The new IBM PowerPC 970 is the heart of the PowerPC Blade. It is based on the 64-Bit Power 4 architecture which is also used in the processors of the IBM eServer pSeries. The 64-bit microprosessor offers full symmetrical multi-processing, has a high reliability (with parity L1, ECC L2 and parity checked system bus) and is manufactured in the latest 0,13 micrometer Copper/SOI CMOS technology. The CPU runs at frequences ranging from 1.8 GHz – 2.5 Ghz, therefore the IBMPowerPC 970 is the fastest PowerPC so far.
It also features onchip 512 KB L2 Cache, Altivec Vector/SIMD unit, 6,4 GB/s I/O system bus throughput. Rumors want that chip to be the next CPU used by Apple when it’s out.
IBM PowerPC 970 Blade Reaches 2.5 GHz in Lab
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Rumors want that chip to be the next CPU used by Apple when it’s out.
Not only rumors want it’s to be in the next powermac, ME TOO!
The PowerPC family has a fantastic design. Its Altivec unit, its clean assembly code compared to x86, its forward-compatibility with 64-bit… And this one is a monster with 6.5 GB/s I/O.
Intel would be very foolish not to venture in 64-bit computing for the workstation market.
All I know I want 1 (or 2)!
Can I use it to replace the 800MHz G-4 in my iMac?
😉
…and being as it is a cut-down POWER4 it should be trivial for SuSe and others to change their current ports (iSeries & Z-series?)to fit.
If there was just a mobo to run it on:(
Too bad the hardware Co.s are too scared of MS retailiation. Both POWER4 and Hammer have running selling Linux ports, but everyone has to wait for MS to catch up.
Those along with the new MIPS and ARM chips recently released (posted here!) could really show the public and corperate America (Sorry rest of the world we can’t show you. Ha, Ha please laugh here) That Linux is a force to be reckoned with! If the rest of the hardware MFGs would realize that Linux means no more bowing to MS (after Xbox success is going to ditch them anyway) Then MS would be the minority OS.
Go get ‘um IBM and AMD!
The chip is reprogrammable to match what the OS needs or want to emulate. If people were to start supporting linux on a large scale they would have their own special playground! Oh yeah, that Guy works there on the internal CPU code. Maybe he could help–they might get a small advantage.
Hehe. I just can’t help but remember the movie Pirates of Silicon Valley where Steve Jobs is all caught up on IBM being the enemy.
Now it almost looks as if Apple & IBM will have a good relationship with one another.
Or atleast I hope they will. Because a PPC 970 would be sweet to have in macs.
Where Apple will be in 2005 when the Power5 derived successor to PPC970 is released.
Apples will once again be the luxury systems, with Power (oh, what a terrible pun) far surpassing that of their puny x86 counterparts.
And despite this, the mentally inept individuals amoung the OSnews crowd, their lone neuron of a brain weeping in its solitude, will continue to insist that Apple needs to switch to x86.
PowerPC is a fantastic processor for every engineer interested in kernel design and computer architectures. It’s simply, powerfull and lack of compatibility layers. Unfortunately users are only interested in the clock speed
I only wish than Apple don’t switch to x86.
Apple and IBM got already a pretty good relationship.
The PowerPC is a result of the teamwork of IBM, Motorola and Apple. And the G3 in the iBook are build by IBM.
Do you mean like this?
http://www-5.ibm.com/de/pressroom/cebit2003/i/highres/bladeprototyp…
“In Lab”?
What’s with all the qualification–I’ve seen more unsubstantiated material posted before. This is a PR for the blades themselves–IBM is specifically saying the blades this Fall will have these clock rates. In the Lab, I’m sure they are running much faster. Yes, people, get with it–IBM was lowballing when they said 1 – 1.4 GHz; they already are assured enough in themselves to know they have substantial enough yields of 1.8 – 2.5 GHz processors at least for their own machines.
“Rumors want that chip to be the next CPU used by Apple when it’s out.” Want? No, some people who are well informed already know that it will appear in an Apple computer this year. This is the first IBM PR to explicitly use the Altivec(TM) trademark–not AV-compatible, not a VMX that can map to Altivec–why do you think IBM licensed something they have said they have no interest in for the last 3 years?
There is nothing more “unsubstantiated” than your own comment.
1. Yes, “in lab”, because YOU also stated, the chip is not out yet. You said it yourself: this Fall it will be released. Until then, it is not out.
2.
> Want? No, some people who are well informed already know that it will appear in an Apple computer this year
Exactly. A Rumor, _no matter_ if that might proven true or not. There are no PRs about it, so that’s what I _have_ to write there: rumor.
You know, I am tired of the internet, really. NO MATTER HOW I write things, people will always find something to bitch about and become rude.
You know anonymous, I can be rude too.
Hell, where will they be when the second rev’s of 970s hit in at the .09 nm process? A 1.2 GHz 970 at .13 nm is rated at 19Ws. IBM apparently won’t even be using 970s clocked this low for their own lines… so how cheap are 1.2 970s going to be for Apple? A 1.2 GHz 970 is still small enough and consumes little enough power to be dropped into a laptop even before the process change. Think about that.
Also note: these should be cheap chips (in the 64 bit class)–they do not have huge caches, they are not much larger than traditional G4s, etc…
Please show me where I was rude?
“Yes, “in lab”, because YOU also stated, the chip is not out yet. You said it yourself: this Fall it will be released. Until then, it is not out.” That makes no sense… This is a PR for the blades that will be released in 4-7 months. IBM is stating the clock rates of the chips appearing in these machines. “In the Lab” I can only interpret as a prototype, pre-production model–these chips have been in full production for over three weeks now. They are well past the “In the Lab” stage. The idea that a chip only exists “in the lab” until there is a product on store shelves is a ridiculous notion, I’m sure you understand that. But maybe we just have a different definition of “lab.”
“You know, I am tired of the internet, really.” Great, then stop trying to respond to every single person who does enjoy it and wants to express their opinion. If you don’t like my opinion, reply to it… if you don’t want to, don’t.
