The Register takes a look back at that day a decade ago when Apple’s announcement of a RISC-based Macintosh, the “Power Mac” really shook things up. The PowerPC was going to trounce the aging X86 architecture, and herald in a new era of fast computing. Problem is, Intel had a few tricks up its sleeve too. Apple’s real achievement, though, was making the transition to a new architecture relatively painless.
I wish it were possible to add X86 instructions to the Power PC. Then IBM could compete with Intel and AMD. It would also alow Apple to transition to industry standard technology.
x86 is not a standard.
so, it’s good that the industry standard is heading for over 100W per chip, and barely gaining speed, where the new 970FX’s(next gen G5’s) are supposed to get around 50W power dissipation at 2+Ghz? doesn’t sound like progress to me, unless you’re intel, amd, or a power company…perhaps a refrigeration company as well… =D
Maybe not in the ISO sense, but x86 is a de-facto standard nonetheless.
I wondered how long it would take to get the “apple should go x86” post, but that was faster than I expected.
I bet it would surprise you to know that most CPUs in the world running computers are NOT x86.
what kind of de-facto standard is that?
and, I am sorry, but X86 is a horrid chip family…100 W for 3 GHz!!!!
the G5 will top out at 3+ GHz and be half, if not less than that in wattage.
heck, at 2 GHz, they have it at 25 Watts.
I bet it would surprise you to know that most CPUs in the world running computers are NOT x86.
Well said. Even the chips going around pretending to be “x86” aren’t really x86 chips at all. If I remember correctly there was an Ars Technica article that explained how modern “x86” CPUs work. They take x86 instructions and convert them into RISC instructions. So deep inside your Pentium 4’s and Athlons, they’re really RISC.
(if I’m completely wrong please forgive me, will try to come up with a link)
@ Dan
I wish it were possible to add X86 instructions to the Power PC. Then IBM could compete with Intel and AMD. It would also alow Apple to transition to industry standard technology.
IIRC there was an IBM PowerPC chip with x86 compatible instructions registers(!) inside.
I bet it would surprise you to know that most CPUs in the world running computers are NOT x86.
If you’re talking about all CPU’s on the planet, including those embedded in cars, trucks, microwave ovens, watches and electric toothbrushes, then you’d be correct.
Including things like game consoles and handhelds, the x86 is barely a blip on the map, historically, though the XBox is certainly more than that if you only count recent years.
As far as general purpose computers used by consumers, the x86 platform is vastly dominant. And when someone mentions “computers”, general purpose consumer computers is what they’re mostly talking about. Not cars, not phones, not satellites or thermostats.
Intel may be less dominant than it was, particularly when you consider AMD, VIA, and such. But those manufacturers are compatible with the x86 architecture for a reason.
It’s a defacto standard for desktops. Saying it “isn’t a defacto standard” is like saying light bulbs aren’t a defacto standard, because most light in the universe comes from stars.
current 2.0ghz g5s do 50watts, the new xserve ones do half that.
Well said. Even the chips going around pretending to be “x86” aren’t really x86 chips at all. If I remember correctly there was an Ars Technica article that explained how modern “x86” CPUs work. They take x86 instructions and convert them into RISC instructions. So deep inside your Pentium 4’s and Athlons, they’re really RISC.
Most – if not all – x86 chips work that way since the Pentium Pro so it’s not something really new.
As I recall, as an fan on the sidelines, the PPC was faster at first. The original PPC designer team was amazing and beat deadlines and expectations with their work. Although I have not researched this, I can say from the sidelines I noticed a lot total BS from management about the PPC. For a long time Motorola, IBM and Apple were trying to agree on a Common Hardware Reference Platform – CHRP – or some derivitative. I think management totally screwed up on this and deflated the PPC momentum. I believe if the three had worked out a reference platform and all of them invested in the desktop then things would have been very different. IBM saw early on that there was little benifit to the PPC w/o the reference and they drew back. Motorola continued to develop PPC mostly because they where using the design in many of their embedded chips so why not make a little profit on the desktop chip? But the desktop chip was not a big money maker nor was it a real priority.
