What, Where, When, Why & How much - Mac OS X 10.6 Snow Leopard

[Northy]

Thanks to t0mat0 of MacRumors for this guide.

This thread is intended to be a collation of information regarding Snow Leopard (abbreviated to SL), both normal and also Server editions. I'll flesh some of this out, and should hopefully add more as time goes on - we've got a whole year go!

I've checked this Forum, and I could not see any specific thread on actually talking about SL (Snow Leopard) in this way - both general and specifically. Hopefully there is enough room to take a thread with a different angle.

For the main part this month, it'll be interesting to collate information from such places as Intel, Nvidia, RoughlyDrafted.com, Apple, and others to have a look at the potential future coming in 2009 and beyond, and what may be in store. We're in a rumour site, and it's fun to see what is not just going to happen, but also the potential of what could happen. It can be just as much used for me as an aide in remembering what the heck I read on Apple from other sources. Google Reader ain't as searchable as a cached macrumors thread via Google search.

Starting from Known knowns: Apple’s Snow Leopard page

When
Shipping ~late Q2, early Q3 2009.

What
“A Quantum Leap”. No solace required. Billed as changing it’s focus, “taking a break from adding new features” and building on Leopard

• Delivering “a new generation of core software technologies” to
  - streamline Mac OS X
  - enhance Mac OS X, including improving quality.
• Reduce the OS footprint
• Out-of-the-box support for Microsoft Exchange 2007 built into Mail, Address Book, and iCal (using the Exchange Web Serices protocol).
• “Grand Central”
  – A set of technologies to improve performance
  - Makes “all of OS X multi-core aware”
  - Optimises Mac OS X it for “allocating tasks across multiple cores and processors”
  - Helps developers, by making it easier for them to create programs that can effectively use the power from multiple cores and processors.
• Extension of 64-bit technology in Mac OS
  - Allowing up to a theoretical 16TB maximum of RAM (No word on what type)
• Quicktime X
  - Streamlined platform for modern media and internet.
  - Optimised support for modern codecs
  - More efficient media playback
• Through Safari, delivering fast Javascript (e.g. implementing this through MobileMe)
• OpenCL (Open Computing Language)
  - A language to help developers use the power of GPUs (graphics processing units) and redirect it for general purpose computing.
  - In other areas, OpenCL is akin to GPGPU.
• ZFS - Not mentioned on the normal SL page, but confirmed for the SL Server edition here
  - Read & write support for the 128-bit ZFS file system
  - Features such as storage pooling, data redundancy, automatic error correction, dynamic volume expansion, snapshots.

Who
Apple, obviously. Intel. PA Semi potentially. But who else? It's a bit of a shoehorning to fit the thread into the WWWWW&HM theme, but it's a decent start. I'd think it'll be interesting to dig into who Apple's paired with in the past for Graphics cards, and GPU chips, and see where the state of play lies there. Seeing as Apple Macs are a more closed system, bar the more accessible Mac Pro, it'll be interesting to see if Apple will open up being able to upgrade the motherboard, memory, graphics card, and CPU(s).

Why
To make money, serve the stockholders, and to create useful Apple product - software and hardware.
Basically, Apple will be sweetening the pot, to get people to move over, just as Leopard had features that Tiger did not, and people have been moving over at a fair rate.

For example, - You have the most recent(4th Generation) MBP:
- You have a multi-touch pad, so you're set for multi-touch.
- You have either a 2.4GHz, 2.5GHz or 2.6GHz Intel Core 2 Duo processor. From wiki - In February 26, 2008 the MBP line was fitted with "Penryn" Intel Core 2 Duo processors.
These processors have multiple cores, so SL could help them with increasing the use of their multiple cores for processing power.
We await some true stats as to what SL will do on current most recent Apple kit - e.g. Mac Pro's, MacBook Pros and MacBooks, but it's safe to say that as Nehalem is released, Apple will have more power to play with, and will be able to deliver more power through Grand Central and OpenCL.

How Much
I'd say something akin to Leopard. I don't think the price will be lowered, which would discount SL's worth, or too much higher, which would detract software sales (which bring the money into Apple, through hardware sales). It's talked about in depth by Daniel here so probably best to discuss it further, once i've started adding precis's of that series of articles.

For breaking that down, currently we have some main topic areas being

• Level of multi-touch in SL
• Usage of SL as a basis for the OS of future iPhones and other "electronic devices"
• Microsoft Exchange 200 support, and how much other Windows support SL will have
• “Grand Central”
  - how it ties in with Intel and their multiple core chips, and also with multiple processors.
  - How it ties in with developing Mac OS X apps, and also iPhone apps
• How 64-bit technology and a 16TB maximum (along with the other mentioned factors) will affect computing
• Quicktime X
• Safari - and Javascript - how this will be deployed. Looking at Air, Silverlight and others (e.g. Prism).
• OpenCL
  - How it will tie in with SL, and Grand Central, and who Apple goes for to supply these GPU chips/ GPU boards for GPGPUage
  - How it will affecting computing, alongside Grand Central
  - It's rivals on the scene, and how Intel will play it's part.

If anyone's interested in this, please say hello, and contribute - i'm all for useful info-packed threads, with as much sources and multimedia thrown in as possible. Unfortunately it's a pain in the RSS to use the macrumor guide wiki section currently, so thread it is.

There's a list of sources on page 2 (the word limit is biting).

[Northy]

The state of play currently



As wiki on the Tick Tock strategy notes:
Tock - Intel Core microarchitecture of 2006
Tick - Shrink/derivative (i.e. Penryn) 45 nm of 2007
Tock - New Intel microarchitecture (i.e. Nehalem) for 2008
Tick - Shrink/derivative (Westmere) 32 nm for 2009
Tock - Future Intel microarchitecture (i.e. Sandy Bridge) for 2010

So Penryn's successor is expected to be Nehalem. Nehamlem will bring back hyperthreading, which will crop up soon enough. SL will release after Nehalem, and potentially around Westmere, and before Sandy Bridge in 2010.


Mac specs past, present & future
(No Mythical Midrange Mac Minitower yet. It's the new Octocore ;)

iMacs: From here and because the wiki doesn't have the useful stats, also from here

A large part of Apple's consumer desktop offerings, and bar the MacBook Pro, probably a favorite to be used in films, movies and commercials - it has appeal (See the excellent film "Inside Man" by Spike Lee, for an example). Introduced 1998, with an evolution to the iMac G3 (egg shaped with a CRT monitor), iMac G4 (hemispherical base +LCD on an arm), to the iMac G5 and Intel iMac (looking like an LCD, with everything behind it). Fair enough to call it the " Gold Standard of desktop computing". It has the Apple aesthetics? The first Mac to include a USB port - and bringing some cross-platform peripheral compatability with it.

MacBooks - From here
MacBook Pros - From here
Mac Pros: - From here

The biggest winner of actual (as opposed to relative) performance benefits. Not as constrained by thermal, nor space issues. It won't just be "non-linear digital editing" for HD video and high end Digital image work that will benefit (ZFS, Grand Central, GPGPUage and 64-bit tech with a huge maximum RAM possibility coming in very useful i'd imagine.)

