I am also interested in this chip, and I'm lucky enough to get 3 pieces of V53A in PGA package. After confirming that the chips are usable, I just stopped testing them because of the cost of the PCBs. However, after watching a video about a self-made 8088 laptop, and many people suggests the uploader to add a 8087 coprocessor, I decided to find out if the V53A supports a 80287XL coprocessor.
Answering the question doesn't seem to be easy, because the method of the connection between the CPU and the FPU isn't available on the Internet. NEC may have published it in the V33 and V53 user manual, but unfortunately, the manual seems only have Japanese version, and NEC hasn't uploaded the manual to their website. Thanks to the Renesas Japanese website, some hardware manuals and application notes about V53A are available in Japanese version. Though the extra information doesn't help much, it's still useful for on-chip peripherals interface design.
I'm lucky enough that I didn't make many mistakes on this. Decoding the NPCS, NPWR and NPRD signal of 80287XL using the V53A bus status signals, and then the 80287 will work. I found this correct solution in my second attempt, and then I'll get a working board, right? Unfortunately, the answer is no. I always get the wrong answer from the 80287XL no matter what instruction I wrote. At first I thought the coprocessor requires 16-bit memory, so I spent some time to get a memory module. But things get even worse, the program can't run at all, the CPU immediately cracked and switched to halt mode. So I have to debug the whole circuit. Finally I find something unexcepted, and I'm sure that causes the problem. I used a 74LVC1G125 for every 16-bit module, indicating a 16-bit operation. However, the '1G125 brings propagation delay as long as 900 nanoseconds, making it unable to read from 8-bit devices. I'm not sure if the '1G125 is fake, or my circuit went wrong, but when I desoldered all the '1G125s and used my 8-bit memory back, everything seems to be solved. The 80287XL starts to output the correct calculation result.
So from the start to now, I'm always focusing on how V53A can be used, instead of its performance compared with 80286 or something else. The performance is still the least important thing I should focus on, since the bus sizing is still a problem. Also it's irresponsible to claim my discovery before confirming everything is correct. However, if someone is interested in designing a PC compatible using V53A, I can say it is totally doable and worth a try.