Moo Bunny Amiga - thread "The real reason for the Apple switch to Intel: $$$"
special effects (2007)

The real reason for the Apple switch to Intel: $$$

Direct link to this message

The real reason for the Apple switch to Intel: $$$

minator. @@@.

(78.105.202.0) [78.105.202.0]

and what could have been...

http://www.9to5mac.com/Apple-IBM-Leopard-Cell-processor


It's interesting to speculate what the machine I'm using (MacBook) would be like with a Cell rather then the Core2Duo. A lot of general purpose day to day stuff would likely be slower - that's the sort of stuff Core2Duo is good at. However the only really processor intensive stuff I do is editing high MP images and 1080p video - the Core2Duo chokes on the images and it can't play 1080p smoothly from a movie editor. A Cell on the other hand wouldn't have even broken a sweat.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

GK. malibannnospam@t-online.de.

(217.113.178.29) 217.113.178.29
On Wednesday, Jun 24, 2009, minator wrote:
>
> and what could have been...
>
> http://www.9to5mac.com/Apple-IBM-Leopard-Cell-processor
>
>
> It's interesting to speculate what the machine I'm using (MacBook) would be
> like with a Cell rather then the Core2Duo. A lot of general purpose day to
> day stuff would likely be slower - that's the sort of stuff Core2Duo is
> good at. However the only really processor intensive stuff I do is editing
> high MP images and 1080p video - the Core2Duo chokes on the images and it
> can't play 1080p smoothly from a movie editor. A Cell on the other hand
> wouldn't have even broken a sweat.

The CELL is a multicore CPU with one 64Bit PowerPC core ( PPE ) and 8 SPE cores. I don't know if the core CPU is fully compatible to the PPC970MP used in G5 Macs but the SPEs are not. Without using the SPEs the CELL is a 'normal' singlecore CPU. The PPC970MP is a dualcore CPU and both cores could run PPC code with 2,5GHz so a CELL running OSX with 3.2GHz might be slower than a PPC970MP.

cu

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(61.29.36.98) dsl-61-29-36-98.request.com.au
On Wednesday, Jun 24, 2009, minator wrote:
>
> and what could have been...
>
> http://www.9to5mac.com/Apple-IBM-Leopard-Cell-processor
>
>
> It's interesting to speculate what the machine I'm using (MacBook) would be
> like with a Cell rather then the Core2Duo. A lot of general purpose day to
> day stuff would likely be slower - that's the sort of stuff Core2Duo is
> good at. However the only really processor intensive stuff I do is editing
> high MP images and 1080p video - the Core2Duo chokes on the images and it
> can't play 1080p smoothly from a movie editor. A Cell on the other hand
> wouldn't have even broken a sweat.
>
The main point of MacOS X 10.6 Snow Leopard's release is running general purpose code on GpGPU(OpenCL).

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(61.29.36.98) dsl-61-29-36-98.request.com.au
On Wednesday, Jun 24, 2009, minator wrote:
>
> and what could have been...
>
> http://www.9to5mac.com/Apple-IBM-Leopard-Cell-processor
>
>
> It's interesting to speculate what the machine I'm using (MacBook) would be
> like with a Cell rather then the Core2Duo. A lot of general purpose day to
> day stuff would likely be slower - that's the sort of stuff Core2Duo is
> good at. However the only really processor intensive stuff I do is editing
> high MP images and 1080p video - the Core2Duo chokes on the images and it
> can't play 1080p smoothly from a movie editor. A Cell on the other hand
> wouldn't have even broken a sweat.
>
Fitting 80 watt CELL into laptop is not a wise idea. It seems this "ex-IBMer" doesn't know about modern X86 having RISCy cores.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

Anon. .

(75.18.203.39) adsl-75-18-203-39.dsl.pltn13.sbcglobal.net
On Friday, Jun 26, 2009, hammer wrote:
> The main point of MacOS X 10.6 Snow Leopard's release is running general
> purpose code on GpGPU(OpenCL).

