This could be due to Mac OS having performance optimizations to run on Mac hardware? As far as I know this has been happening since the first Mac from 1984 (128k) and the release of Windows 1.0 (1985). And in the 1990s, Mac computers usually had slower clock speeds than most standard PC systems,but they were still faster.
This could be due to Mac OS having performance optiminizations to run on Mac hardware?
Artificial benchmarks aside, in real-world applications Macs were generally faster than Windows because the OS was indeed 'optimized' to run on Mac hardware - or more correctly, the hardware was optimized to run a Graphic User Interface. In comparison, Windows had to use whatever hardware was already in the PC, which was generally not optimized for GUI operation.
And in the 1990s, Mac computers usually had slower clock speeds than most standard PC systems,but they were still faster.
Clock speed by itself doesn't mean much, particularly when comparing different architectures. For example a 486SX-25 and 386SX-25 run the same clock speed, but one is several times faster than the other.
The 1990s was a time of very rapid advances in PC performance, so trying to make a valid objective comparison between Mac and Windows in this time period is difficult. However I did find this in the April 1991 issue of Computerworld magazine:-
The article claims that "the Macintosh line, from the entry-level Classic to the high-end Macintosh IIFX, consistently outperformed its PC competitors - often by as much as 70%". Notably the 'biggest leads' were in graphics and publishing applications, suggesting that GUI optimization was a major reason for the better performance.
Naysayers may cite CPU benchmarks in an attempt to prove that the PC was actually faster than the Mac, but this actually reinforces the point that GUI-heavy applications ran faster on the Mac despite its 'slower' CPU. That is the advantage of building the GUI into the system from the start, rather than trying to bolt it on later.
But was there ever a time when the Mac was always faster than Windows on an 'equivalent' PC?
The original Macintosh 128k was released on January 24, 1984. It had built-in 512x342 pixel monochrome bitmap graphics, 128k of RAM, 64k ROM, and a single 400k 3.5 inch floppy drive. What would the equivalent PC be?
The IBM PCjr was released in March 1984. It also had built-in bitmap graphics with 640×200 pixels in 2 or 4 colors, 128k RAM, up to 64k of ROM and a 360k 5.25 inch floppy drive. Surely this is the closest contemporary PC to the original Macintosh?
But no. Microsoft Windows 1.0 required a minimum of 256k RAM and two 360k floppy drives. So the PCjr couldn't run it. If a version of Windows had been produced for the PCjr its performance would been much worse than the Mac, due to the slower 8 bit CPU. The original IBM PC and PC/XT were not much better (a contemporary review of Windows 1.0 in the New York Times described it as "akin to pouring molasses in the Arctic").
So the closest equivalent would have to be the PC/AT. It had a 16 bit 80286 CPU running at 6MHz, 256k RAM (expandable to 16MB), a 1.2M 5.25 inch floppy drive and 20MB hard drive, and could take an EGA graphics card. With those specs it should have been as fast as the Mac if not faster - at least on paper.
Only one problem. The PC/AT was introduced on August 14, 1984, but Windows 1.0 was released on November 20, 1985. So for nearly 2 years the Mac was indeed faster than Windows on any PC, since it did not exist! But had it been available earlier the majority of IBM machines (PCs and XTs) would have run it much slower than a Mac.
Why Mac systems were always faster than Windows in processing performance?
Mind to give any proof to this claim?
According to various Benchmarks, the PCs usually outperform Macs of the same time.
For example a classic Macintosh (68k @ 7.8 MHz; *1) delivered about 0.40-0.52 Dhrystone MIPS (*2), while a PC/AT (80286 @ 6 MHz) scored 0.40 to 0.71 DMIPS - that's close to double the performance of a Mac, depending on compiler used.