…and they’re more expensive than most OSNews users would purchase. Something like the Open PPC board with normal ATX connections that could run this would be dangerous!
Someone at the EFF needs to have a “golf outting” for some Prez and VP’s (cpus, graphics, sound, etc.), maybe some Taiwan guys too. They could invite some Linux Vendors and maybe,[i’m not sure] Robertson. If you could get enough of these guy talking at once (remember the execs are interested in the “industry” and tend to miss the forest for the trees.) They could “gang up” on someone (who I’m sure has experience being stuffed in lockers). This would get the industry back–it would be interesting…
The reason I replied to you was because you did critisized the way I do my work:
“I’ve seen more unsubstantiated material posted before.”
If you haven’t, I wouldn’t reply. I tried to explain to you WHY I chose these words I chose.
You don’t agree? FINE with me, but don’t call my work “unsubstantiated” cause I do take it personally. I don’t bust my a$$ here for free, just to have people like you trolling my work. Show some respect when you outline your opinions and everything would be just fine.
Only interrested in clock speed? More like interrested in good price/performance… But PPC ISA is cleaner and nicer than x86 ISA that’s for sure (not that most programmers need to know the ugly x86 parts).
What is that a link to? i know the link says bladeprototype, but i assume it’s not the ppc970, since it’s running pc133 ram.
…but I’m afraid Eugenia’s going to take it the wrong way and consider it rude and trolling.
So, until she removes her rude remark towards myself, I can’t post the answer.
have faith in IBM. They already made the mistakes that Intel and GAtes are making now.IBM learned. They will almost certainly end up playing a significant role in the CPU business soon. Plus, IBM actually manages to keep its qualified people, something that intel can’t seem to do.
“PowerPC Blade is the prototype of a server blade developed by IBM Development, based on the new high-performance IBM PowerPC 970.”
http://www-5.ibm.com/de/pressroom/cebit2003/en/photos/innovations/i…
…there’s no fan on that prototype board… 8)
http://www-5.ibm.com/de/pressroom/cebit2003/i/highres/bladeprototyp…
Finally a PowerPC chip worthy of the AmigaOne computer. I look forward to computing with all this power on Amiga OS4.
if (Id say when, but whatever) apple uses this chip, how do they upgrade their systems. I mean, current imacs are 1 ghz, will they put a 1.8 gighz chip in the iMac and make all those who bought an old desktop like they shoulda waited?
Or will they slowly update the imac line to use the powermac g4s (1.4ghz), and then use 970s in the powermac line for dual 1.8-2.5s?
With IBM claiming the Power 5 chip (which should come out next year), to be 4X the performer of Power 4…
Errrr. Even a slimmed down version would be killer, unless AMD, Intel, or Sun can come out with anything near it…
I don’t know why people are so fascinated with the PowerPC. It’s really a pretty boring RISC architecture. It’s only 32-bit (POWER is a related, but distinct architecture), at least right now. The instruction set isn’t as clean as MIPS (it was once remarked that how could the PowerPC be a reduced instruction set architecture if it had more instructions than x86
. Even at the same clock speed, it’s not that much faster than an x86 chip, not like an Alpha, which, at 1.2 GHz is only 25% slower than a 2.8 GHz P4 in SpecFP.
Also IBM has G3 CORES running at 100 Ghz in the Lab — yes 100, yes One Hundred
I don’t know why people are so fascinated with the PowerPC. It’s really a pretty boring RISC architecture. It’s only 32-bit (POWER is a related, but distinct architecture), at least right now.
It’s because their assembly language isn’t insane like x86 assembly. Generally they’re very cool processors because they were well designed from the ground up whereas x86 started with a crappy design and it was just hacked to death for years.
Also IBM has G3 CORES running at 100 Ghz in the Lab — yes 100, yes One Hundred
Are you serious?
Finally, some good news on the Apple/PPC front. While 2.5Ghz was no doubt achieved using selective methods, it still bodes well for Apple.
Of course what is more important is that IBM is able to ship enough volume to Apple as soon as possible. Fingers crossed for a late ’03 launch (hey, we can but hope)
LW.
1) POWER is 64 bits and the PowerPC 601-G4 are 32 bits. So you are correct.
2) The PPC970 is a POWER chip and is 64 bits. It also includes the multimedia extentions of the G4 and is backwards compatable with the 32 bit PowerPC.
3) All POWER chips can run 32 bit PowerPC programs.
About that remark about the prototype not having a fan. It’s most likely a 1.2GHz system and at 19Ws who needs a fan since the heat sink can do the job by itself.
If anyone is keeping track… GCC already supports the PPC970 chips. Linux is already running on the System.
I just hope that when Apple releases the 64 bit version of Mac OS X; they fix the OS. They need to lose Mach-O binary format (doable since the 64 bit programs will need a new format anyways). They also need to fix the VM sub-system.
Anyone even stop to consider what a mac housing a couple of these powerhouse chips would cost? I mean, you have to shell out a fair amount of money for a dual g4 today, and with these processors… I do not even want to start counting:)
Gimme gimme!
Most of the PowerPC design is 64 bits. The address space is 32 bits, but its nearly 64 bit ready.
The PPC970 is not a POWER chip; POWER chips are a subset of PPCs. The POWER line is marked by their incredibly large caches–they are intended for servers; hence, IBM never cared for AltiVec, as server operations do not have much demand for vector accelerations.
Aanund–chip costs should essentially be determined by chip size (unless they are having production problems which doesn’t seem to be the case). Since the 970 is only marginally larger than the G4 and since IBM’s production capabilities are larger and better than Moto’s, it should sell for about the same price. Maybe that’s too high for you (current PM prices), but I wouldn’t expect anything like 4000 – 5000–these should hit at 2300 – 3000.
