As astute readers of this blog may know, I just replaced one of my monitors with a new Gateway 24" FHD2401 . It features a very bright picture - much brighter than the screen it is currently placed next to - a cheap Chimei CMV-221D . As a result, the photograph looked extremely different between the two screens. Just dragging windows between the two was pretty bad - they also has a mismatched number of lines and different picture height, 1200 vs 1050 .
I did much online research, and also had the Pantone Huey Pro 3 on loan from a coworker, and was unhappy with it. So, I decided for a slightly more expensive product, the Data Color Spyder 3 Pro. This is a neat little USB sensor device that comes with calibration software.
I bought it last saturday night, March 28 at MicroCcenter. I had the software installed, and my 2 monitors calibrated in about a half hour. The picture looked matched much better. To be fair, there is still a large difference due to the Gateway's brightness , but the colors match fairly well - the tones don't change much when moving one window accross monitors.
I repeated the process with the second computer on my KVM switches, which was also hooked up to the same two monitors . Again, no problem.
I was about to post a glowing review of Spyder 3 Pro, but did not get the time to post on Sunday. That's when I discovered the problem today : the Spyder's monitoring agent process, Spyder3Utility.exe, had consumed over 1.6 GB of RAM, all by itself , after I had run my NSS software QA tests overnight in a loop for 24 hours ! The only job of that agent process is to remind the user to recalibrate - I set it to check daily, and remind once a month - or take ambient light samples, a feature that I disabled. Apparently, there is a huge memory leak. The agent seems to leak some data for each process that is started on the computer. My software QA tests easily ran over a million processes overnight - my PCs don't have quad-core processors for web browsing word processing. If the machine sits idle, there is basically no leak in the agent, but while the QA tests run - which are all text-based programs, by the way - the leak is over 1 MB per minute, which accounts for the 1.6 GB overnight figure .
The good news is that this agent can be stopped without too many negative consequences for me. Stopping it does not lose the monitoring calibration. But the agent may still be needed to set the calibration at startup time. I haven't tried to disable it completely at boot time.
I have reported the memory leak to DataColor.
The agent also doesn't play too well with the KVM switches - when displays disappear or reappear, it gets confused. Another day, another problem.