IBM is disclosing the technical details of a new 64-bit PowerPC microprocessor designed for desktops and entry-level servers. Based on the award winning Power4 design, this processor is an 8-way superscalar design that fully supports Symmetric MultiProcessing. The processor is further enhanced by a vector processing unit implementing over 160 specialized vector instructions and implements a system interface capable of up to 6.4GB/s. Read more here and here. The first article speculates that Apple might find a future in these new IBM CPUs. Update: News.com has an article too.
IBM Expected to Enter the Desktop with a 64-bit PowerPC CPU
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Hey I get the first Post 🙂
Sounds good that Apple might finally get a CPU that will be “similair” in speeds with the PC industry without leaving the PPC. I am fine with that, lets just not keep the Apple 2 years behind the PC curve.
I don’t need to jump to the X86 close world just yet
What’s this baby gonna cost? It may perform…but if it’s a $1000 a pop…it’s probably not gonna help move too many macs…sounds more server oriented than anything…but Big Blue could prove me wrong and besides, Mac users are used to paying through the nose to get decent performance…dual G4 powermac anyone?
When. Soon, like yesterday.
It make sense that Apple ould bethe first to push 64-bit on the dest top. They were the first with a 32-bit line of deskop PCs.
>>What’s this baby gonna cost? It may perform…but if it’s a $1000 a pop
Have you not read the article well:
IBM is disclosing the technical details of a new 64-bit PowerPC microprocessor designed for desktops and entry-level servers
It means not so expensive.
After reading the article(s) on the new IBM chip it just might save Apples Butt. But then again I have seen this before with Apple. So hopefully Apple will do the right thing this time.
–to bytes 256 —
You seem a nice guy why always down on the PPC chip?
This is great news, especially for Apple and people who hate the stupid Intel instruction sets.
PowerPC and IBM chips have always had some of the most well designed instruction sets on the planet that actually made them nice to work with (as opposed to the x86 and itanium), and they always seemed better designed with niftier technology than the other guys. The problem was Motorolas fabs seemed to be geared to the low power/low MHz of the imbeded market.
It would be cool for IBM to make these with it’s new .10 micron fab plant that they just announced (timetable seems about right). Could have some very well performing procs, especially since the Power4 is an ass-kicking unit.
On the side: I would rather see them NOT include the 32-bit compatability. Yes, it would be a bit of an inconvinience, but the reduction in transistors and die size, along with the smaller fab would produce some efficient and fast procs. Plus, it is high time that we swiched to 64-bit. I’ve been waiting for this since the early 90’s and the Alpha chips!
One more reason why it is likely to be the new Macintosh chip, “over 160 specialized vector instructions”, as it happens Altivec is a set of 162 specialized vector instructions in Motorolas PowerPCs that is already widely used in a lot of Mac apps (known as the velocity engine on the apple side it seems). However Altivec is a Motorola trademark so IBM cannot call their set the same thing even if it would happen to be completely compatible.
Steve Jobs would no doubt make Altivec a requirement for a new Mac chip (in fact it was this issue that got IBM out of the PowerPC game a few years ago when they argued that Altivec was a bad idea and that they would have none of these specialized vector things
.
Another thing I seem to recall was an interview with Jobs a while ago where he said somehting along the lines of ‘between Motorolas and IBM’s roadmaps the PowerPC looks like a good idea for us still’ when faced with the x86 question. If I remember correctly one should probably have read more into that since IBM hasnt been terribly active in the PowerPC for some time before this announcement. Not sure however, maybe someone else remembers.
So things are looking good, the Power4 shows that IBM knows what they are doing so it is quite likely that we can look forward to well performing macs in the future.
Good news indeed. However you can bet that they have included 32 bit backwards compatibility, would be insane not to. However this is nothing like patching on 64 bit stuff on an x86, the PowerPC is very clean and largely planned for 64 bits anyway (just look at its big brother POWER). So there will be little problems and lots of advantages keeping 32 bit support really.
I’d like to see them in use on a GNU/linux PPC box or a Amiga4 box ^^
“the PowerPC is very clean and largely planned for 64 bits anyway (just look at its big brother POWER).”
I can’t remember for sure, but I recall reading that the PowerPC and Power4 were just different implementations (for different markets) of the PowerPC ISA/API that was jointly developed by Motorola and IBM. Does anybody remember exactly?
Yep!
What we got here, pardner, is a real computer.