Rayiner, your post is mostly wrong: the 970 is a PPC, not a POWER chip, but it is 100% 64 bit. So are you trashing the G3, the dolphin, or the 970? THe 970 is going to have an 8-way superscalar processing unit, a wide FSB, AltiVec, a longer but not contrived pipeline, run substantially cooler than Intel procs, 64 and 32 bit compatibility… I’d ask you, what’s wrong with that? A 4G P4 doesn’t sound to me like it’s got much going for it, never mind having something on the 970.
The PPC advantage is that it was designed to scale (Moto’s G4 problems aside) the fact is PPCs are in the smallest embedded applications and some of the largest servers,; you can’t say that for x86, or x86-64, or Itanium. PowerPC is a huge line of processors used in embedded devices, cellphones, game consoles, personal computers, blade servers,a dn high end servers.
And just wait for IBM to really start pushing the SMT capabilities of this new 9×0 line… maybe then you’ll drop the: “meh, don’t see anything interesting here” bit.
“POWER chips are a subset of PPCs.”
Wrong, actually PPC is a subset of Power. PPC does not include all POWER ISA, therefore some of the PPC machines running AIX for example have to emulate part of the POWER ISA for compatiblity with older RS6K systems/code. There are some other
subtle differences between POWER and PPC…
BTW this one of the first 64bit PPC implementation, the previous PPC (601, 603, 604, 750 and G4) are 32bits. 64bit floating point support does not make a processor “64bit”, rather the size of the integer register, datapath, and pointer size support (both in size and aryth ops) is what makes a 64bit CPU.
Even though POWER instructions are stripped out of the PPC line, the POWER series is now considered a subset of the PPC family. This is because the POWER was begun back before 1990 and the PPC Alliance was created to push its advantages, but when PPC became a family, they wanted the chip architecture to scale into markets beyond what the POWER was designed for.
As for your 64 bit points: “PowerPC Architecture is a 64-bit architecture. This architecture extends addressing and fixed-point computation to 64 bits, and supports dynamic switching between the 64-bit mode and the 32-bit mode. In 32-bit mode, a 64-bit PowerPC processor will execute application binaries compiled for the 32-bit subset architecture. Because a description of the entire architecture is too large to be addressed here, this paper concentrates on the descriptions of the changes that affect the user-mode 32-bit subset architecture.” Even G3s and G4s are largely 64 bit, but yes, the address registers are 32. Just because the 32-bit user modes have been exploited doesn’t mean these chips aren’t largely 64 bit already.
People should probably read this excellent 2-page story at IBM:
http://www-1.ibm.com/servers/eserver/pseries/hardware/whitepapers/p…
…that the PPC and other RISC chips have more instructions than Intel/CISC chips. Even though RISC stands for Reduced Instruction Set, the philosophy behind the design is simple: “The primary objective of those architectures was to be sufficiently simple so that implementations could have a very short cycle time, which would result in processors that could execute instructions at the fastest possible clock rate.” The philosophy is not to have fewer instructions available. The philosophy being that instructions should be able to be executed per cycle, rather than taking multiple cycles. In terms of “reduced instructions,” the philosophy is to have consistent, well-formed instructeds so that they are simpler to decode, executing the isntructions is reduced.
IBM adding to this philosophy by compartmentalizing certain functions, adding additional processor units to handle special types of instructions. “The POWER Architecture was unique among the existing RISC architectures in that it was functionally partitioned, separating the functions of program flow control, fixed-point computation, and floating-point computation. The architecture’s partitioning facilitated the implementation of superscalar designs, in which multiple functional units concurrently executed independent instructions.”
I think there actually was a 64-bit PowerPC in the last model of the R/S 6000 or something to that effect. I’m pretty sure IBM tried this once way before the average consumer market was ready for it… but the 6000 was a server machine anyway. I’ve actually got a PowerPC programmer’s manual sitting here in front of me POWER and PowerPC by Shlomo Weiss and James E. Smith and it talks all about 64-bit PowerPCs and distinguishes them from POWER. I don’t know. Maybe the book is junk, but it seems real to me
think there actually was a 64-bit PowerPC
Yes, you are correct, and Rayiner Hashem is wrong.
From FOLDOC: http://wombat.doc.ic.ac.uk/foldoc/foldoc.cgi?PPC
The PowerPC standard specifies both 32-bit and 64-bit data paths. Early implementations were 32-bit (e.g. PowerPC 601); later higher-performance implementations were 64-bit (e.g. PowerPC 620). A PowerPC has 32 integer registers (32- or 64 bit) and 32 floating-point (IEEE standard 64 bit) floating-point registers.
Some more on the PPC 620:
http://www.computer.org/proceedings/compcon/7029/70290285abs.htm
The PowerPC 620 RISC microprocessor is the first chip for the application server and technical workstation product line within the PowerPC family. It utilizes a high performance microarchitecture with many advanced superscalar features to exploit instruction level parallelism. It is the first 64-bit implementation of the PowerPC architecture supporting both 32- and 64-bit application software, and is compatible with the PowerPC 601, PowerPC 603, and PowerPC 604 microprocessors.
The PPC970 is assuredly NOT the first 64-bit PPC processor.
ha ha, use it while you can, lameheads.
OSX will still be slow when comparing a 500 MHz PC running XP with a G4 500 Mhz PPC running OSX. So, even if PPC970 run at 2,5 GHz and consequently runs OSX faster, so a P4 at 2,5 GHz does when running WinXP.
Nothing has changed. Lamehead.
How Apple had plans to use the 620, because it could be used to run PC instructions, one of a few attempts to emulate the WinWorld on the Mac, back in the day… Anyway, just thought I’d mention that never-delivered product in light of the recent VPC ness.
Sure, it’s been a little painful, but Apple has had to optimize things quite a bit, and do things like Quartz extreme to shuffle off load to the GPU. If Apple had really fast processors all along, much of this may never have happened if we had these screamining 64 bit ibm cpus all along. things should really scream, with the optimizations and faster chips. Ha, ha, I can’t f**ing wait!!!