Management politics, not design integrity, is what undermined the PPC. Maybe this is turning around – I for one got a G5 and it seems many others are – but the PC market seems a little dry to me and I think IBM has bigger visions to pursue… I’ve followed PPC since the very beginning, I remember reading about it in EETimes (http://www.eet.com/) before it was released and thinking that it was a sweet design. I’ve had a soft place for it ever since.
…. and you’ll probably find the worlds most popular chip is either MIPS, ARM or PowerPC
ARM and MIPS are commonly used in small computers and controller applications which are still computers, just not PCs. embedded systems, etc
PowerPC chips are widely used in telecommunications routers and some games consoles (possibly to include the future XboX2)
That’s really not a fair comparison. You’re comparing a 90nm chip (the PPC 970FX) to a 130nm chips. No duh the latter is going to use more power. A 90nm Opteron would use a lot less power too. And there is no way you’re going to see a 3GHz PPC 970FX at 50 watts. It uses 50 watts at 2.5 GHz. A 3GHz version would be at least 60-70 watts. At that point, its not too far off from the 100W mark set by Intel.
For me it’s in the OS, I like the way OS X multitask,and it does it much better than my XP machine(2800+, 512 ram). Is there a way to make my XP machine behave more like my mac? because, I can notice a difference, XP doesn’t handle heavy graphical multitasking as well as OS X. Maybe now that I’m thinking,,, maybe it’s the chips,,,,because the PPC has some advantages over X86,,,,what do you think?
I would prefere someone who runs both Mac and Windows like I do to answer. I like both systems and use em all the time.
and it does 100W.
1) Intel didn’t start decoding instructions to micro-ops until the pentium pro (which is really a risc core/x86 decoder, as is everything else based on it’s core (P2,P3&variants).)
2) OS X could probably be ported to x86 easily (there are ports of darwin (OS X – GUI) for x86, but apple likes selling hardware (and given the premiums they make, their fans/fanatics, I’m not suprised.)
@spaceboy29: Your problem is Windows. If you want your computer to be more like a mac’s core, install FreeBSD. If you want it to be graphically similar, install a KDE 3.2 using Linux distro. (SuSE or Mandrake come to mind.)
(Similar situation: a friend has an XP2600 (has windows at various points I belive 2000 & XP on it) and a P3700 (gentoo linux/kde), and in terms of reponsiveness of the GUI, the 700MHz machine beats the 2600+.)
The new prescott is a 90nm part, and it’s rated at 103W. The G5 is just much more efficient (as is the Opteron, though heat wise, the Opteron loses to the G5).
Raynier,
the equatiation smaller technology = less power consuming chips is no longer valid. Think of the leakage current, I can’t say a correct number, but today’s chips even if they are not connected to clock, or the frequency of the clock is zero, they consume a large part of the total power. By the way, IBM uses SOI for their chips, this should help a little against the leackage. Intel can’t handle this technology yet (ASFAIK)
Anton
Anton
SOI tech is used by IBM & AMD (I believe it was developed by IBM & licenced to AMD as part of their ongoing work together.)
I remember my first PPC machine, a Powerbook 5300cs. Yes, the machine had its share of problems; bad hinges, bad plastics, fire. Although it wasn’t a first generation PPC, it was my first machine based on the PPC family.
I was impressed with the transistion from the 68K to PPC. I remember Andrew Whelch making a comment that Apple used Maelstrom as one of the 68K emulator metrics. It seemed that Maelstrom was mostly hand written 68K code. If the 68K emulator could handle Maeltrom, it could handle most anything.
Sadly, my Powerbook died during a bad harddrive upgrade. It sits in a box next to my Umax, one of the first Mac clones.
Happy birthday PPC.