This is the computer that John Wayne would have bought. (not really)
Good now Apple has a real alternative. As for compatability I’d argue don’t put it in the chip but rather in the OS. Have 32 bit Apple apps already run in a modified environment, it makes sense to just implement the entire 32 bit system in firmware/software and not pollute the chip for good apps. By keeping the chip pure 64 bit Apple will get amazing performance; most the apps that aren’t going to get ported probably won’t be hurt to much by not getting the boost. This also works well for seperating out the professional line from the consummer line for the next year or two.
Oh and even $1000 for this chip wouldn’t be unreasonable and wouldn’t hurt Apple much on their G4 line. Retail on the 2.5g Intel is about $500 and the Xeons hit $650. A 64 bit chip should cost more and you won’t need as many. A dual 64 bit setup (like a Sun 3500 or the E450s) outperforms a quad Xeon for most server apps.
jbolden1517: You cant just decide to make your OS run 32bit or 64bit apps just like that. Lets imagine a scenario like you described. The CPU is running its 64bit instruction set, then the OS also has to be 64bit. No big problems so far. You want to run 64bit apps, just pass them through the CPU, it’ll execute them flawlessly. But if you want to execute 32bit apps, and your CPU doesnt support 32bit code, then you have a major problem on your hands. Your choices are limited. Emulation is one option, have a virtual PPC running which could execute the code, but thats just plain slow. Really i cant think of other options. Mac OSX already isnt known for its speed, having to emulate a CPU to run apps would only hurt it if you ask me.
A CPU with 32bit compatibility could run faster at 64bit than a 100% 64bit processor can at 32bit. Get my point?
This is indeed very good news, but not quite as good as some of the rumors (“3 times faster than Pentium4 3GHz”). When it comes to SPECbench, the full Power4 is about 35% faster on SPECfp, and 11% SLOWER on SPECint compared to Pentium4 2.5GHz (Itanium is faster than Power4).
Still, this makes a VERY nice chip combined with Altivec, and Apple will finally be able to compete in the high-performance marketl It will not run rings around all other CPUs, though…
Very hard to tell something from the scored of the POWER4, this largely _has_ to be a major redesign at best borrowing some concepts from the POWER4. Mostly because the POWER4 is such an odd architecture (multiple cores per die, gigantic shared cache and more interesting stuff). I would take the POWER4 scores mostly as an indication of what IBM can do with the basic Power design really.
But of course it wont be 300% faster than a 3GHz P4 overall if it really will be a desktop chip, some sanity still exists in the world
Acorn was the first with a 32bit Desktop AND with a Desktop RISC machine (almost 15 years ago) 😉
Thoems
And I always thought that 15 years ago, the only one REAL multimedia multitasking computer and 32bit OS was Amiga 🙂
-pekr-
This should help apple, provided it really happens, gain more shares in the server market. 64 bit capability would probably also be very well received by work station and particularly those SGI creative types.
Hopefully this will help apple get more market share in higher end and high margin markets so that they can decrease their reliance on desktops. Because if that happened then they could make the necessary changes, which i think are largely price driven, to retake a nice piece of the desktop market away from MS.
Didn’t Alpha have some sort of emulation application to allow 32bit Windows apps to run on the 64bit proc? The 32bit emulation had some catchy name which escapes me. If I remember correctly, it ran with decent speed, but had some app compatibility issues.
A.N.
i remember reading something about the G4 being already for 64 bit Apps and OS. Internally the chip already speaks 64 bit, if I remember correctly. The altivec doohicky within the G4 chip works at 128 bits on its stuff so, who out there knows really what the current G4 chip can do already? I sorry if this is very nebulous but I have a lot going on at the moment. Thanks all.
FX32 was the name of the x86 32bit emu on alphas, and i think it was partially hardware?
It make sense that Apple would be the first to push 64-bit on the dest top. They were the first with a 32-bit line of deskop PCs.
This isn’t true at all. Acorn (UK) were the first with the ARM2 based Archimedes in 1988. ARM (Acorn RISC Machine) was specially designed to power the Archimedes (Later RiscPC) line of machines. Macs were still 16bit at the time.
Acorn later spun ARM development off in a joint venture with Apple called Advanced RISC Machines who – to cut a long story short – don’t fab chips they design, but licence them to Digital among others. Intel (the Microsoft of the Hardware world) picked up StrongARM from Digital when they were swallowed by Compaq.
A dual 64 bit setup (like a Sun 3500 or the E450s) outperforms a quad Xeon for most server apps.
