Slashdot Mirror
PowerPC Goes 64 bit
prostoalex writes "ExtremeTech runs a story about IBM planning to introduce a new 64-bit PowerPC architecture for desktops in October at the Microprocessor Forum. The conference agenda tells us that "this processor is an 8-way superscalar design that fully supports Symmetric MultiProcessing. The processor is further enhanced by a vector processing unit implementing over 160 specialized vector instructions and implements a system interface capable of up to 6.4GB/s"." There's also a News.com story.
Is that "160 specialized vector instructions" the infamous motorola designed altivec? Is this the next processor apple will use? Does this mean all that intel talk was well.... just talk?
 
anyone know?
yes i run a goth/punk/emo porn site.
There is also an eWeek story.
If these are as good as they sound, all those speculations and rumors of apple switchin to intel are going to be thrown out the back door.
Later,
Phil
the fact that PowerPC high end is comeing down to the low end well who whould have that coming .....
the intresting part will be a 1.2GHz ARM part from Samsung useing the Alpha technology
(they say its ARM10 but I think thats wrong and its just ARMv5 complient but that sounds bad in marketing speak so thedy said it was like an ARM10(I think I am not sure) )
regards
John Jones
There are a few pages of good discussion here.
The "cue the foo posts in 3, 2, 1..." posts will commence with no subsequent foo posts in 3, 2, 1...
Motorola's 85xx processor, aka G5, is 64-bit, if this article is to be believed.
Is this IBM just coming out with their own 64-bit PPC core? I thought Apple, Motorola and IBM were in an alliance? Seems to me that its quite a competitive alliance, eh?
PowerPC has been available in 64 bit since the introduction of the A10 in 1996.
Here's some proof.
The new multi-code die is very interesting though...
Back in 1993-1994 when IBM was still working on the OS/2 for PowerPC and WorkPlace OS, and Taligent and "Pink" were still on the drawing board, IBM was planning to release the PowerPC 620 Series, a 64-bit version of the PowerPC 604. They intended to use it to run a 64-bit version of OS/2 that ran on the Mach kernel.
The design was scrapped because back then the manufacturing process was way too expensive to be cost effective in mass producing the chip. And we all know what happened to PowerPC OS/2.
http://www.byte.com/art/9411/sec8/art5.htm
POWER and PowerPC are mostly the same, but not quite. There are something like 10 to 15 instructions that either have different semantics or are present in one arch but not the other. IIRC all of those instructions are fairly esoteric supervisor mode instructions, so they'd likely only affect the OS and not user-space programs.
I requested this at IBM's PPC booth at linux world in JANUARY ;-)
what took them so long
Buttsex.
>>a PowerPC is a mac that can run some x86
>>applications and hardware right?
wrong.
PowerPC is a line of CPUs with similar instruction sets and architectures. It is developed by IBM and Motorola, and formerly by Apple. It is related to IBM's Power line. It has nothing to do, and is incompatible with, x86 CPUs. The PowerPC 60x, 750, and 74xx models are and were used in Apple computers.
I believe the current PowerPC instruction set is a subset of the POWER's, with the exception of AltiVec, which sounds like IBM has cloned with their claim of 160+ vector processing instructions. Also, the claim is that 32-bit PPC code should be binary compatible with this 64-bit chip. Sounds like a processor tailor-made for Apple to me, but there's not enough detail yet, and Apple is being very tight-lipped on its future processor plans.
___
Cogito cogito, ergo cogito sum.
Intel won the CPU war on desktop PCs. Look to servers, handhelds, game consoles, etc. for the the next CPU battle worth fighting.
Because the existing Power4 design is an expensive chip to fabricate! Those dual core versions are mucho heavy on the transitor count.
;-)
This is designed to be a "lite" Power4 platform for low-end servers and desktops.
Actually, if truth be told, it's probably being fairly and squarely pitched at Apple as their new CPU
-psyco
I find it amusing that the old i486 machine had
235 instructions. The "RISC" PowerPC originally
had 225 instructions. It now has 160 more
instructions. Compare this to 69 for Sparc
and 94 for MIP-Rx series of RISC processors.
Perhaps we need a new definition for "Reduced"
as it applies to the PowerPC. On the upside, at
least you can't say the PowerPC designers are
stuck on dogma =)
Nope, A PowerPC is what every Mac has currently. G3 / G4 are PowerPC CPUs.
PowerPC and x86 are completely different CPU architectures.
Apple going 64 bit will have almost no impact on developers or the OS other than to improve upon it. The only thing that will have to happen is that Apple, et al will have to recompile all of their binaries using a 64 bit compiler. That's it. If Apple or IBM already has the compiler for 64 bit PPC or at least a working version then it (64-bit Mac) is practically here already.
