Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
Technology

PowerPC Processor Roadmap 94

Posted by Hemos from the mapping-the-future dept.
ezavada writes "Motorola has posted their PowerPC Processor Roadmap. Looks like they expect the G4 to go to 1 GHz, and the G5 to 2+ GHz. There is also a story about this in MacWeek. " The current road map goes out until 2009, while another interesting tidbit is that Motorola expects to continuing making G4s even with the introduction of the G5 and G6-embedded chips perhaps?
This discussion has been archived. No new comments can be posted.

PowerPC Processor Roadmap More

PowerPC Processor Roadmap

Comments Filter:

  • Teasing parallelism out of computer programs is a desperation move when you can't speed up the basic logic of the chip.

    A 2X increase in clock speed (and associated I/O channels) really does mean a 2X increase in performance for all programs, but adding parallelism (extra execution units, VLIW, Vector Processors, etc) only wins for those programs which can be made parallel, and only a fraction of the computing world works that way.

    Everything else is serial, and requires those clock rate increases to get more performance.

    As for Intel [intel.com] "keeping up" with the PowerPC, a friend of mine who used to work for NASA [nasa.gov] is fond of quoting this aphorism:

    With enough thrust, anything will fly.

    With tens of billions in sales, and a 26% profit margin (see Yahoo's financial profile of Intel [yahoo.com]), they've got a whole lot of thrust to put under the obsolete, bloated IA-32 architecture.

    Ah, if it were only as simple as the technical merits...

  • At LinuxWorld San Jose, somebody was showing off quad G3 boards and was talking about G4 boards. Also, troll the mac rumor sites, as I saw a blurb on one of them a week or so ago about a company using IBM's open spec board to create some larger scale SMP boxes with the PPC G4.
  • Please distinguish between ISA and chip design. The x86 instruction set may be antedeluvian but the microarchitecture of the K7 is state of the art (could we expect less from Dirk Meyer and Co.?). And I said before, the G3 and G4 are fine embedded control MPUs and this wasn't serendipity at work either. When Apple twisted the knife in the Mac clone makers (and screwed Mot's own ambitions in this area at great cost $) both IBM and Motorola decided that PPC had to have a more stable market than Apple and that it was embedded control. PPC is going gang busters in embedded control and I wouldn't be suprised if they come in third behind MIPS and ARM in unit volume for 1999. They rule the VME RISC world and both the regular PPC parts and the integrated 860 style devices are hot in telecom. Altivec is powerful and great for DSP work and other hand-coded computational kernels but show me a production C or FORTRAN compiler that can effectively use it.
  • The G3 is actaully a "tweaked" and shrunk 604e, not 603e. The G4's speed is in the fabrication procees, copper not aluminum, and the chip is much smaller. Not to mention all of the AltiVec stuff. Also, the G4 can be set up to parralell process, but inlike most chips, it does not need one particular chip to distribute the work load. The AltiVec allows for this in some way that I don't understand.

  • Re:Microwave processor emissions (Score:4)

    by Jordy ( 440 ) <jordan.snocap@com> on Wednesday September 15, 1999 @06:16AM (#1680421) Homepage
    From what I understand, prolonged exposure to low wattage microwaves can be damanging to exposed soft tissues in your body (such as eyes), but the output of a CPU probably won't do it.

    There are studies on metabolic, reproductive, and neurological changes associated with low intensity EMR.

    Anyway, here are a few facts from the WTO study on EMF:
    • RF Fields above 10 GHz are absorbed at the skin surface, with very little of the energy penetrating into the underlying tissues.
    • For adverse health effects, such as eye cataracts and skin burns, to occur from exposure to RF fields above 10 GHz, power densities above 1000 W/m2 are needed. Such densities are not found in everyday life. They do exist in very close proximity to powerful radars. Current exposure standards preclude human presence in this areas.
    • RF fields between 1 MHz and 10 GHz penetrate exposed tissues and produce heating due to energy absorption in these tissues. The depth of pentration of the RF field into the tissue depends on the frequency of the field and is greater for lower frequencies.
    • An SAR of at least 4 W/kg is needed to produce adverse health effects in people exposed to RF fields in this frequency range. Such energies are found tens of meters away from powerful FM antennas at the top of high towers, which makes these areas inaccessible.
    • Most adverse health effects that could occur from exposure to RF fields between 1 MHz and 10 GHz are consistent with responses to induced heating, resulting in rises in tissue or body temperatures higher than 1C.

    Anyhow, for more information visit the WTO EMF web site. [who.int]

    It might also be noteworthy to know that a lot of cases today probably won't shield well against frequencies above 1 GHz so you might start having problems with phones, radios and the like.

    --

  • My understanding is that IBM/Moto added a few instructions to the PPC specifically to help MacOS along. This was a few years ago in the 603/604 era.