Which ones?
I strongly suspect that a Sun E450 (max 4x480Mhz US-II) would have a terrible time trying to keep up with a quad 2.4Ghz Xeon in nearly any server application. If you restrict the poor E450 to dual processors, it is really going to be in sad shape. The 3500 would do quite a bit better than the 450, but you’d need to stuff eight processors into the system. Both these Sun servers are quite old.
TPC-C is not all that great a benchmark, but here’s a pair of nonclustered results. This particular Sun benchmark is the only TPC-C result currently published by Sun (guess why?)
Sun 4500 (14x464Mhz) SymfoWare VLM 3.0 = 67102.53
Three months later,
PowerEdge 8450 (8x900Mhz) SQL Server 2K = 69901.74
at about half the price.
If you do have any benchmarks showing a large performance gap between Xeons and the older Sun processors, or even newer Sun processors, please post them.
Yours truly,
Jeffrey Boulier
If the chip is actually designed for a “desktop”, then it must be Apple’s next chip. IBM does not make any desktops that use RISC. Some workstations, but no desktops. And IBM didn’t add all those vector instructions for nothing.
The only volume POWER-type platform is Apple.
So Apple gets their performance and gets to increase prices as well. Now that sounds like the Apple I know.
#m
> But if you want to execute 32bit apps, and your CPU
> doesnt support 32bit code, then you have a major problem
> on your hands. Your choices are limited. Emulation is one
> option, have a virtual PPC running which could execute
> the code, but thats just plain slow. Really i cant think
> of other options. Mac OSX already isnt known for its
> speed, having to emulate a CPU to run apps would only
> hurt it if you ask me.
Well three points:
1 – Hardware assisted emulation can be done outside the CPU or at least not done over and over again. 32 bit code when loaded from disk passes through a subsystem and gets converted into 64 bit code. Alpha/NT (another poster mentioned) used a variety of this but further improved it by having the system keep track of additional information during runtime so that the next load was more effecient when it converted. The advantage of doing it outside the CPU is it keeps the CPU simple and thus easy to improve quickly.
2 – Most MacOS apps look will look like:
simple instructions
call to carbon/cocoa
simple instructions
call to carbon/cocoa
simple instructions
call to carbon/cocoa
simple instructions
call to carbon/cocoa
Executing the carbon/cocoa call for most of them will be the bulk of the time. Once carbon/cocoa are 64 bit the speed improvement there will offset any speed loss on the simple instruction component.
3 – Finally you missed a core point in the migration I proposed, the apps that are intensive (in apple’s case the OS and Adobe stuff) go quickly to 64 bit. In addition all the Unix stuff people can download goes quickly to 64 bit. Something like Procite (bibliography software) might stay 32 bit for a long time but who cares if this runs slow (even if I agreed it would, which I tend to doubt due to point 2 above)?
> A CPU with 32bit compatibility could run faster at 64bit
> than a 100% 64bit processor can at 32bit. Get my point?
I don’t disagree that if your goal is to offer 64 bit functionality but mainly run 32 bit apps you want to essentially bundle an entire 32 bit system in with the CPU. I just don’t see that as the situation Apple is likely to face. That is rather than follow the Microsoft / Intel model make a clean break to 64 bit (like Sun, SGI… did) makes much more sense. BTW this clean break in the CPU with outside support was Apple did when they switched from the 68000 line to PPC.
>>>The only volume POWER-type platform is Apple.
Apple represents only 25% of all PowerPC chips manufactured by IBM and Motorola. The ONLY volume PowerPC/POWER platform is the network/communications platform.
The ultimate goal for IBM is ONE single hardware platform and ONE single OS platform — big R&D savings.
They are planning to move their (z-series) mainframes to the POWER6 platform (they already moved successfully their mini-mainframes and mini’s [i-series and p-series] to the POWER4 platform).
http://news.com.com/2100-1001-919579.html
They are also planning to move all their mainframe OS tricks to AIX (all these partition and self-healing things) and then run AIX on all the platforms. They are already running AIX on emulation mode in i-series and they plan a native version for 2004. Once they get that done, they will try to move their mainframe z-series to AIX as well. (Linux can’t scale up to that level and also IBM is not putting those mainframe tricks in Linux.)
http://news.com.com/2110-1001-949007.html
>>(Itanium is faster than Power4).<<
Actually the Itanium is slower than the Power4, but the Itanium 2 is tied with the Power4 since the Power4 outshines the Itanium 2 in integer figures, and the Itanium 2 outshines the Power4 in floating point figures!
