Your corporate training will teach you there is no such thing as stereo types and they are bad, naughty. We all know they are true; it’s just some companies now of days try to force the worker bees to do a flash erase and drop your brain at the door. I never participated in that and as you can imagine it went very well. Dilbert is true…
I am asked many many times which is the best FPGA? After pondering for a pico second, I blurt out Xilinx. Flat out Xilinx has the better offering. Now I realize I lost about 47% of you, I’ll call you the 47%’s.:) Now put on your Chuck Schumer glasses and tilt them down so we can have a smart conversation. Try that look in your next performance evaluation and I guarantee you’ll get your 2.78% raise. Who am I kidding; the bucket was much smaller this year, sorry you owe us money.
I have personally seen resistance to the Xilinx product just because ‘We don’t want to be tied to one vendor’. There is some rational for that as I was burned by PA Semi, who wasn’t That was fun. I suggest that Xilinx is so far ahead of the game with FPGA technology that you must choose Xilinx at this point, they are well established and are going no where. And yes, I have programmed Altera. Xilinx has more resources, great IP and is the leader in the field. When the new node comes out, we hear rumblings’ that this is it for Xilinx but they seem to pull it off each time except for that interesting Virtex-4.
What Xilinx needs to be aware of or even nervous about is TOOLS. No one cares if you have the best IP, or 7 Billion transistors if you can’t route them in a reasonable amount of time. I was in a meeting with Xilinx and told them if Altera comes up with a 2x faster router, I’d trade performance for time if the performance is similar. GPUs’ and now CPUs are slowly eating away the market share pie for FPGA legacy systems because they are expensive to design and then route. Then as you know it’s always why is the FPGA not working or not ready, even though they are the heart of the system. I believe Xilinx’s Vivado HLS and Zynq have pushed them into the next level but they need more tools. They need a tool to answer this question for a group of system architects: How do I size my existing system into the Zynq Processor, help me draw that line. Now that is part of my job but companies need a tool to size the system, a quick rough estimate. Without this it is a tough sell to say, look take that Mercury or CW card and put it into this Xilinx part, really?
On the whole, all FPGA companies struggle to say what exactly they are selling. When they talk to us designers it is pretty good, but you know what? I do not have the pocketbook. They need to convey their message much better to program managers and other bosses how they can benefit their program. That is no easy task because as you know, some of these people at this level have had a lobotomy and must follow process at all times and cannot make a decision unless it is in a manual somewhere. Was that a stereotype? :rolleyes:
lang: en_US
Share this post via:
AI Semiconductor Market