Slashdot Mirror
PPC 970 Confirmed for Apple?
batboy78 writes "In what perhaps is the first 'official' confirmation that IBM's PowerPC 970's will be used by Apple, BusinessWeek claims that IBM has confirmed that it's developing a new set of chips for the Mac: 'IBM says the new Apple chip will be of the 64-bit variety, which means it can process twice as much information per cycle as existing 32-bit chips.'" CT The article has been updated to make the confirmation seem... well, far less comfirming.
Ditto...64 bit means it can address 2^32 times the memory as currently and/or do 64-bit numerical functions in one operation.
The AltiVec engine (which is also available in the 970 though I'm not sure if IBM can call it that or not) is like MMX/SSE/SSE2 where it can process chucks or multiple chunks of data that can be larger than 64-bit (AltiVec and SSE can handle 128-bit, not sure about SSE2).
32-bit and 64-bitness of a processor generally means how much memory it can address and the size of the registers.
The G4 is a 32-bit CPU with a 128-bit vector processing unit (aka SIMD Unit) called Velocity Engine or Altivec. This kind of like (though much superior) to MMX, 3DNow!, etc. The new IBM chip is suppose to be a 64-bit chip with a 128-bit Altivec compatible unit. In the past, the Altivec unit has always suffered from Motorola's slow FSB on the G4. One bonus of the PPC 970 is that it sports a 900MHz DDR bus that can keep the SIMD unit fully fed.
Ok.. I wish people would get this through thier heads. A 64 bit chip is _NOT_ inherently faster than a 32 bit chip. It is able to address more memory space and perform greater precision calculations faster. If you are just working with lots and lots of 32 bit numbers you will see some speed improvement but not close to double. Once you are into the realm of 33 bit and higher numbers which are done with mathematical trickery on 32bit processors, you will see a huge speed increase when working with a 64 bit processor.
Also, the increased memory ceiling helps.
*note: yes, I know this is not technically correct, but I'm not explaining how 32bit and 64bit processors handle thier operations. Maybe someone can reply with that.
Karma: SELECT `karma` FROM `users` WHERE `userid`=138474;
While technically true, saying "64-bit can process 2x faster than 32-bit" is misleading.
64-bit means that each instruction can be 64 bits long, allowing for the native computation of larger numbers. Concurrently, it can process 2 32-bit instructions, but they would have to be instructions such that neither relies on the result of the other.
To sum up: 64-bit is not equal to 2x32-bit, but is much better than 32-bit.
An online Starcraft RPG? Only at
In soviet russia, all your us are belong to base!
Karma: Redundant
Online Starcraft RPG? At
Dietary fiber is like asynchronous IO-- Non-blocking!
They do, but everyone knew about this before today--well, everyone except you, so i guess you have a point. But PowerMac sales were already abysmal anyway.
Oh, and if you want an iMac or a PowerBook, odds are against the new chips premiering in those Macs, so you may have a longer wait than you expect on your hands.
You must have been asleep for most of the 90s when the PPC was kicking x86 butt.
Here is what the article actually says:
"Although Apple (AAPL ) won't talk about it, IBM (IBM ) is developing a new set of chips that Apple will likely use to replace the aging Motorola processors used in its G4 line."
TO REPEAT: "...CHIPS THAT APPLE WILL LIKELY USE..."
In other words, THIS IS JUST MORE FACT-FREE SPECULATION.
It's easy to make up & spread cool- and credible-sounding stuff. Finding & checking hard facts is hard work.
Altivec or Velocity Engine was developed by Apple, IBM and Motorola together (AIM), so Velocity Engine and Altivec are the same thing. The name Altivec is owned by Motorola, but the actual 128-bit vector processing unit is owned by AIM, so IBM can use it in their processors, they just can't call it Altivec.
You might want to check out the MacRumors Buyer's Guide to help you decide when to purchase which Mac model.
For example, right now they recommend purchasing a LCD, XServe, iBook, iPod, or eMac. They're neutral on iMacs, Powerbooks, and Power Macs.
There's no way you'll see a PowerPC 970 in a 12" Powerbook, so don't wait if you want one of those. The iMac is tricky...my guess is that it'll see faster G4's for a while before it eventually gets a processor upgrade. I'd only wait for sure if you want a Power Mac.
Actually, this is wrong. A 64-bit process cannot necessarily process 2-32-bit instructions at once. The number of parallel instructions a processor can process is entirely dependent on the number of pipelines it has. A 64-bit processor with 3 integer pipelines can process 3 32-bit or 3 64-bit integer operations per cycle (in theory) while a 32-bit processor with 3 integer pipelines can still process 3 32-bit integer operations per cycle.
