Q60 and QTOP Index

QLheaven is exceedingly pleased that some of this blog's posts have come to the attention of the QL Forum and have even stirred up some discussion.

The Q60 QTOP-index measurement of 108.88 mentioned in a previous post RPi and QL emulator speeds that was considered to be wrong. This post was actually about the speeds of QL emulation on a Raspberry Pi using either uqxl or smsqmulator. This post showed that these 2 emulators on the modest RPi could perform better than many of the old QL type hardware systems.

Figures for the Q60 and QPC2 were thrown in for context and perhaps some more information regarding the set up of the Q60 could have been given as this was given for the surprisingly very high QPC2 speed which was largely down the PC hardware.

QLHeaven was itself surprised by the relatively low figure of 108.88 that QTOP had produced for the the Q60. But many factors can affect the production of benchmark figures. In the case of the Q60 the setting for the 68060 processors caches will markedly affect this as will the screen size and graphics mode in use. Other software running at the same time as the bench marking software tends to have a significantly lesser effect on the benchmark level produced.Although the slightly better performace of the Q60 with write through caches and 16 bit colour at 1024x512 is thought to be due to improvements in smsq/e over the 18 months or so since the last tests.

QLHeaven will freely confess to routinely using the write through cache setting instead of the faster copyback setting as copyback often screws up compiled SBASIC, such as QLiberated programs making copyback difficult to live with. Secondly the Q60 is almost always used in 16bit colour 1024x512 screen resolution, so it seemed fair to use a QTOP index with these settings. And here is the evidence of the effect of screen resolution,size and cache settings.

                 

                   Q60 80mHz QTOP speed test results

 

Screen Size	Screen Mode	Cache	QTOP Index
1024x512	16 bit	Writethrough	110.635
1024x512	16 bit	Copyback	122.523
1024x512	16bit	Copyback*	138.096
512x256	16 bit	Writethrough	110.006
512x256	16 bit	Copyback	137.756
512x256	mode 4	Writethrough*	148.326
512x256	mode 4	Copyback*	194.845

 

* Only software running - TINDEX

As can be seen the key influences on the TINDEX result are the cache mode, screen resolution and colour mode.