Nestress (and all other test roms I've seen) require a great deal of stuff to be working before they are helpful. They are great for checking high level stuff, like sprite 0 detection, number of sprites on a scanline etc. They are not good at, or intended for, testing lower level things like the correctness of each opcode. I had begun to design such a ROM last summer, but quickly lost interest. The idea was to manually test 3 or 4 opcodes to guarantee that they worked. Then, a set of test ROMS could be run. The first ROM in the set assumed that the opcodes which were manually tested worked, and used them to validate several more opcodes. Each successive ROM in the set assumed that all of the opcodes covered by the previous ROMs worked 100% flawlessly, and used them to test some others. What I found was that the bookkeeping required to figure out which tests required which instructions was way more complicated than I originally thought. Additionally, you need to run several tests for each opcode to make sure the flags are set correctly in all situations. I would be interested in continuing this work if there is anyone who would find it useful. I was originally writing it to test my CPU core, but it turns out that running demos and games tests that stuff well enough to get good results.