Actually the Itanium is slower than the Power4, but the Itanium 2 is tied with the Power4 since the Power4 outshines the Itanium 2 in integer figures, and the Itanium 2 outshines the Power4 in floating point figures!
It’s important to remember that these results refer to the *real* Power4. A desktop derivative will likely have lower performance (although hopefully not too much).
bsharrit:
But Apple isn’t the first to 64-bit desktops, AMD has already beaten them to the punchline, 64bit chip with full support for the older 32bit x86.
Tem:
That 128bit Altivec crap is just marketing hype, basically, it has a 128bit wide datapath to the l2 cache.
G4 is a 100% 32bit processor.
There seems to be some confucsion as to what the difference between 32 and 64 bit processors is.
I doubt very much that supporting 32 bit binarys is going to make much difference to the performance of a 64 bit CPU.
What after all is the difference?
On a 64 bit CPU I would expect it to handle 64 bit integers and 64 bit addressing. An application not compiled for this will not take advantages of these features but will it make any difference to the speed?
You could make a CPU which didn’t support 32 bit integers or 32 bit address and yes it will make instruction decoding simpler and faster but what do you do when the programmer wants a byte to store something? Give him 8 bytes but get the compiler to ignore the other 7? That’ll waste memory and bandwidth.
The big problem I suspect is memory allocation, you’ll need a OS which supports 64 bit addressing for this to work, when you have a 32 bit app you can just map that address space into the 64bit address space.
32 bit versions of widows have been running 16 bit programs for years without any problems so why should it be a problem now?
Apple were not first with 32bit PC
This isn’t true at all. Acorn (UK) were the first with the ARM2 based Archimedes in 1988. ARM (Acorn RISC Machine) was specially designed to power the Archimedes (Later RiscPC) line of machines. Macs were still 16bit at the time.
The Mac II had a 68020 in 1986 and I think PCs got 386s about the same time so they beat the Acorn to 32 bitness by 2 years.
However this is a murky issue, it really depends on the definition of what a 32 bit CPU means. For instance the 386 was a 32 bit CPU but what about the 386SX? It was also considered a 32 bit CPU (addressing, register size) but the difference was only the data bus size (16 bit).
The Mac, Amiga and Atari ST used the 68000 which at the time was referred to as a 16 bit CPU but thats not really true. The 68K have 32 bit registers and 32 bit address registers and a 16 bit data bus – just like the 386SX but the 386SX was always referred to as 32 bit…
On the other hand the 68040 and 486 both added Floating point units and these support 80 bit dual precision floating point types.
The Pentium has a 32 bit registers, addressing but a 64 bit data bus but is still considered 32 bit.
All the 68K series will run 32 bit programs including the 68008 which had an 8 bit data bus. It could be argued that the first 32 bit desktop machine was the Apple Lisa way back in 1983.
On the other hand the Acorn was the first RISC desktop – that can’t be argued with.
Acorn later spun ARM development off in a joint venture with Apple called Advanced RISC Machines who – to cut a long story short – don’t fab chips they design, but licence them to Digital among others.
Indeed, but others being just about everyone in the industry:
http://www.arm.com/arm/semicon_partners?OpenDocument
Intel (the Microsoft of the Hardware world) picked up StrongARM from Digital when they were swallowed by Compaq.
No, they get the ARM when DEC sold Intel a chip Fab to settle a leagle battle, Intel had to manufacture the Alpha for DEC as a result for a while.
Actually an Alpha was the first 64 bit computers. A company in CA was selling them 3 years ago. We are still looking at a non-native OS though.
>> dual 64 bit setup (like a Sun 3500 or the E450s)
>> outperforms a quad Xeon for most server apps.
> Which ones?
> I strongly suspect that a Sun E450 (max 4x480Mhz US-II)
> would have a terrible time trying to keep up with a quad
> 2.4Ghz Xeon in nearly any server application. If you
> restrict the poor E450 to dual processors, it is really
> going to be in sad shape. The 3500 would do quite a bit
> better than the 450, but you’d need to stuff eight
> processors into the system. Both these Sun servers are
> quite old.