A deep unwavering belief is a sure sign you're missing something...
Link
benefits of 64-bit computing
increased scalability
The main benefit of 64-bit computing is increased scalability of your computer and applications. Some applications simply do not fit into a 32-bit computing model. For example, limitations on file size in a 32-bit environment may require database systems to use multiple files to represent a single file. Applications requiring large files, a large number of files, or a large number of users will benefit from 64-bit computing.
increased performance
Any application that is outgrowing a 32-bit computing environment will suffer performance hits. Applications may need large files, large memory, high precision arithmetic, and/or algorithmic accommodations for 32-bit limitations. Applications needing more code or data in memory will benefit from decreased swapping with 64-bit computing. Reduced swapping can make database inquiries as much as 100 times faster (individual performance gains may vary).
Also any way you stack it a 32-bit instruction or a 64-bit instruction is still just one operation.
The G4 really was a supercomputer at launch... but only by the letter of the law. The G4, capable of over 1GFlop, came in north of the Federal definition of a Supercomputer (in relation to the export of arms). So.. you couldn't ship the Macs to any 'enemy' country like Libya... or even to France... at least not right after they were released.
;-) A dual 2.5 GHz machine would be capable of up to 40GFlops (max theoretical) by Apple's calculations. ;-) hehe.
The US Govt. quickly revised the rules. I believe supercomputers are just north of 50GFlops now.... so Apple could get real close again with an SMP 970... if you go by Altivec performance again.
I'm not feeling witty so bite me
I was more intrigued by the "1.8 GHZ per second" claim.
1.8 Billion instructions per second per second. It's about time that somebody made an accelerating chip - way to go, IBM!
If you'll be picky about other people's stuff, you might as well proof-read your own posts.
GHz != Billion instructions per second. GHz is the frequency the clock runs at, and that's all. Depending on the architecture, a single instruction may take several clock cycles. IIRC, Motorola 68HC12 has a few 7 cycle instructions.
-bm
Since when did a BusinessWeek article become official confirmation? Probably, BusinessWeek got their information from rumor posts on MacRumors.com. As well as calling it "official" instead of just official, MacRumors also adds:
No specific executives are quoted, however... so it's unclear from where the information originated.
The PowerPC 970 has been widely rumored and expected to be used in Apple's upcoming Macs, though both IBM and Apple had not made any official announcements about their use.
The PPC 970 only dissipates 19W at 1.2 GHz. The 7455 (the G4 that goes into the current PowerBooks) dissipates 22W at 1 GHz. Those numbers are at full capacity and I got them from official spec sheets.
Don't say that these use more power or produce more heat without the facts to back up your position.
The thing that slows computation down is mostly access to memory. The faster you can yank bits out of memory and slap them back out there, the faster you will compute. CPUs nowadays are highly optimized to make sure that every memory cycle does something useful.
Generally a CPU can compute faster than it can fetch or store, because on-chip memory is faster than off-chip memory. Tricks like caches help to speed things up. Tricks like having wider registers can also help quite a bit, depending on what you're doing.
If you are doing a lot of integer math on 32-bit integers, 64-bit registers aren't going to make any difference. If you are bitblitting images, they can make a difference. If you are doing double-precision floating point operations, they can make a big difference.
You can get similar performance wins by having a wide memory bus, long pipelines and a high clock rate, but the problem with long pipelines is that unless your code is amenable to long pipelines, you wind up doing a lot of pipeline stalls, and all the memory cycles you spend loading the pipeline are wasted, and you don't get much benefit from your faster clock rate. This is a big problem on the Pentium IV, which has a really long pipeline, and is one reason why P-IV performance has been disappointing for a lot of geeks looking for general-purpose performance. P-IV does well with video because video compression and decompression algorithms work nicely with long pipelines.
The bottom line is that there is no one thing that can double your performance, and certainly going from 32-bit registers to 64-bit registers can't double performance in all cases, but it can make a significant difference in some cases. If those cases are cases that Apple's customers care about, then Apple wins.
I know your trying to knock down the 64bit myths.. But in reality, 64 bit math essentially means you are processing 2x as much information per instruction..
While you are right, if you are only processing 32 bit data, then yes, there is only one benefit, and that is 64 bit memory addressing. But if you are processing 64 bit data, then yes, the article is both technically correct, and just correct in general.
Two infinite things: your stupidity and mine. But I'm not sure about the latter. If my sig offends you, I'm sorry.