    This is all from vauge memory, so feel free to abuse me if it's wrong.
  • You used to be able to do this with older NuBus 68k Macs with the Radius Rockets. 6-way SMP on machines with 6 Nubus slots! (One was for the video card)
    --
    "I was a fool to think I could dream as a normal man."
    B. B. Buick
  • Sorry but the facts are against you. The 604 could issue four instructions per cycle. The 603 could issue two instructions per cycle. The G3 and G4 can issue three instructions per cycle but only of one the three is a branch. Thus the G3 and G4 stick with the same 64 bit, two instruction dispatch bus from the fetch/branch unit to the rest of the execution units as the 603 rather than the 128 bit, four instruction wide bus found in the 604. Thus the G3 and G4 are based much more closely on the 603 than the 604 (as far as these things go).
  • I don't care much about increased cruft in the instruction set (e.g., MMX, AltiVec). So what.

    So What? AltiVec is going to make have huge impact on the graphic, video, audio, scientific, voice recognition and gaming world (and probably others). AltiVec is NOT MMX. Here is a quote from MacOSRumors' [macosrumors.com] report for Sept 15th comparing a G3 using OS 8.6 and a G4 using OS 9 (which has an AltiVec optimized version of Open-GL).

    OpenGL-based games and rendering applications simply screamed with even a single 450MHz G4, between 4 and 10 times faster than the same apps running on a 450MHz G3.

    Note that there was NO difference in the applications being run (i.e. the applications aren't AltiVec optimized), the G4 was using a Yosemite motherboard (i.e. the same slow memory system as the G3), the only difference was that Open-GL in OS-9 is AltiVec enhanced.

    The real bottleneck in PCs (a blanket term including Macs) is the bus. No one except SGI and Sun are doing anything interesting about the bus bottleneck problem.

    The Sawtooth motherboards (using MaxBus) TRIPLE the memory speed and DOUBLE the PCI speeds of the Yosemite boards. I would classify doubling and tripling speeds as interesting. According to reports Apple and AIM are on track to fully support PC266 RAM once it's availabale.

  • Well, the big problem right now is that the MacOS isn't even completely PPC-native. First things first, which is MacOSX.

    I'm not aware of the issues surrounding a 64-bit version of OSX, but considering that NextStep ran on a number of platforms, and it's Mach and Unix-based, it's probably pretty portable.

    Don't forget that Apple is a consumer company, and that a 64-bit port right now would be worthless if it only ran on obscure RS6000s and the application vendors (Adobe) decided not to port. However, I would imagine that Apple would have to have 64-bit hardware out in the Merced timeframe (01-02) just to keep their high end market happy.

  • Someone said:

    Stuff, words, faster, MacOS rumours always tells the truth, dude, 500MHz, PC266, zooom, Quake III.

    Sure, and that's all true and good. Really.

    But we near the point when your CPU becomes a microwave oven (as in, EM radiation), and not even a PC hardware expert like Caesar of Ars Technica can deal with that. Some new, smarter design is going to be needed, because we're about to hit a performance wall.

  • PowerPC are a fast, clean and sexy set of chips. More power to Motorola. Let's hope they produce.

    ... and that they come packaged in nicely designed and coloured cases :)

    Stephen

  • MkLinux is terrifically stable, but it's slow.

    In my expierence LinuxPPC can be somewhat less stable. Of course, the last version I used was R4, on a G3 upgraded clone. It might be alot more stable now.

    I don't know about debian/ppc or yellowdog.
  • The Mac version of Half Life is almost done. Logicware, who are doing the port, has sent an early build to sierra for testing.

    You can use just about any PC IDE or SCSI CD-RW on the mac. The driver quality isn't always great, though.

  • Re:Somebody buy those nice people at Motorola a be (Score:3)

    by chip guy ( 87962 ) on Wednesday September 15, 1999 @06:37AM (#1680434)
    The PowerPC family hasn't been maintained as a general purpose desktop and server RISC ever since the Somerset joint design center blew it keeping the 604 on the performance curve and wiffed on the 620. The PPC 750 (G3, or Arthur) is a glorified 603e tweaked to run Mac code better. The G4 is nothing more than a G3 with slightly improved FPU and the Altivec extensions. The G4 is a slick chip for high end embedded control and digital signal processing. But PowerPC hasn't kept pace with microarchitecture developments in the x86 world let alone with its RISC brethren. It has ridiculously short pipelines and rather modest out of order execution resources. The roadmap shows the G4 hitting 1 GHz in a 0.15 um copper process. I should bloody well hope so - the Alpha EV68 and AMD K7 will likely exceed 1.5 GHz in a 0.18 um aluminum process The writing is on the wall for Apple. The PowerPC has gone embedded control. Neither Mot nor IBM want to pretend to compete with Intel for the desktop any more (although IBM is doing some interesting 64 bit Power chips such as Northstar and Pulsar that compete against Xeon in the server market). Mot didn't build Altivec into G4 for Steve Jobs ego. It is there to win sockets in future generations of base stations for wireless services. As the differences between the bleeding edge desktop market and where PowerPC is heading become more and more evident Apple either has to start building products that look less like desktop PCs and more like internet appliances and PDAs etc or chose a new processor family.
  • The MPC7400 equals or beats any x86 chip (including the Athlon) of the same clock speed in integer and FP benchmarks. Its SIMD execution unit is faster and more capable than MMX, SSE, and 3DNow. It's smaller, uses fewer transistors, costs less to manufacture, and uses far less power.