THats all well and good but here is the problem. Apple will be charging 3000 dollars for sure on their Power mac ppc970. So not too many people will be able to get one. I wish that some of the OPPC vendors would release a version of an ATX mobo in the 100 dollar range. Hell I wish MS would release Windows XP PPC64. That would be quite pimp. Why do I say this? Because apple prevents users from using OSX with OpenPPC. And to hell if Im gonna pay that much to run linux as my primary desktop.:)
THats all well and good but here is the problem. Apple will be charging 3000 dollars for sure on their Power mac ppc970.
Uh huh, “3000 dollars for sure”
Even the rumor sites aren’t speculating on the cost of a PPC970 based Macintosh.
I think you should shove that number right back into your ass where you pulled it out of.
One thing is certain… this chip isn’t being made with people like you as the key demographic. (unless you were to *gasp* consider buying a Mac) Keep that in mind…
“Even G3s and G4s are largely 64 bit, but yes, the address registers are 32. Just because the 32-bit user modes have been exploited doesn’t mean these chips aren’t largely 64 bit already.”
Can you address more than 36bit memory space? NO
Can you do 64bit integer ALU ops? NO
Again, for a processor to be considered 64bit, its Integer ALU datapaht must support 64bit operations. 64bit FP ops (IEEE) do not make a processor 64bit. This is standard practice, the fact that you have internal/external 64bit buses doesn’t make a lick of difference, for the same reason that a Pentium can not be cosidered a 64bit processor. The G4/3 do not have a 64bit “user” mode (whatever that means inside a processor) BTW.
ISA wise the PPC ISA is a subset of the POWER ISA. The current POWER4 is actually the merging stone for both families (yes as far as IBM is concerned POWER and PPC before POWER3/4 were sepparated families, not a subet/superset relation), since POWER4 supports most PPC64 instructions, and at the same time it implements some of the deprecated POWER instructions via microcode (ugh) or like previous PPC by also using traps. Earlier POWER did not support certain PPC instructions (for example POWER implements a few FP ops in a different fashion than PPC).
As far as the PPC 620, well… I did not really feel like include vapour (even if it is 64bit vapour) in the review of PPC implementations
. And I think the 620 could not run/emulate x86 in hardware. You may be thinking of the PPC615 which had a x86 core in the same package for emulation.
Ok
What makes you think that their high end machines are gonna be cheap? They are selling piece of shit Dual 1.4 ghz G4’s for $3700. Why cant you buy a single 1.4ghz? So they can charge NEARLY 4000 dollars for a computer.
http://store.apple.com/1-800-MY-APPLE/WebObjects/AppleStore.woa/243…
SO that number will remain right of my ass even though thats where it came from because we both know IT AINT GONNA BE 1500 bucks.
I am not talking about the garbage iMacs I mean full speed G4s.
Tired of waiting “next year” for the G5. I hope this chip shows up in powermacs around summertime…
– Mark
Me too dude. Here’s to a G5 Amiga by 2004.
Firstly you have to replace your $4000 G4 with another (more) expensive G5 64 bit tower. Secondly AMD and Intel aren’t going to sit on their hands and watch this happen. You will still be able to buy a 4-5Ghz AMD/Intel 32 bit processor much cheaper.
A 64 bit AMD/Intel motherboard/processor will still be much cheaper than a whole new Apple.
It will take 3 years at least for substantial numbers to switch to a 64 bit Apple solution.
Just based on Apple’s past performance, they will likely cripple the PowerPC 970 with a slow memory bus, cheap and slow RAM, and other cheap and slow parts.
There is currently no Apple computer made that has top of the line parts in it. It would be foolish to think Apple’s low quality hardware mantra is going to change.
Don’t get your hopes up just because IBM is building good computers with the PPC970. Apple will likely use the lowest quality and lowest speed chips off the manufacturing line and still charge top dollar for them.
The rationale is simple and time-honored. Apple uses the margin from the high-cost low-value PowerMac line to make up for their lower margin iMacs and eMacs.
It will be the dawn of a new day when Apple sells machines based on performance. Don’t believe it until you see it.
Oh great! A 4-5Ghz 32Bit Intel/AMD processor. Will I need to buy a liquid cooler to keep the thing from acting as a space heater, or should I just put up with the deafening fan noise? My Main development workstation is an Athlon 1.4Ghz and I hate the excessive noise caused by the all the fans needed to keep this thing relatively cool.
1 Front Intake Fan
1 Power supply fan
1 rear exhaust fan
1 chipset fan
1 big-ass dragon-orb CPU fan.
I would hate to imagine the cooling requirements for a 4-5Ghz processor.
Apple has a defined market and it ain’t those whose are worried about getting as much Mhz for the least amount of money.
Ok
What makes you think that their high end machines are gonna be cheap? They are selling piece of shit Dual 1.4 ghz G4’s for $3700. Why cant you buy a single 1.4ghz? So they can charge NEARLY 4000 dollars for a computer.
Maybe you should look at that site again. On there they have a duel 1.4 GHz for $2,699 the $3,700 Machine is just one maxed out with more ram and a faster graphics card. Also you can buy one with a single 1GHz for only $1,499.
My Main development workstation is an Athlon 1.4Ghz and I hate the excessive noise caused by the all the fans needed to keep this thing relatively cool.
Look, I’m anything but anti-Apple, but my Athlon XP 1800 is almost silent when compared to my roommate’s dual 1 GHz G4. Even with his door closed I can hear it. iMacs and some of the older models such as the Cube were certainly quiet, but the top of the line Macs no longer are.
Nik: What makes you think that their high end machines are gonna be cheap?
I didn’t say that, I said you were pulling numbers out of your ass. You should wait until the end of the year when one of these systems actually begins to materialize before even bothering to guess on a price.
Anonymous: Firstly you have to replace your $4000 G4 with another (more) expensive G5 64 bit tower. Secondly AMD and Intel aren’t going to sit on their hands and watch this happen. You will still be able to buy a 4-5Ghz AMD/Intel 32 bit processor much cheaper.
