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
PA Semi?
I thought PA Semi were bought specifically to create systems-on-a-chip for portable devices like the iPhone.
Designing a chipset isn't an easy task. Anyone saying that a chipset 'merely' provides access to some features should look at a schematic for an x86 board.
The chipset is the central nervous system of the computer, the CPU is 'merely' the brain. On an Intel-x86 system, even RAM access goes through the chipset. It's nice if your brain and nervous system have the same maker that knows all the peculiarities of both.
Making the chipset proprietary kills native Windows/Linux compatibility unless Apple want to write drivers from now on. It also kills the option to run a hacked OSX on non-Apple hardware, which no doubt provides Apple with new customers through an 'extended test drive' of sorts.
Actually, I'd like Apple to make better use of what Intel provides in their chipsets. Like blindingly fast RAID performance built right in, ten USB ports and six SATA ports! Oh wait, there's no Apple hardware that could make use of all that expandability...
I thought PA Semi were bought specifically to create systems-on-a-chip for portable devices like the iPhone.
Designing a chipset isn't an easy task. Anyone saying that a chipset 'merely' provides access to some features should look at a schematic for an x86 board.
The chipset is the central nervous system of the computer, the CPU is 'merely' the brain. On an Intel-x86 system, even RAM access goes through the chipset. It's nice if your brain and nervous system have the same maker that knows all the peculiarities of both.
Making the chipset proprietary kills native Windows/Linux compatibility unless Apple want to write drivers from now on. It also kills the option to run a hacked OSX on non-Apple hardware, which no doubt provides Apple with new customers through an 'extended test drive' of sorts.
Actually, I'd like Apple to make better use of what Intel provides in their chipsets. Like blindingly fast RAID performance built right in, ten USB ports and six SATA ports! Oh wait, there's no Apple hardware that could make use of all that expandability...
It will be interesting to see what Apple will bring to the table to necessitate doing their own chipset. Whatever it is, I don't think that it will be a conventional reason though.
Things like faster FSB and 3 channel memory aren't necessary because they are already available. A 1066MHz FSB is already sufficient for a dual core processor given that L2 cache is up to 6MB now. And investing in FSB development is a dead end anyways with Quickpath coming soon. 3 channel memory is also no longer a chipset decision with Nehalem using integrated memory controllers. And Mac Pros already support quad channel memory.
I can't really see Xeon MPs being used in Mac Pros since quad core Nehalems with HT will offer 16 logical processors already. Xeon MPs are very expensive and their IMCs will probably require FB-DIMMs which negatively affect their performance compared to upcoming Nehalem Xeon DPs which look to return to registered DDR3 for a performance increase.
Custom accelerator chips are fully supported by Intel since they've opened up their FSB license to accelerator chips and have a project (don't remember the name anymore) to encourage them being placed in PCIe slots. Apple doesn't really need to design a completely new chipset just to add accelerator chips. In terms of expanding the width of SSE, SSE is already going to be supplemented by AVX in future Intel processors which is 256-bit and I believe AVX was designed in such a way to be expanded to 512-bit with minimum effort. Between AVX and OpenCL for GPU, there really isn't a need for other custom accelerator chips.
And integrated graphics is very unlikely since Apple isn't a GPU maker and it'd probably turn out worse and Intel's IGPs. Besides with Larrabee, sticking with Intel for IGPs a bit longer might be worthwhile.
I still think the focus of an Apple chipset design will be on the southbridge. The northbridge currently houses the memory controller and the PCIe links for the GPU so is very performance sensitive and it would make sense for Apple to leave it to Intel since they could tune it in tandem with the CPU. Apple's southbridge though could integrate the functionality that they need like built-in Firewire 400/800 support, built-in gigabit ethernet, built-in multitouch trackpad controller instead of using separate chips hanging off USB and PCIe links as they are now. The southbridge could also be designed to power throttle all the peripherals attached to it more aggressively.
In terms of more forward looking features, I'm hoping for more aggressive voice recognition support and maybe even using the iSight as a source of input, both of which are peripherals that are attached to the southbridge.
Once again, there is nothing Apple could do along these lines that would not simultaneously turn all current Macs into bricks along with the clones. Therefore one would assume that they would not "activate" such a solution for maybe another three years or so, so it makes no sense that a major redesign would be done now just for that.
BTW, Apple has never used a hardware design to prevent cloning. They did add a TPM chip to early Intel Macs but it was never used and later dropped. Cloning of pre-Intel Macs was effectively prevented by putting considerable system code into ROM. Since simply copying the ROM would have been a clear copyright violation, cloners would have had to re-implement its content, which was just too difficult.
They are looking for something extra: extra money savings. They are pretty much stuck with Intel CPUs (at least for laptops) because Intel is miles ahead in terms of performance per watt. Perhaps the chipset is something where other companies are more on par with Intel. I'd bet anything that they are doing the same thing they do with any other purchased part: play the companies against themselves and see who gives them the best deal.
Except, if Apple designs the chipset inside-out, what impetus is there to include good drivers -- if any drivers at all? Part of the reason they moved towards Bootcamp was because hackers were already using existing drivers for much of the Intel hardware (it was a surprising move, but I guess they thought they may as well meet them full way). If Apple does something crazy with the chipset, they have no obligation to actually ship drivers (and you won't be able to use Intel's). That's the real danger here.
Maybe Apple has PowerPC nostalgia and wants to make Macs more different than PCs again?
Except, if Apple designs the chipset inside-out, what impetus is there to include good drivers -- if any drivers at all? Part of the reason they moved towards Bootcamp was because hackers were already using existing drivers for much of the Intel hardware (it was a surprising move, but I guess they thought they may as well meet them full way). If Apple does something crazy with the chipset, they have no obligation to actually ship drivers (and you won't be able to use Intel's). That's the real danger here.
Why would Apple want to prevent users from running Windows on its hardware? The ability to run Windows is now part of Apple's marketing pitch.
I agree, but:
Having the same chipset as the CPU ensures an optimum level of support, performance and economies of scale that you cannot have with separate makers; and even more so in the case of Intel, which has an absolute lead BOTH in terms of performance and fab capacity.
If you are referring to GPU acceleration or parallel processing, fine...for everything else a single, all-inclusive maker is better.
It's like using spare parts for a car that are not supported by the original maker. In other words, NO THANKS...
I think you're right on parallel processing. This is one way for Apple to make advances that are beyond what Intel can do. It also relates to Snow Leopard's advances. Apple has chip engineers on their staff. Their last effort was the memory management chip in the Mac Pro. They will likely do more than can be guessed now; but they could be the only platform that makes is possible for programmers to utilize the multi-core processors effectively. I wouldn't be surprised if they did some work with the graphic chip side as well since it's pivotal to their advances in OS functionality.
Apple has relationships with the manufacturers that can assemble their new boards with competitive pricing. A small premium could bring advances impossible in using the same components as the PC clones.
I think we're all hoping for that any changes make it out in time for the touch promotion(including myself ) but I'm not holding my breath.
) but I'm not holding my breath.The best phrased comments on this topic!
The one known thing that the P.A Semi acquisition brought to Apple was extremely low power management knowledge. If there is one thing that will allow smaller and lighter weight portable devices it is just that.
I don't recall Intel ever shinning in the low-power category. They make great chips, but not outstanding energy-sipping chips and the market is demanding more portability, cooler-running, tighter density, and longer battery life. The first company to bring finished products to market that stands out in this way will have the "first mover" advantage. Those that merely assemble off-the-shelf chip sets into boxes will be at a great disadvantage because they will follow in a fast market.
The PBA was the opening shot in this arena. I expect the next 15 months to include a couple of jolting bumps by Apple to what has been an incrementally evolving portable market for computers. I also won't be surprised if the iPhone turned out to be a stealth portable computer at one end of Apple's complete line of portable computer products.
Who knows? Only Steve Jobs and few people in his "circle of trust" know.
It could very well be that Apple will release new MBPs, MBs, and MBAs tomorrow or a week from Tuesday equipped with Montevina.
Appleinsider is not know for being very accurate with rumors.
Most likely there is some confusion in this current rumor. Possibly unique chipsets are being created for the tablet mac. Or possibly they are making unique chipsets for an entirely new product.
My guess is that Apple will most definitely release updated MBPs, MBs, and possibly MBAs with Montevina chipsets very soon. Soon, as in within 1 month.
After very bad experiences with VIA and SIS chipsets, I only buy PCs with Intel chipsets.
If Apple moves away from Intel chipsets, I'm certainly not going to be a Rev. 1 adopter of those products. They may not be glamorous, but they work.
Now, enhancing Intel's chipset is a different story.
If Apple moves away from Intel chipsets, I'm certainly not going to be a Rev. 1 adopter of those products. They may not be glamorous, but they work.
Now, enhancing Intel's chipset is a different story.
It doesn't really make sense to change from intel for the mpb, especially since the graphics aren't integrated. Perhaps this is only for the macbook? The semiconductor company that apple acquired could make processors for their more mobile platforms, but I don't see them using those processors for the high end desktops or notebooks.
Yes and no. While intel does make the majority of its money from its CPU line, keep in mind its CPU line exists because it creates the chipset support for it. So the two are connected. If they sold a butchered version of the chipset to the rest of the manufacturers (many of which are tied to Apple, believe it or not, such as Foxconn ASUS etc.) and said they were only making premium chipsets for Apple, while at the mercy of Apple's OS X development since that's what the chipset's primary purpose would be, the rest of the manufacturers would throw a fit and Intel would rapidly lose business with them as they would probably shift towards AMD.