    So...remind me again why Apple should switch to a different processor family?
  • The G4 is nothing more than a G3 with slightly improved FPU and the Altivec extensions.

    No, the G4 is the successor to the 604e, NOT the 603e. The latter was intended to be a cheap consumer processor while the 604e's were designed more for servers.

    But PowerPC hasn't kept pace with microarchitecture developments in the x86 world let alone with its RISC brethren.

    How so? I don't know how the G4 stacks up against other RISC chips like the Alpha or the Ultrasparc, but it is far ahead of x86 in many areas. The G4's have Altivec, great FP and integer performance, run cooler with less heat, aren't bogged down by anchient 80's baggage etc.

    It has ridiculously short pipelines and
    The roadmap shows the G4 hitting 1 GHz in a 0.15 um copper process. I should bloody well hope so - the Alpha EV68 and AMD K7 will likely exceed 1.5 GHz in a 0.18 um aluminum process

    It probably will. The PowerPC architecture was designed to have a long life span, so if the G4 doesn't hit 1.5 ghz I'm sure the G5 will. You might even see one in a laptop since the PowerPC's run so much cooler.

    As for your 1.5 ghz K7 system, just take off one of your drive bay covers and use your computer as a toaster oven. Not that the K7's aren't great chips, but damn are those things hot and power hungry.
  • The G4 may very well targeted the 604 but I am talking about what is under the hood. The G4 and G3 internally ressemble the 603e more than the 604. Motorola rates the G4 at 18/18 SPEC95 at 400 MHz with honking large L2. The K7 gets 27/22 at 600 MHz. Please no comments about "per MHz performance". The G4 uses an ancient five stage pipe and will always have a substantial clock rate penalty in a given technology compared to more modern designs like P6, K7, and Alpha. Altivec is a well done SIMD extension but like all SIMD extensions, it will have little usefullness for most applications. You are correct about the relatively low power consumption of PPCs. But this is an issue only for laptops etc. The K7 is circa 50 Watts which is not difficult to deal with in a properly designed system. Current EV6 desktop systems dissipate up to 109 Watts per CPU without heroic effort and even the Exponential PPC based desktop that Apple canned years ago handled a 70+ Watt CPU.
  • My aging PPC601 has the capability to process 64bit numbers - all of its FP registers are in fact 64bits long, even if the int registers are only 32bits. It is true that the powerPC chips apple is using do not deal with 64bit integers, but other varients in the PPC family DO have 64bit int registers.
    So even the G1's are technically 64bit processors, let alone the G2, 3 or 4.
    As a great deal of computations make use of floating point mathematics, Apple isn't incurring such a huge penalty by not writing the core of their OS in 64bit. This is demonstatable -> My 7100/66av (PPC601 @ 66MHz) can and routinely does decode and play 160bit/s MP3's in full 64bit quality without skipping any frames. I'd really like someone do manage that with a i486 @ 66 !
  • The most important part in the computer other than the motherboard is the processor ... But like thigs are going on right now, that might change ... Graphics Cards are becoming huge. New cards looking to a neat 256Mb VRAM is great. But what might be faster sooner is that the GCard takes over the work of the processor.

    Slots are becoming faster now too with AGP 4x coming. Changing processors would be soo easy that anyone could just walk in the store and get themselves a new GCard. With 3-4 Slots free, you could get as much "CPUs"

    One more thing, with the TNT2 with 256Mb, will you have to upgrade your amount of ram to at least 384? I can't imagine a machine with more VRAM than normal RAM


    - Frederik-Jan
  • I'm just trying to figure out how folks like Zenith will deal with export laws when they put an embedded version of a G4 into a TV set. (No, it's not a supercomputer, damn it! Yes, I know the remote control can theoretically do 1 gigaflop, but it's primarly used to change channels!)
  • Please distinguish between ISA and chip design. The x86 instruction set may be antedeluvian but the microarchitecture of the K7 is state of the art (could we expect less from Dirk Meyer and Co.?).

    And if I could only shed the 235 pounds of weight associated with my mass, I could float to the ceiling.

    The microarchitecture of the K7 may be state of the art, but for all practical purposes it's useless when hobbled down by the i86 instruction set.
  • Actually....

    the atari lynx used a 32/16 bit 68000 (32 bit internal registers, 16 bit bus - just like the amiga, 386sx, etc) :)

    smash
  • hrm.. i seem to recall reading somewhere that the 8087 (the math co-processor for an 8086 XT) can handle 80 bit floating point numbers.

    this doesnt make an XT with co-pro an 80 bit system tho... i think "they" only count integer size, as that is what is more commonly used for "normal" data types such as manipulating strings, integers, pointers, etc...

    smash
  • This is a 10 year prediction map. Given the pace at which new technologies arise and can be implemented, I am sure that much of their plans are strictly tentative. The question is, will they change for the better (faster speeds sooner) or the worse (faster speeds later)?