As noted in the wiki, the Power Mac G5 had 4 processors, but fell down on expandability. SL may very well improve on the current level of expandability of the Mac Pro. As it's easier to swap out and put in new, compatiable Intel CPUs in a Mac Pro, it's less useful to talk about what they previously had as stock, so added just for reference. Of note, SL might also see around it's launch, the ushering of an iMac or mid-tower less bulky than the Mac Pro, that allows the user to change it's graphics card.



XServe: Currently: 3 independent hot-plug drive bays with support for SATA or SAS drives (i.e. up to ~ 3TB of hot-plug internal storage)

Apple - Xserve - Technology - System Architecture


CPU
iMacs:
Early 2003: 0.8/1.0Ghz
Late 2003: 1.0/1.25Ghz
Mid 2004: (iMac G5): 1.6/1.8 GHz PowerPC G5
Early 2005: 1.8/2.0 GHz PowerPC G5
Late 2005: 1.9/2.1 GHz
Early 2006: ?/2.0 (Core Duo)
Late 2006: 1.83/2.0/2.33 Ghz (Core 2 Duo)
Late 2007: ?/?/2.8Ghz C2D Extreme
Early 2008: 2.4/2.6/2.8/3.06Ghz* (*Custom Montevinalike chip)

MacBooks:
1st Gen Early 2006 - 1.83/2.0GHz Intel Core Duo (T2400/T2500)
2nd Gen Late 2007 - 1.83/2.0 GHz Intel C2D (T5600/T7200)
3rd Gen Mid 2007 -2.0/2.16 GHz Intel C2D (T7200/T7400)
4th Gen Late 2007- 2.0/2.2 GHz Intel C2D (T7300/T7500)
5th Gen Early 2008 - 2.1/2.4 GHz Intel C2D (T8100/T8300)
5th Gen Late 2008 - ??

MacBook Pros:
1st Gen Early 2006 - 1.83/2.0/2.16 GHz Intel Core Duo Yonah
2nd Gen Late 2006 - 2.16/2.33 Intel C2D Merom
3rd Gen Mid 2007 - 2.2/2.4/2.6 GHz Intel C2D Merom (2.6 came after Nov 2007)
4th Gen Early 2008 -2.4/2.5/2.6 GHz Intel C2D Penryn
5th Gen Late 2008 - ? GHz Nehalem, & Montevina?

Mac Pros:
2006: C2D Xeon
2008: 1/2 sockets of 2.8/3.0/3.2 GHz Quad Core Xeon 5400 (Harpertown 45nm) 64-bit CPUs to give a maximum of 8 cores.
Late 2008: Gainestown likely.

Graphics
iMacs:
Early 2003: Nvidia GeForce 2MX/4MX
Late 2003: NVidia GeForce 4 MX/GeForce FX 5200 Ultra
Mid 2004: NVIDIA GeForce FX 5200 Ultra
Early 2005: ATI 9600
Late 2005: ?
Early 2006: ATI Radeon X1600
Late 2006: ATI Radeon x1600 128MB/NVIDIA GeForce 7300 GT 128MB
Late 2007: ATI Radeon HD ?
Early 2008: ATI Radeon HD 2400 XT 128MB/ATI Radeon HD 2600 PRO 256MB/NVIDIA GeForce 8800 GS 512MB

MacBooks:
1st Gen Early 2006 - Intel GMA 950 graphics processor using 64 MB DDR2 SDRAM shared by main memory
2nd Gen Late 2006 - Intel GMA 950 graphics processor using 64 MB DDR2 SDRAM shared by main memory
3rd Gen Mid 2007 - Intel GMA 950 graphics processor using 64 MB DDR2 SDRAM shared by main memory
4th Gen Late 2007 - Intel GMA X3100 graphics processor using 144 MB of DDR2 SDRAM shared by main memory
5th Gen Early 2008 - Intel GMA X3100 graphics processor using 144 MB of DDR2 SDRAM shared by main memory
6th Gen Late 2008 - ??

MacBook Pros:
1st Gen Early 2006 - ATI Mobility Radeon X1600 with 128 MB (128 MB) or 256 MB of GDDR3 SDRAM and dual-link DVI
2nd Gen Late 2006 - ATI Mobility Radeon X1600 with 128 MB (128 MB) or 256 MB of GDDR3 SDRAM and dual-link DVI
3rd Gen Mid 2007 - nVidia Geforce 8600M GT with 128 MB or 256 MB of GDDR3 SDRAM and dual-link DVI
4th Gen Early 2008 - nVidia Geforce 8600M GT with 256 MB or 512 MB of GDDR3 SDRAM and dual-link DVI
5th Gen Late 2008 - ??

Mac Pros:
2006:
2008:

XServe:
2008: 1 or 2 sockets of 2.8/3.0GHz Quad-Core Intel Xeon 5400

Memory
iMacs:
Early 2003: ?
Late 2003: 256MB DDR333 SDRAM
Mid 2004: 256MB DDR400 SDRAM
Early 2005: 512MB RAM
Late 2005: ?
Early 2006: ?
Late 2006: 512MB/ 1GB RAM
Late 2007: ?
Early 2008: 1/2GB

MacBooks:
1st Gen Early 2006 - Maximum of 2GB
2nd Gen Late 2006 - Maximum 4GB (3.25 GB usable)
3rd Gen Mid 2007 - Maximum 4GB (3.25 GB usable)
4th Gen Late 2007 - Maximum 4GB
5th Gen Early 2008 - Maximum 4GB
6th Gen Early 2008 - ? Probably Max. 4GB due to size issues.

MacBook Pros:
(Throughout 1-4th gen, the memory has been PC2-5300 DDR2 SO-DIMM SDRAM)
1st Gen - Maximum 2 GB
2nd Gen - Maximum 4 GB, but only 3 GB addressable (due to the chipset)
3rd Gen - Maximum 4 GB
4th Gen - Maximum 4 GB
5th Gen - ??

Mac Pros:
2008: Maximum 32GB (8 x 4GB)
2009: Maximum 15TB!

XServe:
2008: Maximum 32GB memory (8x4GB 800MHz DDR2 ECC fully buffered DIMM (FB-DIMM) memory


Future?

Mac Pro Future upgrade: Would Apple risk FB-DIMM memory? A 4 socket Beckton is more for a server, than a powerhouse desktop machine. Apple can get a lot of power from just a 2 socket mainboard, and have decent DDR3 memory, that will overclock much more easily than FB-DIMM2 memory.
As Nehalem chips release early 2009, i'd expect a 2.66/2.93/3.2GHz quad-core processor (Nehalem architecture) lineup. The QuickPath architecture , and the 4 PCI Express slots (capable of running at 8X bandwidth for systems with multiple video cards) would be useful, and give the possibility for high end GPGUage. That'll mean AMD's CrossFireX will be supported out of the box to use extra cards in accelerating 3D on a single display, with the possible addition of NVIDIA's SLI for GeForce cards. Apple may well not go with bog standard components if the graphics side is going to get limited due to board choice.


XServe:
Memory - How could Apple put 16TB of RAM into an XServe? Would there be a feasable way of creating a rackable bolt on that could link with an XServe at a fast enough rate to basically have the RAM outside of the XServe's box? With space inside, with the ability to link XServes together, or maybe create a slave external RAM only XServe has a chance to join the big boys.
Hard drives - ZFS will suit Snow Leopard XServe Servers well. With 4 drive bays currently for Serial ATA, & 1TB drives on the market, (e.g. Barracuda 7200.11) it isn't hard to see the usefulness. (I think ZFS will also be useful with external drives too, making it a lot more simple, and kind of destroying the Drobo marketplace to an extent. Hard core consumers, Mac Pro users etc, but we await to see if ZFS will make it into SL consumer OS). I'd imagine they'd probably go with a 2/4 socket Beckton design, but who knows.

iMac: Will get bumped, but unless they change the design, it'll be still hard to upgrade.

MacBook Pro: With 32GB RAM allowable in already, unless RAM gets smaller, the MBP won't see more RAM, just better graphics, chipset & processor.

MacBook: As the MBP, but less so?

[Northy]

Does Apple have a plan to greatly expand its position?
Mashing up this source amongst other things.



Apple, if it gets ahead of the market in taking advantage of (multiple) processors with multiple cores, they could expand into the server, supercomputer, and also gaming and heavy usage computing fields, whilst bringing the benefits to all consumers. Is Microsoft in such a position?



With the iPhone 3G a few days from release, this platform is stable and just starting to take off. There are hardware and software complaints, and we'll see whether they'll be taken care of (flash for the camera, video, front cam, better Mpixel CMOS camera etc). Can the iPhone and Touch get away with maintenance and evolution, rather than revolution? Quite possibly.

Implication: Apple’s best hardware and software teams may now have time to work on real interesting Mac stuff.

For processors, it's less a MHz race than before, and more about cores & sockets. Beyond the multi-core multi-processor environment, we also have GPUs and custom chips on the scene too.

What's on the horizon?
Raw performance of the upcoming CPUs, the ability to harness it effectively with things like Grand Central, and the potential for either hybrid CPU/GPUs, or GPUs complementing CPUs will bring in a whole lot more processing power. Apple for its part, is wanting to make this tech useful, and easy to use for consumers, easy to implement and access for developers. Snow Leopard is basically blowing the top off the theoretical top end Mac configuration using Leopard. Fully 64-Bit, up to 16TB RAM, and dual socket Nehalem chips to start off with.

Cast
NVIDIA - The graphics chip leader
Intel - The #1 CPU maker
Apple - The #1 OS ;)
AMD - The #2 chip maker
Microsoft - The #>=2 OS provider

Processors - Intel

Remember the Guinness surfer advert? Same deal. Tick follows tock follows tick follows tock. Like Apple, good things come to those who wait!

2007 - Tick - Shrink/derivative to 45nm (i.e. Penryn (aka Yorkfield))
2008 - Tock - New Intel microarchitecture 45nm (i.e. Nehalem (aka Bloomfield))
2009 - Tick - Shrink/derivative to 32nm (i.e. Westmere (aka ?))
2010 - Tock - Intel microarchitecture 32nm (i.e. Sandy Bridge (aka ?))

Core microarchitecture
(An Intel CPU core roadmap is at the bottom which may help)

In 2006, the Intel Core architecture was unveiled. Core chips had Virtualisation Technology (virtualisation support), Intel 64 (Intel's implementation of x86-64) and SSSE3. They were based around an updated version of the Yonah core & and could be considered the latest iteration of the Intel P6 microarchitecture, which traces its history back to the 1995 Pentium Pro).

The Core processor lines - with differing socket use, bus speed, power consumption etc:[list][*] Merom - for mobile computing[*] Conroe - for desktop computing[*] Woodcrest - for servers/workstations

Branding:
Mobile processors - Core 2
Desktop processors - Core 2
Low end Core processors - Pentium Dual Core
Low end Core processors - Celeron
Servers and workstations - Xeon

Just to note - confusingly, Intel processors branded as "Intel Core", e.g. the 65nm Yonah processor and its variants do not use Intel Core microarchitecture despite its name.

Core 2
The Core 2 brand refers to a range of CPUs based on the Intel Core microarchitecture.

The Core 2 processor lines:

• C2S - Core 2 Solo: single-core CPU
• C2D - Core 2 Duo: dual-core CPU
• C2Q - Core 2 Quad: quad-core CPU
• C2X - Core 2 Extreme: dual/quad-core CPU

Branding - (Lower is newer - e.g. Penryn is successor to Merom):
Laptops
Merom (65nm) - Core 2 Solo/Core 2 Duo/Core 2 Extreme(Dual)
Penryn (45nm) - Core 2 Solo/Core 2 Duo/Core 2 Quad Core 2 Extreme (Dual)/Core 2 Duo Extreme (Quad)

Desktops
Conroe (65nm) - C2D/C2X (Dual)
Allendale (65nm) - C2D
Wolfdale (45nm) - C2D
Kentsfield (65nm) - C2Q/C2X (Quad)
Yorkfield (45nm) - C2Q/C2X(Quad)

Servers and Workstations:
(Intel brands servers & workstation Core 2 CPUs as Xeon processors. Xeon CPUs generally have more cache than their desktop counterparts in addition to multiprocessing capabilities. (afaik - it takes peons eons, to work out Xeons...))

Dual-core Xeons:
5100-series Woodcrest
5200-series Wolfdale (45nm)
7100-series Tulsa (65nm)
7200-series Tigerton
3000-series Conroe
3100-series Wolfdale

Quad-Core Xeons:
3200-series Kentsfield - Relabelled C2Q processor
3300-series Yorkfield - Relabelled C2Q processor
5300-series Clovertown - Consists of 2 Woodcrest chips in one package
5400-series Harpertown
7300-series Tigerton - A 4 socket and greater capable

(The fastest Harpertown is the X5482 and is also sold under the name "C2X QX9775" for use in the Intel SkullTrail system).
(The Clovertown X5365 is among the fastest processors, performing up to ~38 gigaflops = 0.038 teraflops )

Future versions:
Whitefield (cancelled)
Aliceton
Dunnington - Last of the Penryn generation - a single die 6 core processor
Gainestown - Based on Nehalem microarchitecture
Beckton - 8 or more core Nehalem processor. Based on Nehalem microarchitecture.

The successor to Penryn is Nehalem, the 32nm shrink of Nehalem is Westmere.
Release schedule: Penryn chips coming out up to ~Q3 08, Nehalem for late 08

Intel seems to be holding some cards to its chest. It has Larrabee (more on other posts) which is an integrated graphics platform that can natively execute CPU x86 code. i.e. when not rendering 3-D graphics, it can donate processing cores for general CPU work.
Intel also has Nehalem coming very soon, then the Sandy Bridge platform in late 2009/early 2010 (which is expected to integrate Larrabee onto a single die with quad core (or more) processors, leading to improved performance).

