This article is part of the Technology Insight series, made possible with funding from Intel.
Earlier this year, Intel launched its most advanced datacenter platform ever, consisting of 3rd Gen Intel Xeon Scalable processors, Optane persistent memory and storage, Ethernet adapters, and FPGAs. The ambitious project seeks to power a broad range of workloads, from cloud to network to intelligent edge.
Getting there, especially during a global pandemic, was a herculean effort. Nobody knows this better than Irma Esmer Papazian, senior principal engineer at Intel and the 3rd Gen Intel Scalable processor’s lead architect. We recently had an opportunity to sit down with Irma and discuss her journey as a woman in technology, her experiences at Intel, and her influences on the company’s newest system-on-chip for datacenters, code-named Ice Lake-SP. It’s a key offering for Intel, as customers begin to increase data center spending.
Above: Irma Esmer Papazian, senior principal engineer and Ice Lake-SP lead architect, holds a 3rd Gen Intel Xeon Scalable processor.
The journey
VentureBeat: Thank you for your time today, Irma. You started at Intel in 1992, giving you almost 30 years of insight into the company’s culture and direction. Can you tell us a bit about the journey in getting to where you are today?
One key lesson I have learned in my life, and from nearly three decades as a microprocessor architect at Intel, is that sponsors and supporters are extremely important; they help provide the direction needed to achieve new heights.
Let me begin with an example from just before I became a teenager. I am one of two children born to an Armenian family in Istanbul, Turkey. One of my teachers recognized my love of learning and urged my mother to enroll me in a private high school. As I finished my first year at St. George Austrian High School, my father, passed away. I was 12 years old. After the initial shock and grief, I remember feeling I could no longer pursue my educational dreams.
I was wrong. My brother, 10 years older than I, had just graduated from college. He took a job and dedicated himself to supporting my mother and me during the next 10+ years of his life. My mother managed a very tight budget, counting every expense very carefully. And St. George High School provided a scholarship that allowed me to stay in school. These hardships motivated me. St. George, gave me a solid foundation, and taught me the self-discipline needed to prepare me for life. By the end of high school, I was fluent in Armenian, Turkish, German, and English.
I listed computer engineering as my top choice in the college entrance exams in Turkey. It sounded cool. I liked math and the sciences. I decided to give it a try, although I hadn’t touched a computer in my entire life. I graduated at the top of my class at university in 1989 and was awarded an IBM PS/2. But as my family lacked the funds to turn it into a game station, the computer was unfortunately never used.
After graduation, a cousin who lived in Chicago, played a crucial role in helping me move to the United States to attend graduate school. He introduced me to his college professor friend who also was originally from Istanbul. He visited us during the summer after my graduation and explained the application process for graduate schools in the United States.
Leaving home to go to the United States was a difficult choice. My mother was quite concerned for me to move halfway around the world. My brother helped her understand the opportunities that I would have if I moved to the United States.. The University of Wisconsin at Madison offered me a teaching assistant position and a generous tuition discount. I began the next chapter of my life.
Dream job in the U.S.
While I had come to the United States with the goal of doing my PhD, Intel came to the campus to interview students. Joining Intel was like a dream come true. So, I chose to stop my education after getting the master’s degree and joined Intel in Folsom, California in 1992.
I was hired into the graduate rotation program, which allowed recent graduates to join Intel without choosing a final job position within the company. This program enabled the employee to network and try up to four different engineering disciplines. At the end of the four rotations, I joined the new product architecture team.
I was the youngest person in my group, and the only woman, which was a bit daunting. Technology-wise, they were clearly ahead of me. They were mentors, who helped me establish myself at Intel.
I learned a lot from the hallway conversations and back-and-forth discussions right outside my cubicle. I thought at first that they were arguing, but then realized this is how progress is made, with no hard feelings afterward. Intel had a culture of “Constructive Confrontation” which created a very collaborative work environment, with no hard role boundaries. This approach empowers junior engineers to get exposure to discussions between their seniors, to learn from them and to actually be part of the conversation when they are ready. I learned so much from them. It opened my eyes, and as a result, I was able to learn faster. I became responsible for microcode development of one of the Pentium Pro processor versions.