    Also, what does IBM have to say about this map?

    All that said and done... Spare a dime to buy a guy a G4?
    --
    Matt Singerman
  • Well if the basic set of instructions of the chips is backwards compatable then linux could really become quite stable on this platform just like the x86 is now.

  • Microwave processor emissions (Score:3)

    by Fnordulicious ( 85996 ) on Wednesday September 15, 1999 @03:25AM (#1680453) Homepage
    So when these latest and greatest processors running at GHz speeds start emitting radiation in the microwave spectrum are we all going to have to start wearing lead aprons like your local radiologist? Or are we just going to have to shield the entire box in a faraday cage?

    Looks like there won't be any more topless, skeletal machines in the future -- "Removing the cover from this machine may void your warranty AND cause severe radiation burns while the machine is in operation."

    fnord
  • The mac has quite a few games that are already ready to be ported or already there. Back a few years ago mac's had supperior sound hardware to some of the PCs that were out. I am not advocating macs it's just that they were aimed for artistically inclined and not the scientifically inclined so therefore multimedia has been quite developed.
  • Could the whole problem be avoided by creating an array of chips instead of ones that could even produce harmful emissions in the first place? Is there any difference between the preformance of a couple of chips with the same chip set and one with the same by twice as fast as either of the linked chips?

  • Re: The PowerPC FAQ (Score:3)

    by Ignis De Maligne ( 49740 ) on Wednesday September 15, 1999 @03:45AM (#1680456)

    OK, since the army of wild ants or whatever won't post news of the PowerPC FAQ, I'll do it here, which seems the most appropriate place to do it.

    Visit LinuxPPC.org [linuxppc.org] to take a gander at the latest revision of the PowerPC FAQ.

    It's my personal project, and I'm (slowly) trying to make it more accurate and timely (as it was previously maintained by someone else who quit).

    Since the industry seems to be interested in having a second affair with the IBM's pseudo-RISC architecture, I hope the FAQ will serve as a good intoduction to the world of PowerPC.

    Comments on it to davenport@access-k12.org [mailto].

    Thanks :)

  • Its not suprising at all to me that G4s would be manufactured for years to come. My company currently embeds powerPC chips in many of its products. (I have two that I'm working on sitting on my desk right now (403s actually)). If you have used a newer laser printer, a RAID array or any number of other devices today, chances are you have used a powerPC processor. If you buy one something like this today, chances are that the processer in it is faster than the processer of your computer was 5 years ago.

  • More Mhz? What else is new. (Score:3)

    by jeff.paulsen ( 6195 ) on Wednesday September 15, 1999 @03:45AM (#1680458)

    Many several of the Mac rumors sites have mentioned future SMP-on-a-chip G4s and G5s. That would interest me much more than more Mhz.

    We've seen PCs go from 4 Mhz to 600 - going up to 1000 Mhz seems like a mere incremental upgrade. The real news about the G4 is the AltiVec unit. The real news about IA-64 is VLIW. (be gentle if I've got that wrong; I don't follow Intel rumors like I ought). I assert that quantitative changes in clock speed are not where the action is right now, and all the interesting performance improvements come with qualitative changes in the way data is moved and operated on.

  • After all, Intel kept doing R&D on the i386 until 1994, long after it was obsoleted by the Pentium. (I heard it was also used in the Atari Lynx... any truth to that rumor? That was about 1994/1995...

    -Markvs

    ..."I see you have a machine that goes PING!"

  • I've been thinking the past week or two that if Apple can keep up their lead on the processor market as they clearly plan to (of course I'm still confused as to why IBM and Motorla make chips for Apple), owning a Mac might not be so bad (even if most of their stuff nowadays looks like it was built by space aliens IMHO). I also thought it was interesting that they were supporting a project, MkLinux [apple.com], with The Open Group [opengroup.org] (not that I've ever been a big fan of TOG).

    Of course, this old article [apple.com] made me laugh a bit. Use MacOS for graphical applications? Why bother with a dual boot? TOG, of all people, should know you can run graphical applications on Linux. After all, they're the ones who put together X.Org [x.org] to maintain the official X [xfree86.org] stuff (TOG getting X.. ugh..). Seems kind of silly to me.

    It seems like a smart move to Apple, though, to support Linux. After all, if they can have Linux running on their machines, and have superior hardware to what's already out there, they could tap some potential markets (especially if they have modems on their computers.. I'm so sick of Linux "desktops" without modems because the computer makers love winmodems, which is something I'm sure that Apple doesn't subscribe to, at least, as their machines obviously don't run Windows).

