One Click on the Banner shows support to my Sponsors
The Source for Mac Performance News and Reviews
Don't forget to check out all the other site features!
Mach 5 (8600/9600) Systems Coverage
This page is the new home for Mach 5 system specific content (Info/Mods/Upgrades/Tips) at xlr8yourmac.com. It will be expanded with as new articles, reader input and performance data is received.
The Mach 5 is a name given to the 250-350mhz 604E CPUs that are used in 1MB inline cache CPU cards of the 8600/250, 8600/300, 9600/300 and 9600/350 Mac models. These CPU cards do not work in any other Mac models. For more info on the Mach5 see Motorola's 604e page which has thermal, power and performance data and a closeup picture of the Mach5 die ( beware - 114K image!).
I've also posted a page on Mach 5 CPU card speed settings. I do not suggest you modify your card. It will void the warranty and is a very delicate procedure. This info is for reference only.
The latest update compares the performance of a overclocked 9600/350 to a overclocked Apple G3/300. I think you'll find the results surprising in many ways. Considering the 9600/350 cost under $1600, it's looking like a better value every day.
Latest 9600 News
Photoshop 5 Tests : I've posted a page showing the PSBench results of the 9600/400 vs a Apple G3/300 running 333/222/83. The results are very interesting. [8/29/98]
9600/400 slower?:A reader said he noted some negative performance in some tests after overclocking his 9600/350 to 400MHz." Hi Mike, I've been running my 9600/400 modified from 9600/350 same way, over a couple of days with no crashes or any fatal problem. I've noticed that FPU was increased by 11% on every test subject and 11% overall. CPU performance had some negative effect from clocking up, most subject has incerased by 9-11% ,but memory R/W was dec. by 8.5%, Bmove/A was dec by 8.4% and instruction has been dec. by 8.2%. After adding all negative and positives together the total increasement was 6% using Symantec's "system info". "
I saw nothing but improvement here, in real apps and MacBench (see below). However one other reader reported odd behavior like losing 200+ points in MacBench's disk test after the modification (which he removed). [8/18/98]
ByteMarks: Mach 5 vs G3: I ran Bytemark DR/3 on both the 9600/400 and my Apple G3/300 (running 333/222). Tests were run with no extensions (backside cache on G3 was enabled).
If you missed it, check the Applications tests scores I previously ran with these two systems. [8/13/98]
9600 Video cards Test: I've posted a Mac Rocket review graphics card review that shows video and movie scores in the 9600/400 system of the ProRez, Mac Rocket, ATI RagePro 3D 4MB/8MB (using driver v3.1 and 3.2) and the OEM Twin Turbo M8 card with the latest driver v4.05. [8/11/98]
Mach 5 Card Speed Mods Explained:More 9600/400 Feedback: Another reader sends info of his 9600/350 mod to 400 MHz:
Due to the many requests for a clear explanation on the Mach 5 CPU card speed settings I've created a Mach 5 CPU Card Info page clearly showing the different speed configurations. [7/21/98]
" Hi Mike,[7/18/98]
I've been running my 9600/400, modified the same way, for over a week now with no problems. Integer [I think he means CPU-Mike] performance is within 5% of my G3/266 but the fp performance is 50% better.
Here's my MacBench numbers:
9600/350 CPU: 793 FP: 904
9600/400 CPU: 839 FP: 1007
G3/266 CPU: 887 FP: 667
[In the above table I changed Int to CPU, as Macbench has no Integer test, it's a more general CPU test-Mike]
I agree, if people aren't used to hand soldering surface mount components then they should have someone do it for them. However, as you know, it will void the warranty.
(name withheld on request)"
9600/400 vs. Apple G3/333:
I just modified a Mach 5 350 MHz card to run at 400 Mhz and ran a subset of my Applications Tests to compare performance. Even before the tests were run the 9600 seemed to be very responsive and I knew it would do well. The results were still suprising in many ways.
The Apple G3 is a Rev 2 motherboard 300 MHz system with 1MB of backside cache, running at 333 MHz CPU speed, 222 Mhz cache speed (via PowerLogix's cache util) with a 66 MHz bus. The 9600/400 is a 350 MHz system overclocked to 400 MHz with a 50 MHz system bus and 100 MHz inline cache speed. Both systems had OS 8.1 and similar extension set. Disk cache was set to 1MB.
