap to get the most out of a G5, ok, that adds to the expense and complexity. A CPU-only replacement which cripples the CPU and reduces the benefits of a native 64-bit CPU. Or a totally new system, A2 or whatever.
I don't think there will be any G5 upgrades for the old PowerMac G4 systems and if there are, I think they will not perform nearly as well as people would expect them to.
But that's just my opinion.
-
HyperionMP wrote:
>The only new motherboard on the horizon is the Pegasos II >which still uses the MPX bus.
You should add "that I know of, based on publically available information".
Otherwise your statement is just plainly wrong.
Amiga MCC NG ??!??
:-o
;-)
-
That sounds like a major problem to me. Just how many people own an AmigaOne Earlybird system anyway?
I do.
-
I don't know exactly what are the possibilities of Amiga manufacturers but a PowerPC 970 with a 64 bits AMiga OS would boost the plateform. Meanwhile the PPC970 needs a new motherboard with huge improvment as you can see on the PowerMac G5 motherboard.
-
It would constitute "hype" if I had mentioned some specifics or endorsed a specific solution.
As it stands, all I was reacting to is an overly static view of the situation.
Having said that, a PPC970 for the Amiga is not something for this year.
An a 64 bit version of AmigaOS is even further away.
-
I don't really get it why all the peopel in here seem to think the G4's trash just because there's a new PPC??? :-? Common, do you use such ressource hungry software all the time? Where do you REALLY gain from it (if you don't use a sick, fscking ressource hungry OS)? Do you encode divx all the time or don't do anything else than 3d modelling? Don't misunderstand me, I'd like to see a PPC970 Pegasos or/and Amiga, but I don't really need it now :-)
-
I'd like to see PPC970/G5 acellerator for A1200, just to show to my friends, that yes, it's the same 10y old tiny keyboardcase machine. but that has no other point than extravaganza:) still wouldnt be able to play Doom3;(
-
Do you encode divx all the time or don't do anything else than 3d modelling?
uhh, that's the plan at least
if you don't use a sick, fscking ressource hungry OS
Im curious how YDL or some-other microkernel variant o/s would do on a G5 or PPC970