When I got engaged, I told my manager of my plans to move from Folsom to the Bay Area. He helped pave the way for me to transfer to the architecture team in Santa Clara, where my first assignment was to model and analyze the memory hierarchy of the first Itanium processor. From 1996 to 2004, I worked on numerous Itanium products, and had two children during this period. In 2005, I started working on the Xeon server products, with increasing levels of responsibility for several Xeon processors.
Above: A wafer of 3rd Gen Intel Scalable processor dies, manufactured on the company’s 10nm process, each with up to 40 cores and 100MB+ of on-die cache.
I advanced relatively quickly and was awarded an Intel Achievement Award in 2009 for my to the Nehalem-EX program. And because my husband supported me throughout my career journey, I was able to balance my career while we raised our children.
“As you climb the ladder, the rules change”
I am sometimes asked if my gender has held me back or prevented me from taking on more responsibilities, but I generally haven’t felt this way at Intel.
There was. However, a point in my career when I felt I should have been promoted to the next level and this naturally didn’t happen. I learned that as you climb the ladder, the rules change. It was no longer sufficient for my direct management to agree that I am ready for the next level. I needed a wider support network, including people at senior levels in the organization. And, what I’ve broadly observed is that women are less likely to network up.
For the first time, I asked to be given a challenge that could lead to new opportunities. At that point in time I was the technical lead and manager for half of the Xeon server performance architecture organization; I wanted a new technical leadership and growth opportunity. I was offered to become the lead architect for the Ice Lake server processor project. I took the offer, while continuing with my existing role with the new role added.
Though I had enjoyed the social aspects of management and solving my team’s problems, I realized my plate had become so full, that was not no longer tenable, and something had to give. I expressed that to my management and given a choice, I embraced the opportunity to focus on technical leadership roles while being able to step away from some of the organizational requirements of management.
I started asking for what I wanted to do, and in that regard perhaps many women are a bit less assertive. Women engineers want to feel that they are extremely well-versed in an area before saying they can handle a job, whereas in my experience, I have seen many men just jump in. That was a new and important awareness for me.
Women in tech at Intel
VentureBeat: According to Intel’s 2020 workforce data, the company demonstrated progress in the advancement of women to senior positions. What do you see Intel doing, from your position inside the company, to advance women into more senior positions?
From my vantage point, I see Intel doing multiple things. One example is that we have an Intel-wide senior technical women’s conference that is held annually. It started as a tiny forum years ago, where you could fit the attendees in a large conference room. The last time we got together, pre-COVID, we filled an auditorium.
This forum has been great to bring women technologists together from all geographies and diverse technical backgrounds, to make sure they can network with each other and with a broad range of senior technical leaders at the company. Also, this forum creates a venue for up-and-coming senior technical women to present their work, which gives them wider visibility and with that better opportunities for career growth.Intel also has mentorship and sponsorship programs.
I think the sponsorships are great. Leadership not only supports the program, but helps with pairing— matching technical women with leaders across the company so they get the opportunities. This is helpful because a lot of the time, I feel that many women don’t make networking a priority and recognize its value.
Above: A manufacturing technician in Kulim, Malaysia displays a tray of Ice Lake-SP-based 3rd Gen Intel Xeon Scalable processors.
VentureBeat: In the same workforce data report, Intel says it saw a decline in U.S. female representation. What are your thoughts on why that was the case?
Well, 2020 was quite unique, right? Anyone who was able to work from home did, which was very different and difficult for many. The boundary between work and personal lives blurred, and many struggled with it. Certainly, in my work experience, I saw that men or women who were helping younger children or other family members at home experienced additional stress. I think it’s no surprise that women struggled to stay in the workforce, not only at Intel but everywhere.
VentureBeat: Understanding that many women are working harder than ever to juggle the demands of their personal and professional lives, do you have any experiences or guidance to share?