Bryce 2 Alexandria II render
Bryce 2 Magical Knights I
After Effects 3.1 Movie
Adobe Premiere 4.2 Make Movie
Had the G3/300 not been overclocked for CPU and Cache speed the Mach 5 400 MHz would have won the overall applications test suite run easily. Even with the G3/300 running 333/222 the difference on these 4 tests totaled less than 10 seconds! As I've said before - there's a lot of life left in those 9600/8600 systems! [7/17/98]
9600/400 MacBench Scores:
Here are the Macbench 4.0 scores for the 9600/400 (Ultimate Rez video card used, 1024x768 thousands color mode display).
- CPU: 852
- FPU: 1028 (No Libmoto)
- Disk: 401 (2MB cache)
- Graphics: 742 (Ultimate Rez, 1024x768 thousands colors)
If I had used the LibMoto extension (as is commonly used for G3 system benchmarks the FPU scores would have been much higher. Libmoto has proven to be practically useless in real world apps however (and detremental to others). Libmoto saved a whopping 4 seconds off a 5+ minute Bryce2 rendering for instance and has some negative effects on other apps as shown on my LibMoto page. [7/17/98]
Kansas (9600/8600 Mach 5) Motherboard Video/Ram Xfer Speeds: Michael Perez wrote with a URL that mentions possible Kansas ram transfer speed issues that are affecting video card performance:
" Hi mike,
Found this on the support web page of a Video oriented software company (http://www.webcom.com/puffin/support/tsdocs/videocards.html) :
'PowerMac 8600/250 and 300, 8600 Graphics Accelerator 1.0 (32k) Apple Power Macintoshes using the Kansas motherboard - all 8600 & 9600 250s, 300s, and 350s - are currently experiencing a performance hit with RAM transfer speeds which limits the playback performance of your video card. Apple is working on a solution to this issue, and the 8600 Graphics Accelerator is first of several drivers planned to eliminate this problem under different configurations. When 9600 solutions are available, we will be sure to post them here, and update the performance results. '
' Twin Turbo 128m8A, 9600 Graphics Accelerator 4.03(P) (128k) The Apple PowerMacintosh 9600 may come with the Twin Turbo 128M8A video card pre-installed. This card uses a different extension - called 9600 Acceleration - than the off-the-shelf Twin Turbo card and may not deliver optimal playback performance. The 9600 Acceleration 4.03(P) extension used in our tests updates the 128m8A to match the off-the-shelf version. This will have a noticeable effect on 9600/200 machines (non-Kansas); Kansas-based 9600s (the 250, 300, & 350) are still limited by the existing Kansas issues. You can determine which Twin Turbo card you are using by a sticker on the card itself, or by inspecting the connectors on the back of your Mac - the 128M8A will have a single port, while the 128 will have two ports. '
Hope it will help you and your 9600 readers concerning the bad video specs.
9600/350 Substitute Video Card Tests: After reading of low scores and the info above, I tested a ATI RagePro 3D and IXMicro Ultimate Rez in the 9600/350 at both 1024x768 thousands colors and 1152x870, millions colors in Macbench 4.0's Graphics and Publishing tests. Although both these cards produced much higher scores than most owners reported with the stock Twin Turbo OEM, it seems the video scores are lower than they should be, regardless of video card used.
I could not test the Twin Turbo as the CDROM seems to have given up the ghost tonight. I'll have to have it replaced, and I'm in the dreaded "Zone 2". That means no on-site service (again). Can someone please send me scores from Macbench 4.0 tests using the OEM Twin Turbo, drivers 4.02a/4.03 at 1024x768 thousands colors and 1152x870, millions colors (75hz)?
Below are the scores I obtained while the CDROM was working (needed for Macbench video tests). Several other cards/systems are show for comparison:
1024x768, thousands colors:
Lo-Res Pub Test
Twin Turbo OEM v4.03a
Twin Turbo OEM v4.03p
ATI RagePro 3D
ATI RagePro 3D
ATI RagePro 3D
OEM Twin Turbo
1152x870, Millions colors:
Hi-Res Pub Test
Twin Turbo OEM v4.03a
ATI RagePro 3D
ATI RagePro 3D
ATI RagePro 3D
OEM Twin Turbo
1) Ultimate Rez PowerTower Pro tests used initial driver release. Other Ultimate Rez tests used latest released version 1.02. Tests run under OS 8.1.
