Mac FAQ Stack 3.62: Processors

The mind's eye?


G3 cards

Subject: Re: Do G3 upgrades require an extension?
From: Matthew Ruben mruben@dept.english.upenn.edu
Date: Sat, Apr 1, 2000 2:32 PM
Message-id: <mruben-E4D311.16320701042000@netnews.upenn.edU>

In article <gozer-7D8B29.11215601042000@supernews.intergate.ca>, Gordon Mulcaster <gozer@mac.com> wrote:

>All G3 cards require an extention to enable speculative processing.

No, they require an extension to _disable_ speculate processing. They require this only for machines that originally shipped with a 604 or 604e CPU, as those CPUs didn't do SP, and so those machines' motherboards didn't support it properly.

Also, NewerTech's cards incorporate a hardware fix for the SP issue, so they require no extension -- at least not for that purpose.

Other cards require extensions to work in the L2 cache slot (like ones made for Starmax clones, the 4400, and the 5400/5500/6400/6500 series).

All G3 cards also require some kind of code -- either an extension/init or a control panel -- to enable the backside cache and (in most cases) to configure the cache's speed relative to the CPU's speed.

As far as the 7500 goes, the extension is required. But more importantly, many 7500s have an unusual condition with G3 cards. Whereas many PCI PowerMacs require that the original L2 cache DIMM (the motherboard cache) be removed for stable operation with a G3 upgrade, some 7500s will not work reliably (or at all) with a G3 upgrade unless a motherboard cache DIMM is _present_. To add to the complication, some 7500s appear not to work with the cache in or out. The solution in these cases is to get a better quality, faster cache DIMM than the factory one and install it. This adds about $50-$100 to the cost of the G3 upgrade, but for these machines it's the only way.

Matt
*

Subject: Re: Do G3 upgrades require an extension?
From: Stephen spasticateN*SP4M@freeuk.com
Date: Tue, Apr 4, 2000 2:21 AM
Message-id: <spasticateN*SP4M-8D64DC.10211204042000@news.freeuk.net>

Not at all, my 7500 runs quite happliy with an XLR8 G3 upgrade and no L2 cache, a 256k L2 cache, originaly pulled from a 7200 didnt like running with the 7500s 50 mhz bus speed and so was long gone before i got the G3, which has worked perfectly since I got it.
As for the extension issue, this machine boots absolutely fine without any XLR8 extensions, although the backside cache is diabled (i cant remeber what the speculative proccessing is set to when i boot sans extensions) This slows things down a bit, but, seeing as I hardly ever do this and the machine is not very usable without any extensions. The extension itself has never cause any problems or conflicts for me. As it dosn't, so far as I can tell do anything particularly after its turned the backside cache on.

Stephen

***

Dual Processors

From: Gordon Mulcaster <gozer@mac.com>
Newsgroups: comp.sys.mac.misc,comp.sys.mac.apps,comp.sys.mac.hardware.misc
Subject: Re: Dual Processor Questions
Message-id: <gozer-65C866.21343416052001@clgrps10.telusplanet.net>
Date: Thu, 17 May 2001 04:34:52 GMT

In article <bsharvy-6BABDD.13334816052001@news.efn.org>,
Ben Sharvy <bsharvy@efn.org> wrote:

> I mean there are some dual-processor 604e machines. I don't remember
> if Apple made one,

Yes, Apple made the dual processor 9500-180MP and 9600-200MP.

> but Umax did. Will an app written to take advantage of those machines
> be able to take advantage a dual-G4 machine?

Yes, it's the same MP library.

> I don't know what SMP means.

Symetrical Multi-Processing. OS 7/8/9 is AMP (Asymetrical Multi-Processing), basically ity has a "main" CPU and a "slave" CPU. The OS and all apps run on the main CPU but can spawn tasks to the slave CPU. In a SMP system (such as OS X) any thread can be run on either CPU, in fact a thread can be switched between CPUs while it's running. SMP is a much better system as it allows the OS to balance the load, swapping the various threads between both (all) CPUs in an effort to balance the work being done by each CPU.