Do the math. PPC970 at 2.5GHz now, most likely outperforming the current top processor (3.06GHz P4) and the Power5 four times as powerful as that.
x86’s has always had one main advantage over their competators: process. AMD and Intel have developed manufacturing processes which Sun, Motorola, and SGI aren’t even near reaching. IBM is essentially ahead of Intel and AMD on manufacturing processes, and now they will soon have a core which is 4 times as powerful as the nearest x86 competator.
I think it’s very likely that Macs will regain the crown as the top performers in the near future.
A 64 bit AMD/Intel motherboard/processor will still be much cheaper than a whole new Apple.
True, but when has price been an issue for Apple customers in the past, who were willing to shell out more money for less powerful hardware? Apple will never be about being the cheapest, but it will be a great boon to their sales if they’re once again the best.
Just based on Apple’s past performance, they will likely cripple the PowerPC 970 with a slow memory bus, cheap and slow RAM, and other cheap and slow parts.
Uhh, what? Cheap and slow RAM? Apple is already putting faster RAM than the chipsets in high end G4s can utilize. The problem isn’t the cost of components, it’s outdated technology.
IBM will certainly develop a high performance chipset for use with these systems (the reference board seems to have a XILINX FPGA where you’d expect the chipset, so I guess it’s still in development)
Apple History: There is currently no Apple computer made that has top of the line parts in it. It would be foolish to think Apple’s low quality hardware mantra is going to change.
Uhh, hello? The high end PowerBooks have… Gigabit Ethernet standard (something Sun Blades aren’t even shipping with), FireWire 800, DVI, Bluetooth. Most of their systems are using DDR RAM even though their current chipsets can’t make use of the bandwidth.
And regardless, there’s no arguing that the PPC970 is going to be an incredible processor. That alone should qualify as a “top of the line” part. Certainly the Power5 derived PPC970 successor will be.
Only 1499 for a new 1ghz machine!!!!!!!!! Jee Whiz thats great!!! To bad X86 can be gotten at 3ghz for that price.
High end machines arent suppossed to be cheap however they arent supposed to be in the 3500 dollar range apple has right now. Thats ridiculous! Bottom line I could build a top dollar pimp AXP or P4 machine for half that price. All I am saying is that Apple hardware is too expensive simply because they have closed themselves off from competition.
http://www.quietpc.com
It really doesn’t cost THAT much to turn off the noise. Some of this might possibly be of interest to those plagued with the “Windtunnel” dual 1.25 G4 towers.
[i]I don’t know why people are so fascinated with the PowerPC. It’s really a pretty boring RISC architecture. It’s only 32-bit (POWER is a related, but distinct architecture), at least right now[i]
This is completely untrue.
The PowerPC ISA is a 64-bit specification, the mainstream CPU’s and microcontrollers we see predominately today are built on an specified 32-bit subset implementation of the full PowerPC ISA. The PowerPC 620, the POWER 3 and the POWER 4 are all techincally 64-bit PowerPC CPU’s. The contiuation of the moniker of POWER on the high end RISC CPU’s is for branding and market segmenttion.
The PowerPC ISA is also the next generation of the old POWER ISA. In fact every POWER CPU since the POWER 3 is also technically a PowerPC CPU in that the two ISA’s have a tremendous amount of overlap. By design. In fact the original PowerPC 601 offered full machine compatibility with the old POWER ISA in addition to the new PowerPC ISA.
And regardless, there’s no arguing that the PPC970 is going to be an incredible processor.
It very well might be a good processor. But even though there are many good reasons to believe that Apple will use it, you don’t know that they will, so isn’t arguing about how great it is going to be a little pointless? I mean, it’s perfectly fine to get excited about new technology, but it gets pretty old to constantly hear about how a speculative product will be so much better than the competition.
I mean, for years we heard that about the G5–how much it’s going to destroy x86 chips–which I think we all know isn’t going to come out. Granted, the 970 does exist and there is a lot of reason to believe Apple will use them, but they haven’t said they will. (And, of course, they won’t, because they’re smart enough to know how people would react and they don’t want to hurt G4 sales right now.)
I have nothing against Apple and nothing against the 970, but your argument is inherently weaker when you say that “I assume the chip will be faster and I assume that Apple will use it, and thus Apple will soon have the speed crown again.” A conclusion built upon merely two assumptions is a dangerous thing. For better of for worse, PCs have two speed advantages right now: x86 chips are unquestionably faster at the moment, and OS X hungers for powerful chips much more than does Windows XP.
I hope everything works out for Apple–I rather like the company–but if I were their executives, I would carefully consider all options before jumping to the 970. Not saying that I wouldn’t use it–just saying that I’d be a little nervous that the same predicament that they’re in now with Motorolla might be repeated later if IBM loses interest in the desktop PowerPC market.
I hate it when people bring that up.
No desktop user today can use gigabit ethernet OK. There’s a reason the Sun systems aren’t shipping with it, and it’s not because they’re stupid. It’s because it’s useful mainly for connecting together high speed server farms.
It’s this kind of thing that turns me off about Apple. Them and their fanboys went around for ages talking about the “Mhz” myth, and then Apple come out new machines that wind up the numbers even when they weren’t being pushed before. Ram? GigaBit ethernet?
Do ppl ever stop to think that a company of the size of IBM have their own agenda with the PPC 970 other than to sell it to Apple?
Fact 1: IBM is investing heavily into Linux and is one of the largest Java-based solutions vendors.
Fact 2: Linux is getting into a state where it is more than ready for a workstation desktop for corporations.
Now, throw the 970 into that and perhaps IBM have an agenda to hit HP and Dell at the corporate desktop with lean, powerful and cheap desktops.
IBM have not been after the consumer market for some years now and quite frankly I don’t see them going down that road any time soon. Media ready computers/consoles/hifi gear will move in there with Sony, MS and others already firmly positioned in this market. Apple is, with their media-hub strategy, giving this market a push but does not have the products or the price level to stop these new products with “simpler” origins from maintaining their dominance.
