History of the PowerPC 603 and 603e

[bigmessowires]

I did some more reading about this yesterday, and there's some (old) controversy about the 5200 and 6200 series and others with a PowerPC 603, and some Reddit arguments involving forum members here. For my dollar, this write-up is the most worthwhile: https://lowendmac.com/2020/the-golden-road-apple-how-i-discovered-that-the-worst-mac-ever-wasnt/

The author found the Performa 6200/75 to be 93% the speed of a PowerMac 6100/60 for integer CPU work. Since the 6200 is clocked 25% faster than the 6100, that means MHz-for-MHz a 603 (as used in the 6200 anyway) is roughly 75% as fast as a 601 at the same clock speed. The 6200's higher clock speed almost but not quite makes up for the 603's comparative sluggishness versus the 601. The Performa 6200/75 is sort of like having a PowerMac 6100/56 if such a thing existed.

The tests were run under OS 7.5.5 with MacBench 4.0. It doesn't say whether SpeedDoubler was used, so probably not. I assume the MacBench CPU tests are PowerPC native and not emulated 68K code.

If synthetic benchmarks aren't convincing, the author also did some real-world tests launching applications and decompressing files. Compared to a Quadra with a 601/66 upgrade card, the 6200/75 performance was nearly identical.

TL,DNR: I'd assumed incorrectly that since the Performa 6200's PowerPC 603 was a newer generation CPU than the 601 and also had a higher clock speed, it would perform better. But the actual performance is comparable to a 6100/60 and is reasonable for the day. Maybe if they'd named the chip the LC601 or something, that might have done a better job of setting expectations.

 

[chelseayr]

not to interrupt but is it just a bit of baseless heresy or weren't the later powerpc performa boards actually basically more or less 68k board designs kludged 'up' a bit to suit the faster cpu/bus/ram overall?

 

[Snial]

I did some more reading about this yesterday, and there's some (old) controversy about the 5200 and 6200 series and others with a PowerPC 603, and some Reddit arguments involving forum members here. For my dollar, this write-up is the most worthwhile: https://lowendmac.com/2020/the-golden-road-apple-how-i-discovered-that-the-worst-mac-ever-wasnt/

@chelseayr it's not a baseless heresy as such: the 5200/6200 series were primed to do badly in reviews because the 603's cache wasn't adequate for the Gary Davidian emulator and Apple had to add that 256kB L2 cache to compensate, which kinda undermined the whole point of a lower-cost PowerPC based home computer.

But it turns out that they weren't really just kludged 68K designs. Particularly the rumour that it somehow had a split 32-bit bus which wrecked access to RAM or I/O, which I believed for a good 20 years or so, until I read the article below (which is linked in BMOW's Low End Mac article). Prepare to be enlightened:

The Mythical Road Apple | Taylor Design

https://lowendmac.com/2020/the-golden-road-apple-how-i-discovered-that-the-worst-mac-ever-wasnt/

The author found the Performa 6200/75 to be 93% the speed of a PowerMac 6100/60 for integer CPU work. Since the 6200 is clocked 25% faster than the 6100, that means MHz-for-MHz a 603 (as used in the 6200 anyway) is roughly 75% as fast as a 601 at the same clock speed. The 6200's higher clock speed almost but not quite makes up for the 603's comparative sluggishness versus the 601. The Performa 6200/75 is sort of like having a PowerMac 6100/56 if such a thing existed.

Indeed, but I think as discussed earlier, the 603 had about half the transistors of a 601, and I think that the design allowed for higher clock speeds (shorter transistor chains for decode / Execute?), so maybe a 60MHz 601 should be compared with a 75MHz 603? The real let-down in the 603 was the Davidian emulator, which... is super-impressive, but not suited to small 8kB L1 caches.

 

[Phipli]

not to interrupt but is it just a bit of baseless heresy or weren't the later powerpc performa boards actually basically more or less 68k board designs kludged 'up' a bit to suit the faster cpu/bus/ram overall?

To be fair... The 6100, 7100 and 8100 are descended from the LC

 

[bigmessowires]

So are the 5200/75 and 6200/75 series of Power Macs the lowest-benchmarked PowerPC Macintosh models ever made? I can't think what else could have been slower. It looks like everything that used the 603e was at least 100 MHz.

 

[Phipli]

So are the 5200/75 and 6200/75 series of Power Macs the lowest-benchmarked PowerPC Macintosh models ever made? I can't think what else could have been slower. It looks like everything that used the 603e was at least 100 MHz.

Some of the lowest spec 603e laptops are in the same ballpark. No L2 cache (unlike the 6200/75).

The PB 5300/117 springs to mind. Terrible machine for several reasons.

 

[Phipli]

This table might be interesting to folks :

Instructions per second - Wikipedia

en.wikipedia.org

So the 601 scores 1.971 MIPS per MHz, while the 603e scores 1.41 MIPS per MHz. So clock to clock in these specific tests, a 603e needs to be 40% faster than a 601, so 84MHz to 60MHz).