> I take it you mean that multiprocessing facilitates multi-tasking, so
> programs will run faster under MacOS without being specifically coded
> for dual-processors, IF there is more than one program running at a
> time. Is that right?

If a singely threaded app is run (such that it can't be split between two or more CPUs) and is using a lot of CPU time (say compressing a video) then the OS is able to move all the other threads to the other CPU(s). If the other tasks aren't tasking the remaining CPU(s) then they will (appear to) be running at "full speed".
*

From: Mark Haase <savar@email.com>
Newsgroups: comp.sys.mac.misc,comp.sys.mac.apps,comp.sys.mac.hardware.misc
Subject: Re: Dual Processor Questions
Message-id: <savar-042E22.00272916052001@news.atl.bellsouth.net>
Date: Wed, 16 May 2001 00:27:29 -0400

In article <bsharvy-5811BF.18284015052001@news.efn.org>, Ben Sharvy <bsharvy@efn.org> wrote:

> Do programs have to be specifically written for dual-processor Macs to
> take advantage of both processors? If so, what are some of the more
> standard programs which are written that way? (PhotoDeluxe? Web
> browsers?)

Under OS9, there is an MP API which has to be specially coded for. The only program I know of that is MP-ready is Photoshop, and though there are others, I'm under the impression that there only a handful of regular consumer apps.

> What about programs written for the dual 604e's, running on a dual G4 or
> G3?

?? Same thing ??

Now quite sure what you mean.. Programs written for MP run on all MPs..

> Is MacOS X written to take advantage of the dual processors? If so, will
> regular applications benefit from an overall speed boost to the system?
> What about other MacOS's?

MacOSX incorporates SMP at a low enough level to make it available to all threads. Thus, the OS can parcel out various programs to various processors, as well as the various threads of any single program (assuming it is multi-threaded).

Other MacOSes do not have this feature.
*

From: "J. Brown" <jeremy-m-brown@home.com>
Newsgroups: comp.sys.mac.misc,comp.sys.mac.apps,comp.sys.mac.hardware.misc
Subject: Re: Dual Processor Questions
Message-id: <jeremy-m-brown-991420.08203117052001@news1.elmhst1.il.home.com>
Date: Thu, 17 May 2001 13:20:32 GMT

In article <Helge.Rebhan-342EE4.09080917052001@com36.dev.esoc.esa.de>,
Helge Rebhan <Helge.Rebhan@REMOVE-THIS.gmx.net> wrote:

> In article
> <jeremy-m-brown-D22677.15592516052001@news1.elmhst1.il.home.com>,
> "J. Brown" <jeremy-m-brown@home.com> wrote:

> > I've always thought it was odd that a lot of people forgot that there
> > were MP macs before the recent dual g4's... they also forgot that
> > quicktime 2.5+ made use of MP, along with a bunch of other apps.

> Really ? I never found information concerning MP support for Quicktime
> or applications like Final Cut Pro. Can you point me to an article for
> this ?

I think the readme file included with the first MP-aware Quicktime mentioned it... That was some time ago, however.

I've heard rumors that encoding on MP Macs was quite a bit quicker than on single processor machines, but alas, I've never had reason enough to buy one, so I'm stuck on a single cpu.

I used sherlock to search the TIL for "Multiprocessing" and came up with a sizable hit list. I don't have time to read all the articles, but there was one about Quicktime 3.0 that had the following blurb:

<<Because QuickTime provides a well defined hardware abstraction layer, applications are isolated from the underlying media technologies. This abstraction means that applications are always ready to take advantage of the latest hardware technology advances such as MMX and multiprocessing.>>

Ciao, jeremy

G3 vs. G4

From: Ben Sharvy <bsharvy@efn.org>
Newsgroups: comp.sys.mac.hardware.misc,comp.sys.mac.system
Subject: G3 vs. G4
Date: Sat, 16 Jun 2001 10:38:08 -0700
Message-id: <bsharvy-D64EAD.10380816062001@news.efn.org>

What exactly does the G4 do significantly better than the G3? I've only heard about faster Photoshop filters. I assume Motorola didn't develop a new processor just to run Photoshop filters. Is the superiority mostly graphical? If so, all graphics? Is it significantly faster for 3D games, rendering graphic-laden Web pages, showing full-screen video, etc.? Assume equal megahertz.