No, hammer, that is not even remotely the main point of this release. Go look at their marketing material at www.apple.com/macosx . OpenCL isn't even mentioned on the front page, and receives equal billing with 4 other things on the 'new technology' page. And at least initially, OpenCL only supports GeForce 8600M or better, or Radeon 4850/4870. This means most Macs capable of running Snow Leopard won't even be able to do GPGPU with OpenCL.

BTW, neither OpenCL nor any other GPGPU system is actually capable of running truly "general purpose" code on a GPU. All GPGPU systems are limited to tiny snippets of code (compute kernels in OpenCL terminology) which have a lot of restrictions. OpenCL just hides a lot more of the esoteric aspects of coding for GPUs by allowing you to write your compute kernels in C.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(60.241.225.101) 60-241-225-101.static.tpgi.com.au
On Friday, Jun 26, 2009, Anon wrote:
> On Friday, Jun 26, 2009, hammer wrote:
> > The main point of MacOS X 10.6 Snow Leopard's release is running general
> > purpose code on GpGPU(OpenCL).
>
> No, hammer, that is not even remotely the main point of this release.

OpenCL is a middleware layer. The focus would be on the higher layers.

From http://www.apple.com/macosx/technology/
64bit
Grand Central Dispatch
OpenCL
QuickTime X
Exchange Support

> Go
> look at their marketing material at www.apple.com/macosx . OpenCL isn't
> even mentioned on the front page, and receives equal billing with 4 other
> things on the 'new technology' page.

Make that 5.


> And at least initially, OpenCL only
> supports GeForce 8600M or better, or Radeon 4850/4870. This means most
> Macs capable of running Snow Leopard won't even be able to do GPGPU with
> OpenCL.

Early 2009 mainstream MacBooks includes Geforce 9400M.

NVIDIA Geforce 8600M GT (32SP), GeForce 8800 GT, GeForce 8800 GTS, Geforce 9400M (16SP), GeForce 9600M GT(32SP), GeForce GT 120(32SP), GeForce GT 130(46 SP).

ATI Radeon 4850, Radeon 4870


> BTW, neither OpenCL nor any other GPGPU system is actually capable of
> running truly "general purpose" code on a GPU. All GPGPU systems are
> limited to tiny snippets of code (compute kernels in OpenCL terminology)

Same thing with SPEs.

> which have a lot of restrictions. OpenCL just hides a lot more of the
> esoteric aspects of coding for GPUs by allowing you to write your compute
> kernels in C.

"They need only rewrite the most performance-intensive parts of their application in OpenCL C" - Apple

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(60.241.225.101) 60-241-225-101.static.tpgi.com.au
On Friday, Jun 26, 2009, hammer wrote:
> On Friday, Jun 26, 2009, Anon wrote:
> > On Friday, Jun 26, 2009, hammer wrote:
> > > The main point of MacOS X 10.6 Snow Leopard's release is running general
> > > purpose code on GpGPU(OpenCL).
> >
> > No, hammer, that is not even remotely the main point of this release.
>
> OpenCL is a middleware layer. The focus would be on the higher layers.
>
> From http://www.apple.com/macosx/technology/
> 64bit
> Grand Central Dispatch
> OpenCL
> QuickTime X
> Exchange Support
>
> > Go
> > look at their marketing material at www.apple.com/macosx . OpenCL isn't
> > even mentioned on the front page, and receives equal billing with 4 other
> > things on the 'new technology' page.
>
> Make that 5.
>
>
> > And at least initially, OpenCL only
> > supports GeForce 8600M or better, or Radeon 4850/4870. This means most
> > Macs capable of running Snow Leopard won't even be able to do GPGPU with
> > OpenCL.
>
> Early 2009 mainstream MacBooks includes Geforce 9400M.
>
> NVIDIA Geforce 8600M GT (32SP), GeForce 8800 GT, GeForce 8800 GTS, Geforce
> 9400M (16SP), GeForce 9600M GT(32SP), GeForce GT 120(32SP), GeForce GT
> 130(46 SP).
>
GeForce GT 130(48 SP).

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

megol. .@..

(90.233.178.99) [90.233.178.99]
On Friday, Jun 26, 2009, hammer wrote:
>
> OpenCL is a middleware layer. The focus would be on the higher layers.
>
Bullshit. What "higher layers" that utilize OpenCL are you refering to? The fantastic Exchange support perhaps?

> Make that 5.
>
Read and understand before writing please. Make that four _other_ like Mr. Anon correctly pointed out. One of which is 64 bit support, truly a technical marvel!

> Early 2009 mainstream MacBooks includes Geforce 9400M.
>
> NVIDIA Geforce 8600M GT (32SP), GeForce 8800 GT, GeForce 8800 GTS, Geforce
> 9400M (16SP), GeForce 9600M GT(32SP), GeForce GT 120(32SP), GeForce GT
> 130(46 SP).
>
> ATI Radeon 4850, Radeon 4870
>
Again just as the OP pointed out. Why write thing with no relevancy?

> Same thing with SPEs.
>
As Mr. Anon haven't claimed that _or_ that SPEs are the main reason behind the newest OS X release I fail to see the relevance.

> "They need only rewrite the most performance-intensive parts of their
> application in OpenCL C" - Apple
>
Compare this to the holy Altivec future Apple promised us. OpenCL (or one of the alternatives) is good for a limited number of tasks not something that accelerates the system as a whole. Most mainstream code is branch or cache limited and not compute limited.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

minator. @@@.

(78.105.202.0) [78.105.202.0]
On Thursday, Jun 25, 2009, GK wrote:

>
> The CELL is a multicore CPU with one 64Bit PowerPC core ( PPE ) and 8 SPE
> cores. I don't know if the core CPU is fully compatible to the PPC970MP
> used in G5 Macs but the SPEs are not.

> Without using the SPEs the CELL is a 'normal' singlecore CPU.

Why on earth would you not use the SPEs?

> The PPC970MP is a dualcore CPU and both cores
> could run PPC code with 2,5GHz so a CELL running OSX with 3.2GHz might be
> slower than a PPC970MP.

Actually I'd expect they would have started with Cell as a co-processor then replaced the PPE with a 970.

Hammer wrote:
>Fitting 80 watt CELL into laptop is not a wise idea.

Huh? The 90nm Cell was replaced within a year and they're making sub 40W 45nm parts now.

Megol / Hammer wrote:

>> "They need only rewrite the most performance-intensive parts of their
>> application in OpenCL C" - Apple
>>
>Compare this to the holy Altivec future Apple promised us.

My understanding is Apple used AltiVec quite extensively.

>OpenCL (or one of the alternatives)
>is good for a limited number of tasks not something that accelerates the system as a whole.
>Most mainstream code is branch or cache limited and not compute limited.

Actual processing tends to be concentrated in small loops ran over and over again. You may have reams of code but very little of it is actually used intensively. The general idea is you target these sorts of things to GPUs.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

megol. .@..

(90.233.188.117) [90.233.188.117]
On Saturday, Jun 27, 2009, minator wrote:
>
> <snip>
> Megol / Hammer wrote:
>
> >> "They need only rewrite the most performance-intensive parts of
> their
> >> application in OpenCL C" - Apple
> >>
> >Compare this to the holy Altivec future Apple promised us.
>
> My understanding is Apple used AltiVec quite extensively.
>
Yes they did. And most of the bigger software companies that was active on Macs optimized using Altivec wherever possible. Still the PPC was beaten in most benchmarks by x86 processors running plain non-vectorized code. General code is not parallelizable enough to get good performance from either vectorization or multi-threading/vectorization (as needed for GPU acceleration).

