Visit our PPC Checker Page and download Michiro Isobe's utility to identify Copper vs Aluminum G3s.

Several readers responded to the MacCentral article 'Motorola's PowerPC's noticeably faster than IBM's? linked in Jan 18th 1999 news:

(Update: For the hopefully final chapter on this saga of conflicting specs (I pray) see Tim Seufert's latest comments the last I will post on this topic - it's gotten more attention than it deserves except for the issue of copper based processors currently shipping only from IBM .)

Ben Chi copied me on a letter to the article's author, Ben Wilson:

" In regards to your article that Motorola 750's are faster than IBM 750's, where did Mr. Klewicki get the performance specifications for the IBM 750 @ 400Mhz? According to IBM's PowerPC performance web page , an IBM 400Mhz 750 gets a rating of 18.8 on the SPECint95 test, not 17.6. This is identical to the number that Motorola publishes. It is possible that IBM had published another number somewhere else... [ see the update below-Mike] after all, these numbers are estimated performance, not actual. Neither Motorola nor IBM has provided concrete data with full disclosure of the hardware and software (especially the compilers) involved. I would take these numbers with grain of salt, even more so than most benchmark results.

It would be interesting and a good piece of journalism if someone would pay the required fee to SPEC to obtain the SPEC CPU benchmark suite source code and ran it with MkLinux, LinuxPPC (both with gcc 2.8.1 and egcs 1.x), and MacOS X Server (with gcc 2.7.2) and g77 on these CPUs. Unfortunately, since IBM isn't shipping AIX on G3's AFAIK, we can't see the results with xlc or Motorola's Reference Compilers.
..Bill Chin "

IBM is Cooler Running Part?:

Timothy Seufert responded as well and notes that the IBM specs show their version runs cooler than the Motorola part (this may be why I've seen IBM parts run at higher speeds based on my testing and reader feedback in the past year).

" Ben, Mike:
Phil Klewicki is mistaken about the IBM 750 having a SPECint95 rating lower than that of the Motorola 750. See the following two web pages for the manufacturers' respective performance claims:

http://www.mot.com/SPS/PowerPC/products/semiconductor/cpu/750.html
www.chips.ibm.com/products/ppc/overview/perform.html

Both the IBM and Motorola 400 MHz 750 CPUs are listed as scoring 18.8 SPECint95 and 12.2 SPECfp95. Both are quoting numbers for a system with 1MB L2 cache clocked at 200 MHz. IBM also specifies an 89 MHz system bus speed, whereas Motorola does not say (but it's probably the same).

The only significant difference between the two manufacturers is the manufacturing process and power consumption. IBM has a better process technology; their 750/400 is rated at 4.1W avg. / 5.7W max. while Motorola's 750/400 uses 5.8W avg. / 8.0W max. This is really only relevant for notebook use; should Apple decide to ship a 400 MHz PowerBook they will almost certainly use IBM's chip instead of Motorola's.
Timothy Seufert "

As I commented yesterday, I wondered if the story explained why my IBM CPU in the Yosemite (B&W G3) 400 produces sub-1300 scores in MacBench 5's CPU test, which is lower than the 1300+ results I've seen posted for the 400MHz Yosemite systems. The CPU mfr is noted on the label attached to the cache chips on the ZIF of the Yosemite.

Update: Phil Klewicki wrote with a note that the lower SPEC98 integer score of 17.6 *was* shown at www.chips.ibm.com/products/ppc/overview earlier today (1/19/98) but now it has been changed to agree with the 18.8 value on the performance page. However he later sent a URL to another page at IBM that still shows the lower 17.6 value for the same part (G3/400/1MB L2/Copper). Phil still questions if the Motorola G3 333-400's are copper CPUs.

Copper G3s: In response to lots of reader mail, I queried some sources in the CPU business to verify that IBM 333MHz and up G3s were copper. I was told they are (not just the 400MHz models as some thought). One source said as far as they knew only IBM was shipping copper based G3s. As I noted yesterday this could account for the lower wattage ratings on the IBM copper G3s (since copper has less resistance than aluminum, it should dissipate less heat).