2) 8500/200 Twin Turbo OEM tests used OS 7.6, driver v4.02 as shown in my Imagine 128 vs Twin Turbo review.
3) All tests run at 75 Hz vertical refresh rate.
It appears that the 9600 has graphics performance lower than it should for a 350 Mhz 604E CPU and 6 PCI slot Mac. Scores were close to a 180 Mhz 604E system running the same card. I'm hoping a reader can send me the OEM TT scores for the 9600/350 with driver versions 4.02a and/or 4.03 at the resolutions above for comparison.
Mach 5 Applications Performance:
To see how a 350mhz Mach 5 604E compares to the G3 CPU Cards, see the Infini-D, Bryce 2 and After Effects scores on myG3 Apps Tests page. The Mach 5 did very well in the Bryce and After Effects tests. It may trail the G3 in Macbench CPU scores, but the strong FPU performance shines in many real world apps. Mach 5 Forum: I added a Mach 5 specific message board to the new forums. Mach 5 system owners are encouraged to share info and experiences there.
More on the L1 Cache size of the Mach 5 Rev 1/Rev 2 CPU cards:
Looks like the cache size is the same, just the Rev2 is shown as the total of data and instruction cache instead of listing them separately. Odd the rev 2 does not note the 300mhz version. Possibly there is some difference in the rev 2/350mhz CPU/Card.
" Dear Mike,
Here's the relevent section on the on-chip cache (from replacement parts catalogue database):
Processor and memory 350-MHz PowerPC 604e RISC processor with floating-point processor and 64K on-chip cache 1MB Apple Inline Cache (level 2) [The cache noted here is the 32K data + 32K instruction cache shown as one total size. No difference in the actual sizes.-Mike]
300- or 350-MHz 604e PowerPC processor Integrated floating-point unit, 32K on-chip data and instruction caches 1MB Apple Inline Cache (level 2)"
Different Revisions of CPU Card/Motherboard?:
I'd heard of the two different revisions of CPU Cards but was not aware that they each had a revised motherboard. An reader who asked to remain anonymous wrote:
" Dear Mike,
Here's a little secret (I think it is cos no one has mentioned it yet) on the 9600/350 that I just found out about while sorting out a deal for the machine.
There are two different motherboards, a Rev 1 and a Rev 2. There are also two different 350Mhz cards; Rev1 and Rev 2 as well. [I've heard about the two CPU card versions, reported here in the past. The early ones are said to run hotter, and have fainter markings according to a reader-Mike]
Apparently, they are not interchangeable, ie rev 1 motherboard uses rev 1 processor card. Interchanging the processor cards can lead to frequent crashes and failure to boot. I just found this out from reading an Apple Tech spec sheet/replacement part manual.
Next bit of gossip. Some 9600/350s and 9650/350s come with the ATI 3D graphics card instead of the IMS TwinTurbo 128. I did not believe it till I saw the internal of a 9600 and a 9650 with the ATI card. There was even an Apple Addendum insert in the document box informing of the change. The two machines are of US build origin and not a local dealer hack job as the documentation proves. "
A later update mentions other differences in the L1 (internal CPU) cache size and a note on different video cards:
" Dear Mike,
Here's another difference I just found out between the Rev1 and Rev2 cards (late night detective work, amazing what some people will let you see if you promise them a beer:-).
Apparently the rev 2 cards has 64k on-chip cache as compared to the 32k in rev 1. [Not sure about that - but I heard that some early lab cards had only 512k of *inline* cache on the card-Mike]. It also appears that the rev2 cards are labeled "Apple Hi performance processor card for home and office use" . It also has the different speeds from 250-400Mhz clearly visible on the side opp to the heatsink with the rated speed shorted by a resistor or capacitor with the rest left bare. "
Note: The 300mhz card I chipped had the (0 ohm) resistors on the speed settings label area - which is a fakeout apparently as all pads were already shorted. Moving that resistor does nothing to change the card speed (as noted here in the past).
More info here as I do more research and receive more info.
For links to Mach 5 CPU card mods, check out my Links Page OC section for a link called 'OC Central'.
Back to XLR8YOURMAC.COM
Your Source for the best in CPU/SCSI/VIDEO card reviews, daily news, and more!
Copyright © 1998.
No part of this site's content is to be reproduced in any form without permission. All brand or product names mentioned here are properties of their respective companies.
Disclaimer: Users must read and are bound by the Site Terms & Conditions of Use.