• Set to be released in 2 flavours - both based on Nehalem CPU Architecture, one being a desktop chip (Havendale), the other a notebook chip (Auburndale).
• Auburndale & Havendale will have 2 Nehalem cores paired with a graphics subsystem. The twin cores will share 4MB of L2 cache and feature an integrated dual-channel memory controller that supports memory configurations up to DDR3-1333 apparently.
• The graphics subsystem will be initially derived from Intel’s G45 integrated graphics. This indicates that neither Auburndale nor Havendale will be for heavy graphics processing, but will be more of an integrated graphics replacement.
• According to Intel roadmaps, the new processors are expected to enter the market in the H1 2009. Which is slap bang when Snow Leopard is pencilled in for release.

Intel's Senior VP Patrick Gelsinger says Xeon MP versions of Nehalem will be up to octocore processors, and will use HT. Most "Enterprise software vendors charge by the socket and not by the number of CPU cores." - If this was to change, we could see quad sockets be more widely available i'd imagine. Until that time, more cores per socket is desired, due to licencing costs... Apple currently only has a dual socket configuration. Could they go hardcore 4 socket crazy?

How does Apple sell all that power to customers? What are the compelling reasons, the killer apps? It's about getting people to start using apps that use this. Video, 3G, The internet in your pocket, a supercomputer at your desk.


Graphics
The main players of the graphics market are

• NVIDIA's GeForce
• AMD's Radeon
• Intel in 2009.

Interesting for the discrete market, a side step for the integrated market. So what's NVIDIA doing? Is it getting into the CPU business? Trying to buy Via - an x86 license and with a line of CPUs...? nVidia makes an offer to buy AMD? ATi?


Custom chips:
Could Apple use either of a graphics card manufacturer's GPU's for GPGUage? Or will it make it's own chipset including GPU(s) for performance?
Apparently, Nehalem could allow Intel to integrate a graphics core into the processor, if it wanted to. So this may very well be where we see the introduction - Larrabee could fit right into this plan.

Other techWiMax, LTE, WiUSB, eSATA, A-GPS, SSD, USB 3.0, FW1600/3200...

[Northy]

Applications for GPGPU
Snagged from the wiki

• Computer clusters or a variation of a parallel computing (utilizing GPU cluster technology) for highly calculation-intensive tasks:
  - High-performance clusters (HPC) (supercomputing) including distributed computing.
  - Grid computing (a form of distributed computing) (networking many heterogeneous computers to create a virtual computer architecture)
  - Load-balancing clusters (a server farm)
• Physical based simulation and physics engines (e.g. Newtonian style physics models) inc. cloth, hair, fluid flow (liquids, smoke)
• Segmentation – 2D and 3D
• CT reconstruction
• Fast Fourier transform
• Tone mapping
• Audio signal processing inc. for digital, analog & speech processing
• Digital image processing
• Video Processing
  - Hardware accelerated video decoding and post-processing (Vista has it. Come on Snow Leopard!!)
  - Hardware accelerated video encoding and pre-processing
• Raytracing
• Scientific computing - weather, climate forecasting, molecular modelling inc. X Ray Crsytallography
• Bioinformatics[4][5]
• Computational finance
• Medical imaging
• Computer vision
• Neural networks
• Cryptography and cryptanalysis

I'd imagine SIGGRAPH 08 & 09 will be buzzing with this stuff.




CUDA - Compute Unified Device Architecture.

Good long read here

An SDK and API - a C compiler and set of development tools for programmers to help use C to code "algorithms for execution" on the GPU. (graphics processing unit). Developed by NVIDIA, it requires an NVIDIA GPU to use CUDA (G8X upwards, including GeForce, Quadro & Tesla lines). It gives developers access to the native instruction set and memory of the massively parallel computational elements in CUDA GPUs. Initially the CUDA SDK made public Feb 2007. So through CUDA, the NVIDIA GPUs can be turned into powerful, programmable open architectures like today’s CPUs (Central Processing Units) simplistically as the wiki says.

What might be helped by this? For the gaming industry, physics calculations - including debris, smoke, fire, fluids. Wiki provides the links to BioMed Central | Full text | High-throughput sequence alignment using Graphics Processing Units and BioMed Central | Full text | CUDA compatible GPU cards as efficient hardware accelerators for Smith-Waterman sequence alignment for the acceleration CUDA gives for non-graphical computation in computational biology/other fields.

Advantages over over general purpose computation on GPUs (GPGPU) using graphics APIs.

• Uses the standard C language, with some simple extensions
• Code can write to arbitrary addresses in memory.
• CUDA exposes a fast shared memory region (16KB in size) that can be shared amongst threads. This can be used as a user-managed cache, enabling higher bandwidth than is possible using texture lookups.
• Faster downloads and readbacks to and from the GPU
• Full support for integer and bitwise operations

Cons:

• CUDA-enabled GPUs are only available from Nvidia
• Texture rendering & recursive functions are not supported
• Deviation from the IEEE 754 standard.
• Potential bottleneck of Bus bandwidth and latency between the CPU and the GPU.
• Threads must run in groups of at least 32 threads that execute identical instructions simultaneously. Branches in the program code do not impact performance significantly, provided that each of 32 threads takes the same execution path; the SIMD execution model becomes a significant limitation for any inherently divergent task (e.g., traversing a ray tracing acceleration data structure).

You can see examples of what CUDA can do here (It's flash based).

Why? From Beyond 3D's article:
- Neither DirectX nor OpenGL are made with GPGPU as their primary design goals, thus limiting their performance
- Arbitrary reads and & writes to memory while bypassing the caching system (or flushing it) is still not supported in the Direct3D 10 API

AMD: Streaming Close to the Metal

CTM's commercial successor is the AMD Stream SDK, released in 2007.
Like CTM, Stream SDK provides tools for general-purpose access to AMD graphics hardware.

Differences:
"The idea behind CTM is that there is efficiency to be gained by giving an experienced programmer more direct control to the underlying hardware.
CTM is thus "fundamentally [an] assembly language. CUDA on the other hand aims to simplify GPGPU programming by exposing the system via a standard implementation of the C language. At this point in time, the underlying assembly language output (also known as "NVAsc") is not exposed to the application developer.

"CUDA exposes the NVIDIA G80 architecture through a language extremely close to ANSI C, and extensions to that language to expose some of the GPU-specific functionality. This is in opposition to AMD's CTM, which is an assembly language construct that aims ot be exposed through third party backends. The two are thus not directly comparable at this time."

Chipsets, graphics, handhelds, desktops, Visualisation, near-time and real-time rendering. Market area examples, for rigid body physics, matrix numerics, wave equation solving, biological sequence matching, finance.

GPGPUS: General-purpose computing on GPUs (graphics processing units)
From the wiki: Made possible "by adding programmable stages and higher precision arithmetic to the rendering pipelines, which allows software developers to use stream processing on non-graphics data."