BTW, the "trouser filling" ibook screen spanning hack does, in fact, work. And it works very well. The only down sides are that I have a low VRAM model ibook (16Meg) and screen spanning appears to disable quartz extreme, even with the 8MB hack thing applied as well; that I can't drive 1600x1200 at anything more than 60Hz, although I do get 1280x1024x75Hz so that's fine; and that when I connect the monitor in the morning it sometimes seems to forget that I like to run at 1280x1024 with the menu bar on the CRT. I think this may be caused by hooking in the monitor while it's powered off, but I've yet to be sufficiently concerned to get all scientific about it.
... personally ... I don't really see the point in the 12" Powerbook if you have the will to apply the hack to an ibook and save yourself a wheel barrow full of money.
All new ibooks have twice as much video memory and probably wouldn't suffer from the QE disabling, so
BTW, I'm typing this screen spanned onto a Sony 19" that has a USB hub in the base. I have a cheap as chips USB keyboard (windows key maps to apple/command, alt to option and ctrl to ctrl) and a standard PC optical mouse hooked into that and when I arrive in the morning I just hook in the usb and monitor connectors and we're away. The ibook's touchpad and keyboard remain active - the only thing that isn't completely duplicated is that I still only have one mouse pointer. It's really *really* cool, and I'm very impressed.
Dave
I write a blog now, you should be afraid.
The MacRumors article has been updated with the following:
Update: The article has been edited with the following correction/retraction:
"IBM did not confirm it was building a chip specifically for Apple, but it does say its new PowerPC
chip will work on Apple platforms"
Not exactly a "Never Mind" but the air of mystery continues.
If you're not living on the edge, you're taking up too much space.
The 'desktop' variant of the chip dissipates 19W at 1.2GHz, which is not out of the realms of possibility for a laptop (many Intel laptop chips dissipate more than this) and the 1.2V version is expected to make do with only 13W. It is entirely possible that the PowerBooks will move to this chip quickly, and the iBook will start to use G4s.
I am TheRaven on Soylent News
Tobin, you're on crack. The 970 at 1.2Ghz is only 19W, and the lower voltage version (~1.2V) will be 13W.
You missed my point, although not as badly as the other guy.
(I have to point out that I was speaking only of laptops -- something that is only clear if you read the parent of my post.)
I didn't say anything to imply that any PC didn't do what it's made to do. But certainly, in the design of these machines (again, speaking of laptops), trade-offs are made. Apple seems to have a good formula worked out that has broad appeal. It's not a Mac vs. PC platform debate -- it's just design mojo that makes users happy. (Indeed, it is part of the functional aesthetics.)
Don't become a regular here -- you will become retarded.
Parent may seem like a joke, but as a long-time Quark customer, I can assure anyone who asks that this is sadly very true.
I've never witnessed another software company that has so much contempt for their customers.
I'd addend that statement by adding:
A: Never. And we don't care if you don't like it, because you've got no choice.
Quark's had a very myopic view of their stranglehold on the Mac publishing platform. They're banking that old-school art directors who have learned Xpress won't jump ship because of the learning curve of a new program.
That's why Quark has invested ridiculous amounts of time and effort (although their coding is outsourced to India now) shoehorning (extremely, horribly sucky) web publishing features into Xpress. Those same art directors would gladly stay in the same environment they're comfortable with publishing for the Web, when they should have been concentrating on PDF workflow, or a print box that doesn't take 14 clicks to print every single time you go to a printer that doesn't have a Quark native-PPD.
With Quark, you pay full upgrade prices for bugfixes, and a complete and utter lack of new features. As a bonus, you usually get printing bugs so bad that your service bureau will refuse to accept the new version's files until the second or third maintenance patch.
As a consultant for several small-to-mid sized ad agencies, I can say that as of right now:
- We're just beginning to transition the smaller shops over to OS X because of Quark's holdout
- The art directors are being actively encouraged to use Indesign *or* Quark for their day-to-day work
- Those art directors are seeing that Indesign is *much* more elegant, and Adobe is actively listening to their customers. Plus it's got a consistent interface, and it's stable. And...
- It's getting increasingly about PDF workflow from comp to film (or even placement, in the case of many publications).
As of Indesign 3 (and hopefully feature parity with Xpress and a good Xpress 5 translator) I'll recommend they drop Quark all together, except for working with legacy files. Granted, that's a year or two out. But I very seriously doubt we'll be up to even Xpress 6.0.1 by then.
Call me a bastard, but after the treatment I've received by Quark for the last 11 years as a multiple-site-license customer, I'll be very happy to see them go under. There's another game in town now, and the chains are eroding.
Quark still wins out today for useability, but I'm confident as of the next rev of Indesign that the game will be over. And good for Adobe.
Quark's made their bed, and I have a feeling that they're so blind that they won't even notice going out of business.