A fool throws a stone into a well and a thousand sages can not remove it.
The original specs for POWER/PowerPC CPU's were 64/32 bit anyways. This was set in stone over 10 years ago.
The great thing is that PPC-64 is that it's natively code compatible with PPC-32. No ISA 'extentions' (like x86-64), or instruction convertion (like Itanium), just a simple processor mode switch.
Apple would be a fool not to jump on this CPU for their high-end workstations or low cost servers.
PowerPC has nothing whatsoever to do with the x86 instruction set. It's just a brand name for a CPU architecture, like "Pentium", "Athlon", or "Alpha". In fact, it isn't even mac spesific.
autopr0n is like, down and stuff.
They're been delayed so long now I'm starting to wonder if they're hanging out with Duke Nukem, or even Prey fer chrissakes.
hmmm did YOU read the article?
" IBM's presentation, for example, says the company's 64-bit PowerPC processor will be designed for desktops and entery level servers. "
Does that NOT say PowerPC? Does it NOT say Desktop?
The chip is based on Power4 manufacturing and tech that's all.. it is a PowerPC.
A fool throws a stone into a well and a thousand sages can not remove it.
There's a fair amount of discussion in comp.arch on this chip. Guesses are that it's adopting the Motorola e500 vector instructions, ones that work on general-purpose registers. Considering that the Power4 has a very, very, very good floating-point unit that works with the memory units to get vector-like performance, I doubt if they need FP vector ops...
Apple no longer benefits from not being x86...cost being the biggest issue
C'mon, you can't be privy to internal Apple component pricing and not share it with us... You are sure they pay more than Intel would charge, even though they buy processors in lots of a few hundred thousand or so, right?
and most of the time now they can't even claim a performance gain.
And the 8-way superscalar 64-bit G4 still won't help, right?
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
This sounds like a beautiful processor, and I'm guessing it's why they contracted a plant in upstate New York to manufacture the .1-micron chips. I'm thinking that this is the mutant brain-child of the Power4 and the G4. Kind of like a miniaturized Power4 with a vector-processing unit, running at 64-bits and possibly with 32-bit PPC binary compatibility built-in. It would be nice if they could apply double the SMP capability to 32-bit code.
Hopefully they'll write a really efficient compiler for it. This could be the chip to launch Mac OS X into the enterprise market.
Karma: Ran over your dogma.
Well, I think it's a little more complicated than that. What about other hardware in the system that uses 32-bits of address space? What about all the driver software that assumes physical addresses are 32-bits wide? Maybe not super difficult, and helped by the fact the IOKit is an object oriented framework, but still harder than flipping a switch.
If I recall correctly, when Nintendo had IBM create the processor for the gamecube, IBM retained the rights to fiddle the their specialized PPC chip, then resell it. I'm wondering if the vector instructions are spillover from the Gamecube chip (I know it does a lot of fast vector math).
You must be a blissful fellow (read: ignorant)....
RISC vs CISC is a concept of design, there is no set amount of instructions for either. Go back to computer science 101.
sin(6cos(r)+5A)
Actually, Apple File Exchange came with System 6 and would let you transfer a file off of an MS-DOS floppy. PITA to use, but it was Apple Software.
Hire a Linux system administrator, systems engineer,
Will Linus soon be praying that AMD and Intel switch to these babies?
Power 4 architecture has an I/O architecture that is WAY too expensive for workstation/low end server use. The PowerPC 64bit, however, would be quite a CPU. To see if Apple is interested, look to see if BSD runs on Power Architectures ...
It's already there, been there for some time, and IBM told me that Apple had Darwin and some GUI running. Apple just needs the market to see that it weould be worth the investment in a new mobo/system design.
dot-sig.
This should be interesting to check benchmarks with. Now we'll be able to look at Itanium vs. Hammer vs. PPC-64. That might be a bit interesting..
PS. Check out my friend's band on this site. They're called Hat Trick of Misery.
-- "Government is the great fiction through which everybody endeavors to live at the expense of everybody else."
Power4 has *huge* cooling requirements, despite being copper-interconnect and all that. (it also has something like 5800 pins, btw, drawing somewhere in the range of 100A worth of current, IIRC) -- I wonder how much cooling needs to be for the 64-bit power PC if they are based on the Power4 design?
My life in the land of the rising sun.
I say to myself. "Self, I wish this damn iBook wasn't burning a whole through my pant leg."