    P.S. Yes I realize the "X" link leads to the XFree86 site. I have my reasons, but I don't want to explain them, ok? :)

  • You'll hit the limit of the drift speed of charge carriers in silicon, and even GaAs, before that.

    Oscillations of the order of 1 GHz are still nowhere near visible light, let alone UV/X/Gamma rays.

    The shielding already availible on cases is more than adequate, they done take in enough power to cause damage.

    Cellphones actively transmit, so there is *some* reason for research in that area
  • PowerPC are a fast, clean and sexy set of chips. More power to Motorola. Let's hope they produce.
  • It's not quite Half-Life, but, I recommend you check out the screenshots of Halo by Bungie (remember the happy people that brought the mac peoples Marathon [2], as well as other fun toys...)

    Check it out here: halo.bungie.com [bungie.com]

    It looks simply gorgeous.
    Almost worth buying a mac for...almost.

  • While another interesting tidbit is that Motorola expects to continuing making G4s even with the introduction of the G5 and G6-embedded chips perhaps?

    Motorola has been very good to embedded designers in making chips for a very long time. 10+ years.

    Heck - where I work still uses a bunch of HC11's.

    This is one reason that the 68k line of processors has been so popular in embedded systems. Not to mention the nifty stuff that Motorola tends to glue together with the CPU all into one chip. This is the reason you will probably still be using PowerPC products 10 years from now - even if you don't realize what that CPU is under the cover of your TV.
  • This is the exact argument that RISC advocates used to declare CISC dead. Yet the Pentium has done a remarkable job keeping up with the performance of much cleaner architectures like the PowerPC. The point is that there are all kinds of ways to get performance improvements, and clock speed, VLIW, RISC, AltiVec, MMX, etc.. are all just some of the ways. It's all just going to keep getting faster....
  • You Apple supporters just don't get it do you? Saying that PPC beats x86 etc at the same speed is a meaningless statement. The five stage pipe used by PPC has to stuff a lot more work between clock edges so it will never catch up in clock rate to a modern microarchitecture desktop CPU. The G4 uses a nice'n'sexy 0.22 um copper process just to go 400 to 500 MHz. The Alpha EV6 can run up to 600 MHz in a 0.35 um aluminum process and over 800 MHz in a 0.25 um aluminum process. Yet the EV6 has the most complex microarchitecture ever shipped (can run with up to 80 instructions simultaneously in flight in its out of order execution core). It is not just the CPU either. Steve Jobs can blather on and on about Gflops all he wants but Apple builds mediocre chipsets to go with their reasonably nice processors. Look at McCalpin's STREAM benchmark scores for effective system memory bandwidth. The best Apple score is a G3/266 system with a STREAM SCALE score of 128 Mbyte/sec. A 440BX system with a Pentium II/350 gets 279 Mbyte/sec (and an Alpha EV6/500 XP1000 gets 971 Mbyte/sec). If Jobs wants a G4 "supercomputer" he had better hire some chipset designers away from Intel. BTW, I won't exactly hold my breath waiting for an Apple G4 system to show up on John Dongarra's Linpack list with 1000+ MFLOP/s in the 100x100 column. ;-)
  • Doubling the clock almost always improves performance more than doubling the number of processors. You can get more from two processors than one that is twice the speed, but that would be for some fairly unlikely case, like if the context switches you save are more expensive than the overhead of SMP or something.
  • You are correct about the relatively low power consumption of PPCs. But this is an issue only for laptops etc. The K7 is circa 50 Watts which is not difficult to deal with in a properly designed system.

    Not an issue for desktops? Dual K7's would take up 100 watts of power all by themselves! You want a quad system with anything else in it, you might as well put in a seperate power supply just for the cpu's!

    There's also the problem that more power means more heat. Apple only puts a small heatsink and no fan on their G3's (don't know bout the G4's) because they run so cool. Unless you have a Kyrotech system, your K7's are going to run hot, even with big 'sinks and fans.
  • I've been thinking the past week or two that if Apple can keep up their lead on the processor market as they clearly plan to (of course I'm still confused as to why IBM and Motorla make chips for Apple), ...

    I'm confused as to what you are implying by this statement. Are you saying that Apple has _anything_ to do with the PowerPC development? The PowerPC is a Motorola/IBM design, Apple is just the biggest customer. I seriously doubt that Apple has the resources or the technical staff to design their own processor. Apple is not keeping the lead on the processor market, Mot is.

    If you were referring to Apple's crippling of the G3 firmware, then I didn't figure that out from what you said.

  • The G4 uses a nice'n'sexy 0.22 um copper process just to go 400 to 500 MHz.

    Do you think that's as fast as it'll go in using that manufacturing process?

    Look at McCalpin's STREAM benchmark scores for effective system memory bandwidth. The best Apple score is a G3/266 system with a STREAM SCALE score of 128 Mbyte/sec. A 440BX system with a Pentium II/350 gets 279 Mbyte/sec (and an Alpha EV6/500 XP1000 gets 971 Mbyte/sec).

