The worldwide pandemic has accelerated the adoption of rising semiconductor applied sciences to satisfy market calls for, which has enabled firms with superior expertise to outperform their competitors. Greater than 50% of firms might want to construct new digital companies to remain economically viable, and restoration from the pandemic will contain everlasting adjustments to many dimensions of a company, together with the tempo at which it conducts enterprise; its core worth proposition; and the expertise.
With digital and technology-driven disruptions making a winner-takes-all dynamic in an increasing variety of industries, solely a subset of organizations is more likely to thrive. In at the moment’s aggressive semiconductor market, the place prime firms have sturdy and expansive expertise portfolios which are all the time evolving, a powerful expertise basis is essential for achievement. The time is now for these firms to make daring and revolutionary investments in superior expertise and digital capabilities.
Results of pandemic on digital ecosystem
The pandemic has amplified the necessity for expertise development and has inspired innovation throughout your entire digital ecosystem, from large information and synthetic intelligence (AI) to cloud computing and Web of Issues (IoT). Conventional brick and mortar retail firms have embraced expertise to stay related and meet the calls for of tech-savvy customers. Huge information has facilitated the digitization of assorted industries and exponential development in eCommerce. Moreover, with journey restrictions imposed practically worldwide, do business from home turned ubiquitous, leading to an unanticipated surge within the uptake of cloud gaming and high-performance computing (HPC) as a service.
Determine 1 The above chart exhibits the compound annual development price for server gross sales. Supply: Teradyne
The server market supporting AI hyperscalers is predicted to develop by 50% yr over yr by way of 2025, whereas cloud gaming CAGR is predicted to develop at an astonishing 72% by way of 2025. Nonetheless, tremendous high-performance methods have particularly difficult energy necessities clocking as much as 10.2 KWatts per server. The rising class of exascale high-performance computer systems—computing methods able to calculating a minimum of 1018 IEEE 754 Double Precision operations per second—and trillion parameter AI fashions for duties corresponding to correct conversational AI require months to coach, even with the processing energy of at the moment’s supercomputers.
Determine 2 In a typical server structure, all processing elements require energy utilization. Supply: Teradyne
Energy administration challenges as a result of larger demand for computing energy
As computing energy will increase, transistor depend per die will increase in tandem. Though course of node counts have decreased through the years, die measurement is rising as transistor counts double each 18 months. So, onboard actual property for energy administration units decreases to accommodate bigger processors. Consequently, rising present draw coupled with lowering availability of board growth space for silicon-based MOSFETs, which provide present to the processors, leads to an attention-grabbing energy administration problem.
Determine 3 The above information highlights battling present necessities and accessible design house. Supply: Teradyne
An insatiable urge for food for larger energy processors, for purposes corresponding to AI coaching servers, drives a considerable enhance in MOSFET drivers. To maintain warmth technology as little as attainable and maximize power effectivity, these units are designed with low RDSON to ship a whole lot of amps of present to the processors they’re powering. Nonetheless, high-volume MOSFET drivers with extraordinarily low RDSON measuring lower than 1 mΩ create challenges for semiconductor take a look at.
Take a look at challenges for reaching exact measurements
Testing semiconductors previous to the set up in remaining software is essential to make sure units meet specified necessities for the lifetime of their use. Sustaining aggressive value of take a look at (COT), whereas offering full take a look at protection, requires precision high-power instrumentation to function precisely and effectively.
Measuring precision voltages throughout the 1 mΩ gate resistance on the MOSFET driver requires tens of amps of present to circulate by way of. Excessive bandwidth, precision and energy devices from automated take a look at gear (ATE) can effectively measure RDSON resistance precisely. Nonetheless, parasitic resistance from the system interface boards (DIB) and a tool take a look at socket’s contact resistance, which might measure as much as 50 instances of the MOSFET driver’s RDSON, pushes the boundaries of sustaining optimum utilization of the take a look at cell.
Moreover, excessive present pulsing could cause magnetic coupling into adjoining traces, compromising measured worth for adjoining websites in excessive parallelism resolution. Sadly, the trade observe of protecting or carefully coupling high-current traces shouldn’t be viable when coping with current-induced magnetic coupling. In an effort to tackle this problem, the high-current traces should be laid out as broadside differential pairs to optimize the magnetic area cancelation. Pulsing substantial present by way of high-contact resistance generates extreme warmth and damages contact pins over time.
Exact RDSON measurements of units may very well be achieved by meticulously designing onboard circuitry supporting software calibration to get rid of path and call resistance. The onboard circuit has a secondary operate to make sure the secure operation of all devices deployed. Optimum throughput may very well be achieved by sustaining excessive gear effectivity by way of a perfect take a look at surroundings.
New energy devices with improved bandwidth, coupled with revolutionary take a look at strategies, assist to delay the contact pin lifespan by shortening pulse width and incorporating an ultra-efficient contact resistance examine earlier than every take a look at execution. Growing the facility instrument’s bandwidth delivers sooner DI/DT, translating to shorter take a look at instances and the opportunity of rising web site counts, leading to larger general throughput. Prolonging the contact pin lifespan additionally reduces consumables’ bills.
Boosting power effectivity with new supplies like GaN
As deep-learning AI turns into extra pervasive, the insatiable demand for computing energy will ensue and supporting energy administration semiconductors will expertise intense development. In the meantime, the carbon footprint from information facilities is attracting consideration and regulatory insurance policies are being enacted to make sure information facilities are geared up with energy-efficient gear. In 2019, information facilities consumed about 2% of the world’s electrical energy, however this quantity is predicted to rise to as much as 8% by 2030.
The effectivity of MOSFETs sometimes maxes out at 95%. To fulfill the rising power consumption of knowledge facilities, new supplies and processes corresponding to gallium nitride (GaN) are being developed to handle the shortcomings of conventional semiconductor supplies. With larger effectivity and switching frequencies, GaN energy provides ship extra energy than their silicon-based predecessors with an analogous footprint. “Turbocharged” GaN energy provides delivering larger energy on an analogous footprint may enhance general server density by as much as 56% on current racks.
Determine 4 The above desk compares materials properties of silicon and gallium nitride. Supply: Teradyne
Gallium nitride energy provides ship three advantages in comparison with silicon-based energy provides. First, current information facilities can enhance their information density. Second, extra environment friendly energy provides translate into decrease working prices. Lastly, the info heart can scale back its CO2 emissions as a part of the worldwide objective to realize net-zero emissions by the yr 2050.
The first trade problem with GaN transistors is the excessive dynamic on-resistance which is troublesome to measure when switched at required excessive frequencies. High take a look at gear producers are working to develop the precision instrumentation required to ensure GaN RDSON specs. Quickly, GaN will exchange silicon as the popular materials expertise for delivering energy, however till GaN’s exhausting switching dynamic on-resistance will be measured constantly and precisely, silicon-based processors will proceed to be utilized.
As demand for high-performance computing purposes, typically delivered as a service, will increase, semiconductor firms should rise to the problem to stay aggressive by adopting new applied sciences and processes. Advances in energy administration and new supplies like GaN will make sure the expertise is ready to sustain with the purposes driving it. Nonetheless, with these new applied sciences come quite a few challenges, each for manufacturing and take a look at. These that may be nimble sufficient to adapt will discover success in these new and rising markets.
Aik-Moh Ng is a product supervisor for analog energy take a look at merchandise at Teradyne.
Lauren Getz is a product supervisor for analog energy take a look at merchandise at Teradyne.
Associated Content material
