It's Christmas in July! I've been saying this for years.
The whole "Road Apple" concept has literally never sat well with me and most of the machines on that list are basically "That way" because they were affordable and aimed at a non-pro market and the compromises Apple made mostly don't actually impact their usage for those markets.
Also, above and beyond this machine, the assertions made on several of the "road apple" machines are literally, provably wrong. The best example of this is the cacheless "Mainstreet" PowerBook G3/233, which
roundly outperforms every predecessor machine, up to and including the PowerBook G3/250 "Kanga". (This mostly comes down to "platform enhancements" - faster bus, graphics, disk interface, etc etc etc, the Kanga is basically a 3400 with a G3 upgrade preintegrated whereas the "Mainstreet" was a full G3-first platform, essentially a miniaturization of the desktop PowerMac G3s.)
The Cacheless G3/233 probably does deserve criticism because of the small passive-matrix display, low overall configuration compared with literally everything else (especially for over $2200 in a moment when the iMac and desktop G3s had better configs for $1299). (The magazines at the time absolutely panned it for that screen, especially given that in the moment you could still buy old-stock PowerBook 1400/166es with better screens, bigger disks, and sometimes more RAM for significantly less than what a PowerBook G3 cost, and get a much nicer machine if your work didn't actually rely on that G3 speed or upgradeability the 1400 didn't have.)
But, it doesn't deserve criticism for its performance, because it had excellent performance for the time.
LEM kept up the whole "road apple" thing for so very long and doubled down on these assertions for literally decades in the face of people writing in to correct the info fairly regularly.
This stuff is why I'm so vocal about the fact that I think spec-dexes should avoid editorializing macs or to be honest, even "contextualizing" them because people who don't themselves fully understand the context get this stuff wrong all the time. (e.g. assertions that the IIvi/vx/P600 "replaced" the IIci, or that the Classic II replaced the SE/30.)
To make this all worse: The fact that LEMisms get taken as gospel here in the year of our lord 2022 when the whole thing was written as a secondary computer buyers guide litearlly 25 years ago is arguably so damaging to the hobby overall. I regularly continue to see people recommend against "problematic" machines that are often significantly easier to deal with and more logistically convenient than "the good ones" because of the misinformation/disinformation and its continued propagation. That leads to people vying for a relatively limited pool of pro oriented machines when their needs (often literally in line with what Performas were for when they were new) would be met by something much simpler and prices could stay lower for everyone in the hobby. That's a significant reduction on the overall dynamic of literally anything vintage as a hobby, but you get the idea. People ignore good inventory and then that inventory goes away and people fight harder over the remains.
fabulous article to link here
Thank you! Yes, here it is:
https://www.taylordesign.net/classic-macintosh/the-mythical-road-apple/
Your summary is exactly correct.
This is all more or less borne out by Apple's own documentation in terms of the developer notes, which have been floating around online for anyone who wants to have, for at least 25 years. The team over at Low End Mac
could have known this back in 1997, but decided not to.
The 476/605 is a cut down 610/650. The 630 is an updated 475/605. The 6200 is a 630 with a PowerPC upgrade pre-integrated onto the motherboard. The same thing happened a few times in the wintel industry around the 486/Pentium changeover, which, as with the Mac (68ks were on sale until mid-1996) took a long time.
In terms of benchmark numbers:
The 6200/75 and the 6100 (both /60 and /66) are neck-and-neck for most things. The 6200 wins disk benches (your own posts on reddit last year show this) and graphics. The 6100 wins a few other things, especially with a cache.
The 6200's weak 68k emulation performance is because of a small L1 cache compared to every other PPC model, 16 vs. 32k - you can mitigate that by running some slightly newer software, using PPC native applications, and using tools like Speed Doubler 8. 7.6.1 + SD8 is a sweet spot for the 6200, in my experience, but 8.1+SD8 should also work well, especially if you need or want HFS+.
Even the 6300, 5300, PowerBook 5300, 1400, and 2300, which share an underlying "it's a 630 with a PPC upgrade bolted on" architecture to the 6200, are typically thought of as "not having this problem". (Because they all have the 32k L1, although some PowerBook 5300/1400/2300 models don't have any L2 cache, while the desktop 6200/6300 and AIO 5200/5300 have 256k of L2 stock.)