Basically expanding the purpose of a GPU from just accelerating parts of the graphics timeline, to using it for general purpose computations, to also accelerate the computer's non-graphics related computations. There are certain restrictions in operation and programming - their effectiveness is suited for solving problems using stream processing - processing things in parallel - operating " in parallel by running a single kernel on many records in a stream at once."

A stream being "a set of records that require similar computation. Streams provide data parallelism."
Kernels are the functions that are being applied to each element in the stream. e.g. in GPUs, vertices & fragments are the elements in streams, with the kernels to be run on them being vertex & fragment shaders.

"The most common form for a stream to take in GPGPU is a 2D grid because this fits naturally with the rendering model built into GPUs. Many computations naturally map into grids: matrix algebra, image processing, physically based simulation, and so on."


Apple's position? OpenCL, on a post below.

[Northy]

A mash up of a series of ongoing articles by Daniel Eran Dilger at roughlydrafted.com

Myths of Snow Leopard 1: PowerPC Support June 16th, 2008

Snow Leopard is going to be the first version of Mac OS X that only runs on Intel Macs. PowerPC Mac users will still be able to use Leopard, which will get updates. The new features in SL are primarily going to impact on
multicore, multiprocessor machines. (OpenCL, Grand Central, and the new 64-bit kernel)

Universal Binaries of the new Mail, Address Book, and iCal could be made, if Exchange support is wanted. Will 3rd party developers keep making Universal Binaries? If it's in their interest - yes (e.g. if they already have their PowerPC code, and it's easy to update the UB).

Of the 27 odd million Mac OS C installed user base, Leopard is on over a third, Tiger over a third, and the rest on an earlier version, so there is still a user base to sell Universal Binaries to, if that user base buys enough software.

The article purports that with the Universal Binary architecture, developers can target both Intel & PowerPC Macs with minimal extra effort. (Adobe being an exception)

Will SL definitely be Intel only? We await to find out - but signs are it will likely be.

Myths of Snow Leopard 2: 32-bit Support June 17th, 2008

Myth 2 - Apple is dropping support for 32-bit Intel Macs because Snow Leopard is 64-bit.

Apple doesn't have a problem, but Windows does have a 32/64-bit conundrum, as Microsoft has to ship multiple architectures of Windows XP, Vista, and Server:

• IA-32, for the 32-bit Intel x86 architecture of most PCs (the same as that of the early Intel
• IA-64, for the 64-bit Intel EPIC architecture developed for Itanium and largely unused by anyone (now Windows Server only).
• x64 (aka AMD64), the 64-bit x86 architecture developed by AMD and copied by Intel after the humiliation of IA-64’s failure. It is the mainstream 64-bit PC architecture, and is used in the latest Core 2 Duo Intel Macs.

Windows' 32/64-bit Conundrum:
Its IA-64 & x64 versions of Windows run from different code bases, and offer poor compatibility with a lot of existing 32-bit software. Also, 64-bit versions of Windows don’t run on 32-bit PCs, meaning that the market for developing 64-bit drivers and apps for Windows is artificially small, and can’t get bigger because there are software barriers to adopting 64-bit PCs that are of Microsoft’s own doing. That chicken and egg problem has no solution outside of Microsoft figuring out how to merge its code versions together, which it doesn’t have the time, inclination, or expertise to do.

Microsoft is betting users will upgrade to 64-bit PCs and yet continue to buy and run the old 32-bit version of Windows until it can manage to clean up the sticky bits and deliver a 64-bit, EFI savvy version of Windows for the mass market. (Daniel noting it's unlikely to happen by Windows 7 in 2010, but perhaps Windows 8 in 2013 will deliver what Mac OS X Leopard did in 2007.)

Apple’s solution? [list][*]Run whatever binary is appropriate to the existing hardware (processor architecture). This has allowed Apple to support both PowerPC & Intel hardware across a user base of ~20-27 million Mac OS X users since 2006. [*]Support the the conventions of running either 32 or 64-bit code.
- If Leopard runs on a 32-bit Mac (e.g. 1st generation Core Duo machines) it runs the 32-bit binary.
- If Leopard runs on a 64-bit Mac (e.g. the latest Core 2 Duos) it runs the 64-bit binary. [*] Universal Binary approach allows it to ship one edition of Leopard and will similarly enable one edition of Snow Leopard. In order to take advantage of 64-bit processor features, apps need to package a 64-bit version of their executable into their Universal Binary.

The majority of Leopard system apps delivered by Apple are 32-bit. (64-bit apps shipping with Leopard being Xcode & Chess only apparently (and httpd (Apache’s daemon)). A move to 64-bit apps helps with apps that need to access large memory spaces e.g. apps working with large files or data sets. Moving to a 64-bit binary can make some apps run slower, as they have more memory to manage. Overall, shifting the entire OS to 64-bits has created an system wide performance. So, all system apps in Snow Leopard will be 32+64 Universal Binaries, unlocking more of the latent performance available in modern 64-bit Macs.

(Tuning these apps to perform as well on 64-bit G5 PowerPC Macs would be a large investment for little upside, benefiting a relatively small number of G5 owners at the expense of diverting resources from optimizing the performance of today’s much faster 64-bit Intel processors. That helps to explain why Snow Leopard is dropping PowerPC support.)

Snow Leopard is also moving to a 64-bit kernel (from Leopard's 32-bit kernel), an essential step in supporting more than 32GB of RAM. Developers will therefore need to deliver 32+64 bit versions of all their kernel extensions and device drivers. All plugins will also need to provide 64-bit support as well, including printer drivers. It'll be interesting to compare how the creation and introduction of these new drivers are managed in comparison to Vista...

By Mid 2009, when Snow Leopard is estimated to come out, Apple will have sold ~8 million more Macs, most of which will benefit from 64-bit software support in Snow Leopard. With 15 million Intel Macs already sold, that would create a ~23 million Intel Mac installed base for which Snow Leopard would be most relevant.
In comparison:
- At the release of Tiger, there was a 16 million Mas in the entire installed base.
- At the release of Leopard, there was a 22 million Mac installed base.
- Currently there is a 27 million Mac installed base, 12 million of which are PowerPC.

Thus the proportion of 64-bit Macs is quickly outnumbering the active number of PowerPC models. Omni Software reports that 83.5% of its customers are actively updating their software from Intel Macs, compared to just 16.5% on PowerPC Macs. A hint as to why Apple can drop PowerPC support in Snow Leopard. Snow Leopard is for where the state of play is in Mid 2009, and where the puck will be after then.

64-bit models - Apple's LP64 vs Window's LLP64
Windows’ 64-bit development model is based on LLP64, which is really a 32-bit model that uses 64-bit addresses.
Apple's 64-bit LP64 model is not only more broadly compatible but is also more powerful.