I do have to say that the iBooks are VERY nice though. Good performance at a great price. My wife loves hers, the only complaint either of us have with it is that it does heat up under the hard drive, and a small fan couldn't possibly hurt to push the air around and out the large vents on the left side where most of the heat builds up.
Sorry, but market forces are now as powerful as performance metrics. Apple no longer benefits from not being x86...cost being the biggest issue, and most of the time now they can't even claim a performance gain.
Intel won the CPU war on desktop PCs. Look to servers, handhelds, game consoles, etc. for the the next CPU battle worth fighting.
Until we have a monoculture in all our products, and have eliminated every trace of competition or choice, everywhere?
You waive your hands at the "invisible hand" of the free market as an argument for competitors to not even try competing for a portion of the marketplace, in effect advocating the replacement of a market with competitors with an intel monopoly.
I suspect you do not even see the contradiction in your argument, so let me spell it out for you. Monopolies are antithetical to a functional Free Market. Without competition the entire basis for capitalism functioning in any worthwhile capacity at all is removed and no free market exists. In short, without competition capitalism dies, and the free market "authority" you are alluding to becomes meaningless.
It astonishes me how people can argue "the market says" with one breath and "everyone should cave and give company X a monopoly" with the next. Indeed, one is forced to wonder if much of the current economic chaos isn't a result of an entire graduating class, perhaps an entire generation, not understanding even a little of economics in any context other than the inflated (and as it turns out largely fradulant) boom of the 1990s.
I won't even get into the fact that free markets are but one force, one tool, necessary for a functioning society or culture, another point often ignored in our western myopia, but that is a discussion for another thread.
The Future of Human Evolution: Autonomy
More liek 99% of the population will never need a full 128 bit integer support. The biggest reason for moving to 64 bit is to increase the amount of memory the system can address. Moving up in size actually has disadvantages, though. Every time you're moving data around, you're pushing twice the amount of bits, whether you need them or not.
BlackGriffen
He'd do better in a Comp. Architecture class, or reading some of Hannibal's stuff over at arstechnica. He's got a knack for explaining modern processors. Of course, neither the PPC or x86 architectures are really purely RISC or CISC anymore, either.
You can only drink 30 or 40 glasses of beer a day, no matter how rich you are.
-- Colonel Adolphus Busch
You're not getting particular great answers to this question. :-)
The PowerPC is a completely different chip architecture (search for explanation of RISC vs. CISC if you want more detail) in comparison to the Intel x86 architecture. There is a lot of complex discussion on this, but theoretically speaking the Power PC architecture is more advanced and efficient at the same processor speed.
At any rate, the Mac OS is optimized for this processor type. I guess it could be rewritten to run on x86, if Apple wanted to do that.
With the help of an emulator, Win32 programs can be executed in the Mac OS environment. (Virtual PC is the one that I can think of off the top of my head.) It's a testament to the architecture of the Power PC that the performance of Windows in an emulated environment is pretty good. (Not a computer science person, but my understanding is that Virtual PC makes Windows thinks that it's on an x86 computer...and it's an elegant hack.)
I think other posts here are discussing how Apple can/will migrate Mac OS X to the 64 bit processor, and whether or not 32 bit programs need to be recomplied/redesigned for the new processor, or if they can run directly on it in some sorta emulation mode.
Mac OS X already takes advantage of multiple processors. Think Xserve and the dual processor PowerMacs.
Actually, the PPC has usually run much cooler than the equivalent Intel and AMD chips. It's what lets Apple get away with some of the case design decisions they make. It also is going to be a plus for them as they get more into the server market. Lower monthly cooling bills is a significant recurring savings.
Whatever 64-bit PPC CPU Apple ends up with, be it from IBM, Motorola, or a new partner like AMD or nVidia, will almost certainly have to natively run 32-bit code.
My bet is that it's either it's reverse engineered or it's licensed, not that it's a variant. Apple isn't likely to want a variant different enough that Motorola's lawyers could claim infringement.
Who'd a thunk it, we've arrived at a day when IBM is the reverse engineering firm!
I think it deserves mention that OpenPPC.org is an IBM-supported initiative to bring OpenPPC-based machines to us commoners.
Get off my virtual lawn, you damned virtual kids!
Yeah, because everybody knows that IBM doesn't have the money to retaliate.
[giggle]
The market caps on these companies are about the same. If you're going to play 'steal the employee' it's always wise to play that game with a much smaller company.