I agree they are quite old and on a mhz level much slower. OTOH I run a dual E450 system pretty regularly on some intensive apps (complex document compositionhttp://www.docscience.com/Prod_Autograph_Architecture.asp) print queue systems (http://www.opserver.com/images/flexserver_arch.pdf) document search and sort (http://www.docsense.pb.com/solutions/sw.html). The document composition system and the document search and sort both have PC versions which I’ve used as well; the print queue system does not but there are zillions of PC print queues (and in particular I’ve used a SCO one pretty heavily). I’ve also run a dual quad Xeon setup (one sql server box one webserver) which serves a website getting a few thousand users a day (pure HTML/ASP nothing tricky like Java). For the pair of Xeon machines we were constantly having to reduce server load under normal usage. The other systems handled spikes of 10 times their normal loads quite well, which is to say the applications ran and produced output at the normal rates.
Similarly with OpenVMS machines running Alphas. I’ve seen a cluster of 4 alpha boxes (6 cpu each) bought around ’98 run hundreds of apps nightly many of which I wouldn’t consider running on a PC (huge databases and some of the worst Cobol code in the history of programming).
I’m not offering any evidence here because for some reason the statistical evidence seems to be totally contradictory to experience. However from your own experience you know that a 386/486/pentium systems were much faster than the 286 systems. A 286-20 didn’t system wasn’t just a little slower than a 386-25; and once the improved motherboards came out with the 386-33’s you were getting something like 5-10x the speed in practice you got from your 286-12.
> TPC-C is not all that great a benchmark, but here’s a
> pair of nonclustered results. This particular Sun
> benchmark is the only TPC-C result currently published by
> Sun (guess why?)
> Sun 4500 (14x464Mhz) SymfoWare VLM 3.0 = 67102.53
> PowerEdge 8450 (8x900Mhz) SQL Server 2K = 69901.74
> at about half the price.
I’ll agree 100% on the price. If you include cost of the servers than Sun relative to Xeon is a nobrainer. Except where there are no competitive Wintel products dollar for dollar Wintel is the better deal. My point was about 64 bit performance.
> If you do have any benchmarks showing a large performance
> gap between Xeons and the older Sun processors, or even
> newer Sun processors, please post them.
Those TCP numbers are good. What I have trouble understanding is why the benchmarks disagree with reality so strikingly. Why do you think people pay $100k for the 64 bit systems when they could get a wintel for $20k?
First off, I do not know anything about the CPU’s. I have come across an article over at the inquirer, http://www.theinquirer.net/?article=4814 . There is a letter response (the 1st letter) that caught my attention.
I was interested in what you guys thought. Does this seem more possible for Apple to go with IBM new Power4 chip and “new” Nvidia GPU and chipset?
tesmako,
You say that IBM hasn’t been active in PowerPC development lately. I disagree. There is a 1 GHz G3 processor, Apple just never used it.
I hope the vector unit in this new 64 bit PPC chip is NOT AltiVec – at least not in the way we know it now. AltiVec right now only accelerates single precision floating point ops, while its competitor SSE2 and 3DNow have been accelerating double precision floating point ops since day one.
ATM
The power4 is slower than the itanium 2.
The current p4 and AMD athlons are almost as fast as the current power4 and itanium.
When next gen p4 and athlons come out they will obviously be faster than power4 if not straight away then very soon.
The power4 for PPC will still have a high poewr consumption and be hot but obviously will not perform like the full power4 so they are going to start off behind x86.
cnet notes that the cut down power4 will not support multi cpu capabilities.
fron cnet “In addition to having Power4 DNA, the new PowerPC will use an eight-way superscalar design, meaning it will be able to issue up to eight instructions per clock cycle.”
ATM my 700 Mhz Athlon does up to 9 .. same as all athlons.
“Actually an Alpha was the first 64 bit computers. ” And i used this on a Win NT box 10 years ago. I guess windows beats apple to 64 bit then.
This sounds like the G5 to me. I read a similar article on TheRegister and it stated that IBM and Motorola worked together in designing the basis of the G5. Maybe IBM took what they learned and now they’re delivering.
>>The power4 is slower than the itanium 2.<<
Actually that is incorrect, they have cancelled each other out on different benchmark numbers, so it’s a tie at this point!
As for the Itanium (better known as the Itanic), it’s a turtle compared to the Power4!
>>”Actually an Alpha was the first 64 bit computers. ” And i used this on a Win NT box 10 years ago. I guess windows beats apple to 64 bit then.<<
And Win NT is also VMS carbon copy, which also ran on the Alpha, BIG DEAL!
Every time i post then realise i havent put a subject line i press back just like the web page says.. and all my text is gone.
Can u please fix this?
