Architectural Power & Performance System Engineer- 118581

The Role:

Ideal candidate will be using a mixture of knowledge of Machine Learning, Computer Architecture, Multimedia Accelerators, Power Delivery, Power Consumption, thermal solutions and their impact on chip performance to deliver future AMD products with industry leading performance while remaining competitive in perf/watt and battery life stats. Their findings and insights will be used directly to impact our power management algorithms within our power management firmware!

The Person:

We are looking for a passionate individual to join our highly visible team to deliver the highest performing and most power efficient and innovative products to millions of users worldwide.

The role is multidisciplinary, and we encourage applying if expertise in ANY of the following areas is possessed:

• Knowledge of CPU micro architecture subunits such as (but not limited to)

  • Pipeline stages of modern x86 or ARM processors or GPUs

  • Familiarity with Rasterizer, Shader, Texture, Sequencer etc in GFX pipeline

• In depth understanding of workings of any following motherboard components and related standards such as

  • Power saving states of on mother board devices like SSDs, PCIe devices, Wifi, WWAN, WLAN chips and USB devices

  • DDRx/LPDDRx Memory controllers

  • Knowledge of JEDEC standard is a huge plus

  • Knowledge of PCIe standards is a huge plus

• Knowledge of APU power consumption, saving and thermal management concepts such as

  • Clock gating and power gating mechanisms

  • Sources of power losses within an APU such as activity, leakage etc

  • Familiarity with performance throttling to reduce activity and dynamic power

  • Reliability and latchup protection mechanism

  • Battery life improvement techniques

  • Understanding of battery peak and rated power and impact of temperature and peak battery power on battery capacity degradation

  • Familiarity with on motherboard power supplies such as bucks, LDOs etc

• Applicant is comfortable developing micro-benches for SOC IPs

  • To exercise DDR reads/writes, multimedia encoders/decoders

  • To initiate IO reads/writes to devices such as USB, PCIe, SSD using various sub IPS

• Applicant understands CPU or GPU ISA (instruction set architecture)

  • Their understanding of ISA enables them to write micro-benches to exercise different Sub IPs within CPU/GPU

  • Applicant is comfortable with RTL

• Applicant has experience improving

  • Perf/Watt

  • Quality of Service

• Applicant has hands-on experience in

  • Use of machine learning for workload profiling/classification and or characterization of any hardware system/IP

KEY RESPONSIBILITIES:

Build desktop and mobile systems with best-in-class performance, perf/watt and battery life. Make use of tools like supervised and unsupervised ML techniques to profile workloads, use targeted micro-benches to exercise a wide array of IPs to collect the necessary data and perform interaction analysis between different algorithms within PM FW. Improve current features and develope new ones. Participate in product definition.

PREFERRED EXPERIENCE:

• Experience with Pytorch, Tensorflow, Keras and Scikitlearn is a plus!

• Strong programming skills: C or C++, and a scripting language (Perl, Python, Ruby, etc.)

• Understanding of concepts relating to:

• computer organization/architecture

• power/thermal management

• Know your way around lab equipment, including oscilloscopes, logic analyzers, etc.

• Strong analytical/problem solving skills and pronounced attention to detail

• Outstanding interpersonal and communication skills

• Must be a self-starter, and able to independently drive tasks to completion

ACADEMIC CREDENTIALS:

• Bachelors, Master’s or Ph.D. degree in Electrical/Computer Engineering or Computer Science with emphasis on computer architecture and/or workload analysis

•