Basically, this creates a 64-bit option for Apple. Very important for a Unix vendor. All you'll lose is the smaller address space :-) If Motorolla really is focusing on the embedded market then letting IBM start producing AltiVec (or something much like it) is less of a shock than when they were looking at the desktop market. Likewise, IBM now sees Apple expanding both their desktop market and their new server line, and an AltiVec optimized GCC 3. Apple, looking onwards and upwards sees Motorolla losing interest, and a line of IBM chips that scale from their fast, cool G3s through the Power line for medium-to-big iron.
Put those three viewpoints together and it looks like Jobs has found that Motorolla alternative he has been talking about.
Veteran, Bermuda Triangle Expeditionary Force, 1992-1951
I had a 500mhz iBook running 10.1, and while the OS was pure heaven, the speed of the iBook was pure hell. I constantly said "damn, I wish this was faster" until I finally gave up when I noticed it'd take me twice as long to browse the web as on a PC. What a drag. I sold it and am looking forward to the day when I can get a decently fast iBook. Maybe the 700mhz model is a lot faster, but I'd be surprised.
$45 per U Colocation Special
Correct. gcc and AIX xlc compilers by default output the common subset - the moral equivalent of -march=486 or so.
Mostly extensions, a few dropped or changed facilities. It's worth noting that the PowerPC is much closer to its immediate ancestor the POWER2. Here are some highlights of POWER -> 32-bit PowerPC, taken from The PowerPC Architecture (IBM, 1994):
The main changes listed for POWER2 -> 32-bit PowerPC seem to be dropping several opcodes for floating point loads/stores, and some semantics of FP <-> int conversions.
"How can you claim that you are anti-crack, while still writing a window manager?" — Metacity README
enough said.
I've seen theoretical arguments for this. As bandwidth in the computer widens a lot of problems that currently are done with an eye towards efficiency can be solved using simple, inelegant brute force techniques. The problem really is that there aren't enough savings to justify the massive extra expense of retooling everything else. After all you wouldn't shoehorn that 128 bit processor into a dinky memory system that would result in a mostly idle processor, now would you? It's all the other components that need to be updated before it becomes practical to take the next step.
Incidentally, Connectix (who make VirtualPC) at one time had a faster 68k emulator on the market. They're Real Wizards at Connectix. VirtualPC doesn't emulate Windows, it emulates an x86 PC at the hardware level. Then you install a standard OEM copy of whichever version of Windows you got. Then you install drivers for the emulated PC hardware. You can also load it up with a basic DOS image, reformat the virtual drive and install Linux or a free BSD. Neat stuff.
Veteran, Bermuda Triangle Expeditionary Force, 1992-1951
Actually, they shouldn't have to recompile anything, unless they want their apps to use 64-bit addressing. (but they will have to modify the OS kernel)
Real 64-bit chips which are new versions of older 32-bit lines are fully backwards compatable.
Case in point, you don't need to recompile 32-bit SPARC applications to run them natively on a 64-bit UltraSPARC.
This was obviously a case of somebody who has been conditioned by the M$ marketing empire to believe that "PC" is a registered trademark of Intel or Microsoft or somebody, and it somehow stands for "Windows Computer."
When he saw that Apple had something called "PowerPC", he immediately assumed that it must have something to do with Wintel compatability, and in his mind tied it to various rumors of various cross-platform projects he overheard while walking past the cubicles in his office.
That, or he knows perfectly well what PowerPC chips are, and he's just trolling.
Information wants to be anthropomorphized.
Is that sort of like the Macintosh being a less powerful version of the Xerox Snap system? How many snaps have you seen? I've only seen one (Voice of America used the system, maybe they still do). There are certain price points where you get quantum jumps in acceptance. Microsoft is starting to hit a nasty one for them with low end PCs no longer being able to afford the MS tax, IBM is trying for a Power4 sweet spot to sell their chips in volume.
PowerPC was designed from the POWER architecture to replace it, and has been designed from the beginning to support 32-bit and 64-bit versions. The architecture is actually designed from the beginning to be a 64-bit architecture, and the common desktop implementations are only the 32-bit subset of the original design.
Here's some additional background info:
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
The sooner Apple starts on the mobos for these the happier I'll be. Besides, nobody ever got fired for buying IBM :-)
Veteran, Bermuda Triangle Expeditionary Force, 1992-1951
One of the reasons Apple has stuck with Motorola's PPC chips vs. IBM PowerX chips is the AltiVec units on the PPC chips. This new chip, while based on the Power4 chip is geared towards desktop use and not server use. Think of a passenger car vs a semi-truck. They both do basicly the same thing but are more optimized for certain uses. The desktop chip needs to be quickly responsive to a wide variety but small concurrent number of processes or threads while the server chip needs to handle large numbers of threads or requests to do pretty much the same thing (file access, database access, ect.).