Results for Whetstones are rather unfair, as already the first PC/XT could feature an 8087 FPU delivering 115 kWhetstones (a PC/AT with 80287 did about 300) where a Mac couldn't deliver more than 15. It wasn't until the 1987 that Apple offered an FPU (68881) with the top line Mac II and the mid 1990s until FPUs were delivered with volume models.
Bottom line: Macs were never faster than PCs of the same time.
*1 - That's any Mac before the introduction of the Mac II in 1987 or SE/30 in 1989.
*2 - Classic Dhrystone 1.1
If you'll forgive the imposition, I've been on a benchmark hunt. These tend to skew later because they assume some sort of convergence of application support. Of course, in sticking with a CPU benchmark I'm assuming that the interesting case is application software that does a lot of processing.
On the early days of the 68000 versus the 286, I found both Motorola's "AN ARCHITECTURAL CONTRAST: The M68000 Microprocessor Family and the 8086/IAPX 286" and Intel's rebuttal, both predating the launch of the Mac. The Intel response is better for citing Motorola's sources, but I was unable to obtain them online regardless, making the whole exercise very he-said-she-said. But my takeaway was that the 80286 would likely have the edge if it were used properly, though that verdict is muddied by a lot of squabbling about register quantities (68000 win!) versus access scheduling and how much effect wait states have (80286 win!).
Closing the 68000 era, I found this on the 68040 versus the 486. It applied various synthetic benchmarks:
- Drhystone (i.e. heavily integer-oriented): the 25 Mhz 68040 was twice as fast as the 25 Mhz 486 (and the 25 Mhz 486 was 20% faster than a 50 Mhz 68030);
- Linpack (i.e. floating pointy): the 25 Mhz 68040 was three times as fast as the 25 Mhz 486 — the article strongly implies it's a 486DX as it lists FPU registers in its mini-spec sheet, but it isn't explicit.
Caveat: the 68040 has a pipeline that's about 50% longer than the 486s, neither does branch prediction, and at least Dryhstones aren't very branchy. So the 68040 probably benefits in a way that it might not with actual application code.
In the mid-PowerPC era, I found http://macspeedzone.com which has a few broken links but still contains the relevant graphs for benchmarks comparing:
- Mathematica, where 400Mhz PowerMac G3 just barely trails a 400Mhz Pentium II;
- vector processing comparing Intel's Signal Processing Library to G4 AltiVec versions of the same functions, in which a 500Mhz G4 trounces a 600Mhz Pentium III;
- SPEC, in which:
- when performing integer arithmetic: a 500Mhz G4 beats a 550Mhz Pentium III but is slightly outpaced by a 600Mhz Pentium III — top of the heap: an Alpha 21264; top of the PC/Mac heap: a 650Mhz Athlon, 25% faster than the G4;
- when performing floating point arithmetic: a 500Mhz G4 beats all of the Intel options and, just barely, the 650Mhz Athlon, but is heavily bested by a bunch of SPARCs, Alphas and PA-RISCs;
- BYTEMark, which skews Mac-happy by:
- integer: rating the 400Mhz G3 as almost 250% as fast as the 400Mhz Pentium II; and
- floating point: rating the 400Mhz G3 at 144% as fast as the 400Mhz Pentium II.
But in net I think there's enough there to disprove the statement that Mac systems were always faster than Windows. It would seem more likely that they started either at parity or slightly behind, overtook during the 486 era, in theory roared ahead during the very early PowerPC period but was back in a tussle at least as early as the Athlon, and then sank slowly backwards as you head up towards the Intel transition.
With the heavy 'in theory' caveat on the early PowerPC days: in the real world it took several years just to get all the 680x0 code out of the OS, never mind applications. Having a fast processor that spends a decent amount of its time emulating another is only so much of a win.
Although the "computer speed" is totally subjective and all those ***stones are just synthetic tests, let's take this question seriously, but from a different point of view.
There are differences between the Intel x86 and the Motorola 68x family. Motorola has a slightly better performance and can do a few more memory operations than an Intel on the same frequency.