Phil's latest Digging: Phil's digging turns up more inconsistencies which are probably (again) just a case of older/newer pages. With a huge site like these companies it is very hard to keep all pages updated I suspect.

" Mike:

Motorola:
http://mot-sps.com/sps/General/chips-nav.html

NOTE: power dissipation @ 400 mhz here is 5.8 - 8.0 watts
AND: the die size is 67 sq. mm / 6.5 million transistors / .22 micron trace width

THEN GO: (Motorola again)
http://mot-sps.com/sps/General/chips-nav.html

NOTE: power dissipation @ 400 mhz is 3.4 - 5.8 watts
6.35 million transistors

SO: On Motorola's web site, there are quite LARGE differences in wattage (power consumption) and a discrepancy in "transistor count".

NOW: (IBM)
www.chips.ibm.com/products/ppc/overview/perform.html

NOTE: .22 micron trace width (same as MOTOROLA)
BUT WITH A BIG BIG ( actually small small) DIFFERENCE: the IBM CHIP IS TINY !!! :
only 40 sq. mm (vs. 67 sq. mm for Motorola as noted above)

BOTTOM LINE:
there are SEVERAL chips being called "400 mhz G3" from both IBM and Motorola!!
This sounds like a "surprise package" for the consumer. Not a good thing.
Regards,
Phil Klewicki "

" First: the different SPECint95 numbers shown on IBM's web site.

I think you explained this already Mike, but basically SPEC benchmark results for a CPU can and do change over time. SPEC is a suite of C programs, so whenever compilers improve, so do SPEC scores. And since SPEC is really a system benchmark which exercises L2 cache and the main memory system, the motherboard the CPU is plugged in to matters.

So, like you said, the low SPECint95 numbers Mr. Klewicki found on IBM's site are probably just old results.

Second: discrepancies on Motorola's web site with respect to power consumption and transistor count.

The Motorola press release Mr. Klewicki found reads: "Operating at a 3.4- to 5.8-watt typical power dissipation range, the PowerPC 750 microprocessors..."

Motorola is saying the lowest speed grade is 3.4 W typ. and the highest 5.8. On the specs page, the 400 MHz CPU is listed as having typical/maximum power consumption of 5.8/8.0 W. The numbers are consistent, since the press release is only talking about typical and not maximum power use.

The 6.5M transistor count is probably just an error. In fact, the page showing 6.5M has a link to a press release which uses the 6.35M figure seen everywhere else.

Third: the difference in size between Motorola's and IBM's chips.

IBM and Motorola both list a 0.22 micron feature size, but feature size isn't everything. IBM's process has 6 metal layers while Motorola's has 5. An extra metal layer means better circuit routing, tighter packing, and a reduction in average wire length. The resulting chip is smaller and uses less power.

Finally, Mr. Klewicki's conclusion:

BOTTOM LINE:
there are SEVERAL chips being called "400 mhz G3" from both IBM and Motorola!! This sounds like a "surprise package" for the consumer. Not a good thing.

If there is anything bad for the consumer going on here, it's that somebody is irresponsibly trying to spread the false idea that there are big performance differences between the Motorola and the IBM PowerPC 750. A 400 MHz 750 performs exactly the same whether it's from Motorola or IBM. I would be willing to bet money that they would be identical right down to a cycle-by-cycle analysis of the CPU's internal state while executing a program. The 750 was one of the last joint designs by the Somerset design facility, so Motorola, IBM, and Apple all codeveloped it. The only difference between an IBM and a Motorola chip is that each company has to do its own layout, since their chip fabrication processes are different.

In the future, we may see significantly different chips from IBM and Motorola since the joint design concept has been abandoned (the Motorola "G4" is the first step in that process) [I thought there was a movement now to ensure Altivec was in the IBM G4 as well, from a EE Times article-Mike] , but you can bet that they won't be giving them the same model numbers now.
Tim Seufert "

Other Site Links of Interest:

• CPU Upgrade Articles (G3, G4, 604, reviews and related articles)
• Systems Articles (G3, G4, PowerBooks, older Mac articles/reviews/tests)
• G3-ZONE Hub of articles on Beige and B&W G3 info/reviews
• B&W G3 Tips, Comments and Specs