The new Prescott does use a lot more power than the G5, but it also has to power a cache that’s twice as large as the G5’s. So again, its not really valid to say that the PPC architecture is inherently more power efficient. Rather, power usage is more a function of the overall design of the chip. The P4 is designed to sacrifice IPC for clock speed, so it has a high power usage as a result. The G5 also makes this trade-off to an extent (it has a relatively long pipeline) and thus has a relatively high power usage too.
When it comes down to it I can narrow my decision down to three things:
1. Macs are more expensive, although I’ve heard they have higher quality hardware then perhaps a Dell or HP computer so I guess thats ok for people who can afford it but that doesn’t include me.
2. I’ve never tried Mac OS X and I doubt I’ll ever get the chance without buying a Mac. Apple’s idea of marketing the product is providing all of three screenshots and a page and a half of fluff targeted at business types and compulsive buyers hence I don’t know in detail what the OS and its apps are really like or how well they work.
3. There is a shortage of alternative operating systems for the PPC architecture, what few Linux distros there are for PPC are ones that I didn’t like on my x86 (ie: Gentoo, Mandrake and Debian based distros) so if I don’t like Mac OS X there is nowhere to go.
Anyway this is just constructive criticism so feel free to correct me where I might be wrong.
Check and see if they have an apple store near by your house. If that’s not an option, I’m quite sure *someone* who lives near by you could show you a Mac.
Now regarding the price, and alternative OSes: You get a mac for one reason, the software. The ability to run Photoshop side by side with gaim or <insert your favorite x11 app> and still use a “real” terminal. There really is no reason to run alternative OSes for the software packages if that’s what you wanted to run them for, because most of them are already ported to run in OS X just fine or are being actively ported. You can install software easily via darwinports or fink through a pretty interface or the terminal.
According to IBM, the 970FX uses 24.5 W @ 2.0 GHz. Perhaps you are confusing the single CPU numbers with the dual CPU numbers from Apple? The 970 does use 51 W @ 1.8 GHz though.
Lets face it, Opteron, G5 and the latest P4’s all process 32-bit instructions about the same speed (and only people running *BSD, a beta Windows or Linux on the Opteron and G5 take advantage of it. (OS X does a tiny bit of stuff (just memory) in 64-bit)).
From what I have seen, the order of Computing power is Opteron, G5, P4, in general stuff, but in terms of Computing Power, they are roughly equivilent. The fact that a P4 has to have twice as much cache to keep up is irrelevant.
Electrically, the G5 wins hands down. 51W @ 1.8 GHz Opteron is somewhere around 60-80, and P4 Prescott is horrible: 103W @ 3.2 GHz
I missed these (mid-March shipping) a 30W @ 1.4 GHz & a 55W @ 2 GHz, so Opteron isn’t nearly as far behind the G5 in power consumption as I thought.
Numbers on the 848 (8-way capable, etc) 89W @ 2GHz
If you look at desktop Macs, yes they do seem overpriced compared to desktop PCs (given any OS).
But look at the Mac laptop line. The new G4 iBooks come equipped with an ATI Radeon 9200 graphics chip and dedicated video memory. A similar PC laptop would be the Centrino based ones (if we compare by heat/power consumption). Most Centrino laptops available now use Intel’s integrated (855G?) graphics with _shared_ video memory. And seldom do I see a PC laptop with a firewire port that’s 6-pin instead of 4-pin.
ok in regards to the Intel bashing:
http://arstechnica.com/cpu/004/pentium-m/pentium-m-5.html
ull see clearly that the centrino uses less watts, and has more transistors than the IBM970 (G5) at the same clock speed. enough with the FUD about x86
So, I guess it would kill people to buy a used iMac off of eBay or something.
With enough ram, a vintage 1999 G3 350 provides acceptable 10.3.x performance.
Sorry but you seem to be all over the map on this one.