    ...and the Sawtooth G4 Macs haven't even been tested. Their claimed throughput is 800MB/sec. We'll see what the actual scores are.

  • Isn't the cost of chip in the design rather than fabrication? Why don't they just snag the *designs* of ibm's pwr series of chips? That stuff has got to run orders of magnitude faster than ppc chips and still remain relatively compatible. Actually speaking of which, i'm *really* surprised someone at Apple hasn't worked out that it would be great pr campaign if someone ported MacOS (w/ complete compatibility on binary apps) for one the pwrX chips.. Sure it would cost $50k to buy but they'd "win" the PC game :)

    -avi
  • Speeding the CPU and associated I/O channels is NOT ENOUGH, it won't do you any good if you spend all your time waiting for the memory...

    That's why companies are putting lots of memory cache into the CPU...

    MHz alone is a poor measure of CPU performances
  • Hobbled? AMD is shipping production quantities of 700 MHz K7's to distributors as we speak. The G4 is built in a better process (0.22 um CMOS+Cu vs 0.25 um CMOS+Al) yet is unlikely to go beyond 500 MHz without a shrink. And the K7/650 yields 63% higher SPECint95 and 24% higher SPECfp95 numbers than the G4/400 despite its smaller L2 cache and lack of a Intel style compiler tweaking tiger team. Not bad for inferior silicon technology and the massive deadweight of x86 ;-)
  • Yes, Apple worked closely with IBM/Mot during the development of the G3. It's optimized for MacOS code.

    Here's some info. [tidbits.com]

    -F.

  • Erm... Intel is nowhere in the embedded market. Motorola already owns about 32% of it.
  • I don't know all that their is to know about updating the out-of-date stuff. Try visiting the old one at Mot's site.

    I see you understand the terms "community" and "in development".

    Ass.
  • This is the exact argument that RISC advocates used to declare CISC dead. Yet the Pentium has done a remarkable job keeping up with the performance of much cleaner architectures like the PowerPC. The point is that there are all kinds of ways to get performance improvements, and clock speed, VLIW, RISC, AltiVec, MMX, etc.. are all just some of the ways. It's all just going to keep getting faster....

    Agreed. It's just that clock speed isn't where the interesting breakthroughs in performance are coming from today.

    Alternatively, it could be that clock speed goes up, we're all used to it, we expect it. Like compound interest, it's beneficial, but in any given small period, it's like watching paint dry.

  • The Atari Lynx used a 65C02 (8-bit).
  • Having used Linux for about six years, it's always been on the Intel platform. For my next computer, I'd like to switch to something non-Intel, perhaps the PowerPC.

    Does anyone have a good collection of hardware vendors which actively support (possibly even ship) Linux on PowerPC machines? I wouldn't mind a nice 2 processor SMP PowerPC-based machine, but I don't know where to go to get it.

    Also, my Intel machines in the past have been relatively backwards compatible. Do the typical PowerPC architectures typically maintain backwards compatibility?

    Thanks for any info.

  • Egh. Sorry about the busted link.

    Try www.linuxppc.org/powerpcfaq/ [linuxppc.org].

  • Maybe the frequencies are bad, but it seems like the power would be pretty low. You might be able to cook a steak with the latest kilowatt pentiums, but PPCs? Nah, PPCs are useless in the kitchen, except for computing purposes.


    ---
    Have a Sloppy day!

  • Ok, so, aside from the fact that I'm sleep-deprived (having three days off in a row when you work the graveyard has "interesting" consequences), no I wasn't trying to say that Apple had anything to do with PPC dev. Just edit out any misleading connotation of that sentence in my post. I know that PPC is a Motorola/IBM design. I was just thinking about the fact that they (Apple) use those chips in their computers and starting babbling some utter nonsense.

    And no, I wasn't trying to say anything about Apple crippling any firmware.

    All I was trying to say was a little commentary on the nifty things you could do with a Mac, Linux, and those nifty chips.. mainly in regards to Mac since they (Apple again) are the biggest customer of PPC. Ayup. Hopefully that made more sense. So.. I'm going to shut up now. :)

  • all the interesting performance improvements come with qualitative changes in the way data is moved and operated on.

    Shrinking the die, clocking the chip up, adding a bigger cache and bundling it with a whopper heatsink/fan are just ways of extending a processor's life.

    This is an absolutely necessary activity, though--how else is a company to fund their new processor research? The Merced's been under development for years and Intel won't even begin to make money on it for [insert time frame]!

    The fundamental shifts that are going to win 4x and 10x speed increases cost a lot of money to research and implement.

    A multi-core processor is a real paradigm shift. Whoops, I used a buzz-word! But it's true.

    VLIW (Very Long Instruction Word) is also a seriously different way of thinking.

    These sorts of innovations take a long time to bear fruit, the 33Mhz and 50MHz speed increments that we see along the way are what makes them financially feasible.

