Hello everyone,

I am a nerdy guy and hope to become a digital designer sometime in the future.

I am wondering which side has more career opportunities, given the current market situation?

I am thinking that it might be ASIC, especially for the trend in AI chip design.

Thank you

 

I’d urge you to keep learning about a broad range of devices rather than focus too early on a specific discipline. Semiconductor manufacturers will continue to develop ASICs where the application sees sufficient market growth to favor an integrated solution. As processor speeds continue to increase they encroach on applications where only an FPGA was fast enough. But the demand for higher speed will continue. When/if you decide to specialize pick what interests you most

 

at the start of your career, oh the fun.
you have 50 odd years of learning ahead of you
Just think, what technology you would have learnt 50 years ago ?
I started with DTL logic, valves, delay lines for fast digital logic of a few Mhz..
im retired now , was designing fpgas with Ghz io, and 500 Mhz clock, and built in Ghz CPUs
none of that was even thought about when I started .

the point is , most fpga and asic work is system level,
we learnt about how signals propergate on wires and cables, fluid dynamics, field theory, calculus , numerical approximation techniques , set theory, que theory , thermal modeling,how to machine and material theory and a bunch more .

turning the handle of writing code is a very minor part of what one does , and with AI that part of the job is diminishing fast.

Guided AI can code really well, and its getting better.
I look back on compilers for say C , when it first came out, we had to do all sorts of tricks to make it efficient.
I.e. multiply by 15, we used to do as a shift and subtract, now the compiler will do that if its better than using a built in multiplier.

 

FPGA vs ASIC

If you can do one you can do the other. There is not that much difference.

 

Hello everyone,

I am a nerdy guy and hope to become a digital designer sometime in the future.

I am wondering which side has more career opportunities, given the current market situation?

I am thinking that it might be ASIC, especially for the trend in AI chip design.

Thank you

I'm assuming when you refer to FPGAs that you are referring to FPGA-based designs, as opposed to designing the FPGAs themselves. Please correct me if I'm wrong.

I did both for many years. My primary job was as a full-custom mixed-signal ASIC designer, but I also did a lot of FPGA designs, mostly for test systems or the next level up from our ASICs in the customer's system.

While @ronsimpson said that there's not much difference, I would beg to differ. From a certain perspective this is true -- there's a lot of overlap. But there are major and fundamental differences, particularly for many kinds of ASIC work unless you are just gluing IP blocks together that others (a.k.a., real ASIC designers) created. Now, if you are talking about designing the FPGAs themselves, an FPGA basically is particular type of ASIC.

Since you asked specifically about career opportunities, I think that goes to the FPGA side of the ledger pretty firmly. How many companies do ASIC design and how many ASIC designers do they employ? How many new ASIC designs are created every year? Now look at how many companies do FPGA designs and how many new FPGA designs companies. Lots and lots of people design FPGA-based solutions as freelance engineers. Very few people do freelance ASIC design, particular as a one-person endeavor.

 

There are many sub-fields within the 2 broad categories you've defined. Designs have become so complex that you're unlikely to be doing an entire ASIC design yourself and possibly not even an FPGA by yourself. The sub-fields often end up broken down along the lines of the tools you'll be using to do the job, but will generally break down into logical design, verification, and physical design.

ASIC design also breaks down into gate-array, standard-cell, and full custom with complexity increasing accordingly.

I've done gate-array ASIC, standard-cell ASIC, and FPGA designs, but not since retiring ~15 years ago so my perspective may be flawed.

On the technical level, full-custom ASIC design is serious engineering with a lot more opportunity to flex your brain as (as I understand it) modern nanometer designs require understanding of the entire range of physics involved. This decreases somewhat as you go to standard-cell and gate-array designs.

In terms of design, the difference between FPGA design at one end of the spectrum and full-custom ASIC at the other is the amount of freedom you have. FPGA design is largely defined by dealing with resource constraints while full-custom ASIC is a blank sheet.

Verification is generally the biggest portion of the project with 70% being the rule-of-thumb in my era. I may take flack for this, but in my opinion it is a pure software function. The toolset may be different, but in the end you are writing non-production software to exercise another design.

Now that I've muddied the waters, I'll give you my opinion: FPGAs are where the work is, full-custom ASICs are where the money is.

ASICs have become ruinously expensive so that the number of companies using then has dropped considerably as a proportion of the market. Mask sets now cost in the million$ so the cost of making a mistake is huge. I suspect that this means that much more time is being spent on verification.

FPGAs are like writing software -- you can iterate your design an unlimited number of times as long as you don't need to change the PCB. This makes it far cheaper to incorporate into designs, especially when specifications are as "fluid" as they often are. Unless/until something else comes along, I see them as making up the vast majority of the designs. High-volume/high-performance designs will likely still be ASICs, but the crossover point keeps sliding higher as time goes on.

The downside to all of this is that AI is an unknown factor. I won't even try to guess how that will affect the future market except to say that your job is probably safer the more creative brain power you need to expend to do it.

I hope that helps a bit.

 

In general, I would pretty firmly second what @I_luv2work@nothin_all_day stated above. As with all things, there are exceptions, of course. Whether those matter depends on what you want to do and whether seeking out those exceptions is worth it to you (plus, you might just fall into one of them without realizing it). I worked for a very small company (never had more than sixteen employees while I was there and only had four (six shortly thereafter) when I started -- and that included the president). We did array-oriented lunatic-fringe designs, mostly SPRITE (signal processing in the element) arrays. The president did most of the analog pixel design and then usually one of us would do everything else -- and I mean everything. Interfacing with the customer, designing the rest of the circuitry, drawing the schematics, doing the simulations, laying out the chip (most of which were 100% hand laid out except for the pads and, sometimes, some random logic that we had a very simple standard place and route tool (that usually required a lot of manual rework to clean up) because the vast majority of the circuity had to mate to the array column and row ends. Then we did the design-rule and layout-vs-schematic verifications, followed by the database preparation and submission to fab. While in fab, we did the test system design and if that need a PCB made, we did everything for that, too. Then we tested the parts once they got out of fab and prepared the documentation for the customer. It was nice not being pigeonholed doing the same small aspect of design over and over, plus you really had ownership of the chip and were in a great position to help the customer work through any problems they had. But, it did mean that there was friction in the works since you might use a given tool, such as that standard place and route tool, once or twice every couple of years, and so you spend a fair amount of time refamiliarizing yourself on how to do basic things.