The Windows/DOS combination, at least in Windows 1 you are mentioning, runs in real mode, with almost no benefits of protected mode or 32bit architecture. It locked the whole Windows into the "640kB world, segmented". On the other side, the 68k family was "32bit by nature" with full memory access, so it could do some operations faster.
The other thing is "user feeling". Mac OS had a better-optimized UI, so the user did not notice annoying slowdowns (as in the Windows did).
Are we comparing oranges to oranges?
A difference in the amount of RAM could explain the performance difference.
I recall years ago learning a story that would help explain this.
Macs typically had more memory than their PC counterparts in many colleges(/universities).
Many colleges figured out that PCs were cheaper than Macs. And so they gravitated towards using PCs.
However, certain college programs like "graphics design" and "desktop publishing" reported that Macs provided better performance, and this was really needed for the tasks used by these specific industries. The key reason that the Macs performed better was simply that the Macs tended to have beefier hardware, including notably higher amounts of RAM.
The colleges were able to afford the higher prices for Macs, as long as they weren't buying too many of them. So the college programs that needed the better hardware were supplied with the more expensive Macs which were supposedly necessary, while other departments got cheaper PCs since they weren't crying so loudly about an apparently-justified need for the more expensive hardware. The PCs that ran word processors and spreadsheets could often get by with less RAM, and so the colleges saved money by buying computers with less RAM for the general-purpose computer labs.
The end result is that Macs got a reputation for working better on graphics programs in college. Then, the college graduates went into the industry and told their employers that they were Mac-trained because Macs were better, and so many companies cooperated by getting Macs for the departments that were going to be heavily staffed by such trained graduates.
So, if you want to have meaningful comparison between brands (like Apple vs. others), make sure you know how much RAM were in the machines. If you don't know, then that could very well have likely been the cause of the performance difference.
The users' perception of "processing performance" on Macs vs. Windows PCs is likely influenced by 3 unmentioned differentiators:
- The relative performance of all versions of MacOS (macOS) vs. all the contemporary versions of Windows OS.
- The particular applications that the users are actually running.
- The extent to which the rest of the system has been adequately equipped to handle the particular applications.
For #1, it should not be particularly controversial that the OS shipped with Macintosh computers has, for the most part, maintained a performance advantage over contemporary Windows PCs when viewed through the subjective lens of user experience. An OS "feeling responsive" is mainly an issue of system integration; hardware and software working well together. Since the Macintosh OS has always been able to target a narrower set of hardware variations, this is easier and better achieved than for PCs that must target a much broader range of hardware. Just as is alluded to in the question, "optimizing for the hardware" requires the developers to know specifically the hardware that is being targeted. Since Apple is actually in the computer business, this was always part of their software development "DNA".
Number 2 and #3 are best illustrated through examples, I think. Take Desktop Publishing for the Mac vs. Computer-aided Design (CAD) for PC's. An early adopter of Desktop Publishing applications had a relatively easy choice for a Macintosh. Aldus' products ran acceptably on Macs, and extremely well on Macs specifically expanded for that application, usually with higher resolution displays and plenty of RAM. The exact same can be said for AutoCAD users and PC's. Note that in the AutoCAD case, Windows was not initially a consideration.
The point is the choice of application dictates everything else about the system, including the right OS and the right hardware add-ons. Most AutoCAD users also chose specific graphics cards that included accelerated 2D drawing capabilities and very high-resolution frame buffers in order to get that optimized "processing performance" they needed. Such graphics cards were readily available for PCs to feed this market.
In summary, computers are not just CPU's, which compute certain quantities at certain speeds. They are complete systems of Hardware+OS+applications that are (optimally) tuned for the application workloads that their users are demanding from them. That's sort of why these are called "Personal" Computers to begin with, and why it should be obvious that comparing them based on something as mundane as MIPS or clock speed or numeric benchmarks is just a curiosity and not a recipe for deciding whether a given system is best for a given user.