> >OpenCL (or one of the alternatives)
> >is good for a limited number of tasks not something that accelerates
> the system as a whole.
> >Most mainstream code is branch or cache limited and not compute
> limited.
>
> Actual processing tends to be concentrated in small loops ran over and over
> again. You may have reams of code but very little of it is actually used
> intensively. The general idea is you target these sorts of things to
> GPUs.
>
Yes that is true. But still: most code is not suitable to convert to OpenCL (or alternatives) as the GPU still isn't a general purpose processor. Wordprocessors/browsers/email programs/databases*/compilers** etc are not parallelizable enough to warrant the (sometimes extreme) redesigns needed for running on GPUs or even multithreaded. Some tasks are: Video encoding, audio encoding, protein folding, ray tracing etc.
(* by doing complete rewrites a simple database could get better performance. Doing it with ACID support is much harder...)
(** some compiler passes could theoretically get good performance when running parallel code. No real compiler tries to do this yet AFAIK)
IMHO of course.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

GK. malibannnospam@t-online.de.

(79.216.251.115) [79.216.251.115]
On Saturday, Jun 27, 2009, minator wrote:
> On Thursday, Jun 25, 2009, GK wrote:
>
> >
> > The CELL is a multicore CPU with one 64Bit PowerPC core ( PPE ) and 8 SPE
> > cores. I don't know if the core CPU is fully compatible to the PPC970MP
> > used in G5 Macs but the SPEs are not.
>
> > Without using the SPEs the CELL is a 'normal' singlecore CPU.
>
> Why on earth would you not use the SPEs?

SPEs are ( AFAIK ) not binary compatible to the PPC, normal tasks, threads or processes could not use them, they are only useful for small specialized code parts.

> > The PPC970MP is a dualcore CPU and both cores
> > could run PPC code with 2,5GHz so a CELL running OSX with 3.2GHz might be
> > slower than a PPC970MP.
>
> Actually I'd expect they would have started with Cell as a co-processor
> then replaced the PPE with a 970.

That doesn't sound like a good idea, using different CPUs as coprozessors was done on many platforms but it was not really sucessfull.
( Examples: x86 PC with i960 or 68k, Amiga with x86 or PPC )

> >Compare this to the holy Altivec future Apple promised us.
> My understanding is Apple used AltiVec quite extensively.

I doesn't know if the Altivec was used directly or via API calls that would redirect the tasks to non Altivec programs when running on a G3.

It is a problem to handle the coprozessor support:

1. Application developers could compile a program or library to use a specific co processor. ( Example: different programm versions of Amiga prorams )

2. Application developers could use system APIs that use a coprocessor if available. ( Example: if you use Amiga Floting Point Libraries, AFAIK also done on the Mac )

3. Application developers jumps conditioally to the code part with or without co processor ( Example: IBM MMX extension )


> Actual processing tends to be concentrated in small loops ran over and over
> again. You may have reams of code but very little of it is actually used
> intensively. The general idea is you target these sorts of things to
> GPUs.


And if someone uses a different GPU you have to compile the code for his GPU ... that doesn't sound like a good idea.

And there is still a comunication problem between the main CPU and the GPUs.

I have my doubts a special co processor, with own memory could be managed correctly by the main system that was designed to di SMP.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(61.29.36.98) dsl-61-29-36-98.request.com.au
On Saturday, Jun 27, 2009, megol wrote:
> On Saturday, Jun 27, 2009, minator wrote:
> >
> > <snip>
> > Megol / Hammer wrote:
> >
> > >> "They need only rewrite the most performance-intensive parts of
> > their
> > >> application in OpenCL C" - Apple
> > >>
> > >Compare this to the holy Altivec future Apple promised us.
> >
> > My understanding is Apple used AltiVec quite extensively.
> >
> Yes they did. And most of the bigger software companies that was active on
> Macs optimized using Altivec wherever possible. Still the PPC was beaten in
> most benchmarks by x86 processors running plain non-vectorized code.
> General code is not parallelizable enough to get good performance from
> either vectorization or multi-threading/vectorization (as needed for GPU
> acceleration).
>
> > >OpenCL (or one of the alternatives)
> > >is good for a limited number of tasks not something that accelerates
> > the system as a whole.
> > >Most mainstream code is branch or cache limited and not compute
> > limited.
> >
> > Actual processing tends to be concentrated in small loops ran over and over
> > again. You may have reams of code but very little of it is actually used
> > intensively. The general idea is you target these sorts of things to
> > GPUs.
> >
> Yes that is true. But still: most code is not suitable to convert to OpenCL
> (or alternatives) as the GPU still isn't a general purpose processor.
> Wordprocessors/browsers

Browsers's Flash FLV.