Intel is the Microsoft of hardware - they've dictated hardware development for the last 20+ years. The ATX form factor, USB, PCI-E, etc. might all have come from different interest groups but who ultimately dictated whether they would go mainstream in the end was Intel. Take the SLI licensing for example. Intel wanted SLI on their motherboards, Nvidia said no. Intel refused to issue the QPI (QuickPath Interconnect) license for Nvidia, essentially locking them out of making Nehalem chipsets. In the last month, it was confirmed that SLI would go to the X58 chipsets for Nehalem after all. And Nvidia is no slouch in the high tech industry - their influence on GPU's is far reaching and profound.
There's no doubt that any chipset Apple decides to make on its own that uses Intel CPU's will be done at Intel's discretion / cooperation. But Intel knows its own business model and it will certainly not allow any chipset to overthrow its own ability to market its other chipsets to the market. As i said, if this chipset made its other chipsets too weak to compete in the market, thus lowering its ability to increase costs and create bidding wars among the many other major OEM board makers, it would rather pull the plug on this magical chipset than cannibalize its sales elsewhere. If you don't believe Intel would do this, then look no further than its days of Pentium 4 Netburst where it sold ridiculously overpriced CPU's at $1000+ that were barely better than the next iteration. Actually, that still occurs today where Intel locks the multiplier on its CPU's and only allows unlocked multipliers if people are willing to pay $1000+ on the Core 2 Extreme processors (now the QX9650 and QX9770) despite that at stock, performance is marginal.
From what I've read of Grand Central is that it's basically increasing the multithreading ability of the OS. In other words, it's meant to optimize the ability of OS X and its software to support multicore applications. They do this by making these changes at the kernel level so that programmers do not need to make the coding multithreaded at the code level but instead the multithreading will be taken care of at a lower level.
This is expected though since multicore CPU's have become the norm in the past couple of years. Windows 7 is supposed to take advantage of multicore optimizations as well so its the norm. Intel isn't going to stand in the way since it has every reason to want to be on that path. With HyperThreading, Nehalem quad cores already offer 16 threads and at the end of next year, with 32nm, Nehalem will have MCM'd 8 cores allowing 32 threads. And by 2012, after Westmere and Sandy Bridge 32nm and 22nm, we're expecting 16+ cores for 64+ threads.
GPU computing doesn't allow for the same complexity as CPU's since CPU's are more general purpose and can do a lot of arithmetic that GPU's cannot. GPU's, however, have many more "cores." They like to market them as cores now, but are better known as shaders. For instance, the GTX280 from Nvidia has 240 shader processors so 240 cores each which can do a MADD operation (2 FLOPS) + a MUL operation (1 FLOP). ATI's RV770 architecture has 800 Stream Processors / cores each capable of 2 FLOPS as well. However, they aren't capable of doing double precision to the same efficiency as CPU's so for ATI's scenario, you divide their FLOPs by 5 to get double precision FLOPs since it requires a full 5 shaders to do 1 DP operation. So in the case of the ATI 4850 which has 1 TFlop of single precision, it does ~ 200 GFlops double precision.
The issue is that GPU computing has only been pushed recently. Apple and AMD/ATI are onboard for OpenCL (Open Computing Language). Nvidia is much further ahead with CUDA for their GPUs. Microsoft might hold the trump card with DirectX 11's Compute Shade which makes any DX11 compatible card have the ability to do computation with the shaders. We might be seeing a repeat of DirectX vs. OpenGL vs. Glide in the future, although this time its DX11 vs. OpenCL vs. CUDA.
However, as stated, GPU computing (or GPGPU) requires these instruction sets to mature and they tend to be limited to running specialized instructions. For instance, Nvidia demo'd the encoding of a video at 300x faster than a CPU with their GTX280. However, it's because it was coded for it. And the CPU was still relevant since it needs to actually put everything together in the end and run the OS as well.
GPGPU's are coming along, and coming along fast, but it will not replace CPU's and chipsets for a few years, especially if intel has anything to say about it
I liken the marriage with Intel similar to manufacturers married to Windows. They provide the performance and flexibility needed to increase your market but don't kid yourselves, they have the muscle to force you to do what they want in the end if they feel you're out of line.