    (I'm not arguing with the poster... just adding to his point)

  • PowerPC as an embedded solution (Score:1)

    by Anonymous Coward
    I think its great that embedded markets are getting more of a choice as to the processor they wish to run. The only viable solution to date with long life cycles is the socket 370 intel chips. Intel has added one or two processors to their embedded program and the life cycle is 6 years. This is great for large industrial solutions integrators that ship products world wide for very specific applications(i.e. power plant control systems and Human Machine Interface systems). I'm an industrial engineer and i would love more choices for long term projects. In addition i feel companies such as Texas Micro and Axiom need to start developing CompactPCI solutions for use with these new chips. Untill now CompactPCI, as robust as it is, has always been slow due to physical size limitations of the chips and motherboards. If anyone can make a fast and affordable CompactPCI solution they would inject life back into the truely hardcore industrial and telco markets.
  • Altivec has been just as much Apple's baby as Motorola's.

    I've talked to some Apple engineers about this at last WWDC and Apple indeed has had quite an influence on the AltiVec design, at least as far as architecture is concerned.

    Let's face, if you have programmed anything in Assembly that had to be hand-optimized you cerainly had a long secret wish list, ranging from some new opcodes to something that directly supports your latest buzzword-compliant algorithm in hardware... so Apple, as AIM's largest customer, is in the unique position of getting this sort of wish granted. Rumors were that they also had serious influence on the architecture of the 68040 processor, especially regarding the MMU.

    And IMHO they did a great job. AltiVec is very powerful... with the new bus architecture Intel will be pressed to keep up.

  • Also, what does IBM have to say about this map?

    IIRC, IBM and Motorola got together a few months ago and decided on a standard architecture for future PowerPC chips to maintain compatibility between their different designs. That way they wouldn't have to negotiate every time one of them came up with a new feature (The need for such a standard was evident when they had a little disagreement about Altivec). It's called "Book E," and although it's not mentioned in the roadmap, I'd guess it'll be implemented from the G5's and on.
  • CISC obviously isn't quite dead, but is it GOOD? Desktop CPUs aren't often judged on any qualities besides raw speed, but I think efficiency is important, too.

    The PowerPC chip uses far less power and thus produces less heat than the current Pentiums. I would call this a performance advantage, or some kind of advantage.

    That means you need less ventilation power (resulting in a quieter machine), and less muscle in your UPS (you DO have a UPS, right?).

    Electricity may be pretty cheap, but we could all use a machine that used less juice, especially when we leave them on all the time.

    Does anyone know how Merced might compare in power consumption?
  • The G4 uses an ancient five stage pipe and will always have a substantial clock rate penalty in a given technology compared to more modern designs like P6, K7, and Alpha.

    I'll provisionally agree with this about Alpha (which grew in a somewhat different direction, architecurally), but both the P6 and the K7 aren't necessarily "more modern"... personally I prefer the G4's architectural simplicity. Any CPU that still executes the creaking old x86 instructions necessarily has to jump through all sorts of weird hoops to get any speed at all... PowerPCs aren't shackled to an obsolete instruction set. And with a cleaner, simpler set, efficient backside caches and now AltiVec, a deeper pipeline and (in most cases) even a faster main bus is simply unnecessary - and would give diminishing returns if implemented.

    Altivec is a well done SIMD extension but like all SIMD extensions, it will have little usefullness for most applications.

    This is a popular fallacy, even some Mac developers think so... probably considering the MMX fiasco. AltiVec speeds up many operating system tasks, even memory block moves have a 50% to 100% improvement, anything that draws to the screen or to a graphics buffer will see huge improvements, even TCP/IP and Ethernet packet processing gets speeded up. Don't forget that one of the markets Motorola is aiming at here is routers. I hear 3Com and Cisco are converting all of their stuff - which formerly used 68K CPUs - to embedded PowerPCs, including the G4's for high-end routers and bridges.

    Also, everybody talks about the AltiVec accelerating floating-point stuff - but remember it also processes integers! Either 4 32-bit integers, 8 16-bit, or 16 8-bit integers in one cycle, with useful stuff like multiply-accumulate and "pin to min/max" options. This obviously won't show up in standard benchmarks since you're not allowed to recode them, but anything that processes huge amounts of data will benefit.

  • Well the 21264 actually has 7 pipe stages for non-FP instructions. BTW, the G4 uses 4 pipe stages for integer instructions.
  • First of all the Alpha EV6 has a 7 stage pipeline not 15. Secondly, it has a branch misprediction penalty of 7 or 8 cycles, not 200. Thirdly, you obviously don't have a clue what "superpipelining" is. And the EV6 can issue four instructions per clock (including up to four integer) while the G3 and G4 can only issue three (and one must be a branch) so you tell me which is more superscalar. Finally, am I a buddy of Groves? Nope. In fact the only reason I wouldn't p*ss on Andy Grove if I had the chance would be if he was on fire. BTW, what the heck is a 20164?
  • Um, those new graphics cards from nVidia are 256bit, not 256MB of VRAM. IIRC, they support up to 64MB of VRAM.
    --

  • The five stage pipe used by PPC has to stuff a lot more work between clock edges so it will never catch up in clock rate to a modern microarchitecture desktop CPU.