IBM's just bringing down a high horsepower diesel and gearing it for passenger use. They're also throwing in AltiVec type of processes which Apple really likes/needs. One more reason why going to X86 chips would be difficult; no AltiVec.
I drank what? -- Socrates
First off, I know almost nothing about the technical aspects of processors, so if this sounds like a clueless question, my apologies. However, I am aware that programs compiled for 32-bit processors won't work on a 64-bit processor. If that's the case, what happens if Apple should jump to this chip? Does that mean we Mac users have yet another OS 9-to-X-type wait while developers drag their feet updating their apps or is there some way that a 64-bit processor can also handle 32-bit apps? As far as I can see, that's the only problem with Apple going to this chip. I'm not sure how eager Apple will be to annoy users who are finally seeing the light at the end of the OS X tunnel.
--Rick "If it isn't broken, take it apart and find out why."
They can, they can. There are plenty of good reasons not to:
In short, provide even one application domain where having 128 bits of addressable memory, or a convenient 128-bit word size, would come even close to offsetting the inherent architectural costs compared to a 32- or 64-bit design. I can't think of one.
NO, IPv6 isn't a valid answer! (: Word size hasn't been a significant obstacle for current implementations.
"How can you claim that you are anti-crack, while still writing a window manager?" — Metacity README
NOBODY HAS SAID THAT APPLE WILL USE THIS CHIP. Everybody *suspects* that they *might* use it. But everyone is blowing smoke and guessing until we actually hear news from inside Cupertino.
IBM will use this processor in their machines. We have no evidence whatsoever that this chip was designed to be compatible with Apple software. It might be.
There are no trails. There are no trees out here.
If I remember what I've read before correctly, POWER is little-endian. When Apple/IBM/Motorola started working on the PowerPC, one of their goals was to run AIX and MacOS (though not simultaneously). And, as we know, 680x0 processors were big-endian. So, the PowerPC was given an endianness switch, and the firmware was tasked with flipping it into the required position during boot.
There's more detail about the PowerPC's heritage (including its relation to Motorola's 88000 processor) in this Wikipedia page.
In Soviet Russia, Jesus asks: "What Would You Do?"
IBM has alread licensed Altivec from Motorola as far back as 1998-2000, allegedly to help with the design of the PowerPC variant used in the Gamecube. The PowerPC 405 embedded processor in the Gamecube contains 38 additional instructions for vector FP math (vs. the 162 in Altivec). A glance at this PDF file from the web makes me pretty sure that these aren't just lifted from Altivec. Instead of Altivec's 128-bit 4 32-bit FP vectors, Gecko adds instructions for fitting 2 32-bit FP numbers in a single 64-bit FPU register and working with them. It also adds some odd but interesting MMU features.
Anyway, I know it's been licensed because back in 2000 there was a lot of conspiracy theories that Motorola was preventing IBM from selling faster clocked PPC chips to Apple than they could produce via an obscure clause in that license. Both parties denied it, of course. I don't really believe that was the case. I think it was just bitter rumor-mongering by Mac users who were rightfully angry at Motorola for pissing away the performance (and MHz) advantage that PowerPC had on x86 chips back in the 603/604e and Pentium/PPro days.
Oh, admittedly, the MHz advantage went away as Intel/AMD extended their pipelines for that explicit purpose earning theirselves increased performance penalties for mispredicted branches and requiring increased CPI for many instructions, but I still miss the days when PPCs were faster per clock AND had higher clock rates. Now the clock rate advantage is so extreme that the PowerPCs' better performance per cycle doesn't catch up for the most commonly executed code. Once again, though, I digress.
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
You've got the wrong kind of "proprietary" in mind. Compaq used proprietary boards so that you couldn't upgrade with standard parts, were locked into their upgrade path. For a standard user, this did not offer any additional functionality.
Apple uses proprietary boards so they can offer features like autoswitching networking (just plug in an ordinary ethernet cable between two macs, and the two computers show up on each others local network), target disk mode (use a scsi or firewire cable between two macs, and one computer becomes a HDD on the second computer), instant dynamic network configuration (change your IP/ or configure multiple network devices with just a few clicks, no restarts), dynamically driving multiple monitors with multiple cards (I can plug in two graphics cards, and two monitors, then tell the mac which monitor to drive with which card, while its on.), and USB/Firewire plug and play ease that's still years ahead of windows (oh look, it's the windows hardware manager, again...). On the portables, multimonitor/external monitor support is so slick, it's enough to make a Wintel laptop user cry.
