I have a ton of time on my hands due to my recovery from Achilles surgery so I went on one of my favorite engineering subreddits and asked what subjects or topics new grads would be interested in. The response was overwhelming. I took my best shot at answering their questions and collected just a sample of the Q&A to post here on my blog.
It’s been a while since I was a student and I forget how much mystery and uncertainty there was in wading out into the real world. I recall just kind of flying by the seat of my pants and hoping things worked out. I put together this Q&A in the hope that some young engineer will find a speck of value in our Q&A exchange and it will help he or she along their path.
Question from Anonymous Engineer 1
I would be interested to know what areas you think are currently and will be cool and exciting in the coming years.
One of my biggest fears graduating soon is becoming a code monkey.
Response from PadoPado
Hey. Thanks for the q. Off the top of my head the hardware/OS interface is a current trend that I see as becoming increasingly important. In the past our CPUs/GPUs did very little in terms of communicating with the OS. The OS of course has a bunch of information that the hardware would LOVE to use like scheduling information and its own notion of busy-ness that can be communicated to the hardware. It’s a tough area because it spans a large swath of topics from hardware to OS to actual applications and its hard to find people that are good in all of these areas.
Another trend is the use of machine learning algorithms in our CPU/GPU firmware that detect and adapt to workloads. Architecturally, MCMs or multichip modules are the big thing right now and the tech seems to have legs but latency is going to be a big question mark as core counts or module counts increase. Also, you only become a code monkey if you want to be.
Question from Anonymous Engineer 2
A few years back, I had a chemistry teacher that had come over from Intel doing chip design. He wasn’t very articulate about exactly what he had done at Intel, but what I understood was that he mainly worked all the way down at the gate level, trying to optimize compute times.
The kind of work he did really turned me off from companies that *make* silicon, as opposed to company that *uses* silicon.
Can you talk about the different types of work that goes on at a company like Intel? Is it really a bunch of obsessive people working on one single chip?
Response from PadoPado
Wow, this is great. I take for granted the years I’ve spent in the industry and grown a bit blind to what is and what isn’t commonly known for a new grad. Your chemistry teacher sounds like he was doing some sort of physical design or layout work which is very very tedious stuff… apologies if I offended any layout folks… The good news is that working on chip design is incredibly diverse and I think it will make for a good topic.
Off the top of my head some of the very diverse sub-categories are layout/physical design, RTL, analog design, process engineering, product development (high volume manufacturing), CPU architecture, SOC architecture, GPU architecture, high speed interfaces, verification, emulation, power, post silicon validation, firmware, hardware/operating system interaction etc etc… I think I’ll need a bit of time to organize the different types of work that go on and do it justice but there really is a lot of creative work and tough and exciting problems just as there are routine and boring parts.
Question from Anonymous Engineer 3
I looked at your background in your blog post, you’ve had a cool career! Three things for me, in order of interest:
- Do you have any thoughts on ultra-low leakage/’sleep-mode’ semiconductor design?
- When you ran into a difficult problem you could not solve yourself, what was your strategy for finding resources? Any thoughts on how/if this changed over the years?
- I’m looking for a quality resource on signal integrity, is there a book/app note/something you came across you think deserves unique attention? If so, a quick review of what makes it so good would also be appreciated.
Thanks in advance for any input, and good luck with recovering.
Response from PadoPado
Hey! Thanks. I don’t have any particular expertise regarding ultra low-leakage semiconductor design aside from the notion that low leakage usually means slower transistors due to higher vt among other things and our low leakage transistors are usually placed outside of critical speed paths. For “sleep modes” I think you are referring to what I call “idle states”. That’s when you place a particular block in a sleep mode (clock gated, clock + power gated, collapse the entire voltage rail). The policies that dictate when to sleep and how deep to sleep is a particularly fun optimization problem because different workloads or circuit behaviors will have different sleep requirements.
With regard to solving problems I was lucky to have great mentors along the way and some of the best advice I got was that when I pick my mentors brains I should have a self imposed “question budget”. Too many questions and I am dangerously close to being annoying (lol) and am not doing the requisite legwork by myself. This concept really helped my mind to systemically arrive at the root of what I was really stuck on and often times I was able to answer the problem myself. I got the notion of the question budget about halfway into my career from a 1:1 chat with an engineering VP.
For signal integrity anything by Bogatin is great. He has this great exercise where you visualize being a signal as it travels through a medium that really clicked with my learning style. I have a link to a pdf and am not sure of the legality… but here it is. It’s an older edition so the frequencies we see today in modern designs may have shifted up a bit compared to the text book but the core concepts have not changed. Maxwell’s equations still staying strong after all these years.
Question from Anonymous Engineer 4
I have read your blog and i found it interesting, your first article. I think the topics that most of us would be interested to hear about would be regarding challenges and intriguing problems you faced and how you tackled them, how you approached the problem, what type of knowledge helped you and you accumulated while solving the problem.
Response from PadoPado
Hey, I was wracking my brain for a bit and I realize that most of the technical problems that I faced heavily involve trade secret or proprietary information. If I abstract away the details into safe generalities then the anecdotes kind of lose their punch. That said, I’ll try… Learning how to speak in other people’s “languages” has been my biggest tool in solving problems. By “languages” I mean having enough working knowledge to understand issues in terms of different sub-categories of EE and being able to tie it together. Analog guys have hammers and look for nails… digital guys have philips screwdrivers and are looking for screws… FW guys have a ruler, etc etc. I don’t have any of these tools but have a good idea of when to use each.
I will say that as I’ve moved up the engineering ladder the most difficult problems to solve have been human to human in nature. It’s one thing to produce a technical solution or an algorithm or methodology and it’s a whole other beast in getting the right people on board with your idea. Sometimes the best idea doesn’t always win out. This is something that I haven’t quite got the hang of yet but am making improvements.
Question from Anonymous Engineer 5
I have an ECE background but currently, and happily, work in software engineering at the moment. I would like to take my mix of hardware/software knowledge to the silicon area at some point. What are some interesting areas in firmware or digital logic that you see in the semiconductor industry?
Response from PadoPado
Power. It’s super in demand and is at the intersection of hardware and software with a little physics mixed in. A lot of power management features are controlled via firmware and EVERYONE wants to reduce power. This is true from mobile devices all the way on up to data centers where they care about power density.
Pado Pado 