There are some well-known chess test suites the CT800 had to crunch, and ELO 2100 seems realistic. You can also see the detailed results (TXT). All benchmarks were performed without overclocking.
Colditz suite: measures the tactical strength. It is aimed at players around ELO 1900, which means strong hobby chess players. The result is ELO 2344, which is not realistic. The conclusion is that this test is too easy to be meaningful, but still, the tactical abilities are clearly above hobby level. So the CT800 really fits the expert class.
BT-2450 (German only): these 30 positions were analysed in analysis mode for 15 minutes each. The recorded time is how long it takes for the correct move to appear and then stay, or 900 seconds if that does not happen. The formula for the playing strength then is ELO = 2450 − time/30, which is the total time in seconds. This test is considerably harder than the Colditz test, and the result is ELO 2135.
BT-2630 (German only): a variant of the BT-2450 where the positions 13, 15, 16, 21, 23, 28 and 29 were changed, and the base value has been raised to 2630. That modifies the formula to ELO = 2630 − time/30, but the other conditions are the same as in the previous test. The CT800 did not solve any of the replacement puzzles, and the result is ELO 2144.
A final mini benchmark for the computing speed is using the analysis mode on the initial position. Here is how long it takes to reach different depths, given as plies / seconds where the depth refers to the brute force base: 6/1, 7/4, 8/20, 9/56, 10/203.