Period-accurate benchmarking tools?

[This Does Not Compute]

I'm working on a project where I need to benchmark a couple of Macs from the 1993-1995 era to see how they compare, at least in terms of computational performance. I never really got too much into this during that time period, so I can't remember many benchmarking utilities. Can anyone recommend something appropriate that would work on both 68k and PPC?

 

[Kai Robinson]

I would definitely recommend Macbench 4.0, as this was the last version to support 68K, but also supports PPC.

Link on Macintosh Garden: https://macintoshgarden.org/apps/macbench-40

 

[alxlab]

Here's a few that I've used:

Speedometer 4.0.2
MacBench 4.0
Apple Personal Diagnostic v1.1.1 - 1.1.3
Norton Utilities 3.1.x

Traditionally I've used Speedometer for benchmarking since it's small I've saved records from doing benchmarks from many different computers.

I like the Apple Personal Disgnostic as well. It works on my Mac SE and I guess it would on a Plus as well. It has benchmarking but I use it mainly to test the system hardware and to check if the ram sticks I get are good.

Norton Utilities also has diagnostics and benchmarks but it's a lot bigger application than Apple Personal Diagnostics.

Haven't used MacBench 4.0 too much but it seems to be another benchmark used commonly.

 

[ScutBoy]

I have Speedometer handy, and while it's basic it tells me what I want to know. Plus, I figure if I use the same tool all over, the results should be fairly consistent within that universe of data.

 

[This Does Not Compute]

So I tried Speedometer 4.02 on both machines and it works great -- the results it generates feel accurate to me, considering the systems. MacBench, however, was a no-go on one of them. I tried versions 3 and 4, and for both, they just hung at the splash screen. I even wiped the machine and reinstalled Mac OS 8.0, no change. I suspect it's simply incompatible with this specific machine because of some of the upgrades it has, which is strange but perhaps would be asking too much considering what MacBench's main use case was. Speedometer reported the wrong CPU (like, by a lot) in the same machine. (You'll see what machine it is in an upcoming video.)

 

[Patrick]

are you also paying attention in WHAT you are testing?

that is, for powerpc, are you testing the 68k emulation or native powerpc.

 

[JDW]

I would recommend MacBench 3.0 (when it works) because you can then run it on an SE/30 too. If you go with MacBench 4.0, you can't use it on the SE/30. Maybe that doesn't matter to you, but that's my personal rule of thumb. Admittedly, I don't use OS 8 on my vintage 68K Macs, opting for speedier System software like 7.5.5 and below, which is perhaps why I've never had an issue getting MacBench 3 to work. I also prefer MacBench to other benchmarking utilities because it is the most extensive, taking almost 2 hours to complete. No other benchmarking app takes that long or does as many tests.

I do like Speedometer, but keep in mind version 3 and version 4 yield different results. So if you start with 4, be sure not to fall back to an older version 3.x because the results will be way off.

Also be sure to use Norton System Info too, as a part of Norton Utilities 3.5.x.

It's best to set colors to 256 (on color or grayscale Macs), and the cache to 128K, before testing.

There are other benchmark utilities too, like SCSI Director Pro 4.0, but that is mainly useful if you wish to focus on hard drive (or SCSI2SD) performance.

 

[Brutus75]

I am using MacBench 3.0 on my Mac IIci but it always sets the comparison to a Power Mac 6100/60. Is there a way to change this?

 

[JDW]

MacBench 3.0… it always sets the comparison to a Power Mac 6100/60. Is there a way to change this?

I don’t have the software open in front of me but I’ve used it enough to remember. In one of the menus on the right side, there’s a command that lets you remove any Mac displayed in the results. I always use that menu to remove the bothersome PM6100 results.

 

[phipli]

So my thoughts are that it depends what your purpose is.

As you say 1993 to 1995 that means you don't really need any of the ones that support System 6 and older, especially as they're often 68k code and don't benchmark PPCs with native code (some do, I'm not that bothered, I have other dislikes about them like where one popular solution seems to get "Math" and "CPU" backwards).

Really this leaves MacBench and Norton System Info.

My thoughts on this is that if you want a "Performance Quotient", MacBench is probably for you. It tries to give "real world" performance, with is a better indication of how a machine feels (especially if you work in publishing), but, my issue with that is it is opaque and carries any specific prejudices of the tests over. So if the scoring leans slightly towards some type of test that a specific machine does badly in, that machine will score badly even if it is good in your usecase (e.g. poor disk performance might taint the publishing graphics test, but your use case is games that run from RAM once loaded and depend on raw pixels to screen).

On the other hand, I always like Norton System Info because of the granularity - it tells you exactly what the individual test scores were, allowing you to work out how and where a machine scores well (as well as giving overall easy to read scores). This is generally what I use because when I'm comparing, I tend to be considering memory performance, or cast to integer, or trigonometric functions, or ability to translate an image vs draw a rounded rectangle or whatever. I like to understand the "why" and "what", not just a score. I realise this isn't what everyone is after. System Info also has an easy way of bulk exporting results which is super useful if you're doing an analysis of several machines and want to plot results or something.

As System Info doesn't immediately make it clear how to do things like drill down into individual tests, select which graphics card, or export results, I made a small writeup a while back. It might help you : https://elephantandchicken.co.uk/stuffandnonsense/?p=1790

Oh, Norton System Info also has cat pictures. That also endears it to me.

 

[phipli]

One more thing I use because it gives a very good comparison for a specific case, although it was a 1997 release, so a little out of your date range and only works on PPC, so isn't actually, in any way, meeting your criteria... but anyway, it is a useful thing for your toolkit!

The game "MDK" includes both a FPS counter, but also a graphics benchmark. It does a load of raw pixel manipulation that is really good for judging a Mac's ability to do raw, non-quickdraw pixel bashing. Basically, it is a good indication of how well a computer will do at non-hardware-accelerated 3D games like Doom, Descent or software rendered Quake or something.

It works on any PPC.

Here is the first video I could find I took of it, which is weirdly of it running on a CPU upgraded (66MHz 601) IIci running Mac OS 8.5, so possibly the single worst computer it will run on... but ignoring that...