To put it another way, what does the G4 NOT do much faster than the G3?
*

Subject: Re: G3 vs. G4
From: Alan Somers <NsOoSmPeArMs@home.com.invalid>
Newsgroups: comp.sys.mac.hardware.misc,comp.sys.mac.system
Message-id: <160620011451540360%NsOoSmPeArMs@home.com.invalid>
Date: Sat, 16 Jun 2001 19:51:59 GMT

Integer operations and branching. Floating point operations are not any faster if using a regular floating point unit. However, if floating point computations are optimized to use the Altivec unit, they are considerably faster. So, the only apps that are going to see considerable gains on a G4 are those that are floating point intensive, e.g. image processing and video compression, assuming that they are optimized for Altivec.
*

From: Gordon Mulcaster <gozer@mac.com>
Newsgroups: comp.sys.mac.hardware.misc,comp.sys.mac.system
Subject: Re: G3 vs. G4
Message-id: <gozer-8B3511.13544416062001@clgrps11.telusplanet.net>
Date: Sat, 16 Jun 2001 20:54:05 GMT

IIRC the G4's floating point unit is based on the FP unit in the 604(e) and the G3's floating point unit is based on the FP unit in the 603(e), which means the G4's floating point unit is faster than the G3's.

It also supports 60x/Maxbus for faster memory throughput, supports up to a 2 MB backside L2 cache (the G3 supports a maximum of 1 MB), and supports all 5 MERSI states (important for multi-processor support).

The "new" G4s also include a 256 KB onboard 256 bit wide, full speed L2 cache and support (IIRC) a 4 MB L3 backside cache, the Altivec unit has also been upgraded.

> However, if floating point computations are optimized to use the
> Altivec unit, they are considerably faster. So, the only apps that
> are going to see considerable gains on a G4 are those that are
> floating point intensive, e.g. image processing and video
> compression, assuming that they are optimized for Altivec.

Altivec is particularily good at what is called signal processing (which covers image/sound editing/(de)compression), it's particularily good at (en/de)cyption and 3D image creation, and it's also very good at moving memory around.

Altivec is not limited to floating-point work it is also well versed at integer work.

For more info I'd suggest pointing a browser at http://www.mackido.com/Hardware/
*

Newsgroups: comp.sys.mac.hardware.misc,comp.sys.mac.system
Subject: Re: G3 vs. G4
From: russotto@wanda.pond.com (Matthew Russotto)
Message-id: <d7JY6.109059$Ne5.4274792@e420r-sjo3.usenetserver.com>
Date: Fri, 22 Jun 2001 15:03:37 GMT

In article <bsharvy-5980FD.17295021062001@news.efn.org>, Ben Sharvy <bsharvy@efn.org> wrote:
>So, in a nutshell, the G4 is better at one kind of math, and that's it.

Sure. But given that processors don't do much aside from branching, math, and I/O, that's pretty major.

--
Matthew T. Russotto russotto@pond.com
"Extremism in defense of liberty is no vice, and moderation in pursuit of justice is no virtue."
*

From: tacitr@aol.com (Tacit)
Newsgroups: comp.sys.mac.system
Date: 22 Jun 2001 16:37:08 GMT
Subject: Re: G3 vs. G4
Message-id: <20010622123708.12049.00001199@ng-mq1.aol.com>

>So, in a nutshell, the G4 is better at one kind of math, and that's it.