    It's all about tradeoffs. The short G3/G4 PowerPC pipeline limits the clock ramp, but it sure makes for much faster branch recovery on a mispredict. The Pentium Pro/II/III architecture only makes up for its longer pipeline by expending much more effort on predicting branches (and the work goes on: at least two papers in this year's ISCA conference from Intel research groups which work on branch prediction-related techniques).

    Moto/IBM get to expend less chip area on all these expensive prediction techniques, have a much simpler pipeline, and hence get smaller, cheaper chips which use less power (advantageous when you look at Intel's problems putting their high-speed Pentiums into laptops).

    On the other hand, it's hardly as though Motorola/IBM are ignorant of the advantages of superpipelining. You can bet your bottom dollar that the Power4 (design goal = 2 CPU core @ 1GHz) has more than a 5 stage pipe, and Motorola is about to announce the details of the G5 pipeline in three weeks at the Microprocessor Forum.

    Check out the first day's sessions [mdronline.com].

  • A must-visit companion URL for PPC roadmap info (Score:3)

    by John Siracusa ( 4209 ) on Wednesday September 15, 1999 @04:49AM (#1680504) Homepage
    Check out this older PDF on Moto/IBM's "Book E" [ibm.com] architecture.

    "Book E" is the "extensible architecture" mentioned in the light purple G5 block of Moto's PPC roadmap.

  • I don't care much about increased cruft in the instruction set (e.g., MMX, AltiVec). So what. The real bottleneck in PCs (a blanket term including Macs) is the bus. No one except SGI and Sun are doing anything interesting about the bus bottleneck problem.

  • Don't worry.

    The PPC family tends to run in the 10-20 W range (sometimes lower, sometimes higher). Comparing that to a 1000 W microwave oven would make me think that it would take a loooonnnng time to cook your eyes looking at it. Plus the chips and case try to shield against as much EM and radio interferance as they can usually. I doubt we'll have any worries soon.

    I'm just wondering why no one who's burned their fingers on a hot Pentium has tried to sue Intel or their OEM yet. Seems just as reasonable as that crazy old McDonald's coffee lawsuit.
  • If you're patient (and have a lot of money), you can wait a little for the rumored dual-processor G4 PowerMacs from Apple. You could also get a PowerMac and add a dual-processor G4 card from any of the vendors like PowerLogix, Sonnet or Newer Technology, but that will (obviously) get very expensive--though I *think* they will take trade-ins on ZIF chips.

    Other possibility: you could try getting in touch with Groupe Bull or Motorola Computer Group (at http://www.mcg.mot.com/ [mot.com]) about that. Both make PowerPC-based boards and computers, though AFAIK they only sell to vendors and not to little saps like us. ;-)

    For that matter, if you wait a little, IBM's open hardware specs ought to also generate some interest amoung hardware manufacturers.

    Last recommendation: Get in touch with Yellow Dog Linux [yellowdoglinux.com] or LinuxPPC [linuxppc.com] about it. They'd be happy to help you out in finding something, I'm sure. Both are pretty quick in supporting new hardware, as well; though AltiVec is not yet supported (the code has to be "vectorized" first), meaning only that its advantages won't yet make a difference in Linux, that development is already starting to get underway.

    Oh, and if you're looking for news and info about Linux for PowerPC Macs, check out my site at http://linux.macnews.de/ [macnews.de]. :-)

    Ethelred [surf.to]

  • AIM Consordium -- Apple, IBM, Motorola.

    Apple does have a small role in the design of the PPC. Their main contribution to the project is in compiler writing. The MrC optimized-as-all-hell PPC C compiler had heavy Apple support, and Apple has been working with Motorola on Altivec from the early days when the idea of putting VMX (a proposed but never implemented SIMD archetecture for the POWER family) into the PPC design was discussed. They helped add the Altivec support to the MrC compiler.

    As anyone who knows about chip design knows, the compiler writers are just as essential as the people who map out how the silicon's going to look. (Heck in the case of Merced, they're even more important, but I digress.) Altivec has been just as much Apple's baby as Motorola's. That's why there will be so much support for it in the MacOS in the future. It's no marketing gimmick like MMX was; Apple wanted it to be something they could USE everywhere they could.
  • At Chiba (Japan) World PC Expo

    IBM's Mr. Akazawa created a buzz with the announcement that a 700-MHz G3 chip based on SOI technology would be released this year. Meanwhile, Interware's Mr. Nakajima showed a road map of his company's product line and said the company plans to release 1-GHz G4 upgrades by the end of 2000 and will offer a G4 upgrade card for the iMac.

    Quoted from http://macweek.zdnet.com/1999/09/12/tokyo.html

Slashdot Top Deals

"The four building blocks of the universe are fire, water, gravel and vinyl." -- Dave Barry

Close