Hacker News Reader

Comment by LeoPanthera | Hacker News Reader

> The iMac was underpowered even at introduction. But it showed that computers had already reached the point at which they have more processing power than needed to accomplish almost anything you needed a computer for.

...is that actually true? Because it launched with a 233MHz G3, and although my memory is hazy, I remember people specifically pointing out that on most available cross-platform benchmarks it beat whatever Pentium II was current then.

That situation didn't last, but I have specific memories of "This is a cheap home computer that's faster than Dell's highest end tower".

View on HN

As someone that bought a 603e based all in one around that time, I remember the introduction of the G3 being almost comparable to the intel->apple silicon transition.

That was definitely my impression one generation later when they released the G4. To this day I still miss that chassis design.

Sometimes I wonder how different the world would look today if Apple had become large enough to justify the investment the G5 required and macs were running on those.

I suspect even if the Intel transition hadn't happened, assuming the Mac survived it probably still would have been destined for Apple Silicon once the ARM platform became powerful enough.

The G3 was a beast compared to the 603. IIRC it was about 30% faster clock for clock. By going with a G3 in the iMac seriously beefed up the capabilities of their consumer machines. The iMac was more than twice as fast as the similarly priced Performas sold the day before the iMac's introduction. It was even more powerful than even higher end Performas approaching twice the price of the iMac.

One of the more tragic examples, from the perspective of laptop buyers, was the WallStreet PowerBooks. They started at $2300 and was easily 30% faster than the high end 3400s that were $4-5k. The original G3 PowerBook (Kanga) was over $5k. It was a huge boost in power for half the price and honestly much better industrial design. The 3400 was a chonky boi.

Was the performance true on Apple marketing benchmarks or in actual use?

Apple has a looooong history of cherrpicking benchmarks, especially when their arch wasn't as fast as competing Intels.

Actual use. You’re probably thinking of the G5/PPC970, which was on introduction more of a case of “the fastest, provided absolutely the only thing you care about is vector operations”.

I'm mostly asking because as much as I remember the G3s weren't really noticably faster than my (then) Windows machines... but memory can mislead.

It may also have changed while you owned one: the stuff which was optimized native code ran quite well but there were also plenty of rushed ports or things running under Rosetta. I don’t think it was Photoshop but I remember some of the designers I knew commenting that one app shipping a true optimized port was like getting a new computer.

Two big things that affected the cross platform experience and perception of speed at the time were the twin pillars of the late 90s web, Internet Explorer and Flash, both of which had superior optimization for Windows and x86.

The introduction of Safari is an underrated milestone in changing that, and I would suggest the Mac didn’t really match the Windows browsing experience inclusive of Flash until the introduction of Intel Macs.

Stock iMac had 64Mb RAM, 5400 RPM based hard disk, and a poor 2D ATI based video card. And Mac OS 8.1. I actually purchased it (I was young and naive), and the Mac version of Unreal (game) to network play against my homegrow PC (Pentium 200, with Voodoo Rush video card), and the game performance was night and day (in favour of Wintel). MacOS 8.1 was so bad, I ended up installing YellowDog Linux on it, collected dust, then eventually gave it away.

The G3 was more powerful per cycle than a Pentium II.

Ditto with G4's and similar cycled Pentium III's.

But intel (and AMD) scaled up MHz/GHz a lot further. Motorola was stuck at 500MHz as the Athlon and PIII were already racing to 1GHz.

The G4 at 800 MHZ was still much more powerful than a Pentium III, or at least on par.

"much more powerful" is not anything like "on par". So what if that beautiful and expensive chunk of metal and plastic was slower in games and such?

I had a dual-866MHz G4 (MDD) and in most things my Tualatin P3-1200 at the time absolutely beat it. But the G4 was worth using until it literally caught fire just after the warranty ended.

Motorola?

Motorola has a long history of making processors, such as the 68000-series. They powered Macs (and Amiga)

From Wikipedia:

> The PowerPC 7xx is a family of third generation 32-bit PowerPC microprocessors designed and manufactured by IBM and Motorola (spun off as Freescale Semiconductor bought by NXP Semiconductors)

Supplier of the CPUs we're talking about (alongside IBM).

Motorola was one of the companies behind the PowerPC, but those found in the iMac was made by IBM.

No one saying it competed with high-end machines of the day. But it was quite powerful compared to PCs targeting a home audience.

The other thing to remember, is that computers were only barely powerful enough back then. Powerful models felt powerful precisely because we were desperate for every last CPU cycle we could get.

Barely powerful in early mid-90's. From 1996 an the multimedia PC and a bit later witht the Pentium MMX, the power gap from a 286 based DOS PC from 1991 was something else.

It's completely wrong that the hardware was underpowered.

The bottle neck was always macOS 9 which was a pretty rudimentary OS by today's standards.

It didn't have a SIMD unit. Although x86 only had MMX at the time, which wasn't much better.

Also MMX was basically the first mainstream use of SIMD so adoption was glacially slow since you had not only limited compiler support but also no libraries and little programmer familiarity with the style. That took a long time for knowledge and tools to improve, especially in the nascent internet era where people couldn’t just look things up on a comprehensive reference site or get help on GitHub.

I'd say there's still limited compiler support for MMX. Because of the strange way you have to use it (it trashes the floating point registers until the next `emms`) compilers won't use it even if it's beneficial, so no autovectorization.

A lot of the performance complaints came from software that was built for the 68k and ran under emulation. As software built for PPC came out, that complaint went away.

What is funny is that macOS 9 was so snappy on the GUI side but you had to start employing RAM disks and given different names to the same application to run them simultaneously. It could really scoot but occasionally a dialog box might bring everything to a halt until hitting OK or Cancel.

The CMD640 IDE chipset used was utter trash and most of the bottleneck. It wasn't until the G4 series that apple stopped slapping it in everything. A system with the CMD640 only used for cdrom, and a bootable SCSI PCI card performed a lot better.

https://www.penny-arcade.com/comic/1999/04/28/silly-rabbit-q...

It was a bit of a sore spot For a while there.

No it is not true. Especially relative to Macs. The original iMac also discontinued serial and SCSI with the use USB.

I think mostly people think of "games" when they think of performance.

>...is that actually true?

It is not.

Reply on news.ycombinator.com