There are plenty of things you _can't_ do as a result of proprietary HW, such as move as quickly with the industry as new HW comes out (lets see how long it takes apple to get AGP 8x...*roll*), but the main differences in functionality between MacOS9 and Win9X/ME/XP is the hardware tweaks that you don't realize by using "open" HW. In the Wintel world, the peripherals people don't work with the OS people, who don't live on the same continent as the BIOS ppl, etc. etc.
Mac admins live for these tweaks, since it means hours less frustration and "pointless clicking" to set up an office of computers, or get them networking just so, using external devices, while adapting the systems to individual work flow and idiosyncrasies.
This is where I get my recommended daily allowance of "Foot in Mouth."
I've been running into a new acronymn the last few weeks
:)
;)
;)
VMX.
Its a PowerPC vector instruction ISA...
Anywho, the most interesting run in was in an IBM publication where they referred to AltiVec/VMX as the vector instructions on a Motorola 74XX 'G4' CPU
Anywho, point is, the only people in the entire world I've heard referring to AltiVec as VMX are IBMers
But IBMers like to use alternative words to the rest of the computer industry
Anywho, my prediction is these chips feature a VMX unit
It is a very IBM sounding acronym, and matches up with other vector ISA names... MMX, SSE, SSE2, VIS, MDMX etc
---
Live Long & Prosper \\//_
CYA STUX =`B^) 'da Captain,
Jedi & Last *-fytr
Why not use the proven high-performance IBM power4 design instead. Why design something brand new that will have "less" performance that the already existing 64bit single/dual core power4 design.
From the register article:
It needs to remain competitive with entry-level workstations against the likes of Sun and HP's Alpha, where the size and heat dissipation of the mighty POWER4 have kept it out of systems below $12,000. IBM's desktop workstations still run POWER3
IBM would need to design this chip even if Apple didn't exist, simply because the current Power4 cannot be produced cheaply enough for a $10,000 workstation, much less a $800 macintosh.
"(Man) tries to live his own life as if he were telling a story. But you have to choose: live or tell." --Sartre
The firmware :)
http://www.openfirmware.org/
---
Live Long & Prosper \\//_
CYA STUX =`B^) 'da Captain,
Jedi & Last *-fytr
Correct in concept, wrong on some details. RS/6000 and Mac were both big-endian.
POWER chips were big-endian only. POWER had a few instructions with a big endian bias, i.e. the used computed values as big-endian indices into registers.
PowerPC removed the biased instructions (and made other changes too) resulting in an "unbiased" architecture.
I don't know what caused PowerPC to become endian-agnostic. Maybe it was a desire for elegance. But it was not a desire to accomodate both MacOS and AIX 3 heritage; they were both big-endian.
- Clem (POWER/AIX user since about 1992)
OTOH, an IBM 750FX @ 800Mhz uses less than 5w and performs _better_ than a Motorola 7455 per clock (unless the Altivec hardware on the G4 comes into play)
Based on the award winning Power4 design, this processor is an 8-way superscalar design that
fully supports Symmetric MultiProcessing," the description says. "The processor is further
enhanced by a vector processing unit implementing over 160 specialized vector instructions and
implements a system interface capable of up to 6.4GB/s.
It's the 8-way ss + new vector instruction set that's new. The 8-way would drive the overall bandwidth requirement.
Clearly this is no longer a RISC design because the original PPC instruction set had what, 174 instructions? So add 160 more and you have 334.
If this chip is going to be used in iSeries systems, then it's actually going to be a 65-bit CPU for real, at least when run in tags-active ("OS/400") mode. Careful what you joke about!
actually, a 700Mhz Sahara should be able to smack a 600Mhz Coppermine P3 around quite handily.. not only is it based on the original 4-stage G3/G4 pipeline (which is higher IPC than Motorola's current 7-stage G4's) but it also has a half meg of L2 cache on the die with a 256Bit link to the processor core.
If Apple ever moved to Intel, they would be crushed.
You imply that moving to x86 (Intel) would mean Apple no longer maintains a separate platform from Microsoft, but this isn't necessarily so. Just because Apple currently uses one processor architecture in their machines doesn't mean they can't switch to another while maintaining their separation from the PC world with proprietary ROMs and the like.
The idea of Apple moving to x86 doesn't mean they would release a boxed OS X that runs on a standard Dell box. More likely they'd just come out with new Macs that are functionally identical to the current machines, but with an Intel or AMD chip inside instead. That way they maintain their focus as a hardware company first, while gaining the ability to run Windows apps natively and the economies of scale that result from using CPUs that are practically commodity items.
The Mac isn't about being the fastest machine on the block, it's about being the best designed, easiest to use, most useful machine on the block.