Not quite. It's better at a particular kind of math, but it's also better at accessing memory quickly (it offers higher memory bandwidth than a G3), and it has a larger cache and support for a L3 cache, both of which make a huge difference (since a trip out to main memory to fetch an instruction can waste 500 clock cycles(!) or more).

So what it comes down to is it's better at math and better at accessing memory, which describes the bulk of what a processor does. :)

------
Literary forums; Onyx, a game of sexual exploration; Xero, the industrial magazine of art, fiction and photography; fine-art photo gallery--all at http://www.xeromag.com/franklin.html

G3 Upgrades & RAM

Author: Benjamin Bunny (---.cdesd.k12.or.us)
Date: 07-28-02 23:39

My PowerLogix upgrade card (for a 7500) is unstable when RAM is interleaved.

I cannot return it to the vendor, because PowerLogix won't authorize it, saying the card isn't defective, and Other World Computing (OWC) won't take it without authorization from PowerLogix.

PowerLogix blames the RAM. I talked to OWC, the seller/marketer of the RAM, which is also the vendor who sells the PowerLogix card. OWC says those Powerlogix cards often don't work with interleaved RAM. It is known. Nonetheless, PowerLogix refused to help, and insists on blaming RAM (or the logic board). The RAM worked fine with the slower G3 card I had before.
This is an area where customers seem a little exploited. An upgrade manufacturer cannot demand that the machine meet the requirements of its upgrade. The upgrade must meet the requirements of the machine. My interleaved 7500 runs flawlessly with its intended processor, with a 120 MHZ 604, and with a 266 MHZ G3 (XLR8). It only fails with the 400 MHZ processor from PowerLogix.

If a machine runs flawlessly at the speed for which it is designed (100 MHZ, in this case), there are no grounds for saying it is defective.

Re: PowerLogix Complaint, and Exploiting Us!
Author: 000101001010100 (Registered User)
Date: 07-28-02 23:52
I had the same issue- I was unable to get a Powerlogix 450Mhz G4 card running properly (or running at all for that matter) with my 128MB DIMMs in my 8600- and this problem was acknowledged by Powerlogix tech support, and I was told to get rid of my 128MB DIMMs for 64MB DIMMs, because the 128MB DIMMs don't function properly (Ok, that's why lots of places sell 128MB FPM/EDO DIMMs for my machine...). I returned the card to Outpost.com, and bought an XLR8 ZIF Carrier and couldn't be happier. It runs fine with my 128MB DIMMs, and not only runs fine at the rated speed, but lets me go to 550Mhz with my OWC Mercury ZIF (voids your warranty though...).
Do yourself a favor and buy XLR8 next time!
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Life's a bowl of punch, go 'head n spike it!" -Nick Hexum
PowerMac 8600 @ 520Mhz (Summer sucks) G3 XLR8, Radeon 32MB DDR PCI, 384MB RAM
Pavilion 6730 Celeron 600Mhz @ 745Mhz, 3D Prophet 4000XT PCI, 256MB RAM

Re: Actually..
Author: OWC Larry(remote) (---.macsales.com)
Date: 07-29-02 02:09
These machines are very capable... far more than some would like since you can upgrade an older machine and put off the need to buy a new one. Heck, we even put Xpostfacto out there free of charge so you can run OS X on many unsupported PRe-G3 machines. Now, with all that said... Somethings you have to tweak a little and the 7500 is VERY picky...

More so because it was only designed to support up to a 45MHz bus with processors higher than a 601. Just like the 7600..... Most will do the 50MHz required for a 500MHz G3 or 450MHz G4 (G3 is up to 10X bus, G4 is up to 9X bus speed for the multiplier). Your slower G3 ran the logic board bus speed slower, within or nearer its rated speed. The G4/450MHz maxes you to 50MHz and the memory interleaving is what appears to be suffering. You will have a greater gain from the G4 than from the interleaving..
The Powerlogix card is adjustable. It is likely that if you adjust the speed down just a little, you may be able to drop beneath the point where the 7500 can't do interleaving with what is an overclocked bus.
There is a ton of confusion with the 7500 and I know if I don't say this now, 50 people will say that well it has a 50MHz bus with the 601 it came stock with. Very true, but it was not rated for higher than a 45MHz bus speed with 604 and later processors.