Apple really needs a new CPU and the PPC 970 looks nice enough but we have yet to see a solid PR that confirms that Apple will be able to use it. UNTIL then, Eugenia is absolutely correct to call it a rumour. Many rumours turn out to be correct but far to many doesn’t. Let’s hope this one is true.
Misc comments:
steve wolff: “Also IBM has G3 CORES running at 100 Ghz in the Lab — yes 100, yes One Hundred”
No they don’t! If they could run the G3 core at that speed then y are we not seeing higher speed G3s say at 20Ghz? I think you are mistaking this for a report of IBM having a in the lab constructed transistors switching at that speed. Now ONE transistor does not make a G3 core.
There’s a reason the Sun systems aren’t shipping with it, and it’s not because they’re stupid. It’s because it’s useful mainly for connecting together high speed server farms.
The Gigabit Ethernet and DDR Ram in Apples computers are simply marketing things. Trying to sell a computer today that uses PC 133 RAM as a high performance computer would be laughable regardless if DDR RAM is needed or not. The competition would just point at the PC133 RAM and compare against their own PC333 RAM as a “fact” shoving their own product superior. The same is true about Apples use of Gigabit. When there will be a need for Gigabit Ethernet then there’ll be Gigabit ethernet on all the computers that needs it.
SUNs computers are not the fastest on the market CPU wise but SUN knows what’s needed and that is throughput and at that they excel. Now where’s the Gigabit on their workstations? Could that mean that it is not needed yet?
Exactly who is going to buy these G5 Macs? People make the BMW/Hyundai comparison all the time. The reality is people pay a premium for a BMW because it is a superior piece of engineering to a Hyundai. The reality is that a high quality PC is a far better piece of hardware than a Mac. There is no point saying that PCs suck when you are comparing a bargain basement PC with a G4. In Australia a dual 1.4 gig G4 box costs AUS$8009. A similar specced (and faster)high quality 2.8 Athlon costs less than AUS$3000 with XP Pro bundled.
The Mac users don’t care about price argument is rubbish…Apple market share has fallen from 20% to less than 3% in a decade. Plenty of people would buy a Mac if they were reasonably priced and fast…the reality is you are buying something that has the looks and price of a Ferrari and the performance and build quality of a Hyundai.
I remember reading a Mac mag in about 1997 with the PowerPC roadmap…I think it is about 3-4 years behind by now.
A 64 bit mac will arrive 6 months later, 30% slower and 25% more expensive than predicted.
(unfortunately)Pretty soon Macs will lose mindshare and slip slowly beneath the waves. If you don’t think it will happen you don’t remember Commodore Business Machines.
The 64 bit capabilities would be extremely useful for conquering the work station, creative, and server markets. It is quite possible that any G5 would also provide another potential pricing segmentation. Mr jobs could then offer 32 bit machines and 64 bit machines. The former might be the consumer oriented ones. The latter could be targeted to the creative/industry applications, at least initially
So mac haters, just give up. OK?
What is the system architecture of the PowerPC970? Does it use a shared FSB like Intel (Pentium 1-4), a point-to-point FSB like Athlon (and Alpha 21264?), or is memory connected directly to the CPU with CPU interconnects like Hammer (and Alpha 21364?). Hammer has me really excited because of the ease of building multiprocessor systems and increased theoretical scalability (esp. with NUMA kernel optimizations). I’m really curious to see how it performs.
Regardless of its system architecture, the PowerPC970 seems like a good chip for Apple (I doubt it will blow away the x86 side performance wise, but all it really NEEDS to do it be competetive – unlike current G4). I imagine IBM will be a much more reliable supplier of chips than Motorolla.
“No desktop user today can use gigabit ethernet OK”
What? I have three macs at home and they all talk to each other at blizzard speeds over the gigabit port. I think you meant to say, “No PC desktop user today can use gigabit ethernet OK”
javi, no, POWER is a subset of PPC. Just because POWER includes instructions outside of the PowerPC instruction set, doesn’t change the language used by the AIM alliance. The PowerPC alos contains instructions outside of POWER chips. Yes, the two families will not be completely bridged until POWER4 and 5 completely replace existing POWER chips. But the concept of the PPC famly is that, in fact, POWER chips are also in the family and are a subset.
G5 talk? What would have definitely have been the G5 is cancelled. Calling the 970 is inappropriate. (I hope Apple calls the 9×0 family the G64, but whatever.) What I would point out is this is further evidence of Apple’s use of the 970. (I personally know that they will use it, but yes, my sources are people under NDA saying vague things but that I will be happy and soon.) Although people got sucked into rumors of the G5, the story is becoming more and more clear that Apple rejected the G5 because of Moto’s insistence on bus architecture. And that shortly thereafter, IBM began design of the 970. Also, yes, IBM does plan to use the 970–but Altivec code is just being added to Linux, zero Linux apps have AltiVec optimizations, IBM has been uninterested in licensing AltiVec… Do people still doubt where this chip is going? Whatever, some people need to put their fingers into Jesus’s wounds…
“IBM have not been after the consumer market for some years now and quite frankly I don’t see them going down that road any time soon.” Who says they are? This is the next great moment in the PPC line–this is the merger of the highend with the workstation class chip at the cost of a consumer chip. Yes, IBM is going to sell these chips, and if attitudes here indicate anything, PPC Linux is going to become very popular, and Apple WILL be shipping these too. Yes, AMD and Intel might get some serious competition from a completely different architecture because they are going to have the consumer-accessible 64 bit lead without any compatibility (Intel) or support (AMD) issues.
Yes, Spishak, Mike’s silly. Many small offices, freelancers, home users are able to take advantage of Gigabit ethernet for little cost.
