ray@dsiramd.dsir.govt.nz (Ray Brownrigg) (07/27/89)
1) Does anyone have any measurements of the relative performance of colour vs. monochrome SPARCstations? I have some data which imply that on a colour 3/60, performance improves by a few percent (as much as 8% for the whetstone benchmark) when not running in the suntools environment. This was running *non-graphics* applications (specifically the C and Fortran compilers, and various CPU benchmarks). Thus one might expect to get some of this performance improvement for a monochrome workstation when compared with a colour one. More importantly, of course is the performance of graphics applications. I have heard of someone who selects monochrome operation to obtain noticeably better performance of a graphics application (S-PLUS dynamic graphics on a 3/60). Anyone have any data? 2) Related to this, I presume that the GX graphics accelerator is designed to enhance colour graphics, and so is inappropriate to consider for a monochrome workstation. Anybody care to elucidate? Ray Brownrigg domain: ray@dsiramd.dsir.govt.nz Applied Maths Div, DSIR ACSnet: ray@dsiramd.nz[@munnari] PO Box 1335 System: OLIVETTI/AT&T 3B2/400B+, System V R3.0 Wellington, New Zealand "unx -rules -OK"
khaw@parcplace.com (Mike Khaw) (08/16/89)
[Sorry to waste net bandwidth, I couldn't reach the sender in NZ] >1) Does anyone have any measurements of the relative performance of colour >vs. monochrome SPARCstations? I have some data which imply that on a >colour 3/60, performance improves by a few percent (as much as 8% for the I have one measurement on hand for a color 3/60, diskless, 8 Mb RAM, SunOS 4.0: 1) suntools -n $options # default gray background, see $options below 2) open 1 shelltool 3) swin -c # set click to type 4) cd / 5) run a C program 100 times that executes the same code as the "Redisplay All" root menu selection. when $options = "" # use /dev/fb; i.e., menus in overlay bit plane # windows in 8-bit-deep color planes elapsed time = 29 sec "-8bit_color_only -toggle_enable" # use 8-bit deep color fb for everything elapsed time = 30+ sec "-overlay_enable -toggle_enable" # use 1-bit deep mono only elapsed time = 17 sec In all 3 cases, CPU time = 0.4 user + 9.5 sys. By way of comparison, a diskless 3/50, 8 Mb RAM, SunOS 3.2 (mono, of course) did the same thing in: 0.6 user + 11.0 sys. 18 sec elapsed time. In other words, with a slower CPU, the 3/50 is faster than the color frame buffer on the 3/60. The reason is that the low-end (10-bit deep "cgfour") color frame buffer is just dumb memory, so the CPU has to move all the bits around. I don't have the specs on the Sparcstation frame buffer, but it sounds like it's just like the cgfour. You could always try what I did on my 3/110: live in a mono desktop and when you really need color, start up suntools -8bit_color_only and use adjacentscreens or switcher to move between the 2 desktops. Having 2 desktops didn't seem to slow things down noticeably, but then the 3/110 had a local 140 Mb shoebox with tons of swap space (a so-called "dataless" configuration that had just /, minimal /usr and swap, and got all user filesystems from NFS servers). Mike Khaw -- ParcPlace Systems, 1550 Plymouth St., Mountain View, CA 94043 415/691-6749 Domain=khaw@parcplace.com, UUCP={uunet,sun,decwrl}!parcplace!khaw