No one is exploiting anyone. The 7500 is still a great machine and with a little extra help it will go another extra distance for you.

Re: PowerLogix Complaint, and Exploiting Us!
Author: ka jot (Registered User)
Date: 07-29-02 12:04
The 7500 has been problematic with many G3 (and presumably G4) upgrades. It is by far the pickiest of the various machines I've upgraded. My 7500s have not been able to handle Firewire, either. However, I have used Powerlogix upgrades in 7500s with interleaved RAM.

I assume you have no cache DIMM installed on the motherboard.
If you want to keep this machine going, consider buying a 7600 logic board on eBay. It'll cost you $10 or so. My 7600 has worked great with interleaved RAM and a couple of Powerlogix cards.

Re: PowerLogix Complaint, and Exploiting Us!
Author: Benjamin Bunny (---.cdesd.k12.or.us)
Date: 08-11-02 22:12

I was able to solve the problem by changing the settings on the card, so now it is a 420 MHZ card running a 46.5 MHZ bus. The 50 MHZ did indeed seem to be the hangup.

Miscellaneous

From: Jeff Walther <trag@io.com>
Newsgroups: comp.sys.mac.hardware.misc
Subject: Re: Question on 604e CPU Card
Date: Mon, 02 Jul 2001 16:55:21 -0500
Message-id: <3B40EDC9.C4D42522@io.com>

Paul McGrane wrote:

> The CPU cards from 8600 and 9600s that run from 250 to 350Mhz used an
> *inline* level 2 cache, which was sort of Apple's stopgap between the
> lookaside caches of the previous Macs and the backside cache of the G3
> architecture.

> I believe in these Macs (weren't they
> codenamed Kanga? Yeah I think I studied those MacWorlds too much back
> then :^) the L2 cache runs at 100Mhz.

The Apple PowerMacs which had PPC604e CPUs running at 250 MHz or higher used the Kansas motherboard (8600 Enhanced, 9600 Enhanced) and used the Mach 5 CPU card with the In-Line cache.

The Mach 5 CPU cards will only work in Kansas motherboards and no other motherboards. The original poster is lucky he didn't damage his 8500.

Note, however, that Power Computing, Umax and some third party upgrade vendors produced 250 MHz PPC604e CPU daughter cards with the original specs which will work in the x500 family of PCI Macs.

So *Apple* 604e CPU cards at 250 MHz and above won't work. Clone maker 604E CPU cards at 250 MHz will work.
***

From: Ernie Leimkuhler <stagesmith@mindspring.com>
Newsgroups: comp.sys.mac.hardware.misc,comp.sys.mac.misc,comp.sys.mac.apps,uk.comp.sys.mac
Subject: Re: Clock Chipping 6100?
Date: Mon, 02 Jul 2001 16:03:14 -0700
Message-id: <stagesmith-12CC96.16031402072001@news.mindspring.com> Posted-And-Mailed: yes

In article <1evxppw.1g264i2ctt4sgN@usr1604-cro.cableinet.co.uk>,
michaeldines@NO_SPICED_HAMcableinet.co.uk (Michael Dines) wrote:

> Anyone clock chipped a PPC6100/60?
>
> Two questions:
> Which on the motherboard is the clock chip?
>
> Did you use a Mach10 Jet clip-on crystal? If so, could you tell me if it
> matters which way up it goes?
>
> tia.

Clock chipping is a really bad idea unless you have a couple of spare motherboards. I fried my 6100 doing that and swore it off after a long discussion with a repair tech about heat and chip life.

Go for a cheap G3 card. Much much better.

--
"Everything's already been said, but since nobody was listening, we have to start again."

-Andre Glide
top of page

Valid HTML 4.01! Valid CSS!