First off the P4 sacrifices CPI not IPC, two very different concepts. CPI is associated with the length of the pipeline (in which case the P4 is significantly longer), vs. the IPC which is a function of the width of issue, in which case the P4 is comparable to the P3 family.
And yes the PPC architecture is inherently more power efficient because it needs a smaller amount of control logic overhead vs. the comparable x86 machine, since the PPC is a pure RISC machine whereas the new x86 has to pretend to be a RISC machine. I.e. you can achieve the same amount of computing with far less number of transistors, give a same technology… the fact that one architecture needs fewer transistors it makes it inherently more power efficient.
<< The new G4 iBooks come equipped with an ATI Radeon 9200 graphics chip and dedicated video memory. A similar PC laptop would be the Centrino based ones (if we compare by heat/power consumption). >>
Yes but the G4 iBooks are at 800mhz and come with a 12inch screen. unless u take into consideration the 14inch iBook which gives u NOTHING more in terms of resolution, and an extra 133mhz. The centrino laptops come in 14.1″ screens as the norm, also running at 1.3+ghz. so SPEEDWISE the centrino based laptops win, with battery life to boot
<< Most Centrino laptops available now use Intel’s integrated (855G?) graphics with _shared_ video memory. And seldom do I see a PC laptop with a firewire port that’s 6-pin instead of 4-pin. >>
first of all, USB 2.0 i think is pretty much what the industry is backing. Firewire is good and all, but like i said, everyone’s jumping the USB 2.0 route, so for most users, its a pointless argument. Oh and did i forget to mention that centrino laptops require NO EXPENSIVE ADDON card to have wireless LAN functionality?
Oh and this is to all you zealots that seem to think Apple has untouchable “quality” hence they’re allowed to charge a premium cause “they dont choose cheap components”…. To that i say: i think the recent iBook logic board problem speaks for itself. The part thats most disgusting is how long apple too to acknowledge the problem, and how it treated its users during the iBook and iPod (batter life) fiascos.
Way to go apple!
I have a 12in ibook G4 and if you want a bigger screen it sort of defeats the purpose of being ultra portable.
I get consistently longer battery life that my friend’s centrino. And the “ultra expensive” add on card for wireless is $70 (talking educational discount here).
As for the “fiascos” you mean send my laptop and get it back 48 hrs later, yeah that is a terrible way of treating customers, not! And the ipod battery outrage, I guess you are refering to those dumbasses who made a vide about the outrage of Apple not wanting to fix something for free that was out of waranty. I mean i am sure IBM will replace a 3 yr old thinkpad w/o warranty no questions asked, right?
In any case if the centrino machine is a better match for you needs, by any means buy it and be happy. Everyone should chose the right tool for their job, I am tired of people making blanket uninformed statements.
From eBay,
PowerMac G4
Memory (RAM):
Processor Type:
Hard Drive Capacity:
Processor Speed:
400 MHz
Screen Size:
Operating System:
Mac OS X
Condition:
This is a great machine. Each of the two hard drives has 38GB for tons of storage and ease of use. Upgraded to 576 SDRAM. Loaded with OS 10.3 Panther which runs quickly and beautifully. The condition is very nice, with only a couple of tiny scratches on the side. 2 Firewire ports, 2 USB ports. CDROM/DVD drive and ZIP drive built-in. Tower only. You will be very happy with performance! I think this will cost $35-$55 to ship, depending on where it is going. Thanks.
Winning bid:
US $330.00
You can find deals like this on ebay every day, so there’s no excuse for not trying a mac because of price. Try it a couple of months, then sell it without losing a dime.
http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=2790910525&categ…
“first of all, USB 2.0 i think is pretty much what the industry is backing. Firewire is good and all, but like i said, everyone’s jumping the USB 2.0 route, so for most users, its a pointless argument.”