Michael had an excellent point about IBM making a desktop chip – and even more so as IBM says it’s getting out of the desktop business.
I2 is prerelease so far.. its Integer marks are speculated to increase before its officially released. If so the I2 beats the power4.
IBM announced opteron servers.. opteron beats I2 and same clock freq (and comes in higher speeds), so IBM’s fastest computers wont have IBM chips.
“As for the Itanium (better known as the Itanic), it’s a turtle compared to the Power4! ”
Yeah the first itanium sucked.. not surprissing since its a completely different design to other cpus.. the power4 is a hyped g3.. how long can much more can they pump it? It allready takes 2 super pumped G3s (and 500W) to get to the current top speed. (along with I2 and just above p4 and AMDS).
Michael didn’t have an excellent point at all. The PowerPC chip was/is/will be primarily targeted at network/communications systems. IBM’s PR people and the news reporters might call it a “desktop” chip, but 75% of all powerpc chips (both IBM + Motorola) go to network/communications systems.
Even in the CNET article, they wrote:
“Network equipment and other communications gear is the most likely destination for the new PowerPC, as the bulk of existing PowerPCs are used there. However, IBM is also wooing Apple Computer, sources familiar with the chip said.”
The Mac angle is an after-thought secondary goal for this new PowerPC chip.
I’m sure it is primarily intended for networking, but, who knows? Maybe this is why Steve Is Serene <g>.
>>Michael didn’t have an excellent point at all. The PowerPC chip was/is/will be primarily targeted at network/communications systems. IBM’s PR people and the news reporters might call it a “desktop” chip, but 75% of all powerpc chips (both IBM + Motorola) go to network/communications systems.<<
Actually if you go to both Motorola and IBM’s website, you’ll see that they specify the PowerPC to be for Portable and Desktop computing as well as what you mentioned!
http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC…
Example Applications
• Networking and telecommunications infrastructure
• High-performance computing (scientific, medical, etc.)
• Desktop and portable computing
http://www-3.ibm.com/chips/products/powerpc/consumer.html
PowerPC 7xx High Performance Microprocessors
• Desktop PCs
• Notebook PCs
• High-End Printers
Percentage doesn’t mean squat as long as the technology is geared to what it was intended!
Can you remember that Microsoft wants to support 4 different 64Bit architectures?
1. Itanium
2. Hammer
I think, we can extend the list now:
3. IBM Power4 “lite”
Now what will be numbre 4 ???
It make sense that Apple ould bethe first to push 64-bit on the dest top. They were the first with a 32-bit line of deskop PCs.
Apple wasn’t the first to push 32-bit, IIRC. Try Acorn. Besides, if they can do it before the end of this year, they could beat HP to it with Hammer-based PCs, but that seems unlikely, Jobs said he is opened to such a thing after 20% of the Mac market is using OS X.
On the side: I would rather see them NOT include the 32-bit compatability.
Great, than moving to x86 would be as easy as moving to IBM’s new stuff.
However this is nothing like patching on 64 bit stuff on an x86, the PowerPC is very clean and largely planned for 64 bits anyway (just look at its big brother POWER).
POWER isn’t a PowerPC processor. And PowerPCs was designed when 64-bit was unheard of, just like 128-bit right now.
A dual 64 bit setup (like a Sun 3500 or the E450s) outperforms a quad Xeon for most server apps.
I’m sorry, could you tell me why Itanium was invented?
The only volume POWER-type platform is Apple.
Amiga could be a possiblity.
And Win NT is also VMS carbon copy, which also ran on the Alpha, BIG DEAL!
Windows NT was designed by DEC enginners, I agree. But a VMS carbon copy? Any prove to back this?
Every time i post then realise i havent put a subject line i press back just like the web page says.. and all my text is gone.
Use Opera, it does it fine 🙂 Haha!
IBM announced opteron servers.. opteron beats I2 […]
No benchmark proved that, just some numbers AMD provided use with.
Percentage doesn’t mean squat as long as the technology is geared to what it was intended!
If majority of their customers come from one market, they would most probably target more resources towards that market. It’s business common sense.
Have a look on other CPUs to be annouced on the Microprocessor forum.
In the Embedded section:
quad-CPU system on the chip – Wow !
Halla – 1.2 GHz Alpha chip for embedded system .
IBM’s PowerPC4 is in desktop section – that would be intersting to have 8-way SMP enabled desktop.
