Good point! I forgot about x86-64. Still, though, x86-64 is actually yet another extension to x86.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
ISSCC 2004: IBM on PowerPC 970FX
- Thread starter MacRumors
- Start date
- Sort by reaction score
Good point! I forgot about x86-64. Still, though, x86-64 is actually yet another extension to x86.
The other mode does have the x86 baggage though (64-bit OS + 32 bit application) so again x86 style code. For the full benefit you have to have 64-bit OS + 64-bit application AND 64-bit drivers then and only then can you avoid x86 and it all runs nicely and quite fast I should say.
The only problem, then, is getting x86-64 deployed widely. So far, AMD is the only chip maker (that I know of) making x86-64 processors. I believe Intel is going to join in the future, though. That step will speed along deployment of x86-64. However, Microsoft still needs to make a version of Windows that fully supports and uses x86-64 for it to be worthwhile to deploy.
Intel is going to strap on 64-bit entension to the existing pentium 4 hence all the flaws with the pentium 4 are carried along with it, this is NOT what AMD did the Opteron is not an Athlon XP with some extensions. Intel is taking the quick way out, they were relying on Itanium to provide the 64-bit platform. IMO the current prescot in fact has these extensions allready but disabled, remember the 125M transister count? so thats an extra 70M for 512KB of cache? those extra transistors are the extensions!!
army_guy, your post just proves to me how short-sighted Intel can be. Intel clearly got their plans wrong; now they'll pay a high price by losing marketshare to AMD.
Forget about Intel now, there out of the game in both 32-bit and 64-bit aka the Itanium. If only the Itanium was this good 5 years ago then I people would consider it but too little too late I say. AMD is gaining ground in the server segment especially with the 800 series have just been priced slashed and compared to Itanium well you get the picture. The rumur of QUAD Opteron workstations is floating about so Intel has lost in that market as well. As for the gaming market well theres the FX to cover that and performs way better than the P4. And with all the major OEMS except DELL of coarse using Opterons well wheres INTEL. I should also mention CRAY they have a cluster on the way (2-4 WAY node based) custom built with CRAY interconnects and then theres REDSTORM.
technically you cant have full 64 bit os right now
Technicaly you cant have a fully compiled 64 bit os right now. THere are NO compilers on the market that can to that in this day and age, although some come close. It seems like when ibm released their g5 compiler alot of people said the same thing, there is no full 64 bit compiler yet.
Technicaly you cant have a fully compiled 64 bit os right now. THere are NO compilers on the market that can to that in this day and age, although some come close. It seems like when ibm released their g5 compiler alot of people said the same thing, there is no full 64 bit compiler yet.
Re: technically you cant have full 64 bit os right now
Just curious - what does a compiler have to have to be fully 64-bit? I don't see any reason why a 64-bit program couldn't be compiled on a 32-bit machine, then moved to a 64-bit machine for testing (or if you're lucky enough to have one of those machines that can run 32-bit and 64-bit code natively, like the G5, compiling the program/OS in 32-bit mode and testing it in 64-bit mode).
Just curious - what does a compiler have to have to be fully 64-bit? I don't see any reason why a 64-bit program couldn't be compiled on a 32-bit machine, then moved to a 64-bit machine for testing (or if you're lucky enough to have one of those machines that can run 32-bit and 64-bit code natively, like the G5, compiling the program/OS in 32-bit mode and testing it in 64-bit mode).
Re: Re: technically you cant have full 64 bit os right now
I never said 64 bit compiler, i just said no compiler can compile FULL 64 bit code yet, this is an impossibility. And plus I know someone on the darwin team so I know wut i am talking about
. He said within half a year to a year there should be full compilers that can do full 64 bit code, he also said within 2-3 years when apple gets their whole line 64 bit they are going to release a 32 bit os x and a 64 bit os x.
I never said 64 bit compiler, i just said no compiler can compile FULL 64 bit code yet, this is an impossibility. And plus I know someone on the darwin team so I know wut i am talking about
. He said within half a year to a year there should be full compilers that can do full 64 bit code, he also said within 2-3 years when apple gets their whole line 64 bit they are going to release a 32 bit os x and a 64 bit os x.