Why did Microsoft chose the LLP64 model? Microsoft essentially hoped to add 64-bit pointers to allow apps to access more RAM while retaining 32-bit integral types for compatibility with all of the legacy operating system constructs in Windows that made assumptions about 32-bit code. However, Microsoft should have been fully aware that 64-bit computing was coming a decade ago when it was doing its work on porting NT to the 64-bit Alpha and again in its efforts to port Windows to the Itanium IA-64. Why the great compromise of tacking on partial 64-bit support as an afterthought today?

Whatever the reasons, Microsoft's choice of using "an oddball, legacy-limited version of 64-bit computing" whether the "result of malice or just plain incompetence" is opposite to Apple's choice.

Apple's 64-bit model LP64 seen in Leopard and Snow Leopard use the same as the 64-bit versions of Linux, SGI IRIX, and other commercial distributions of Unix. That ensures broad compatibility with the 64-bit applications and libraries already available. Kinda useful.

In "64-Bit Programming Models" (here, representatives of groups with expertise in 64-bit computing (including Digital, Hewlett-Packard, IBM, Intel, Novell, NCR, the original Santa Cruz Operation, Sunsoft, & X/Open) delivered a joint explanation of why LP64 is a better model for the future of computing than the LLP64 model Microsoft chose.

The short answer Daniel gives: LP64

• Supports easier porting of existing code
• Supports interoperability between 32 and 64-bit computing environments
• Has industry standard compliance for cross-platform interoperability
• Has better performance
• Gives a smoother transition from existing systems.

The upshot? Microsoft and also developers trying to work with Vista, & Windows 7 will struggle with the transition to 64-bits. Apple will be furthering its lead in deploying 64-bit computing to mainstream consumers in a highly interoperable, no compromise strategy that can backwardly support existing 32-bit hardware. For end users, Snow Leopard will simply make everything faster when running on the latest 64-bit hardware. Apple is hiding a lot of planning and work under the guide that Snow Leopard is just about taking stock and refining Mac OS X.

With this in mind, the 3rd myth is easier to unravel.
A message seems to be that Apple has a superior design for 64-bit. However, they have yet to create 64-bit apps.

[Northy]

Myths of Snow Leopard 3: Mac Sidelined for iPhone June 19th, 2008

It's not like the iPhone is getting much attention, is it? With limited comments on Snow Leopard (due mid 2009) there is a myth that Apple is de-emphasizing the Mac as it focuses attention on the iPhone.

Snow Leopard is marketed currently as “taking a break” from adding major new marketing features. But this is just a supposed lack of new features (maybe just a good case of unde sell, over deliver). "Software sells systems" ...

Daniel's angle? "Apple postponed Leopard’s release on the Mac in order to prepare for the iPhone debut, not because it decided Mac sales weren’t important, but because Mac sales were through the roof and didn’t need Leopard to accelerate them."

In comparison, look at bad-selling, over-selling and fraudulent mis-selling of Vista, e.g. the "Vista Capable" PC problem, an ongoing legal case currently, or XP sales going down as Vista (you can "buy" Vista, and then downgrade" to XP).

Apple did have record unit sales in 2007. And Apple doesn't make too much money on OS sales, it makes money on systems, hardware. The iPhone might well have been rushed out, and needed more attention, but that hasn't taken Apple's long term focus off the Mac side of Apple either. With the iPhone released, Leopard sold well, and was as the author of the article said, "fashionably late".

iPods and iPhones have helped sales of other Apple products, helped finance retail store rollout, and widened the potential audience for the Mac. They have driven buyers to the iTunes Store, and to Safari. iPod sales are still high, despite the addition of millions of new iPhones. No need to talk about cannabilism of sales.

Also, the iPhone has also been reaching out to persuade Windows users to consider the Mac platform. Daniel didn't provide hard numbers, but i'd imagine they aren't too hard to come by - the number of people converted to Mac is growing, and the hardware sales are on decent profit margin devices (e.g. the cornering of the >$1,000 pound laptop market, the mp3 player market)

Another effect is that as consumers and execs become iPhone users, this increases the audience for the upcoming App Store, which in turn feeds into the appeal for developers to work on apps for iPhone, giving them a taste of Obj C, and Apple’s Cocoa development tools.

The R&D from the iPhone, iPod and iMac have all been "cross pollinating" in terms of technology as well. Knowledge can be passed between the device areas, and integration can be created.

Apple's development of an entirely new interface paradigm for the iPhone OS, may well yet feedback to OS X. I'd think that the iPhone’s UIKit, and SDK will benefit the Mac OS X AppKit (e.g. adding the modern convention of properties as a way to simplify the class interfaces for the iPhone, and then adding properties to the desktop AppKit in Leopard.)

Daniel comments that QuickTime X (on Snow Leopard) is another example of repurposing code retooled for the iPhone to provide a highly efficient media playback. The extensive work on developing push support for Exchange Server on the iPhone, will be included in other ways, as Exchange support baked into Snow Leopard. MobileMe's mayb well be helpful in terms of research done/tools etc. for Snow Leopard Server’s push services.

Apple’s new Push Notification Service, allows iPhone & iPod touch users to set up server side notification alerts that don’t require any mobile applications to stay running in the background. Along with Bonjour discovery, PNS will keep iPhones wirelessly connected in all sorts of sophisticated ways that third party developers imagine in their applications.

Why couldn't this be used in Snow Leopard too? The point is, the technologies Apple is working on is flowing back and forth within the business it seems, helping not only share and build upon existing ideas, but also create combined value that is greater than the sum of its parts.

We'll have to wait for Snow Leopard. But seeing as it's slated to arrive before Vista's successor, and the fact that Apple at any time will be able to drop more hints to the consumer, and beta versions to developers, no-one should be worried Apple is forgetting about the Mac platform. Apple's aim is to do a few things, well.

[Northy]

Myths of Snow Leopard 4: Exchange is the Only New Feature! June 20th, 2008

Myths of Snow Leopard 4: Exchange is the Only New Feature!

Well, to start, we can just look at the Snow Leopard and Snow Leopard Server pages on Apple.com, and see what is publicly listed as features...

It helps Apple keep its work under the radar for a bit longer, and simplifies current marketing. Apple it seems has several reasons to promote the idea of "no new features", whilst promising overall improvements in how Mac OS X works under the hood (in a kind of "don't tell me how it works, just show it works" way).

Apple has the opportunity to improve its code through:
- code refactoring (Wiki definition: Code refactoring is the process of changing a computer program's code to make it amenable to change, improve its readability, or simplify its structure, while preserving its existing functionality. - Martin Fowler has apparently written in depth about refactoring)
- Beyond code refactoring in it's strictest sense, optimising the code
- adding new features

From the sounds of Quicktime X, Apple will be doing a mix of things. It has the opportunity to make 64-bit versions of apps, optimise the apps, add new features, and also pare the app size down).

(Aside in the article: Bill Gates was a big fan of "new" rather than "better" as can be seen by quotes from him- in an interview with Focus magazine in 1995, he explained why his company cared more about adding new features than refactoring code to fix bugs:

“The reason we come up with new versions is not to fix bugs,” Gates said. “It’s absolutely not. It’s the stupidest reason to buy a new version I ever heard. When we do a new version we put in lots of new things that people are asking for. And so, in no sense, is stability a reason to move to a new version. It’s never a reason.”

