Retro: Benchmarks through the ages
1994 hard disks were connected with an early version of the IDE (integrated drive electronics) interface. With a bandwidth of 33MB/sec, it was more than sufficient, since only the fastest hard disks could approach 10MB/sec transfer rates. Incidentally, hard disks were all formatted with the FAT16 filing system, meaning you couldn’t have a single partition larger than 2GB. This wasn’t a terrible burden, as the largest hard disks were only around 540MB anyway.
These days, the serial-transfer paradigm has sensibly taken over from the old parallel way of doing things, and bandwidth is frighteningly high. Issues of cross-talk and clock-skew meant that parallel interfaces had hit the buffers and extremely high-speed serial interfaces are now the way forward. Hence, we have SATA and PCI Express, both based on the same fundamental high-speed serial transfer technology. Where the PCI bus of 1994 could transfer 33MB/sec, the PCI Express interface of today can push 8GB of data from north bridge to graphics card every second. And where the PATA-based IDE hard disk interface of 1994 topped out at 33MB/sec, SATA can now shove 300MB/sec down the pipe. This is just as well, since a single high-end hard disk such as Seagate’s 15K Cheetah can now achieve almost 100MB/sec, and formerly stupendously expensive RAID multidisk capabilities are routinely built into motherboards costing £30. Just hook up a few consumer-level drives costing £50 or so and you’ve got 100MB/sec, no problem.
A quick glance at Windows XP Task Manager shows the Firefox browser occupying seven times more memory than a 1994 PC had in total.
Testing the old guard
It’s saying something in itself that it’s simply impossible to test a 486 system, circa PC Pro issue 1, in a meaningful way against a modern PC. With an average complement of 8MB of RAM against the 2GB of today, you couldn’t even begin to load up Windows XP. For a bit of fun, try right-clicking on a clear area of your Taskbar and selecting Task Manager. Hit the Processes tab and then click on the Mem Usage column to order your running programs by the amount of memory they’re consuming. When we did this, we found that a single instance of Firefox was using 56MB – seven times the total memory complement of a 486 PC. Windows Explorer was using 47MB and the Word software we used to write these very words had 27MB all to itself. Enough said.
But by December 1998 and issue 50 of PC Pro, PCs had reached an average of 80MB of RAM – in our Labs group test, the entrants were a mix of 64MB and 128MB systems. Average hard disk capacity was around the 6GB mark (the FAT32 filing system allowing larger partitions) and the favoured processor was a 333MHz Pentium II. This is the point at which it becomes possible to at least attempt to load up Windows XP and run our present-day benchmarks on an old system. So that’s exactly what we did.
Remember that in our current benchmark suite, the results are normalised to our 3GHz Pentium D reference system. This in itself is getting long in the tooth, and new PCs now routinely score over 2.00, meaning they’re twice as fast as the reference.
The front-end application to our benchmarks suite, understandably, doesn’t normally expect to see scores of less than 0.1 and got a little grumpy, so we resorted to trickery. We managed to get both our 3ds Max and Photoshop tests going, although the rest of our tests proved too much for the 128MB of RAM on offer. These two are actually good tests to run, though, since 3ds Max is almost completely CPU-bound, whereas Photoshop is a memory hog, thrashing the hard disk. It turns out that a PC from this issue is on average about 50 times faster than one of issue 50 vintage; a frame of our 3ds Max test completes in an achingly slow 28 minutes, against the 50 seconds or so of a modern PC. And Photoshop sits and grinds for over four hours before completing a single run of our tests, compared to the five-and-a-half minutes or so of a new machine. That, ladies and gentlemen, is progress.