Whose going to buy this? Umm, all the people that have been holding off on the PMs for the last three years. You are talking about lots of patient people who are dying for it. That’s a lot of pent up demand–I bet that no matter how good the volumes are, Apple has difficulty keeping up with orders for many months. These are going to sell like hot cakes (do hot cakes sell well?).
What? I have three macs at home and they all talk to each other at blizzard speeds over the gigabit port.
Could you please elaborate a little about the configurations of your Macs and the tasks you’re doing? I’d be more than a little surprised to hear that a home user’s hard drives are so fast and needs are so vast that 100baseT is actually a bottleneck.
“It’s this kind of thing that turns me off about Apple. Them and their fanboys went around for ages talking about the “Mhz” myth, and then Apple come out new machines that wind up the numbers even when they weren’t being pushed before. Ram? GigaBit ethernet?”
Obviously, PC Fanboy, you have never set foot in a design or graphics shop, but then playing the latest 3rd Person shooter doesn’t exactly expose you to the real world.
Try transferring some enormously large files all day long on a network running 10/100, and then get back to me as to how this doesn’t matter.
Being a PC Fanboy, I’m sure you think that Firewire 800 is useless also. Of course, you Fanboys were the same ones who suffered with USB 1.1, while Mac users were using Firewire 400 for years before.
What you might need in a desktop is not the same as everybody else, what a shocker, huh?
What a putz.
Off topic but neccesary for fast cpus.
As for that PPC mobo picture, the fans could be mounted from the case, this would keep the weight & vibration off the cpu mounting.
Several people are mentioning fan noise.
I have had many Macs & PCs all types. Fan noise has always been an issue with most of them.
Over time all fans get noisier, they loosen in their fittings and start to rattle. Leaving fans open to dust probably speeds this decay quite a bit, & I often have the lid off.
On my LC >>10yrs, the tiny fan would shake itself to death since it was just pushed into a wimpy plastic molding. My PPC6100 wasn’t any better, same silly design.
On both my 1GHz Ahlons the original fans from the crappy dealers were quickly replaced with ThermalTakes with impressive looking curvy fins. Over 2yrs those turned into rattle boxes, I was getting sick of all the noise from all the fans, PSU, HDracks etc. They were quiet only at a distance of maybe 50ft.
Solution,
1st I replaced the fixed voltage TTs with PCtoys fans that were bolt on Zalman heavy duty radial copper finned beasties $40 each. That price include speed/noise adjuster & alot of copper. Since these are slower cpus far below capacity of the fans, I set the voltage to silence (about 7V). Believe me they are basically silent at 1ft. The 1Ghz is now at 55c & the 800MHz is nearer 40c. I will reuse these fans when I upgrade to something 2x faster. At max level, the fans do have 2x speed and alot more noise, but don’t give much extra cooling.
The HD racks I stopped using, those 1″ fans are mostly noise makers, I put the HDs back on the PC floor and let it warm a little. Don’t use tiny fans.
Next I went inside the PSU and rewired the fans 12_0 connections to be 12_5 instead so now that fan gets 7V & is also whisper quite. This should only be done if you are comfortable with that & it has a rating atleast 250 or 300 and is lightly loaded. Better would be to insert a Zalman $6 voltage adjuster to control that. Better still to replace PSU with a silent model but those can be expensive. I admit I did fry 1 PSU coz I didn’t put case back on and airflow over PSU switcher was 0. That spare fan was put to use as an exhaust fan again at 12_5 for silence.
On the 1GHz I also used the silencing material $20, it makes a small difference, lining 4 sides of the case internally.
Some of this could apply to noisy PowerMacs too, since they probably use many ATX like parts, but I’d be alot more careful there.
Silence is beutifull!
I don’t know why people are so fascinated with the PowerPC.
Well, first of all it’s aimed at a niche which conerns us all, namely the desktop. Secondly, it’s seeing ongoing development in that direction.
Compare this to other also-rans, such as the ARM (now aimed squarely at embedded apps), the MIPS (after SGI’s let-down aimed at embedded apps) and the Alpha (murdered (yes!) by HP) and the SPARC (SUN’s pet and quite an odd processor in itself), and you’ll see why the PowerPC is important.
Not to mention that it required the killing-off of at least two processors (the m68k and the m88k).
Me? I’d like to see an Intel-like monster 68090. If Intel could push their piss architecture this far, Motorola could have, too.
ARM is not an also ran, it is possibly the most succeseful cpu in the embedded space, certainly in ASIC IP. Its life on the desktop though was doomed by x86. the PPC is only now being discussed for ASIC licensing from semi houses other than IBM/Moto so ARM is 20 or 30 companies ahead & solidly entrenched. One reason this is happenning is because PPC is in the Xilinx FPGA Pro fabric & ARM is in the Altera highend fabric and folks want to turn those FPGA protos into low cost ASICs.
Also 68K (on of my favourites of all the cpus out there) didn’t die either, it was risced into the ColdFire architecture. I believe most 68K code runs on it, and I think those parts run at a 300MHz or so but I haven’t checked lately, so it had a focus shift.
The idea of a 68090 isn’t so crazy, the cpu I work on in my spare time borrows many features from x86, 68k, ARM, and the Transputer of course (perhaps the mother of multithreading). But it will never see GHz silicon, just FPGA.
IBM wouldn’t have included Altivec if they didn’t intend for Apple to be one of their customers.
Of course this doesn’t mean that IBM has many other plans for the chips, but this nonsense that they didn’t have Apple in mind when they produced this chip is just that.
And Apple is not only going to use the 970, but they are going to have top-end rigs with Quads and even Eight processors per box.
Apple intends to own the high-end of the market.
What PC Fanboys fail to realize is that the speed title has traditionally gone back and forth, it’s only been in the last couple of years that Motorola’s complete imcompetence has put Apple in the hole.
IBM is a whole ‘nother ballgame, though, and by the end of 2003, the speed title is going to be Apple and IBM, and it might just be a butt-kickin’.
“Apple intends to own the high-end of the market.”