Ouch. New features were easier to sell than the concept of good software, so Microsoft took the low road. Touting features, keeping schtum about any lack of improvements under the hood.

Consumers perception is part of the problem - Consumers happily pay for hardware, but hate having to buy software. "They are well aware that the hardware they buy will soon be replaced by a faster model with more RAM at perhaps a lower price, but when it comes to software, every new release that “only” fixes bugs is regarded as something that “should have been” offered for free." It is also typically much harder to track down and eliminate bugs than to simply tack on more new features.

It is possible to sell quality to the consumer though, and in part, Apple can do this by not relying on OS sales for money, and not bothering too much about piracy of the OS - as Apple has a tight reign on the hardware. Apple will be able to inform users as to how Snow Leopard will be a better quality product, and show the doubting Thomas's the proof of the pudding - they'll be able to go into a store and see the improvements, and hear about them in reviews.

Apple has the luxury of doing such things, as it isn’t facing an immediate need to out-feature Windows Vista. The company has announced that Snow Leopard will involve a lot of code refactoring to tighten up performance, improve reliability, and slim down disk consumption. The only new feature, according to Apple, will be new support for push messaging with Exchange Server. That isn’t exactly accurate however.

In some respects, many of the new features in Snow Leopard can be regarded as a form of code refactoring because they will only improve how things work, rather than adding extensive new features. But there will also be a lot of new features that are just plain new.

Apple will be hard at work driving home the point that the "just works" feeling on the iPhones and iPods also extends to the experience when using Macs running OS X 10.5/6.

Why would someone want an upgrade, for something that works reasonably well? Well, with Snow Leopard, the new OS will be able to potentially show a decent performance benefit solely from the OS change, without needing any hardware updates, whilst also showing the performance bar of it's hardware as being raised significantly. Makes a change from people actually paying more, to "downgrade" to XP...

Through iLife 09 and other applications, Apple can bring in many more features which will link in well with iPods, iPhones, and Macs.


Myths of Snow Leopard 5: No Carbon! June 24th, 2008

Is Apple killing Carbon so all apps will be Cocoa only? Not exactly.


Carbon and Cocoa both compete and complement. Should it be ripped out, or slowly faded out? By shifting Carbon out of the frame, Apple can deliver a cohesive, consistent, and potentially more stable user experience while focusing its development efforts around a single strategy.

Currently, Carbon apps include iTunes, Final Cut Pro, Photoshop, and huge assortment of other important apps. Many apps are a mix of both. Whilst pure Cocoa apps can offer a more consistent user interface using less code, and benefit from other features - i.e. represent better technology, the transition from Carbon to Cocoa isn't an overnight one.


Cocoa is the modern incarnation of the object-oriented NeXTSTEP Objective-C (Obj-C) frameworks. Carbon is the extension of the classic Mac OS Toolbox; it was developed by Apple in order to pacify the complaints of existing Mac OS software authors during the development of Mac OS X after they rejected the move to Rhapsody, which would have essentially shifted Mac development to Cocoa in one great leap forward.

It not being feasible to convince developers circa 1997 to write all their software over largely from scratch using a new approach and tools that demanded a significant investment in mastering new concepts. Also, NeXTSTEP’s desktop development tools and frameworks had been sitting in cold storage from around 1994 through 1997 as NeXT worked to repurpose its core technologies into developing web server applications in WebObjects - so developers would have had a hard time using those tools from a cold start.

Apple needed to overhaul and modernize NeXT’s frameworks just as it needed to bring NEXTSTEP’s core OS foundation up to date with the latest software technology that had been delivered by the BSD development community over that period.

Existing Mac developers obligated Apple to spend much of its efforts getting Carbon up to speed first before prioritizing updates to the new Cocoa frameworks. A large amount of functional overlap between the two APIs resulted in a hybrid model where most of the shared foundational core of Mac OS X was written in Carbon-like C/C++ libraries, and exposed as modern, object-oriented APIs using a layer of Cocoa frosting.

In addition to software that had originated on the classic Mac OS and had been ported to native Carbon libraries, Mac OS X can also run POSIX software developed for Unix or Linux. Some of that software has an X Window System user interface (aka X11), which looks rather ugly and out of place on the Mac desktop, but can run just as it does on Linux thanks to integrated X11 support.

Unix software without any GUI can be given one using Cocoa. That includes huge libraries of highly regarded code from OpenGL routines to the GNU FFmpeg media decoding libraries to BSD firewalls. When Apple developed Safari, it used an off-the-shelf, open source HTML rendering engine from KDE to produce WebKit, which it then wrapped in a Cocoa interface to deliver Safari as a Mac application. That modular design has enabled third parties to port WebKit to Windows, Linux, and even Nokia’s smartphones.

[Northy]

Apple has also hinted at technology that would allow developers to access Windows DLLs to rapidly port device drivers or other specialized software to the Mac with little effort. The ability to take foreign software, whether open or proprietary, for use in creating native Mac OS X apps offers a look at how Carbon apps can migrate their user interfaces to Cocoa, resulting in user interface consistency and other benefits for users while resulting in less code for developers to maintain.

Apple last year announced it would only be implementing a 64-bit Cocoa architecture, and not implementing a 64-bit Carbon architecture.
Developers who need a 64-bit user interface will need to use Cocoa. This line in the sand enables Apple to focus its resources on developing a single object-oriented user interface API for the 64-bit future. Developers such as Adobe and Microsoft will need to either stay in the past or move decisively into the future (See Adobe's CS4 suite, Microsoft Office etc.).

However, Apple plans to support and maintain the 32-bit Carbon Human Interface Toolbox well into the future, although it will not be adding any significant new features to those APIs. Snow Leopard will lead Carbon developers to Cocoa with carrots rather than just sticks - HICocoaView enables Carbon apps to add Cocoa features as an incremental step; Carbon apps will be required to adopt a Cocoa user interface entirely, and whilst doing so, Apple will encourage deveopers to consider adopting the Cocoa frameworks for other parts of their apps as well.

An example is Apple's own Finder in Leopard - largely a Carbon app, but it makes use of HICocoaView to embed Cocoa NSViews, such as when displaying CoverFlow. For Snow Leopard, the Finder would therefore apparently require building the entire user interface in Cocoa. Apple has indicated that will be happening in Snow Leopard. Therefore, the company is well aware of the effort needed to move to Cocoa, and is starting to lead by example.

There is still a murky area, neither Carbon, nor Cocoa it seems (Core Video, Quartz), amongst other things. So there is still work to be done on these areas too. Slowly raising the bar on what amount and parts of an app should be Cocoa at minimum, Apple is clearly pushing developers toward Cocoa. Meanwhile, Apple continues to support legacy code. Office 2004 was written as a PowerPC CFM app, which requires Apple to host it on top of CFMApp, which itself runs on top of Rosetta on Intel Macs. It will continue to work as expected in Snow Leopard. Anyone who likes to say that Apple “doesn’t support legacy” hasn’t looked too hard at what Apple has done to jump through hoops so Adobe and Microsoft wouldn’t have to bring their old code into the modern world.

