Dell Dynamic Power Mode - PUG Forum - Precision User Group - Dell Community

Dell Dynamic Power Mode

Precision User Group

Precision User Group
A description has not yet been added to this group.

Dell Dynamic Power Mode

  • Hey I'm the Lead Performance Engineering for Precision Workstations. Most of my days are spent tuning our platforms to deliver the best performance and helping the product development team design for performance. I wrote a paper about one of the things we do to maximize system performance: dynamic power limits. 

    Depending on your application, performance is often determined by the component where you app spends most of its time. If an application spends most of its time performing operations on the CPU, increasing CPU performance will usually make your application perform better. CPU performance is often determined by CPU frequency, so increasing CPU frequency correlates back to better application performance. But, increasing CPU frequency requires more power, and power generates heat, so you need to deal with that waste heat.

    As the CPU starts to get hot, Dell Dynamic Power Mode adjusts CPU power down and then back up (very slightly) to maintain a desired temperature threshold. That way, your system can run right up to its limits and deliver better application performance.

    One of the coolest things about this (see what I did there?) is that, because it's dynamic, it reacts to your environment. Dell Dynamic Power Mode will make adjustments to power limits based on real-time temperature measurements, so the cooler your environment, the better your performance.

  • Hi AleX,

    the link is broken. Could you please re-paste'it.

    THx

  • Updated the link. Thanks for the heads up! Apparently this discussion forum engine doesn't like Chrome. IE worked.

  • I received a question about whether Dell Dynamic Power Mode is only applicable on mobile or also on fixed.

    Right now, Dynamic Power applies to mobile platforms, but it has been shown to be effective in fixed platforms as well and we'll continue to evaluate and enable it in fixed if and where it makes sense.

    The reason is, mobile designs are typically not limited by the max operating temperature of the parts inside it like the CPU and GPU, unless the design is a super-thick mobile system. The thinner the mobile system, the more limited you become by skin temperatures. Thus, the more performance benefit you can carve out by monitoring temperatures and dynamically adjusting power to stay within the temperature limit and not go over.

    Contrast this with fixed platforms (e.g. desktops, all-in-ones), where you usually have plenty of cooling and aren't limited by skin temperature. There's still benefit in that you can drive higher power for short periods of time to push the limits of the temperature rating of the part, but the improvements are much smaller here. The exception to this are fixed designs where you're constrained - for example airflow (thermal resistance) or the size of the heat sink (thermal capacitance). If you make a little desktop system or one that is passively cooled, you might want to monitor temperature and reduce power very gradually over time as it heats up, but in the short term when the part is cool, you can burst power up higher for short durations and yield higher performance in interactive applications, but still stay within the constraints of the design as the parts inside start to heat up.