How is Apple going to build marketshare then? There’s no way for them to double their marketshare based on selling a small number of high-end workstations.
As current Apple PowerMacs are 2X to 4X the cost of the equivalent performance in PC world, if Apple goes higher-end with the new 64-bit machines, it would seem that while the machines might be very speedy, no one will be able to afford them and the PC will still offer significantly better price/performance.
I certainly hope Apple’s 64-bit machines do not cost any more than the current 32-bit machines. Apple has a real opportunity to revitalize the personal computer with an affordable 64-bit machine, especially as AMD has blown getting their 64-bit processors to market on time.
Now, im not any sort of analyst… But doesnt it seem likely that IBM wants to sell computers, they have a PPC design, they can garrantee business to apple, and sell their own Linux/AIX workstations at a lower cost (for higher performance, something I think Linux users want), and not rely on intel for cheap linux workstations. Also MS said theyll support 4 64bit arch. well, maybe IBM will sell ppc windows workstations too.
This also lets IBM stop selling low margin commodity hardware, but they probably wont stop entirely. Vertical markets are more profitable.
It makes a lot of sense from every angle, but I am a mac user, so I focus on the Mac issues.
I want one in my AMIGA ONE !
pX
Will they ever get it?
“it would seem that while the machines might be very speedy, no one will be able to afford them and the PC will still offer significantly better price/performance.”
Any long time Mac user can point to numerous quarters, even entire years, where Apple couldn’t meet demand with their limited production capabilities… The first 970 Macs will be one of these machines to be very scarce and hard to come by… Maybe the most desired in Apple history. Mac users are not concerned if they are going to sell, Mac users are concerned whether or not they are going to have adequate volumes to satisfy everyone.
You cannot please everyone. Just chill.
“Could you please elaborate a little about the configurations of your Macs and the tasks you’re doing? I’d be more than a little surprised to hear that a home user’s hard drives are so fast and needs are so vast that 100baseT is actually a bottleneck”
Glad to explain. Developing a MIDI-over-Ethernet app/system where I’ve coded the server and the client. MIDI data(messages) are stored in memory (client) or created on the fly (by the click/drag of a mouse) and passed over Ethernet to a waiting memory queue on the other end (the server). I do not write to harddisk as it would increase the latency over the socket. Obviously, the faster the ether connections, the better my sequencers sync time will be.
Thanks for hearing me out, are you surprized? It is too bad you are too narrow minded to realize that there IS good use for a fatter socket pipe over ethernet.
Uh huh. Here’s some more information for you to ponder. There’s much more to MIDI over Ethernet than top speed. You might find an isochronous 100BaseT connection a much better solution than a non-isochronous 1000BaseT connection.
From Emagic’s discussion of why they use MIDI to interface to Logic Control vs. Ethernet:
2. Why is MIDI being used for data transfer? Isn’t Ethernet faster?
MIDI is one of the best technical solutions available and the speed of the MIDI protocol is more than fast enough for this application. From a musical standpoint, MIDI is more than capable of carrying timing-sensitive note information from master keyboards. Controller movements, by comparison, are much less critical.
MIDI offers fundamentally more bandwidth than is necessary for the Logic Control System. To move a fader to its required position requires only the same amount of data as a single note event. In addition to this, MIDI was initially conceived as a real-time protocol; the evolution of the software and hardware over many years has resulted in a system capable of delivering especially low latency. Latency and transmission speed are two different things. Depending on the situation, MIDI is capable of providing better performance than an Ethernet connection, especially in respect to latency. It is not without good reason that there’s currently no Ethernet MIDI interfaces.
In any case, this discussion about transmission speed does not apply in connection with motorized faders. Logic Control features modern, high-end motorized faders from Penny & Giles — in the opinion of many experts, the quickest and best motorized faders in the world — which are themselves, as mechanical components, much slower than MIDI.
As a practicality, we recommend that all those interested make a direct comparison with other available devices. It will then become abundantly clear that Logic Control is one of the fastest controllers available, even in comparison with Ethernet-connected systems costing 10 times as much.
–ms
Fact 1: IBM is investing heavily into Linux and is one of the largest Java-based solutions vendors.
Observation 1: OS-X treats Java as virtually equivalent to Objective-C – most OS-X APIs can be used from Java, and Java applications can be started by the user like every other application.
Observation 2: The OS-X Developer Tools come with jikes as well as javac.
Is it possible that OSX will be ported to run on high end IBM machines? I imagine that OSX on IBM would be popular in certain situations such as high powered workstations. I have read that OS9 would run on IBM PowerPC machines but were forbidden by the Apple EULA.
OSX would need to have major work done before it is ready for any really high-end hardware. The current OS X is not pervasively multi-threaded nor can it be split into multiple processes.
Apple has done a bit of work getting it to support two processors but admits there is plenty more to be done before OS X even supports two processors effeciently.
Depending on IBM’s commitment, Darwin is open-source and any company who wants to optimise the foundations of the OS is free to. IBM have more Unix and Linux code knocking around than anyone else…
“Why is MIDI being used for data transfer? Isn’t Ethernet faster”
You are missing the point entirely, my app syncs MIDI over ethernet, be they in the same room, or in another room, or in another building. Noone was talking about making a MIDI interface over ethernet. My app will let me sync my sequencer with a sequencer at a mates house. It also lets me play, say an OXygen 8 keyboard, here at home and have the MIDI messages go over ethernet to trigger a sound modules anywhere in the world.
So when it comes to somthing HERE IN THE SAME ROOM AS MYSELF, data over a MIDI interface is more than fine, but when I want to sync up with someone elsewhere in the world, you’ll need to connect SOMEHOW right? For now, that is over ethernet. Do you finally understand or are you going to quote something else that is irrelevent, Michael?
Just give me a PPC 970 enabled AmigaOne and the whole world will be happy. Pffft, just forget about the Macs man! Hell Yeah!
The AmigaONE barely run the Motorola G4 as/is, what makes you think that broken northbridge can handle the 970?