Re: Re: Re: technically you cant have full 64 bit os right now
Okay, perhaps I should rephrase my question. What in the heck is full 64-bit code? I thought there was 32-bit code and 64-bit code and that was that (no code of mixed bitness). Are you referring to code that is part 32-bit and part 64-bit?
Originally posted by NusuniAdmin
I never said 64 bit compiler, i just said no compiler can compile FULL 64 bit code yet, this is an impossibility. And plus I know someone on the darwin team so I know wut i am talking about
Okay, perhaps I should rephrase my question. What in the heck is full 64-bit code? I thought there was 32-bit code and 64-bit code and that was that (no code of mixed bitness). Are you referring to code that is part 32-bit and part 64-bit?
Re: Re: Re: Re: technically you cant have full 64 bit os right now
Normally, you have a choice at compile time. Say you want 64-bit integer math, but you don't care about addressing memory past 2GB, you can set up your compile to use 64-bit instructions, but keep 32-bit pointers. You can also do it the other way around, with 64-bit memory pointers, but 32-bit instructions. At least that's the way I've seen it implemented in the past; I haven't played around with the G5 compiler modes, yet. It's difficult to do without a G5
Re: Re: Re: Re: Re: technically you cant have full 64 bit os right now
Okay, that makes sense now. What, then, is holding up compilers that can make pure 64-bit code?
Originally posted by daveL
Normally, you have a choice at compile time. Say you want 64-bit integer math, but you don't care about addressing memory past 2GB, you can set up your compile to use 64-bit instructions, but keep 32-bit pointers. You can also do it the other way around, with 64-bit memory pointers, but 32-bit instructions. At least that's the way I've seen it implemented in the past; I haven't played around with the G5 compiler modes, yet. It's difficult to do without a G5
Okay, that makes sense now. What, then, is holding up compilers that can make pure 64-bit code?
Re: Re: Re: Re: Re: Re: technically you cant have full 64 bit os right now
I've read that freeBSD, upon which Darwin/OSX is based, is 64-bit clean.
I guess I'll have to save up for a G5 and try it.
I don't know. I know you can produce 64-bit code for Sun's ultraSPARC processors, etc with gcc. I'm sure IBM's Power4 compiler can do the same. I haven't investigated whether IBM's xlc compiler for the 970 will gen 64-bit code.
I've read that freeBSD, upon which Darwin/OSX is based, is 64-bit clean.
I guess I'll have to save up for a G5 and try it.
All i know is that Itanium compiles FULL 64-bit code, I think you need to hardware compile Itanium code meaning that its not done at the software level like gcc but using a dedicated mahine thats another reason why porting to Itanium is expensive and produces efficient code. Has anyone compiled an Itanium piece of software?
ya
I can't remember exactly why it cant be compiled for 64 bit right now, but it has something to do with the compilers, i know they would have to rewire the boot strap for 64 bit ppc asm.
I can't remember exactly why it cant be compiled for 64 bit right now, but it has something to do with the compilers, i know they would have to rewire the boot strap for 64 bit ppc asm.
Delays on the PM G5s etc from IBM?
I think that the reason we haven't seen the updated PM G5s (and perhaps PB G5s) is IBM.
Take note of the following facts:
1. Apple's xserve G5s are shipping in "6-8 weeks".
2. InfoWorld quotes an IBM spokesman, Feb 13, 2004 (http://www.infoworld.com/article/04/02/13/HNibmblends_1.html) that says "Samples of the chip have already shipped, and production volumes are increasing." Note, they are producing, but if they were producing at full capacity they'd have said so and not used the language they did.
3. There was an article here (http://www.reed-electronics.com/electronicnews/index.asp?layout=siteInfo&doc_id=30982) that said this:
"IBM is about to deliver its very first volume production of an SOC on 90nm, said Reeves. This will be a design for Apple."
Note that it says "about to".
So, I think it is safe to say they are producing them in some volume, but it is not yet enough to handle xserve G5s, PM G5s (updated with the 970fx of course), and PB G5s all at once.
Thus, I would expect Apple to catch up on the xserve G5s (or at least know *when* they will be able to do so) and then announce upgraded PM G5s at that point to ship as soon as they can after xserve G5 orders.
Until they can handle demand for those two, I doubt you'll see PB G5s. There may be other factors delaying the PB G5s, but 970fx chip supplies may be one of them.