Interestingly, something like a 1400/166
is meaningfully slower than, say, a 7600/120. The 7200/7500 likely also outperform it, and something like the 6360/160 should also outperform it. Again, all because of that cost-saving architecture. Although, ultimately these speeds were all "basically fine" for their markets.
I have a 6100 and a 6200 and can likely pull them out in the next few weeks, my results in MacBench 4 are also on vtools. I believe I also did some testing in a 68k native app, that's in Public on vtools.
Do you have a link to the IDE drive testing you mention? I couldn’t seem to locate the post.
I deleted my reddit account a few months ago, but I seem to recall that literally your own testing showed this.
That said, one of LEM's newer authors got their hands on a 6200 and tested it, their numbers show the 6200 marginally beating the 6100 at disk performance, that article is here:
https://lowendmac.com/2020/the-golden-road-apple-how-i-discovered-that-the-worst-mac-ever-wasnt/
However - I knew of someone that purchased one of these new - for just under < $4000 AUD which included a small RAM upgrade and matching CRT, this wasn't a budget price
I'm curious as to what the comparables and other Macs were selling for at the time.
In the US, the 6200 and 5200 typically sold for $1800-2500 or so. I think I'm quoted somewhere a few years ago saying $1200 but I think that may have been, like, edu pricing on an LC version. So, you can have bought like three 6200s and some LocalTalk gear for what a single 9500 cost. 6100 variants (including the Performa 6100 variants) were selling at the upper end of that range or above it and the PowerMacs 7200 and 7500 were solidly above that line.
Though, like, it's worth mentioning that the Mac retail supply chain was usually LOADED with old inventory for YEARS after a model had been discontinued, so if something like a Quadra 650 or 800 was being firesold alongside a 6200, there could have been plenty of reasons to consider going for the 650 or 800.
Apple had hilarious ("bad") rules around displaying prices and who, specifically, was allowed to even sell Macs at the time so we don't have an awful lot of print pricing examples for most Macs from that era, that starts being a thing in like 1997 or so.
The other thing is -- I've basically literally always heard Apple's pricing is unfavorable globally (you can read this as: "literally everywhere outside the United States") between things like not updating prices as exchange rates and picking the same numbers even when a currency is worth significantly more than the USD, which I believe has hit Canada and Australia hard sometimes.
6200 was released thirteen months worth of horrific fat binary ports to PPC.***
FAT binary should have been fast on the 6200, and any PowerPC Mac. Although, it's worth noting that system 7.5 on PowerPC was an almost completely unmitigated disaster and the x100 hardware was, in reality, often not all that hot even running native code. UniversalBinaries during the PPC -> Intel transition also performed excellently, above and beyond the Intel Macs being so fast that they usually outperformed the machines they were replacing even in emulation.
You also, really, "should" arguably be comparing like for like in terms of price
erformance and whether or not something is a "worthwhile" upgrade. So, compare the 6200 not to the 840av, but to the 605/630. Although that does fall apart a tiny bit because 68k had such a long tail, and the 630 really is such a good performer for its price tag, at least if you ignore the fact that it postdates the powerMac x100 family.
2. It used the 603 CPU which really needs a cache card like the 601 to really shine. I think the Performa 6200 shipped without the cache.
Sort of! I mentioned this above but it bears mentioning again: The 603 in the 6200 has 16k of L1 cache all 5200/6200 models and variant had 256k of L2 cache as stock. That's globally. As far as I have ever seen, there was absolutely no "Cheap International Version like the cacheless 6400/180" as was common on models both before and after the x200/x300 groups.
The best explanation I've seen for the root of why 68k emulation in particular is so slow is because the 68k emulation code thrashes in and out of the L1 cache. I don't know what specific trick Speed Doubler 8 does to make this happen but it works around that problem somehow. Speed Doubler 8 is supposed to also help basically all PPC Macs at 68k emulation, especially in the 7-8.1 era. 9 itself also had an updated 68k emulator from Apple that is likely better than SD8's but on the machines where it'd be relevant (6100/6200) 9 is really too heavily to actually use. (Although IME 9 does win benchmarks on the 6200, I'll have to pull out the macbench results but 9 on my 6200 turns in bench results a few percent faster than 7.6.1, even though it's markedly slower to actually use.)
The 6300/5300, and the PowerBooks that share this architecture (5300, 1400, 2300) all have the 32k of L1 cache, but some of those PowerBooks do not have any L2 cache, which
does impact those models performance, sometimes significantly.