[Northy]

Myths of Snow Leopard 6: Apple is Out of Ideas! June 27th, 2008

An article touching on aspect Daniel at Roughlydrafted.com's already talked about in previous articles.

Snow Leopard doesn't indicate Apple is out of ideas for new applications and features - it indicates it's not willing to promote and advertise features and applications it doesn't want to talk about yet.

Marketing. Jobs and Apple aren't giving away their grand views of the road ahead, unlike Microsoft.

Another aspect is the strange notion that having a list of new applications and features is better (maybe a hangup from drinking Microsoft Kool Aid) rather than wanting features and applications only on merit - only if they're useful, and worthy enough to be included.

From what can be read between the lines of the known confirmed Snow Leopard information thus far, Daniel makes the assertion that Apple has laid out a cohesive strategy for strengthening Snow Leopard’s performance and its suitability for running the next generation of software on the next generation of hardware. As Daniel says, clearly "Apple is being lead by engineers, not just clever marketers."

OpenCL, Grand Central, LLVM, ZFS, CUPS, Quicktime X to name the advances that have been published -marketers can get to work after the work is done...

Apple's continuing investment in enabling technologies seems to be going to pay off again, when Snow Leopard rolls out. "When viewed within context of technology cross pollination with the iPhone, Apple’s Pro Apps, its consumer app suites, and its expanding role in online subscription software, it’s clear Apple is not running short of ideas. As for Snow Leopard, there’s still a lot to be revealed."


From the comments:
"OS X made the iPhone possible. The iPhone feeds OS X both financially and in feature demands. OS X matures further still as a desktop and a handheld platform par excellence. What’s not to like!"

A line of thought coming from the comments and other articles, is the possibility of Apple championing portability of the OS with Snow Leopard - e..g to have the OS on an SSD to give performance gains. The possibility of 10.6 being a (mini) "code review".

"It’s nice to hear, read, and see Apple increasingly described as an engineering firm within the “Halo Effect” realm. The end-user products naturally garner the deserved accolades — design aesthetic, ease of use, ergonomic attention, stability, and the intangible sensory experience — from its consumers."


Myths of Snow Leopard 7: Free?! July 1st, 2008

Why $129?

Selling Snow Leopard for Less Would Make Selling 10.7 at Regular Price Rather Difficult.
- If Apple sold Snow Leopard at a steep discount as an apology for not adding fluff features, it would deflate the perceived value of Apple’s operating system software.
- The main group to benefit from Snow Leopard will be owners of recent, 64-bit Macs who are likely to willingly pay full price to fully unlock the power of their existing hardware.
- Everyone else is just as likely to just wait for Snow Leopard until they buy their next new Mac and are able to take full advantage of its advances.
- Keeping the retail price of Snow Leopard unchanged wouldn’t help set any new sales records for a reference release of Mac OS X, but would help induce sales of new Macs, because buyers would think of new systems as including an additional $129 of software for free.

Apple doesn't make much money from OS software sales. Apple, unlike Microsoft
- does not sell bundled licensing to other hardware makers.
- is forced to actually deliver a product that is good enough to convince the market to go out of its way to choose to buy it.
- can't coast on a software licensing model like Microsoft’s (which has allowed MS to continue making money on sales of Windows XP for years despite minimal feature enhancements over the last half decade.)

- has to work harder to add value and differentiation to the company’s OS software.
- has to work hard to trumpet the retail interest in Mac OS X at every release

4Q 2007 - Apple brought in $9.6 billion, almost entirely from Mac and iPod hardware. It "only" earned $170 million from sales of Leopard that quarter.
1Q 2008 - Retail box OS sales quickly dropped down to $40 million.

There are no compelling reasons to lower the price of Snow Leopard - Apple doesn't need to induce volume sales to broaden its installed base, and has no direct rival that it has to compete against.

Apple would rather you buy a new computer, than give away Mac OS X. Most of Snow Leopard’s features announced so far exploit the potential of new and forthcoming hardware. The primary purpose of Mac OS X is to distinguish Mac hardware from PCs. Selling it at retail only helps Apple pull in some extra revenue from users who are not ready to buy new hardware.

The alternatives to buying a Mac OS X upgrade at retail?
- Not upgrading at all
- Buying a new Mac
- pirating a copy.

It now makes no sense for Apple to give away its development work - Mac users who aren’t going to upgrade unless the software is nearly free are not worth Apple’s attention. They are likely to just steal it anyway.

We see it with Microsoft, and it happens with Apple. But Apple doesn't go all WGA on us. Apple doesn't really police Mac OS X licensing with DRM, activation procedures, or spyware because it only sells to premium customers rather than trying to tax the entire PC market.

The majority of Microsoft’s customers are thieves that would only pay for Windows if they had no choice. A fair percentage of Apple's customers probably use a non-licenced version of the OS too.


The key benefit Apple has marketed in Snow Leopard so far is Exchange Server support. But there will be more benefits to come, and any current Mac Pro, or MBP, or any version prior to Snow Leopard will be definitely able to receive a decent performance boost from it, if Apple's enhancements bear fruit.

Exchange - Microsoft charges Mac users $500 (a whopping $350 premium over the regular version) for the version of Office 2008 that includes support for Exchange. Why? Microsoft knows that the organizations who have chosen Exchange are not price sensitive. Those customers already pay absurd licensing costs for its server and client access licenses, so they are likely to also shell out crazy amounts of money for a slightly less awful version of the Entourage Mac email client.

If Microsoft can get away with charging businesses and education users $500 for Exchange support in Office 2008, Apple will have no problem selling those same customers an overhauled operating system that adds Exchange support for Mail, iCal and Address Book for just $129.

What about home users who have no need for Exchange? Outside of those that want to buy every new release, that segment of the market is unlikely to buy Snow Leopard. We know this because they largely didn’t pay for Leopard.

Who Bought Leopard?
Only a minority of Mac users will actually upgrade at retail. Then a number will upgrade via a nbon-licenced copy, and a large number will upgrade via hardware purchases.

(Consider the Leopard launch. Apple’s $170 million in Leopard revenues reported in its debut quarter is only enough to buy 1.3 million copies at retail price. A third of retail packages were family pack versions, meaning Apple actually sold fewer boxes than that at full price. Of course, lots of those retail boxes where sold to retailers at lower wholesale prices and then marked up by the retailer.)

Apple reported selling 2 million copies of Leopard in the first weekend. It did not continue to report how many additional copies it sold after that initial figure because Apple didn’t want to highlight the fact that most of the people who bought Mac OS X in the quarter did so over the first weekend. That weekend figure also probably included shipments to stores, further padding the number with marketing muscle.

[Northy]

More recently, the company indicated that of the 27.5 million installed base of Mac OS X users, 37% are running Leopard. That would be 10.1 million Macs running Leopard. Apple has sold roughly 4.6 million new Macs in the last three