And if a Mac had a blazing fast x86 chip inside it'd STILL be the best designed, easiest to use, and most useful machine on the block. It'd just be fast as shit as well. Not that I'm not happy with the speed of my Mac, but if the speed gap gets any wider Apple would be crazy not to consider switching.
All movements for social change begin as missions, evolve into businesses, and end up as rackets.
Apparently the single core lite versions are just multi-core versions where the other cores failed testing...
:)
;)
of course I have no idea if this is true or not
Have you seen the size of those Power4 MCMs?
I can't really imagine an iMac with one
---
Live Long & Prosper \\//_
CYA STUX =`B^) 'da Captain,
Jedi & Last *-fytr
I've written some highly optimized code for MMX/MMXExtended and altivec.
:)
;))
A 500Mhz G4 is about 30% faster than a 1.3Ghz Athlon when running these image filtering functions...
And the Athlon is using DDR, while the G4 has only a 100mhz bus... and the functions place a huge demand on both the integer units and the memory interface.
Basically my point is...
I'm not convinced that a 2.5Ghz P4 is faster than a DualG4... when running highly optimized vector code.
And that's of course the reason people need 2.5Ghz PCs
(don't tell me you need that much power for a typing letters
BUT, I haven't written SSE2 versions yet, and SSE2 does seem to be fairly competitive with AltiVec (not as good, but competitive, AltiVec makes MMX 'feel' like a cheap toy)
---
Live Long & Prosper \\//_
CYA STUX =`B^) 'da Captain,
Jedi & Last *-fytr
intel said the Xscale would do 900MHz but all the test chips and produvtion has been 400MHz
so really unless you have benchmarked it (A chip which I havnt seen) you can say that
compare a 400MHz G3 against a 400MHz Xscale and then we can talk
in terms of apps like telephonery and video then the ARM wins because of the DSP extensions in terms of floating point add a VPU and then the ARM is not to shabby
regards
John Jones
The difference is that each of the 142 instructions had multiple modes, multiple formats, multiple argument lists and where compeltely different from each other
RISC means reduced complexity, not reduced instructions, all the instructions can roughly be divided into
load/store
processing
branching
that's more or less it, and they're all the same in each category, they just do different things
so you have loads, they all load from memory to a specified register, there are different instructions for loading a 8 bits, 16bits, 32 bits etc, and some different ones to handle signextension etc.
meanwhile the processing instructions are basically all of the form
C = A + B
where A, B and C are any registers, B could be a number instead of a register. The + of course can be one of many operations, and thats the difference between teh instruction
All the instructions might update the condition registers, or might be word, byte, or halfword variants... but there aren't very many base instructions but, if you combine the base instructions with the 5 combinations of instruction for every instruction, then you get a fairly large number.
But because the ISA is so elegant its very easy to learn. And its very easy for the CPU to process it.
In relation, to a CISC architecture (which is a retroactive name), any RISC design is a model of simplicity.
---
Live Long & Prosper \\//_
CYA STUX =`B^) 'da Captain,
Jedi & Last *-fytr
Ahem...
;)
:)
The big ones (IBM's) use PowerPC chips...
They just call them POWER3 and POWER4
IBM doesn't make big ones which don't use PowerPC's as far as I know.... all the ones which use Intel's chips aren't.... big
---
Live Long & Prosper \\//_
CYA STUX =`B^) 'da Captain,
Jedi & Last *-fytr
WTH are you talking about?
Apple uses 64bit FPU operations on the 64bit FPU...
BUT the altivec unit is only capable of 32bit FP operation (of course it does 4 of them at a time)
---
Live Long & Prosper \\//_
CYA STUX =`B^) 'da Captain,
Jedi & Last *-fytr
That all 160 of these vector instructions take that format? I do find that a bit hard to belive.
Also, load store instructions don't take that form.
autopr0n is like, down and stuff.
What, you're going to blame this fiasco on one age group being myopic
Absolutely not. The wrongdoers and politicians span several generations.
My point was how commonplace the rhetoric of "free market says, so everyone should just cave and do" has become. This is a relatively new development, one that flies in the face of economic theory, free markets, free societies, and a functional society, and one that would have been laughed into submission a few short years ago.
These days one rarely hears a rebuttal, so I begin to wonder why that is and speculate that there is perhaps a group of people who do not grasp capitalism, perhaps because the only form of it they've been exposed to is the corporate perversion of it we've had throughout the late 80s, 90's, and early naughties.
Nothing more was implied or intended than that.
