The Detailed Explanation of GPU Turbo

Under the hood, Huawei uses TensorFlow neural network models that are pre-trained by the company on a title-by-title basis. By examining the title in detail, over many thousands of hours (real or simulated), the neural network can build its own internal model of how the game runs and its power/performance requirements. The end result can be put into one dense sentence:

Optimized Per-Device Per-Game DVFS Control using Neural Networks

In the training phase, the network analyzes and adjusts the SoC’s DVFS parameters in order to achieve the best possible performance while minimizing power consumption. This entails trying its best to hit the nearest DVFS states on the CPUs, GPU, and memory controllers that still allow for hitting 60fps, yet without going to any higher state than is necessary (in other words, minimizing performance headroom). The end result is that for every unit of work that the CPU/GPU/DRAM has to do or manage, the corresponding hardware block has the perfectly optimized amount of power needed. This has a knock-on effect for both performance and power consumption, but mostly in the latter.

The resulting model is then included in the firmware for devices that support GPU Turbo. Each title has a specific network model for each smartphone, as the workload varies with the title and the resources available vary with the phone model. As far as we understand the technology, on the device itself there appears to be an interception layer between the application and GPU driver which monitors render calls. These serve as inputs to the neural network model.  Because the network model was trained to output the DVFS settings that would be most optimal for a given scene, the GPU Turbo mechanism can apply this immediately to the hardware and adjust the DVFS accordingly.

For SoCs that have them, the inferencing (execution) of the network model is accelerated by the SoC’s own NPU. Where GPU Turbo is introduced in SoCs that don’t sport an NPU, a CPU software fall-back is used. This allows for extremely fast prediction. One thing that I do have to wonder is just how much rendering latency this induces, however it can’t be that much and Huawei says they focus a lot on this area of the implementation. Huawei confirmed that these models are all 16-bit floating point (FP16), which means that for future devices like the Kirin 980, further optimization might occur through using INT8 models based on the new NPU support.

Essentially, because GPU Turbo is in effect a DVFS mechanism that works in conjunction with the rendering pipeline and with a much finer granularity, it’s able to predict the hardware requirements for the coming frame and adjust accordingly. This is how GPU Turbo in particular is able to make claims of much reduced performance jitter versus more conventional "reactive" DVFS drivers, which just monitor GPU utilization rate via hardware counters and adapt after-the-fact.

Thoughts After A More Detailed Explanation

What Huawei has done here is certainly an interesting approach with the clear potential for real-world benefits. We can see how distributing resources optimally across available hardware within a limited power budget will help the performance, the efficiency, and the power consumption, all of which is already a careful balancing act in smartphones. So the detailed explanation makes a lot of technical sense, and we have no issues with this at all. It’s a very impressive feat that could have ramifications in a much wider technology space, eventually including PCs.

The downside to the technology is the per-device & per-game nature of it. Huawei did not go into detail about long it took to train a single game: the first version of GPU Turbo supports PUBG and a Chinese game called Mobile Legends: Bang Bang. The second version, coming with the Mate 20, includes NBA 2K18, Rules of Survival, Arena of Valor, and Vainglory.

Technically the granularity is per-SoC rather than per-device, although different devices will have different limits in thermal performance or memory performance. But it is obvious that while Huawei is very proud of the technology, it is a slow per-game roll out. There is no silver bullet here – while an ideal goal would be a single optimized network to deal with every game in the market, we have to rely on default mechanisms to get the job done.

Huawei is going after its core gaming market first with GPU Turbo, which means plenty of Battle Royale and MOBA action, like PUBG and Arena of Valor, as well as tie-ins with companies like EA/Tencent for NBA 2K18. I suspect on the back of this realization, some companies will want to get in contact with Huawei to add their title to the list of games to be optimized. Our only request is that you also include tools so we can benchmark the game and output frame-time data, please!

On the next page, we go into our analysis on GPU Turbo with devices on hand. We also come across an issue with how Arm’s Mali GPU (used in Huawei Kirin SoCs) renders games differently to Huawei’s competitor devices.

The Claimed Benefits of GPU Turbo: Huawei’s Figures The Difficulty in Analyzing GPU Turbo
POST A COMMENT

64 Comments

View All Comments

  • GreenReaper - Wednesday, September 5, 2018 - link

    > all members of the invited press to the show, typically around 500-2000, are sampled

    Curious.... so, what you're saying is that a Chinese company is going all-out to provide devices designed to monitor personal data to every possible tech journalist - and they can now coincidentally no longer become root to investigate them...?
    Reply
  • Smell This - Wednesday, September 5, 2018 - link

    Creative Engineering

    (i.e., imaginative and innovative marketing ___ to cheat)

    ;-)
    Reply
  • yhselp - Wednesday, September 5, 2018 - link

    If GPU Turbo is actually capable of smoothing out frame-time, it might be a huge win for consumers. It doesn't matter how pretty a game looks, or how fast it renders, if it has an inconsistent frame-time, making the controls sluggish. Any person who cares about gaming on a smartphone, especially something like PUBG and MOBAs, should find responsive controls preferable over higher image quality and sheer framerate. If Huawei manages to deliver smoother gameplay on its devices, it would be a huge win, despite the lower image quality.

    Hopefully they make tools to test frame-time easily available. We can't expect Digital Foundry to manually test every supported Kirin device and game.
    Reply
  • s.yu - Friday, September 7, 2018 - link

    It's a huge loss for consumers in the long term if they get away time and again with their deception. Reply
  • Achtung_BG - Wednesday, September 5, 2018 - link

    I love anandtech, good job! Reply
  • Flunk - Wednesday, September 5, 2018 - link

    "Up To" always makes me suspicious, always. I could claim that adding my sticker to your phone could offer up to a 200% frame rate increase and still not technically be incorrect. Reply
  • s.yu - Friday, September 7, 2018 - link

    I agree, how about citing the 1st percentage of performance numbers, read a spike for a second and stick it on the ad. Reply
  • mazz7 - Thursday, September 6, 2018 - link

    Huawei is surely moving fast in this mobile arena :) look forward into the new device that use Kirin 980 Reply
  • abufrejoval - Thursday, September 6, 2018 - link

    Can't say that I care that much about mobile gaming any more, did some on the tablets but since those only receive left-over hardware these days, I'll just stick to the desktop for gaming.

    I could be attracted to buy a Kirin 980 device, simply to play with the NN accellerator, but with a locked down bootloader, they locked themselves out of my MVP.

    Hopefully a HiKey980 model will be available shortly, that doesn't have this crazy limitation, I was getting very close to buying the HiKey970, when the Kirin 980 appeared.

    I still fail to see the rationale behind the lock-down decision and I wish you could have drilled a bit into the engineers in Berlin to find out why they changed their policy. If it's all about hiding the architectural weaknesses you so regularly expose, it doesn't really seem to be working.

    It's a little irritating that HMD/Nokia is following the same path and it's easy to see why phone for the Chinese domestic market would be locked down by order of the government to ensure any government trojan is properly protected from removal.

    But EU/free world imports without unlockable bootloader should really be banned.

    If it's all about compensating for the lack of a secure element or HSM on the phone e.g. for payment, it's just a very bad design choice. I understand why Google didn't want the Telcos charge rent on SIM cards, but an approach like the L4 kernel on a dedicated CPU chosen by Apple seems a better alternative.
    Reply
  • s.yu - Friday, September 7, 2018 - link

    So it's settled then, by Anandtech as always. Great work! Reply

Log in

Don't have an account? Sign up now