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.
Apple to Adopt Alternative Chipsets for Next Laptops?
- Thread starter MacRumors
- Start date
- Sort by reaction score
If locking down OS support is even a goal of this rumored alternative Apple chipset, I don't see it preventing the use of Windows on Macs, but rather preventing the use of OS X on anything but Macs.
Montevina uses Socket P (478). Mobile Nehalem uses mPGA 989.
That first iteration is Bloomfield (high-end desktop) and Gainestown (DP server) only. Mobile Nehalem's first iteration is Q3 2009.
The G5 perform pretty well considering its at 90nm, slower RAM, older GPU in these tests.
http://www.anandtech.com/mac/showdoc.aspx?i=2816&p=1
http://www.anandtech.com/mac/showdoc.aspx?i=2816&p=1
An updated PowerPC from IBM would trounce Intel. Not to mention Intel is NOW moving away from FSB and adopting an Interconnect. But again, no reason Apple can't have both Intel and Power on the same Logicboard.
Actually, why not?
The PowerPC processor was designed by a consortium made up of IBM, Motorola, and Apple. IBM stopped producing the PPC chips in about '02, leaving Motorola as the sole remaining manufacturer of the chips. Motorola chose to no continue R&D on the PPC chips after touching the 2Mhz 'barrier.' Why?(Third Base!) I don't know (Shortstop!
Now that they're on the x86 platform, it was only a matter of time before somebody would try to 'clone' the Mac onto a non-Apple motherboard. As long as it was by hobbyists for their personal use, they didn't complain. On the other hand, Psystar make it quite obvious that there were those willing to do it for money.
Strangely, at just about the time of the Psystar launch, Apple buys a company known for making its own version of the PPC chip. Coincidence? Could they be planning to go back to the PPC but use an in-house manufactured chip? Or are they instead using this company to build some sort of security chip designed not only to prevent unauthorized 'cloning' but also for better hardware security? Maybe even something designed to block the sorts of malware now trying to make itself felt in the Macintosh communicty?
Now we hear of Apple perhaps adopting a new, non-Intel chipset. The coincidences build. And what of AMD? It wasn't that long ago that AMD produced processors as fast or faster than Intel's chips of the time and proved multiple times that their processors were more accurate in floating-point processes. AMD developed a reputation as the processor to use. In return, Intel began flooding the market with lower-cost chips; going so far, allegedly, as to actually sell specific chips at near zero price point in order to influence the OEM market.
What is Apple doing and why are they doing it? It looks to me like there is no one correct answer.
PPC is alive and well. Heard of Xbox 360, perhaps Wii or maybe even PS3? Xenon, Broadway, and Cell are all variants of the PPC architecture, all designed, or co-designed by IBM, and definitely still in manufacturing.
No it isn't.
IBM already has a chip better than X86. It is called Cell.
No it isn't. Nehalem is not going to be pin nor electrically compatible with the current socket 478. As a matter of a fact I am pretty sure that Nehalem is going to have more than 478 pins (mobile variant, desktop is 775 pins). Why? Cause they have to integrate the memory controller. That will add pins all by itself.
You are completely wrong.
Socket M was once accepted because Intel transitioned from Core -> Core 2. However, Penryn itself (the codename for 45nm mobile CPUs based on the Conroe architecture) uses Socket P, different from M.
Furthermore, Nehalem uses mPGA989 which requires totally different pins. Nehalem feature a completely new architecture with features incompatible with previous architectures, meaning that Montevina and currently existing sockets ARE NOT compatible!
For desktops, its no different as Nehalem is moving Intel CPU's from LGA775 (existing Core 2 Extreme, Quad, and Duos) to LGA1366 and LGA1160. The #'s indicate pins FYI so its clear they have a lot more going on as Nehalem is a tremendous step forward from the pre-existing Intel architectures as they are moving away from FSB to Quick-Path Interconnect as well as supprot for IMC, hyperthreading, etc.
OK, sorry. My mistake. I didn't realize that Nehalem was using a different set of pins and required a different socket.
Thanks.
What? The G5 loses in just about every benchmark. Its slower clock for clock. And that's considering that those were the initial Core 2's. We're already on Wolfdale/Yorkfield/Penryn which are 45nm, run cooler, and clock for clock are 8-10% faster than equivalent Conroe/Kentsfield/Merom. Furthermore, they clock much higher than those PPC CPU's - their overclocking headroom is ridiculous.
And so what if Intel is just moving to QPI and using the Interconnect. AMD had it with K8 a while ago and it stomped Netburst but then Intel came back with a P3 / Pen M derivative in the Core 2 seriers and it's thoroughly stomping AMD. They might've been behind technologically in terms of features, but what matters in the end is real world performance, and Intel delivered even with fsb.
And Anandtech has already shown that Nehalem is 10-30% faster clock for clock than current Yorkfields which is scary considering they'll be on 32nm by the end of 2009.
x86 is just an instruction set - Cell itself is a procesor.
And x86 is still in charge despite being old - the x86 is far less specialized than Cell and thus it might not win in THEORETICAL FLOPS (oh boy, PS3 puts out 2Tflops, but in actual practice, its inferior to CPU throughput, much less GPU output) but FLOPS are not created equal. Furthermore, the x86 allows a wider set of instructions, which means a lot more stuff can be run on it.
Hence devs for PS3 are annoyed at how complicated coding for it is since all they care about is peak computational throughput, and not ease of coding. And if the software writers don't want to support the hardware, the hardware won't fly.
Lets not even forget that the Cell architecture runs programs without a focus on latency because of how its instruction set runs in line rather than a mroe distributed focus like x86. Thus while Cell can run through programs specialized for it fast, it can't run a whole lot at the same time.
There are TONS of talks about this but the fact of the matter is, Cell has gaudy marketing numbers, but in actual practice, x86 still rules.
The thing with Cell is code has to be written specifically for it mostly due to the SPU's and also due to it not being an OOO CPU. OOO optimized code runs poorly on Cell. When you optimize for Cell, it's performance starts to shine. If CPU's were so much faster than Cell you wouldn't have seen Stanford make a Folding client for the PS3 (which really does have a higher FLOP rating than most modern CPU's).
The coding problem is the same thing facing multi-core development. The only main difference is all the cores are the same so it doesn't matter where your code runs and they can run code out of order. Apple is supposed to have solved all of this with Grand Central though. So I see no real reason to not use Cell (or its derivative) in some form.
Of course Grand Central still just sounds like a scheduler to me but we will see.
Not quite. Nethalem uses a completely different interconnect technology in quickpath.
I never said the OLD PPC chips were faster then current Intel chips. I'm just saying that for old technology they are still competitive. Intel should be blowing these chips out of the water but they're not.
If IBM pumped out 32nm g5's with the same ram, hdd, and gpu spec'd system Intel would be crying. Notice I said G5. Just a shrink to 45nm or 32nm. Would be enough.
Cell is an inorder cpu not out of order different from your general CPU. I see it as an add on.
Compared to the C2D? Not really. My MacBook can probably beat the best G5 iMacs (in terms of CPU). And it's a laptop. You never had a chance in hell of seeing a G5 laptop. And Nehalem is 30-50% faster than what we currently have.
Compared to the G4, which was what was in Apple laptops before Intel, the move to x86 and Core arch brought MASSIVE performance improvements. How well do you think a 1.4Ghz G4 would compare against my dual core, 2.0Ghz C2D?
Apple sell more laptops than anything, laptop chips are what matters and Intel laptop chips blow the old PPC chips out of the water. Nobody could argue that Apple's PPC based laptops are "competitive" with their current offerings.
C2D is 65nm
G5 is 90nm
At the same die size the PowerPC chip is much more efficient because the ISA is cleaner and it does not have the overhead found in the X86 ISA.
Nehalem is another die shrink + an interconnect (quickpath) think hyper-transport so yes it will see performance gains.
As for the G4 vs C2D. G4 is a single core. A more comparable match would be something like the 2ghz PWRficient PowerPC chip since its dual core vs the C2D.
A die shrink would help the G5 attain higher clock speed for a given thermal envelope. But I don't know if it would ever be useful in a notebook at comparable speeds. I mean we are hitting 2.6Ghz in notebooks now. That is close to the fastest G5. But in a notebook.
Although I will admit that Intel does have the advantage with die sizes. Not many fabs can crank out 45NM dies like Intel can (note Penryn is 45NM).
No they won't try it either. It's not a matter of which OS is better. Clearly OS X is. But there are TONS of business apps that companies use that are Win only and will always be win only. If people can dual boot or use virtualization then they can better justify buying a Mac otherwise they will just suffer with their Win machine.
Architecture means more than just shrinking die.
And anyways, Intel fab tech is far ahead with 45nm high K already introduced to market, and was tested well over a year and a half ago. They already have 32nm Nehalem prototypes supposedly and will be at 22nm in 2011 or so. They're on a roll.
It's always an IF IBM can do it - and the fact is, they're already quite behind. Look at AMD, Intel's closest rival. K10 is about a half gen behind performance wise, and architecturally, and their fabs are another half gen behind as only now they're about to introduce 45nm, and they're arleady strugglign massively.
As time goes on, the gap only gets larger between Intel and the rest.
And anyways, Intel fab tech is far ahead with 45nm high K already introduced to market, and was tested well over a year and a half ago. They already have 32nm Nehalem prototypes supposedly and will be at 22nm in 2011 or so. They're on a roll.
It's always an IF IBM can do it - and the fact is, they're already quite behind. Look at AMD, Intel's closest rival. K10 is about a half gen behind performance wise, and architecturally, and their fabs are another half gen behind as only now they're about to introduce 45nm, and they're arleady strugglign massively.
As time goes on, the gap only gets larger between Intel and the rest.
Exactly, and that's why they won't drop the support, not so soon at least, unless they want an intense backlash. So unless Steve is on the verge of comitting corporate suicide, it's not going to happen.
So it sounds like you guys are admitting that Apple needs Windows to be successful in the market. And I think that is sad. OS X should stand on its own. After all it is better.
Why would Apple do something like this? Why would Apple want to waste a large chunk of money designing and manufacturing their own chipset when the Intel chipset is already designed, certified, etc... I hope Apple doesn't go with VIA or (shudder) an NVIDIA nForce chipset--these chipsets always have horrible drivers laden with severe bugs (in addition to the frequent hardware bugs). If the "mature" Windows chipset drivers from VIA and NVIDIA suck so hard, imagine what prototype beta drivers will do to Mac OS. Also, why is Apple purposefully trying to piss off Intel like this? It makes no sense to go with an Intel chip and then try to bolt it on to an AMD/VIA/NVIDIA chipset, particularly for a laptop.
And I think we can let the "Apple is switching back to PPC" rumor die; this would be the worst possible move that Apple could make and would destroy all the headway that they've made in the last 3 years.
And I think we can let the "Apple is switching back to PPC" rumor die; this would be the worst possible move that Apple could make and would destroy all the headway that they've made in the last 3 years.
As much as I like the PPC ISA... this really isn't an accurate statement for current generations of Intel cores. It would take more then a process shrink to bring a PowerPC core into the power/performance ratio of current Intel processors. The PowerPC ISA isn't a magic bullet for power/performance.
Mainstream desktop and mobile Nehalems will not use QuickPath.
As the Athlon 64 would tell you.
Apple must think the performance/power/features improvement of their own chipset outweighs the cost in money.
Intel's chipsets are pretty so-so actually. Awful integrated graphics. Poor substandard wireless (I actually just think this story is about the wireless chipset provider, as Apple has historically chosen to avoid Intel's embarrassment of an offering with Centrino) and everything else is so-so.
VIA do suck of course, and stopped offering anything about 4 years ago, they are a non-runner.
AMD could offer Intel chipsets via their ATI part which has done Intel chipsets in the past. This would have excellent graphics (for an integrated part) but wireless and ethernet would have to be sourced separately.
NVIDIA also offer reasonable chipsets (good graphics, good ethernet, good performance, a bit power hungry however, and recent issues aren't encouraging).
I don't see Apple upsetting the Apple cart that much. Intel CPU, Intel Chipset, non-Intel wireless, discrete graphics option on *all* laptops, if not standard.
Actually Intel have had a few well publicised shortages of chipsets in the past few years.
That's when they haven't been releasing chipsets that have had drivers that are incredibly buggy, especially relating to the integrated graphics.
AMD's chipsets are made on 55nm, Intels have just moved to 65nm. If there's anything that AMD are doing okay, it's with their ATI related stuff. I still don't see Apple getting them to make a new Intel chipset however.
It entirely depends on your definition of "success." Apple was successful for decades without supporting Windows, at least to the extent that they stayed in business. Now because Macs do support Windows Apple is more successful than it otherwise would be. Quantifying exactly how much more successful is probably something only Apple knows internally. But Apple has come to embrace "peaceful coexistence" rather than active hostility in its cold war with Microsoft, as Steve Jobs himself admitted at one point.