Oh, and if any one generation were to take a big chunck of the blame for the philosophical myopia and malaise that has engulfed so much of the American psyche over the last twenty years, I would most certainly rank my generation at the top of that list, though, as you rightly point out, the blame is a little more widely spread than that.
The Future of Human Evolution: Autonomy
I don't know the technical details, but from my own experience:
VPC doesn't just make windows think it's on a PC, it actualy emulates the x86 environment, right down to the hardware. This means you could install any x86 based OS to your mac. It uses (dynamic?) disk images to serve as the harddrive, and runs the machine in it's own window. As near as I've been able to tell, the only thing VPC can't do well is video games. Oh well, that's why I bought a PS2
T Money
World Domination with a plastic spoon since 1984
Correction:
This: "Intel has *not* monopolized CPU sales quite yet."
Should read: "Intel has *not* monopolized x86 CPU sales quite yet."
Actually I should also toss the "quite yet". I cannot predict the future after all.
Table-ized A.I.
gelfling wrote
and, earlier, autopr0n wrote
RISC is about reduced complexity and designing architectures for best use by optimizing compilers.
If anyone actually wants to learn something (rather than simply spouting ignorant marketing propaganda) you can check out John Mashey's periodic RISC vs. CISC article from the comp.arch newsgroup.
The essence of the RISC philosophy can be summed up in this quote from Hennessey & Patterson: "Make the common case fast and rare case correct." The question is, however, how do we know what is the common case? The answer from H&P is to look at the occurrance of various instructions in actual programs. Since most programs are generated by compilers, this amounts to looking at what is used most commonly by compiler code generation engines.
Once you have made an inventory of instruction usage, what you find is that lots of time and instructions are spent moving data between registers and memory because you don't have enough registers, and over 80% of the addressing is done with only a couple of addressing modes. In other words, compilers are simply no good at making use of most complicated instructions common in most pre-RISC architectures (including the 80x86).
When you design an architecture that caters to the kind of usage a compiler likes, you come up with an architecture that has lots of registers and performs all arithmetic operations on values in the registers, only accessing memory with simple load and store instructions supporting a few simple addressing modes (absolute address, register direct, and, possibly, register direct with immediate index). Along the way, in order to make various CPU bookkeepping tasks as simple as possible, you will use a fixed length instruction word with only a few instruction formats, and ensure than any single instruction can only cause one fault condition (memory access fault, instruction operand fault, etc.)
This general description is a pretty good match for almost all, so-called, RISC architectures: MIPS, M88000, ROMP, POWER, PowerPC, PA-RISC, Alpha, AMD29000, Sparc, ARM, Clipper, etc. The fact that all of these architectures share so many common features says something profound about the descriptive power of the term 'RISC' which cannot, easily, be said of the converse term, 'CISC'.
Just to head off the other popular mis-conception concerning RISC vs. CISC: RISC has nothing to do with the implementation features of a given processor, only with the programmer visible architectural features (number of registers, instructions, instruction encoding, etc.). Hence, no matter what name Intel or AMD use to describe the current dynamic micro-coding scheme used to salvage another generation of bloated, power guzzling, 8-bit microcontroller follow-ons, the 80x86 will never be considered a RISC processor.
It actually emulates the hardware of a PC so Linux x86, *BSD, etc will also run fine. One of the reasons that VirtualPC is surprisingly not shitty is that the PPC has a few little tricks to help. For example when emulating x86 code, VirtualPC throws the PPC into little endian mode (it's natively big endian). This saves lots of byte swapping instructions. The PPC was designed to be a great emulator.
Duh! Apple never talks about unreleased hardware this far in advance. It's a slam dunk that the sales force that is going to be tasked to sell this will be looking at Apple as a large client for the chip and that they will be pressuring the engineering team on making their SIMD system Altivec.
I suspect that the engineers aren't going to resist that too much.
(OSX is way too slow now, maybe it just needs a couple of procs for "eye-candy"...)
I'm getting sick of hearing people say OSX is slow. I run OSX on an old 300MHz G3 iBook with 128MB ram. I find it more than usable.
OSX 10.1.5 gave hardware accelerated support for my ATI video which made it suddenly much quicker visually. 10.1.5 is much quicker than the 10.1.3 I have on CD.
OSX keeps getting quicker and plenty of RAM helps too, which is why I'll be upgrading this old machine to 320MB soon. Jaguar should be even quicker again.
Now having said all this, I imagine running Jaguar with RAM maxed out on any high end G3 or G4 should be very nice.
Coming from a Lintel/Wintel background, I am loving the combination of excellent stability and extreme intuitive usability.
War crimes, torture, lies, illegal spying... Would someone give Bush a blowjob, already, so he can be impeached?