Baum’s customers find it extremely important to run a complete scenario on the entire design to uncover and fix “power bugs.” Using other solutions limit them to only a part of a design or just a small window of a scenario because of limitating capacity constraints or slow simulation performance. With Baum they don’t have to compromise.
Here is a good example of how Baum’s products helped one of our customers find and fix a major power problem in their mobile chip design. The design centered around a third-party graphics processing unit and they wanted to optimize power consumption early in the design cycle. In previous designs, they would run test scenarios to find peak toggle rates to estimate power around these peaks but were not satisfied that this was a robust approach because of the risk of missing power optimization opportunities. These concerns were confirmed after they added PowerBaum to their methodology.
Using PowerBaum power models to analyze the dynamic power behavior of a complete scenario they uncovered a serious design issue that they would have missed with other tools. Looking at the top half of the figure below in the “Before” situation, they found that two different idle periods (“1” and “2”) had very different power consumption. Looking closely, you can see that the second Idle period consumed almost twice the power as in the first idle period (38mw vs 20mw). This had significant impact on their product because the device spent a significant portion of time in the idle state.
They optimized their power gating circuitry based on information provided from PowerBaum’s power models. With the newly optimized design they re-ran the PowerBaum power analysis and achieved the power profile in the lower half of the figure (the “After” situation). As you can see, after fixing their “power bug” they were able to reduce power consumption by 90% in the second idle period and it matched that of the first idle period. Overall, they reduced overall average power consumption by 23% for the complete scenario.
An important takeaway in this example is that if they had simply looked at the window of peak activity when the GPU was actively processing data they would have completely missed the opportunity to reduce the idle power. This illustrates the importance of analyzing dynamic power waveforms of the complete scenario.
Ensuring the integrity of the power distribution network of a device is a key component in design closure. Identifying the conditions of peak power is critical to understanding the IR drop affects to the device and ensuring that it will operate reliably.
In this customer example, Baum’s customer ran many realistic scenarios to understand the exact conditions their design operated under peak power conditions. The figure below shows two power windows. The top waveforms show the complete scenario being analyzed, while the bottom waveforms show an expanded view focusing on the details of the peak power. In each case the red power waveform illustrates the original power waveform prior to optimization. They used this peak power profile to feed their IR drop analysis tools and discovered that they had to drastically reduce the peak power in this situation to ensure that the device would operate reliably.
Using the information provided by the PowerBaum model power analysis, they modified their design to reduce the peak power at the conditions identified by the Baum power analysis. After modifying the design, they reran power analysis using PowerBaum power models and reduced the peak power by 45% (as shown in the green power waveforms). Feeding this back into their IR drop analysis tools they confirmed that the design met IR drop requirements after optimization and would ensure the device would run reliably.