a co-worker of mine has one, and has been generally happy with its performance. Seems to me like it can't be super accurate, but if you repeat your acceleration experiments and carefully measure the results, you should be able to get a reasonable idea of how your performance is affected by various changes.