-
Notifications
You must be signed in to change notification settings - Fork 20
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Feature request]: MTK decoding latency semi-solution #113
Comments
Reports say that use Warp Drive or Warp 2 drops the latency on these MTK chips to 5ms. Have you tried that? |
Maybe that works for less powerful chips or chips w/o lowlatency codec, but for me warp drive introduces +2ms decoding and unstable decoding latency, warp 2 is more stable, but still somewhere at 22-24ms. Even Antutu split-screen is not helping them so much (10ms). |
They're on 9300 iirc, and it's using the correct low latency decoder. Package name wise you have to modify it yourself though, it's not possible for an app to change package name on its own. |
Those power limitations is crazy. If I do CPU Throttling Test in split screen delay goes from 22ms -> 8ms.
Oh, then... there are probably two cases: codec related (when Warp is efficient) and power management related (when background work is effective, but no impact from settings/codecs) FIX: |
Detailed CPU monitoring info:
|
Modified Artemis apk with ApkTool to mimic Antutu. |
Hi, i have an Alldocube Iplay 50 mini pro, with an mtk g99. |
For me, with Lenovo Tab 12.7 2025, Warp2 gives unlocked FPS, so when it goes up to 200fps - I get 60ms at overlay, but it feels like 300ms delay. FPS is locked at the options, btw. |
Try the latest Apollo, it now has frame rate limit enabled by default. |
Well done, now Warp Drive/2 is stable, but 20ms - still my maximum without background CPU load for waking cores up. |
Sadly you need a way to set all the cores to perf governor, or atleast put your top app/foreground cpusets to the big cores. If you are rooted you can do it with smarpatpack kernel manager, or with any kernel manager wich gives you the option to add custom tunables. Or, you can just find the Max fps that will give you the min latency (in my case is 160hz) and create a custom resolution with the desired Hz. Im using a hdmi dummy to do that, i dont know if you can do it with the virtual display. |
Poco x7 pro here. The codec being used is low latency, and at 60fps 1080p 20mbps, I have 20ms of delay. At 120fps 40mbps, the delay is 7ms when I touch the screen, but if I use the gamepad, it goes back to 20ms. In my opinion, this is a bad reading; even with 20ms, I don't experience input lag compared to playing locally on a monitor. |
Mmh i didn't explore the powersaving path. On my Xiaomi 14T pro i have a better decode time when i put the "ultimate" performance mode in battery settings. So perhaps there is the decoder issue but perhaps also a powersaving thing... What's weird, though, is that quickly bringing up and closing the notification bar lowers the decoding time! Frankly, now that mediatek are the #1 chipset on sale (they've overtaken qualcomm), wouldn't it be possible to put pressure on them to look into this bug? |
We have to put pressure on. Yes.. But how? |
However I repeat. I don't know if the problem is real (at least for those who have low latency working) or is a latency calculation error. I don't see a big difference, in fact I don't see it at all, from playing directly on a PC. |
There is a big difference on fast paced game like FPS. Test a overwatch or a valorant and you can't even hit. |
Post your apk please |
Here it is. Probably right version, but probably not. |
Nothing.. same latency for me |
I want to consider the well-known problem of increased latency on mtk chips.
I have Lenovo Pad 12.7 2025 (Mediatek Dimensity 8300), the average decoding delay - 20 ms (no difference in codecs, settings and other. h264 is faster by about 2 ms, which is insignificant).
Device supports c2.mtk. “codec_name”.decoder.lowlatency.
I noticed that the latency is software limitation, namely, if you do some work in the background (split-screen), it can easily go down to 14-15 ms.
But most importantly, if you open the initial page of Antutu Benchmark in the same split-screen, the latency drops to an incredible 6-10 ms (i.e. two or three times less than initially). Closing the benchmark, we get the standard and unpleasant 20ms.
The influence on power consumption is unexplored.
Maybe it is possible to bypass these artificial power management limitations?
Describe the solution you'd like
Is it possible to change the Artemis package name to mimic the benchmark utility?
But of course it would be interesting to invent some advanced tweak.
Describe alternatives you've considered
Split-screen with benchmark app.
Screenshots
No response
The text was updated successfully, but these errors were encountered: