I hate to sound like an angry jerk, but I already explained the historical reason why it was there. I explained my own reasoning why I think I need it. I am not a FPGA guy, so may be he is right about being able to do everything he claims, but I don't want to risk removing parts and have to make an ugly mess of wire soldering it on later. I do know the PLL part a bit and play with their software so I know exactly what it can do. I want to try out a part in a prototype.
I planned it as it was a cheaper solution to make 3 clocks, (cost saving on 2) but it turns out the layout doesn't work that way. When everything is done, the cost difference isn't much. By then my parts are already ordered for checking my layout.
I had to make a decision, do I waste $8 shipping to buy a $1.77 cost part and rip out a section of the layout just to "save" $0.38? Do I remove a part that a competitor put on his board and risk not being to take advantage of the FPGA code with minimal changes if he actually releases it on a cold day in hell? They are also limited resources. So if the other design release a core that uses 5 PLL (they have 4 internal +3 external). May be they want to be jerks just to piss you off or really they found out they do need that. May be I *do* need to preserve that on chip PLL for my high speed BLVDS high speed serial expansion port for an accelerator that the other guys don't have.
Can you magically be able to make do with just 4 internal PLL without recoding it? Or that people could/would complain about not having that as a marketing feature? Talk is cheap when you don't have take the full responsibility for making that decision.
Not enough memory, lack of colour resolution those I can understand. They might not be my target market, but I do have ideas how I can cover them.
Sure people can be jerks and say it suck because it doesn't compete with a 20 years old high ended product aiming at a rich crowd, but at least that's telling me there are some marketing potentials. I looked at what is being offered in the current baseline products and the amount of memory they are offering. Do they say those products suck too?
Do I think 128MB is going to make a difference? Is 256MB the magic number? What is the cost to implement that to make the 10% people happy? By having the minimal # of chips to keep the signals clean so that I can run the chip as fast as possible. I am using 32-bit wide memory bus so that I can compete with DDR design without paying an extra $50 for my proto PCB. More memory bandwidth means that at some point I can get higher resolutions out of the FPGA. They probably never thought of that. May be they don't care. Feature creeps is one of the major killer of projects. I have real deadline and budget to keep.
Believe it or not, designing this board is my political statement to the existing product. I don't go on their forum and complain that their product suck. I don't care about retro gaming. I want networking, internet, HDD, USB, RTC and future expansion in the baseline product.
Sure you can speculate and talk all day. Has anyone actually done something about it? I fire my warning shot by showing that a guy can design a competing product at $200 in his home lab. By the way it is small and sexy too. Whether or not I actually follow up and make this a commercial product is a different matter. Until then, they would have something to think about at night.