>>>>Actually if you go to both Motorola and IBM’s website, you’ll see that they specify the PowerPC to be for Portable and Desktop computing as well as what you mentioned!
Of course they, as hardware manufacturers, are going to promote their wares as capable as possible. This pill will cure headaches, allergys and cancer. My set-top box dev. kit will run wince, qnx, vxworks, psos, lynx, linux and embedded dos.
>>>Percentage doesn’t mean squat as long as the technology is geared to what it was intended!
Percentage means that if they claim that this thing will be good on both big distributed routers and desktop pc’s, then it just means that they design this thing for the 75% of their customers (big distributed routers) and the pc function is just a tag-along.
“Can you remember that Microsoft wants to support 4 different 64Bit architectures?
1. Itanium
2. Hammer
I think, we can extend the list now:
3. IBM Power4 “lite”
Now what will be numbre 4 ???”
Well, with respect to 32 bit processors and MSFT:
Back 8 years ago or so Apple, Mot, IBM, and MSFT were planning to build “CHRP” Common hardware reference platform machines, that would run Copland (Apple’s deceased pre-acquisition of NeXT next-gen OS), Win NT, and AIX. MSFT pulled out, fearing that people who could run Mac and NT on the same box would soon become pretty disenchanted with NT. At that point, there was little incentive for IBM and AAPl to stay in the project. But good stuff came out of it. The modern Mac, unlike Macs of old, contains few non-industry standard parts. We use the same RAM, Hard disks, buses (remember NuBus?) etc.
I’m glad to see Apple may have another chip to fall back on and IBM is a top-notch hardware and processor designer. I would hope that the Power5’s or whatever they call them will be able to compete head on with the Clawhammer and Itanium, I have no doubt they should. MS has ditched Win95 and are almost getting rid of 98 & ME since NT is a better 64 bit platform, albeit ppl will need to purchase another OS. Linux will have 64 bit Wintel support in upcoming kernels.
So why shouldn’t apple move to 64 bit? If ppl still want OS9 App support than they can keep the old computers. Many Businesses and individuals still have old computers. I still see ppl with 486s and 386s because it does the job for them.Now I could be wrong here with this point, but Apple is syncing with the FreeBSD 5.0 cvs. Won’t this newer kernel also support Wintel 64 bit processors? If there’s better 64 bit proccessing in the kernel wouldn’t this allow easier ports along with all the other goodies in the 5.0 kernel?
The prices of the chips will inevitably fall with time as they get “older” (read: out for more than 2 months) and are produced in larger quantities.
>>And Win NT is also VMS carbon copy, which also ran on the Alpha, BIG DEAL!
Windows NT was designed by DEC enginners, I agree. But a VMS carbon copy? Any prove to back this?<<
http://www.win2000mag.com/Articles/Index.cfm?IssueID=97&ArticleID=4…
There you go… I have to admit that I like VMS very much (much more than UNIX in some cases), but NT doesn’t quite match VMS in terms of reliability (though NT is more reliable than the other Windows/DOS). Too bad Microsoft couldn’t have implemented VMS’s CLI (DECterm) command infrastructure, I enjoyed the VMS commands (syntax) better than UNIX, but you know that is all but a vivid memory now 🙁
NT a CARBON COPY of VMS?…lol…it was INFLUENCED by VMS…but it was also influenced by UNIX and OS/2 and DOS and Windows 3.1 and maybe even CP/M…but to say that NT is a CARBON COPY of VMS is total friggin bullshit CattBeMac and you know it!!!
So they hired some DEC engineers…yeah so what…is .NET friggin Delphi b/c MS hired Andres Hijlsburg?
-bytes256
>>…but to say that NT is a CARBON COPY of VMS is total friggin bullshit CattBeMac and you know it!!!<<
Now I know you’re full of it since OS/2 was the beginning of NT’s birth… DUH!!
Once you get your facts straight, then come back and run your mouth!!!
no actually i think you should get your damn facts straight…NT 3.1 was supposed to be OS/2 3.0…but IBM decided to use an internal rewrite instead and thus divorced themselves from MS…so guess what…I would say that NT being a fucking DERIVATIVE of OS/2 would probably mean that OS/2 had a little bit of influence upon NT…just a guess there
>POWER isn’t a PowerPC processor. And PowerPCs was designed
>when 64-bit was unheard of, just like 128-bit right now.
