"... since when does a moving sign display need gigaflops?"
It doesn't of course. The argument Intel makes is the same one they have made in the past which is "here is a box your generic engineer can program that is cheap enough to run a sign."
The argument is based on a time/cost to market one. Lets assume you make 1,000 digital signs, and the total cost to the end customer is $10,000 per sign. So you have $10M in potential (but unrealized) revenue sitting there. You can either spend a year developing the software for this sign using an expensive embedded systems programmer for a year at $150,000, or you can spend two months developing the software with a cut-n-paste coder fresh out of college (annual salary $45K so 2 months is $7,500. So your RaPi solution costs $50 each (when you include a case and cables) vs $400 each. So over 1,000 signs that is $350,000 in 'parts' cost extra to use the Intel solution but you also save $142,500 in labor costs, and you are in the market 10 months earlier.
So is that time to market advantage (and the cheaper programmer) a win or not? And since you're looking at $10M in revenue $142,500 is about 1.5% of your margin your giving up.
And finally there is the point that if you do decide to make digital signs out of RaPi's and you can't get the crappy blob-of-bits Broadcom tosses in for their GPU control what are your options? You've got a volume of 10,000 and Broadcom could care less. With Intel they will, with proper documentation, give you every piece of information you need, they already have the support team in place to support their huge OEM business.
When you come at the problem that way it looks a bit different.
>So is that time to market advantage (and the cheaper programmer) a win or not? And since you're looking at $10M in revenue $142,500 is about 1.5% of your margin your giving up.
It's about 1.5% of your revenue. What it is of your margin depends on what your other costs are (and it's likely much higher).
True, although with the Intel box you also have PSU, cooling and cabling and you find that multi-headed video cards and distribution cables to drive lots of display end up costing far more than putting a single board Linux box in the back of each TV.
But the real argument against a Wintel sign is to head over to thedailywtf.com any day and see pictures giant billboards showing BSOD or asking for a driver
Seems like that second point is about the "win" part. Surely a PC board is going to have much better support running Linux than an ARM SoC. The linked article is about hardware, not windows.
The first can be true or not. There are Intel PC boards available at all price ranges down to $80 or so. It's only the very bottom of the SoC range (i.e. the Raspberry Pi, a system so bleeding edge you still can't actually buy one anywhere) where you start to see price competition.
Yes it's only the rPI that actually delivers on the CHEAP linux SoC so far -- otherwise you can get a VIA x86 board for the same $100 as most ARM.
What is interesting is that rPi finally delivers "the computers will be just bumps on cables" prediction. With HDMI out and powered by USB it becomes very nice for a distributed video screen display compared to the normal multi-headed PC solutions.
It's just odd that if Intel thinks the future is competing with ARM in tablets with lower power Atom it produces something that is essentially a high end desktop replacement
machine.
It doesn't of course. The argument Intel makes is the same one they have made in the past which is "here is a box your generic engineer can program that is cheap enough to run a sign."
The argument is based on a time/cost to market one. Lets assume you make 1,000 digital signs, and the total cost to the end customer is $10,000 per sign. So you have $10M in potential (but unrealized) revenue sitting there. You can either spend a year developing the software for this sign using an expensive embedded systems programmer for a year at $150,000, or you can spend two months developing the software with a cut-n-paste coder fresh out of college (annual salary $45K so 2 months is $7,500. So your RaPi solution costs $50 each (when you include a case and cables) vs $400 each. So over 1,000 signs that is $350,000 in 'parts' cost extra to use the Intel solution but you also save $142,500 in labor costs, and you are in the market 10 months earlier.
So is that time to market advantage (and the cheaper programmer) a win or not? And since you're looking at $10M in revenue $142,500 is about 1.5% of your margin your giving up.
And finally there is the point that if you do decide to make digital signs out of RaPi's and you can't get the crappy blob-of-bits Broadcom tosses in for their GPU control what are your options? You've got a volume of 10,000 and Broadcom could care less. With Intel they will, with proper documentation, give you every piece of information you need, they already have the support team in place to support their huge OEM business.
When you come at the problem that way it looks a bit different.