I've not seen this second fan so I do not know, but is it because the unit is getting too hot? What is too hot for this type of tech anyway?
110C but that is internal junction temp. These type of chips tend to have poor heat transfer so the internal temp can be 20-30C higher than surface temp. Still the cooling design seems to be a sub-optimal and my guess is the result of the company already buying hundreds of cases, fans, and other components for a 20W which will never exist (well outside of flawed simulations).
Someone correct me if I am wrong but I would imagine a
high density pin heat sink on the two FPGA and then a larger fan mounted on the side blowing across both chips would be more effective. You are getting a higher airflow, and it is better directed at the actual heat source. The unitary airflow would also allow more even cooling of both chips.
So IMHO :
The bad news is they are salvaging their existing inventory of parts resulting in sub-optimal cooling.
The good news is is more evidence that they are just poorly run enterprise not a scam.
The super good news is for people who like to tinker one probably could come up w/ a custom cooling solution which is more effective.
A side note:
The higher heat load and higher MH/chip makes watercooling interesting. Putting a $50 waterblock on a 8W FPGA (Spartan-6) is kinda silly as even a tiny fan can keep temps at optimal. Watercooling isn't going to improve thermals much, and since you are only getting 200MH/chip that adds a hefty $0.25 per MH in cooling gear. A higher output chip, especially one running hotter makes water cooling more "interesting".
Honestly my impression of the product improves as my impression of the company declines.
The reduced specs, delays, plausible sourcing (BFL drop that "blend of ASIC & FPGA" from your website it just makes you look stupid), and "bolt on fans" seems more inline with the type of enterprise which could source some last gen chips than one that would be making custom ASICs or current get sASICs.
So I may buy one just to fool around with the watercooling potential. Of course they would need to start shipping some product first so likely I just have to wait another .... (wait for it) .... 4-6 weeks.