That’s WAY off. To start with, there’s no real shortage of choice with either one. The difference, especially when talking about that the ‘industry’ is backing, comes down to what will be used in the future. FireWire is being improved (FW 800 currently, even though it’s not of much use to a lot of people, and in the future, well check the patent sites and you’ll see that Apple has registered the name GigaWire) and USB 2 is being replaced by UWB.
<< 400 MHz >>
<< You will be very happy with performance! >>
plz tell me ur kidding me? at school (graphic design) we got 1ghz eMacs with 768mb of RAM, and my ATHLON SMOKES IT in everything i do (photoshop, illustrator, quark)…
i was also the owner of a 800mhz eMac with 1gb of RAM, and a 1.6ghz G5 with 1gb of RAM (cause i thought finally a machine fast enough to run OSX)… and i sold them both, i mean for how much i payed my G5, i got a pc here that DEMOLISHES my G5 in ANYTHING i throw at it.
unless this guy only wants to surf the net and write emails id have to go against what ur saying here. 400mhz is very low to recommend to run OSX on top of.
i ain’t bashing MAC’s or the processors, but OSX (AQUA) has a lot of overhead. i just wish u were able to turn some AQUA features off, expecially when on the job.
And yes the PPC architecture is inherently more power efficient because it needs a smaller amount of control logic overhead vs. the comparable x86 machine, since the PPC is a pure RISC machine whereas the new x86 has to pretend to be a RISC machine.
The problem with your argument is that PPC ain’t pure RISC (things like MIPS are probably closer to that “ideal”). It’s probably more similiar to current pseudoCISC processors from AMD and Intel (RISCish CISC vs. CISCish RISC
). For instance, G5 CPU has a decoding stage, on which the instructions are broken down into simpler ones. Have a look here: http://developer.apple.com/hardware/ve/g5.html
The G5 does instruction cracking and microcoding. A number of PowerPC instructions are divided up into two or more internal instructions called microoperations (µops). Those that are broken into two are said to be cracked. Such instructions take up two spaces in the dispatch group. Those that are broken into three or more are said to be microcoded, and take up an entire dispatch group unto themselves. No AltiVec instructions are cracked or microcoded, though vector stores have a vector and LSU component that is visible in the simulator. These only take up one slot in the dispatch group, but take up two issue queue slots as per other stores — one in the LSU and one in the vector store unit.
First, please type normal sentences. That shorthand may be cool to your 15 year old buddies, but adults tend to use more or less full sentences.
Second, Apple components are high quality. There are always some components with problems on pretty much every machine. But, you go try to price a comparative PC to, say, a Powerbook and see what happens. ATI mobility is a high quality part. Firewire is an expensive feature. High speed wirless is an expensive feature.
I didn’t believe it at first either, but I went and tried to make an *equivalent* PC (one with as many of the *same* components as possible) and found that the PC is the same. In fact, I think the Dell I priced out was even a little more expensive.
I didn’t say otherwise. I said it uses 50 watts at 2.5 GHz. According to IBM’s press-release on the 2.5GHz part, anyway.
CPI is a useless measure on any pipelined CPU. A G5 uses about 17 clocks per simple instruction. A P4 uses 20. A Prescott uses over 30. But it doesn’t matter — all of them retire simple integer instructions at a rate of one instruction per cycle (or 2 instructions per cycle, in the case of the P4’s double-clocked ALUs). IPC is a much more valid measure on current CPUs. IPC measures the average number of instructions executed per clock cycle. The P4 has a low IPC, because its long pipeline is prone to having bubbles. Also, the P4 is a narrower design than most modern CPUs. Whereas the P3 had 3 integer units, it has 2. Whereas the Athlon has 3 floating-point units, it has 1.
As for the inherent performance overhead of translation, Kobold is absolutely right. The extra transistors needed to implement the decoder is minimal relative to the size of the CPU. That’s why the Pentium-M manages to be competitive in terms of power usage.