> /email programs/databases*

Database's bitmapped indexes.

> /compilers** etc are not
> parallelizable enough to warrant the (sometimes extreme) redesigns needed
> for running on GPUs or even multithreaded. Some tasks are: Video encoding,
> audio encoding, protein folding, ray tracing etc.

Anti-virus scanner e.g. Kaspersky SafeStream.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(61.29.36.98) dsl-61-29-36-98.request.com.au
On Friday, Jun 26, 2009, megol wrote:
> On Friday, Jun 26, 2009, hammer wrote:
> >
> > OpenCL is a middleware layer. The focus would be on the higher layers.
> >
> Bullshit. What "higher layers" that utilize OpenCL are you refering to? The
> fantastic Exchange support perhaps?
>
To quote Apple,

"OpenCL stands for Open Computing Language. It’s a C-based programming language with a structure that will be familiar to programmers, who can simply use Xcode developer tools to adapt their programs to work with OpenCL. They don’t have to completely rewrite applications to use OpenCL"

"For example, sophisticated financial modeling techniques can be incorporated into desktop accounting software and personal finance software. Media applications can perform complex, intensive operations with larger video and graphics files. Games can have more accurate physics simulations. And scientists and researchers can tackle far more challenging problems using their everyday Mac computers"

Most of which are already available in PC world e.g. Mathlab, PowerDirector 7, Adobe Photoshop CS4, SciFinance, Office FX(PowerPoint Direct3D plug-in), MAGIX Video Pro X, Adobe Acrobat 9, PhysX enabled games (e.g. Mirror's Edge, UT3, Sacred 2, 'etc'), LightStage/Otoy(RT Raytracing)

> > Make that 5.
> >
> Read and understand before writing please. Make that four _other_ like Mr.
> Anon correctly pointed out. One of which is 64 bit support, truly a
> technical marvel!

If you are going to point to http://www.apple.com/macosx/
Make sure the bullet points matches Apple's.

> > Early 2009 mainstream MacBooks includes Geforce 9400M.
> >
> > NVIDIA Geforce 8600M GT (32SP), GeForce 8800 GT, GeForce 8800 GTS, Geforce
> > 9400M (16SP), GeForce 9600M GT(32SP), GeForce GT 120(32SP), GeForce GT
> > 130(46 SP).
> >
> > ATI Radeon 4850, Radeon 4870
> >
> Again just as the OP pointed out. Why write thing with no relevancy?

The 2009 baseline MacBook includes 9400M.

> > Same thing with SPEs.
> >
> As Mr. Anon haven't claimed that _or_ that SPEs are the main reason behind
> the newest OS X release I fail to see the relevance.

The next MacOS release enables general applications to use GpGPU.

> > "They need only rewrite the most performance-intensive parts of their
> > application in OpenCL C" - Apple
> >
> Compare this to the holy Altivec future Apple promised us.

The competition has SSE.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(61.29.36.98) dsl-61-29-36-98.request.com.au
On Saturday, Jun 27, 2009, megol wrote:
> On Saturday, Jun 27, 2009, minator wrote:
> >
> > <snip>
> > Megol / Hammer wrote:
> >
> > >> "They need only rewrite the most performance-intensive parts of
> > their
> > >> application in OpenCL C" - Apple
> > >>
> > >Compare this to the holy Altivec future Apple promised us.
> >
> > My understanding is Apple used AltiVec quite extensively.
> >
> Yes they did. And most of the bigger software companies that was active on
> Macs optimized using Altivec wherever possible. Still the PPC was beaten in
> most benchmarks by x86 processors running plain non-vectorized code.