The 6200 (note 603, not 603e) Vs 6100 tests seem to be closer than that, but I don't have visibility of the details of the Wikipedia or benchmark setups.

 

[bigmessowires]

That actually agrees pretty closely with those other benchmarks. He found that a 6200/75 was 93% the CPU performance of a 6100/60, so MHz-for-MHz a 603 is 74.4% as performant as a 601 (0.93 times 60/75). The Wikipedia MIPS numbers are 1.41 versus 1.971, which is 71.5%. Close enough. Exact numbers would depend on the specific benchmark used.

Some of the lowest spec 603e laptops are in the same ballpark. No L2 cache (unlike the 6200/75).

The PB 5300/117 springs to mind. Terrible machine for several reasons.

From my searching, the slowest 603e portables were the PowerBook 5300/100, PowerBook 5300c/100, PowerBook 5300cs/100, and PowerBook Duo 2300c/100. I'd be curious to see some benchmarks of how they rate compared to the 6200/75 and 5200/75.

 

[Phipli]

From my searching, the slowest 603e portables were the PowerBook 5300/100, PowerBook 5300c/100, PowerBook 5300cs/100, and PowerBook Duo 2300c/100. I'd be curious to see some benchmarks of how they rate compared to the 6200/75 and 5200/75.

I didn't realise there was a 5300/100, I thought the 117 was the slowest, my bad.

 

[Snial]

I didn't realise there was a 5300/100, I thought the 117 was the slowest, my bad.

Oh yes, the 5300 (greyscale)/100 was the slowest! I had one, I knew from first hand experience!

But with 40MB of RAM it was usable, despite being a downgrade from my PowerMac 4400/160. I completed the first draft of my MPhil thesis using it, running Mac OS 8.1, which feasible because it had a second monitor display card & I had a 15” monitor!

The biggest problem wasn’t the speed, but the 500MB hard disk. I had to supplement it with CodeWarrior on my Zip drive. In spring 2000 the drive started to fail so I bought a tangerine iBook/300 (with 96MB of Ram & Airport - oooh). I also used it at work to help develop (probably) the first WiFi drivers for a Symbian OS webtablet.

 

[cobalt60]

I plan to put a 603e in my 9600 too, but that is going to take some careful soldering and research, as well as a spare 604 BGA card to swap chips on.

Power Computing PowerBases used 603e on the same type of CPU card. Some G3 upgrades for 7300/8500/9600 etc had a switch to use with the PowerBase. This leads me to believe it may be somewhat easy to modify one of these cards to use in a 9600.

 

[Phipli]

Power Computing PowerBases used 603e on the same type of CPU card. Some G3 upgrades for 7300/8500/9600 etc had a switch to use with the PowerBase. This leads me to believe it may be somewhat easy to modify one of these cards to use in a 9600.

I'm just waiting on a spare card so I can have a similar 604e and 603e card. The chip pinouts are compatible, but I'll need to spend some time with the datasheet to work out the needed core voltage and modify the card to correctly supply it. Project on hold is all.

The other thing is I need to make sure the heatsink is the right height and perhaps modify it with a shim or machining, because these eBay reballed chips often have the wrong sized balls and I don't want to crack the die.

I have the 603e chip sat waiting.

 

[Unbounded_Villain]

I just want to chime in on the 6200 performance issue. I had a 6205 as my first PPC Mac. Connectix Speed Doubler worked well enough to correct a lot of the issues with 68k emulation.

The bigger issue was the problems with the serial ports and the Comm Slot. The serial ports were limited to 9600 baud because there wasn't hardware handshaking, but one or two external modems had the hardware that could actually do 56k/v.90 and I had one. It was a Global Village something or other. Furthermore, if there was a lot of network access on the modem port, typing sputtered. You could get around this by using a Comm Slot modem, but they never went higher than 28.8 and Apple's modems topped out at 14.4. They'd move on to PCI and Comm Slot II and better designed machines starting with the 6360. Apple has never cared about supporting machines they don't currently sell, so that was that.

Ethernet Comm Slot were RARE. Ethernet was seen as something for businesses and universities until the rise of broadband in the 2000s. It was basically impossible to get Comm Slot and even Comm Slot II Ethernet cards unless you were buying education machines. However, using a Comm Slot Ethernet card disabled the printer port.

The best thing about the 6200 series was the AV cards that could be added. I used to watch cable TV on my Mac while using AOL all the time. It was truly awesome. I spent so many years watching TV and then DVD video in small windows on my Mac when PiP came around for streaming, it was like, "HALLELUJAH!" However, the IDE drive was so slow you couldn't reliably record video without lots of dropped frames. There was just no winning with that computer. Luckily I only had it for a year before I got a 7500, which I still have to this day.