And for all the talk that Apple can restart PPC - first of all, doing so would abandon x86 since PPC does not have an x86 license, thereby torpedoing support for Windows and other programs that have been ported due to x86 instruction sets. Second, one does not simply just jump start CPU technology and expect to be competitive. Look at AMD - they went from domination with K8 to near bankruptcy now simply because they are a half generation behind in CPU architecture with K10 and a half generation behind with fabs. And don't even think that Intel is going to fab anything besides their own stuff for you. AMD being behind Intel (the clear #1 in fab technology in the world) in fabs is nothing to snicker at either - the fact that they have worse yields on their own chips means lower margins, lower performance (thus lower clocks), and each delay only makes the gap larger. Also, Intel already produced prototypes of 32nm Nehalem while AMD hasn't even gotten its first 45nm chip out to retail yet (in a few months is the current roadmap). And we're talking about AMD, the only company out there that has competed against Intel in the past 10 years at any sort of competitive level.
As far as Apple going GPU - forget it. A company of Intel's size struggles with making IGP solutions competitive with Nvidia and ATI (now AMD), what makes you think Apple, a company that hasn't designed GPUs in its history, would be competitive? And Intel's been making IGPs for its chipsets for a very long time. If they're bringing GPU compute technology onboard, its going to be in some form from AMD or Nvidia or Intel Larrabee+derivatives.
VIA proved indeed since the stone-ages of personal computing that they make the worst computer-chips ever. Before and after I turned to Mac (2001), I've handled a lot of PC motherboards, and every PC that had the most trouble/defects besides the standard Windows crap, it always had VIA crap on them. That company made me throw PC's out of the window way too many times. The money I wasted on that, all down the drain.
So Apple... if you really do need different chips? Get the IBM stuff back or make your own designs, but NEVER EVER use that cheap no-good crap from VIA & co.
Its possible though from the tone of the article, unlikely, that Apple could look into the Puma platform from AMD. At mobility levels, the Core 2s are not that much faster than AMD's CPU offerings. But the biggest gripe people seem to have is the terrible IGP on the MacBooks and the AMD Puma IGP is way better than the Intel IGP. Also, they support Universal Video Decoder so that might finally mean a Blu-Ray player built in...
Seems likely they want to but at the same time, the question was raised how would they support older Macs without cutting them out from future OS versions?
The biggest issue would be the hard drive controller. If it's not backwards compatible, you'd need to install the drivers during the Windows install with F6. Vista isn't so bad since it can read the drivers from a USB device, plus it already includes newer controllers. XP however normally requires the drive controller drivers on a floppy or slipstreamed into the install media.
So depending on how different this chipset is, they may need to drop official XP support and/or add a few steps to the Boot Camp install.
Contrary to the belief of many of the users in this thread, Intel is not the only competent chipset manufacturer in the world... and they're in a different league to Apple. Intel has a target audience of the mainstream PC market, and consequently they have to use mainstream backwards-compatible techniques. Apple, however, has the ability to restructure everything at the drop of the hat if it's necessary, and this ability to innovate is the reason they've been able to go from almost bankrupt to a thriving genuine threat to the competition so quickly.
If, for example, Apple's chipset (I do NOT think Apple will go with AMD or VIA) provides extra CPU to GPU communication functionality and OpenCL-specific optimisation, this does not affect the ability for Macs to run on Windows, but it does however provide something that cannot be replicated on PCs, resulting in better performance to Mac users, a more impressive "leading edge" (Intel is currently giving people the impression that a Mac is just a PC with a different OS), and a way to stop the Hackintosh industry without suing everyone. Win, win, win for Apple... Win, win, (who cares?) for everyone else.
Edit:
Remember, Apple is not competing with Intel on sales, so the concept of collaborating and extending a platform like Montevina is not out of the question.
If, for example, Apple's chipset (I do NOT think Apple will go with AMD or VIA) provides extra CPU to GPU communication functionality and OpenCL-specific optimisation, this does not affect the ability for Macs to run on Windows, but it does however provide something that cannot be replicated on PCs, resulting in better performance to Mac users, a more impressive "leading edge" (Intel is currently giving people the impression that a Mac is just a PC with a different OS), and a way to stop the Hackintosh industry without suing everyone. Win, win, win for Apple... Win, win, (who cares?) for everyone else.
Edit:
Remember, Apple is not competing with Intel on sales, so the concept of collaborating and extending a platform like Montevina is not out of the question.
Here's my take.
1. I love where Apple is at now with Intel.
2. I have had a few AMD systems and I was not fond of them at the time. in fact the HP I just de-comissioned was an AMD Athelon 64 bit (3 years old).
However, I know a lot of gamers love the AMD, so maybe Apple will use the Intel for "normal" laptops and desktops and AMD for a gaming Mac, since so many people were crying for one.
Also, aren't AMD chips cheaper (I mean less expensive). Hence the article on transitioning lines and lower margins. Could Apple be bending to the demand of the masses and sacraficing quality? Lower priced? Gaming.
I hope Apple just does not become another computer company. I finally switched, my wife finally switched, and the nitch factor is great.
JustsSay a few prayers Rev!--))