I think that the reason we haven't seen the updated PM G5s (and perhaps PB G5s) is IBM.
Take note of the following facts:
1. Apple's xserve G5s are shipping in "6-8 weeks".
2. InfoWorld quotes an IBM spokesman, Feb 13, 2004 (http://www.infoworld.com/article/04/02/13/HNibmblends_1.html) that says "Samples of the chip have already shipped, and production volumes are increasing." Note, they are producing, but if they were producing at full capacity they'd have said so and not used the language they did.
3. There was an article here (http://www.reed-electronics.com/electronicnews/index.asp?layout=siteInfo&doc_id=30982) that said this:
"IBM is about to deliver its very first volume production of an SOC on 90nm, said Reeves. This will be a design for Apple."
Note that it says "about to".
So, I think it is safe to say they are producing them in some volume, but it is not yet enough to handle xserve G5s, PM G5s (updated with the 970fx of course), and PB G5s all at once.
Thus, I would expect Apple to catch up on the xserve G5s (or at least know *when* they will be able to do so) and then announce upgraded PM G5s at that point to ship as soon as they can after xserve G5 orders.
Until they can handle demand for those two, I doubt you'll see PB G5s. There may be other factors delaying the PB G5s, but 970fx chip supplies may be one of them.
Everyone here seems to be worried about the speed of a new G5. I don't see what everyone here is complaining about. The G5 is the fastest out there right now, so if apple is smart they will wait awhile for the competition to catch up before releasing another. You have to think in the marketing and sales aspect here. They just released the G5, people are not going to want to upgrade already, and if product updates come out to quick, no one will ever buy. It's a lot nicer to the consumer if when he or she buys the product, he or she has the top of the line for at least 4-6 months. In addition, for the powerbooks, I don't think those will come out any time soon. If anyone here has a G4 they know how hot that runs. The G5 would melt the powerbook. The wattage is still to high. 50 watts would drain the battery to quick. As far as price drops, I wouldn't count on any of that. Apple is known for extreme prices. Look at the new ipod, it's the same price as a 15 gig and a half inch smaller. Who ever buys that is stupid. They have the same battery life, yet one holds a bit more songs than the other. I don't know what Apple was thinking with that. My predictions would be a new G5 by next december at the latest and a new powerbook in a year.
Did you even read any of the posts in this thread? The 970fx that Apple will probably use in the PowerBook G5s does not dissipate 50 watts. That's for a 2.5 GHz running at maximum. The PowerBook G5 will be in the 1.4 GHz - 2.0 GHz range and will use much less maximum power at that speed (and even less under typical use).
And you should also go read the iPod mini thread (https://forums.macrumors.com/showthread.php?s=&threadid=60828) for many reasons why the mini is not such a stupid idea after all. You don't remember how many people thought the original iPod was the dumbest idea and was doomed to fail either, do you?
IBM is surely cranking out the wafers. I am excited for new Powermacs so i can get a g5 for a little less $ !!!
Really! I could not agree more. Yes, sir, we only have 100,000 in pre-orders. I think we should cut our losses; this thing's a real loser. Of course, we don't want those PBs running too (note: too) fast, either. We should wait for the competition to get a leg up first. What a crock.
an engineering class might be warranted if folx thought apple was going to come out with a 50w processor in a new PB anytime soon.
What people were saying (and what you were apparently oblivious to) is that the new 970fx is going to have slower chips that dissapate in the neighborhood of 12-15w. This has been mentioned several times in this thread. What, you can't be bothered to read what people are saying before you respond??
You don't have to put the fastest/hottest chip in the PB just 'cuz it exists. Instead, you might wanna put in the chip that is still blazing fasts yet works with the objective limitations of the device you're creating.
That's basic engineering. Perhaps you should take a class??
That was what I was trying to get to. I have read the posts here and it seems that more people are interested in the faster processor not a slower one for the PB. Besides that, I can bet most people think the proc speed is the biggest difference, when in fact it is not. If apple was to slap a 1meg+ cache onto a 1ghz G5, it would run close to a 2ghz with a 512 cache. So yes, I have read the books and understand how these things work. A good example of this would be the intel centrino package...the processor runs at a much slower speed than the p4, but because it has a 1meg cache, it runs very close to it but uses much less power.