Wrong. The actual PowerPC ISA is included in a document called “The PowerPC Architecture: A Specification for a New Family of RISC Processors.” The PowerPC ISA is a simplified/stripped down version of the original POWER architecture, and I assure you that the designers of the ISA were completely familiar with 64 bit architectures. The base ISA contains a complete implementation of a 64 bit specification of the ISA. However, the early chips were implemetations of the 32 bit subset of the complete ISA. However, that doesn’t mean that the original specification wasn’t created with 64 bit firmly in mind.
Eugenia, sorry to tell you, but your “rah rah go go x86” crap is so off-target it’s laughable, and this is the beginning of the stuff that will show you how wrong you were.
>>no actually i think you should get your damn facts straight…NT 3.1 was supposed to be OS/2 3.0…but IBM decided to use an internal rewrite instead and thus divorced themselves from MS…so guess what…I would say that NT being a fucking DERIVATIVE of OS/2 would probably mean that OS/2 had a little bit of influence upon NT…just a guess there<<
Before you get all bent out of shape, go to the link I provided and have a read on the subject. Don’t forget OS/2 was the beginning of NT’s life, which was a spin off of VMS. The rest is history…
CattBeMac: There you go… [..]
The article you gave just told that Microsoft hired 21 men from the original VMS team. But it isn’t a carbon copy (there were many differences between the two platforms), and Win32 was very different from VMS’s APIs. In other words, the design and ideas may be copied by Microsoft, but it isn’t a carbon copy. VMS, I admit, is the biggest influence on NT, but isn’t the only influence. For example, the shell they used is similar to that in DOS.
Eugenia, sorry to tell you, but your “rah rah go go x86” crap is so off-target it’s laughable, and this is the beginning of the stuff that will show you how wrong you were.
Erik J. Barzeski: x86 remains to be a possiblity, and to my opinion, the best option for Apple. Sure, migration would be hard, but then with Prescott and Hammer, things would be way different. Plus, x86 processor evolve much more faster than PPC processors, and unless Apple opens the hardware market, it would remain that way.
CattBeMac: Before you get all bent out of shape, go to the link I provided and have a read on the subject. Don’t forget OS/2 was the beginning of NT’s life, which was a spin off of VMS. The rest is history…
bytes is correct. NT 3.0 was a complete rewrite of OS/2, influence by VMS primarily, but also by UNIX, DOS and OS/2. NT 3.0 had practically no OS/2 2.0 code, and was suppose to be OS/2 NT. OS/2 3.0 and NT 3.0 was ery different. NT 3.0 was closer to being VMS than OS/2 3.0.
I agree VMS is a very huge influence over NT, but the differences between the platforms are as apparent as BSD and System V (okay, not so apparent…). NT is certainly NOT a carbon copy. The same way VS isn’t a carbon copy of Borland even though many of Borland C++’s developers went over to Microsoft.
Plus, x86 processor evolve much more faster than PPC processors, and unless Apple opens the hardware market, it would remain that way.
yeah but evolved kludge is still kludge.
>>I agree VMS is a very huge influence over NT, but the differences between the platforms are as apparent as BSD and System V (okay, not so apparent…). NT is certainly NOT a carbon copy. The same way VS isn’t a carbon copy of Borland even though many of Borland C++’s developers went over to Microsoft<<
You take my ‘caron copy’ statement too literally, but I think you know what I am getting at. The whole point of the discussion what that NT ran on Alphas, which I think was no big deal since NT had a lot in common with VMS (which also ran on Alphas, and of course VAX). It is said that NT is VMS written in C. Your explanation of the DOS shell being used instead of the VMS (DECterm) shell isn’t a good argument, OSes (BeOS being an example where POSIX is concerned) can run various shells if implemented properly! And of courrse Shells and the low level OS itself are 2 different things in many cases!
>>I agree VMS is a very huge influence over NT, but the differences between the platforms are as apparent as BSD and System V (okay, not so apparent…). NT is certainly NOT a carbon copy. The same way VS isn’t a carbon copy of Borland even though many of Borland C++’s developers went over to Microsoft<<
You take my ‘carbon copy’ statement too literally, but I think you know what I am getting at. The whole point of the discussion what that NT ran on Alphas, which I think was no big deal since NT had a lot in common with VMS (which also ran on Alphas, and of course VAX). It is said that NT is VMS written in C. Your explanation of the DOS shell being used instead of the VMS (DECterm) shell isn’t a good argument, OSes (BeOS being an example where POSIX is concerned) can run various shells if implemented properly! And of courrse Shells and the low level OS itself are 2 different things in many cases!