Athlons includes 3-way AGU/ALU and better bus e.g. EV6 or HTT.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(61.29.36.98) dsl-61-29-36-98.request.com.au
On Saturday, Jun 27, 2009, minator wrote:
> On Thursday, Jun 25, 2009, GK wrote:
>
> >
> > The CELL is a multicore CPU with one 64Bit PowerPC core ( PPE ) and 8 SPE
> > cores. I don't know if the core CPU is fully compatible to the PPC970MP
> > used in G5 Macs but the SPEs are not.
>
> > Without using the SPEs the CELL is a 'normal' singlecore CPU.
>
> Why on earth would you not use the SPEs?
>
> > The PPC970MP is a dualcore CPU and both cores
> > could run PPC code with 2,5GHz so a CELL running OSX with 3.2GHz might be
> > slower than a PPC970MP.
>
> Actually I'd expect they would have started with Cell as a co-processor
> then replaced the PPE with a 970.
>
> Hammer wrote:
> >Fitting 80 watt CELL into laptop is not a wise idea.
>
> Huh? The 90nm Cell was replaced within a year and they're making sub 40W
> 45nm parts now.

Have they reduced the power supply to 65-to-90 watts?

For comparison, Mobility Radeon HD 4650 consumes 25 watts e.g. under ASUS N81's MXM-II spec.

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

hammer. hammer@amiga.org.

(61.29.36.98) dsl-61-29-36-98.request.com.au
On Saturday, Jun 27, 2009, minator wrote:
> On Thursday, Jun 25, 2009, GK wrote:
>
> >
> > The CELL is a multicore CPU with one 64Bit PowerPC core ( PPE ) and 8 SPE
> > cores. I don't know if the core CPU is fully compatible to the PPC970MP
> > used in G5 Macs but the SPEs are not.
>
> > Without using the SPEs the CELL is a 'normal' singlecore CPU.
>
> Why on earth would you not use the SPEs?
>
> > The PPC970MP is a dualcore CPU and both cores
> > could run PPC code with 2,5GHz so a CELL running OSX with 3.2GHz might be
> > slower than a PPC970MP.
>
> Actually I'd expect they would have started with Cell as a co-processor
> then replaced the PPE with a 970.
>
> Hammer wrote:
> >Fitting 80 watt CELL into laptop is not a wise idea.
>
> Huh? The 90nm Cell was replaced within a year and they're making sub 40W
> 45nm parts now.
>
What PS3 model number would that be?

Model CECHPxx/CECHLxx/CECHMxx has Typical Power Consumption of 110 watts.
Model CECHAxx has Typical Power Consumption of 180W.

Via RMclock's numbers and while playing a game, my gaming laptop** consumes around 50 watts.

**Capable to match PS3/Xbox360 multi-platform games in detail and resolution e.g. Prototype, Fallout3, UT3, Mirror's Edge, 'etc'

Reply to this
Direct link to this message

Re: The real reason for the Apple switch to Intel: $$$

GK. malibannnospam@t-online.de.

(79.216.208.181) [79.216.208.181]
On Monday, Jun 29, 2009, hammer wrote:
> On Saturday, Jun 27, 2009, minator wrote:

> >
> > Hammer wrote:
> > >Fitting 80 watt CELL into laptop is not a wise idea.
> >
> > Huh? The 90nm Cell was replaced within a year and they're making sub 40W
> > 45nm parts now.
> >
> What PS3 model number would that be?
>
> Model CECHPxx/CECHLxx/CECHMxx has Typical Power Consumption of 110 watts.
> Model CECHAxx has Typical Power Consumption of 180W.

AFAIK the GPU in the PS3 consumed more Power than the 90nm CELL ...


Reply to this

Short list view

Previous thread - Next thread

Back to Moo Bunny

It is now: 11:57:21 GMT.
Moo Bunny brand Message Boards
[General Stuff (10:30:54)] [Space (Aug 2009)] [Anime (2007)]
[WalkingFish (Oct 3)] [Cabinet of Curiosities (2007)] [Portland (2008)]
Retro and Exotic Computing
[Atari ST (2007)] [WeirdMac (Jan 2009)]
Amiga Boards
[Amiga Section (11:37:14)]
John Shepard's Projects
[Rastport (Tuesday)] [Squid's Amiga Page & Rumor Mill (n/a)] [Art by John Shepard (Sep 2009)]