Just bought an iBook a week and a half ago, and this thing is NICE! The educational discounts makes it pretty darn affordable too. It’s no speed demon, but if you want a fast computer, you can get a desktop machine for a few hundred bucks that’ll rip apart any laptop. 12″ is exactly what I was looking for, who needs a huge display on a notebook? I only found one PC notebook that was 12″, had comparable features than the iBook, and a lower price. I chose the iBook because the other was some noname brand, and I found numerous bad reviews about them.
So far, I’ve been very happy with my purchase, and I’m a financially strapped student, so don’t tell me Macs are too expensive. They do have some fairly priced ones.
<< Most Centrino laptops available now use Intel’s integrated (855G?) graphics with _shared_ video memory. And seldom do I see a PC laptop with a firewire port that’s 6-pin instead of 4-pin. >>
yuppers… and that’s one of the few props I give to mac laptops. I don’t think you’ll ever seen a PC laptop with a 6-pin. sucks.
<< first of all, USB 2.0 i think is pretty much what the industry is backing. Firewire is good and all, but like i said, everyone’s jumping the USB 2.0 route, so for most users, its a pointless argument.>>
good thing my digital camcorder has a USB 2.0 interface. just kidding. good luck finding a company to jump that route.
I’m not very inclined to go with a use Mac nevermind an old one as I hear that Mac OS X uses a fair amount of system resources and I need a faster computer for compiling my programs anyway.
My P4 2.6 with 512 Mb of DDR ram, an 80 Gb hard drive with 3yr warranty and an NVidia GForce FX 5200 was less then $1000 CAD whereas the price of Mac’s in Canada is rather high.
http://store.apple.com/1-800-MY-APPLE/WebObjects/canadastore
Personally if I go for a Mac it’ll have to be an iBook because then I get something I wouldn’t mind paying the price tag on and I would finally get the chance to try Mac OS X.
plz tell me ur kidding me? at school (graphic design) we got 1ghz eMacs with 768mb of RAM, and my ATHLON SMOKES IT in everything i do (photoshop, illustrator, quark)… >
You never say how fast your Athlon is.
BTW, my personal friend Eisner Award winning colorist and graphic designer Laura Martin colors on a 450 mhz 1999 powermac under OSx.
So, I’d say that an in-demand colorist still using that hardware qualifies as fast enough.
1999’s machine is not the latest and greatest but it’s Cheap and useful and an easy way to see if OS X is an Os you want to keep using.
Acceptable for whom and for what? Not much.
Getting an old G4 from 1999 or 2000 for $350ish would not be a bad idea at all. Especially considering that if you thought you liked Mac OS X enough to pay more for the hardware, you could buy a 1.3GHz G4 processor upgrade for $395 from OtherWorldComputing and probably get an additional 100-150MHz overclock out of it.
$350 for the G4 + $395 for the 1.3GHz upgrade = $750ish with the possability of a 1.4+ overclock. Not bad.
You are all wrong – Apple should immediately switch to the VAX ISA – the only ISA with a future! x86, yesterdays news, PPC – old school all outdated by now, MIPS – ridiculous simple, ARM – I have never heard of a more twisted architecture… VAX – the future!!!
Finding a chip manifacturer with the nescessary foresight may prove difficult, but a VAX produced at 1.2u will outrun even the fastest Intel processors*.
* if sufficently underclocked
Not that this is my specialty, but it did bring to mind my Architecture class from last spring.
You state that the PPC architecture is not truly RISC because it breaks instructions down into multiple microoperations. This is true, but it was always my understanding that this was part of the fuzzy line between RISC and CISC.
I humbly suggest that there is a distinction you are missing. To my understanding, RISC does not imply a one-to-one correspondance between machine code instructions and microoperations. That is the territory of EPIC (Explicit Parallel Instruction Computing), where each instruction corresponds precisely to one microoperation. RISC is simply a philosophy guiding the design of a microprocessor’s ISA.
Again, this is far from my realm of expertise–I’m a code monkey by trade. If I’m mistaken try not to flame too hard.
KOMPRESSOR