Prioritize. Think what is important to you? Personal prioritization comes first. After that, I think there is a need for clear communication. Once you talk to your manager, your coworkers, and those at home, and you know what you can do, everything gets better. And when you’re happy, it will show in your work.
Don’t be afraid to reach out, either. After doing the due diligence and putting in hard work on a given challenge, don’t wait too long to ask if you think you need help to solve a problem. If you think you asked for help but nothing happened, I would encourage you to think through if you were clear enough. Did you list the exact help you needed? Did you say what would be the consequence of not getting the help? Were you clear about the tradeoffs that would need to be made? If not, do that first. Instead of bringing problems, bring solutions and alternatives to your leadership’s attention.
Once you clearly communicate a request for help, the managers and technical leads around you should be able to help you find a solution. If they cannot, they will be happy to guide you further. In my experience, disappointments are most of the time due to unclear communications.
I often ask my mentees if they have a role model. If they do, what are the qualities of that person that they really like? Are there any they dislike? I found for myself that the characteristics in a person you look up to will change as you grow in your career. Understanding your next set of goals with clarity will put you more in control of your career.
Ice Lake: Composing, then conducting
VentureBeat: You recently served as the lead architect for Intel’s Ice Lake-SP system-on-chip. Today’s data center-oriented processor architectures are massively complex, involving the very latest in hardware, software, and process. With so many pieces in play, what does a lead architect’s average day consist of to keep a handle on something so sophisticated?
As the lead architect, you wear many hats. Right at the beginning, during the initial phase, you are like a composer. You have a somewhat blank sheet in front of you. You have high-level requirements and targets in mind, and you architect to them.
After you create the design, you then become the conductor. You need to know each instrument, but you don’t need to be a master performer of each. And you need to think critically about how each instrument is used and coordinate to make sure everybody’s playing with the right intensity, color, transitions and harmonies. Whether an orchestra or an engineering team, that’s where the greatness comes from. You become the glue between all the IPs, accelerators, and cores in the SoC.
Later during execution, you are still the conductor, refining the performance of each instrument, solving new arising issues, culminating in the optimal complete performance.
VentureBeat: That’s a good analogy. As a trumpet player, I understand the challenge of controlling intensity. Most conductors also have instruments they’re trained on. In that regard, is there any aspect of the Ice Lake SoC that you had more technical influence on than the others?
My background is SoC performance, so scalability was my key goal. We added PCIe Gen4, we added more memory channels and controllers, and we added more cores. In the end, I wanted to make sure each SKU could hit the highest performance levels and maximize efficiency. So, I would say my biggest role was solving those SoC bottlenecks, making sure the coherence flows are optimized, and when the whole processor is active, we have a fair distribution of the limited resources.
VentureBeat: With Ice Lake-SP now shipping in production, were there any learnings you’d like to share as a leader on that project?
Definitely. In every program, we learn new things. The use cases for more sophisticated SoCs are evolving and changing. Now we have a greater focus on AI and deep learning. At the same time, there is an increased need for general-purpose computing. You also have higher security, virtualization and networking demands.
So, at the end of the day, for new technologies we’re putting into our SoCs, we need to be proactive, building the test vehicles and software enablement ahead of the hardware. Sometimes this happens head-to-head. Sometimes the software isn’t available when the hardware is ready. And those are the times when we take longer to validate. I’d say we’re always working to improve the pre-silicon validation capabilities for software and hardware together.
VentureBeat: What’s your next project?
My transition to the next program actually happened gradually during the last phases of Ice Lake’s execution. Today, I am 100 percent focused on the next-generation Sapphire Rapids program, and I’m looking at the performance architecture of that program.
In general, when a program ends, we do postmortems and try to identify learnings that we communicate to our architecture and design communities. After that, some folks move on to the next CPU architecture. I was wearing the performance architecture hat for Ice Lake; I will continue to focus on performance architecture for Sapphire Rapids. I can’t wait to talk with everyone about the innovative features we’ll deliver in Sapphire Rapids – oh, and of course, the amazing performance.
Source: Read Full Article