Benchmarks won't tell you anything meaningful.
Benchmarks will only show you peak performance on batch tasks, they can't accurately model user interaction. A benchmark can tell you what your IO is like, or what your peak CPU performance is, but it can't tell you what your average event latency is, or how quickly it can respond to user input.
Pick the device up, use it. See how quickly the user interface moves in response to your inputs. See how often the UI drops a frame or two. Leave it be for about 10 or 20 seconds, then see how quickly it becomes responsive again. Load a webpage, scroll it up and down. On equal hardware (WP7 is single core 1Ghz Snapdragon) they should be similar in terms of responsiveness. They're not. (I'm no fan of WP7, btw, but it gets a lot of unwarranted abuse, normally from people who like Android)
" What I am driving at is that at least 128, 256 mb used to be quite common on cheap Android tablets, phones and even small netbooks. You still can get 256 mb tablet on newegg. "
And these low-memory devices are what got Android its reputation for having nasty user experience on cheap hardware. Once you installed a few apps with background services, the OS would spend so much time swapping and/or doing GC cleanup to free some RAM that it couldn't keep up with even basic user interactions. Games, which often pre-cache lots of graphics in RAM, suffer especially badly on this kind of hardware.
Actually, true multitasking like Android's (or before it, Symbian's) is actually bad for gaming, as other tasks can steal resources from the game at inopportune moments. In this respect, the dumb tasking of iOS or WP is better... but, frankly, only in that very specific use-case.