[Impeller] Severe performance degradation on Adreno 840

Steps to reproduce

Clone the repro project

https://github.com/edTheGuy00/bubble_perf_test

Run the app with Impeller enabled on a Snapdragon 8 Elite Gen 5 device

flutter run

Observe 5 - 20 FPS
Run the app on the same device without Impeller

flutter run --no-enable-impeller

Observe results

Cause

Rendering Animated bubbles using RadialGradient shaders and TextPainters, from my testing it works fine on all my devices:

60 - 120 FPS

Xiaomi 12 - Snapdragon 8 Gen 1 Adreno 730 ~ 120 FPS
iPhone SE 2 ~ 60 FPS
iPhone 13 mini ~ 60 FPS

But there is severe performance issues on the brand new OnePlus 15

5 - 20 FPS

Device affected: OnePlus 15 (Snapdragon 8 Elite Gen 5, Adreno 840 GPU)
Android version: 16
Flutter version: 3.38.4 (Impeller renderer)

CONFIRMED FINDINGS:

Impeller on Adreno 840: ~5-20 fps (BROKEN)
Skia on Adreno 840: ~120 fps (WORKS)
This is NOT a debug vs release issue - both have the same problem
Works fine on older devices (Snapdragon 8 Gen1 / Adreno 730) with Impeller
Works fine on same device with Skia renderer
Works fine on Simulator and Emulator

All recommended Flutter performance optimizations are implemented:

Shader caching: RadialGradient.createShader() cached per bubble, only recreated on radius/color change
TextPainter caching: Reused across frames, only recreated on style change
TextStyle caching: Stored to prevent identity comparison failures
Paint object reuse: Single Paint instance per bubble
RepaintBoundary: Isolates painting from widget tree
Position-independent shaders: Canvas transform used, shaders created at origin
Settled state detection: Stops repainting when bubbles settle
Small bubble optimization: Solid color for bubbles < 20px (no gradient)

Additional context

The Adreno 840 (Snapdragon 8 Elite) moved from Tile-Based Rendering (TBR) to Immediate Mode Rendering (IMR):

Higher sensitivity to frequent shader state changes
Different memory/cache behavior than older TBR GPUs
Penalizes per-frame allocations more heavily

Expected results

Newer devices with the Snapdragon 8 Elite Gen 5 should see smooth 120 FPS

Actual results

5 - 20 FPS

Code sample

https://github.com/edTheGuy00/bubble_perf_test

Screenshots or Video

Screenshots / Video demonstration

[OnePlus 15]	[Xiaomi 12]
Record_2026-01-14-12-39-40_6591e8d729632762c0d2c9c34b8ad961.mp4	screen-20260114-123310.mp4

Logs

[Paste your logs here]

Flutter Doctor output

Doctor output

$ flutter doctor
[✓] Flutter (Channel stable, 3.38.4, on macOS 26.2 25C56 darwin-arm64)
[✓] Android toolchain - develop for Android devices (Android SDK version 36.1.0-rc1)
[✓] Xcode - develop for iOS and macOS (Xcode 26.2)
[✓] Chrome - develop for the web
[✓] Connected device (5 available)
[✓] Network resources

$ flutter --version
Flutter 3.38.4 • channel stable • https://github.com/flutter/flutter.git
Framework • revision 66dd93f9a2 (6 weeks ago) • 2025-12-03 14:56:10 -0800
Engine • hash 360877569ab6632a564a0a8a815b2e0fe5ae294a (revision a5cb96369e) (1
months ago) • 2025-12-03 20:56:06.000Z
Tools • Dart 3.10.3 • DevTools 2.51.1

@edTheGuy00

Hi @edTheGuy00, what version of flutter were you using initially?

Can you test on the master/beta channels to see if this still reproduces over there?

Unfortunately, all my devices use a Mali GPU and I can't reproduce this issue on them.

@danagbemava-nc

Hi @danagbemava-nc , I'm Running my attached sample app https://github.com/edTheGuy00/bubble_perf_test

seeing the exact same results across the board only on my OnePlus 15:

Impeller ON ~ 5 - 15 FPS

--no-enable-impeller ~ 90 FPS

Flutter 3.38.6 • channel stable
Flutter 3.40.0-0.2.pre • channel beta
Flutter 3.40.0-1.0.pre-570 • channel master

@edTheGuy00

Thanks for the update.

@edTheGuy00, one final thing, from what version of flutter did you start experiencing this issue?

@danagbemava-nc

@danagbemava-nc I started seeing the issue when I got my new OnePlus 15, I guessing the issue was always there but I Was building on Flutter 3.38.4 at the time.

@danagbemava-nc

@danagbemava-nc I suspect this issue might affect all Adreno 800 series chips, I added this fix-around on my app for now to see 120 FPS on my OnePlus 15 and for anyone else experiencing the same issue.

    companion object {
        private const val TAG = "ImpellerConfig"
    }

    override fun createFlutterFragment(): FlutterFragment {
        val shellArgs = FlutterShellArgs.fromIntent(intent)

        if (isAdreno800Series()) {
            // Force Skia (disable Impeller) for this specific hardware
            // Adreno 800 series uses IMR (Immediate Mode Rendering) instead of TBR (Tile-Based Rendering)
            // which causes severe performance degradation with Impeller's rendering approach
            shellArgs.add(FlutterShellArgs.ARG_DISABLE_IMPELLER)
            Log.w(TAG, "Adreno 800 series detected. Disabling Impeller to avoid IMR performance degradation.")
        }

        return FlutterFragment.withNewEngine()
            .flutterShellArgs(shellArgs)
            .shouldAutomaticallyHandleOnBackPressed(true)
            .build()
    }

    /**
     * Detects if the device has an Adreno 800 series GPU (Snapdragon 8 Elite / Gen 5).
     * These GPUs use IMR (Immediate Mode Rendering) which has performance issues with Impeller.
     *
     * Known affected GPUs:
     * - Adreno 830 (Snapdragon 8 Elite, SM8750)
     * - Adreno 840 (Snapdragon 8 Gen 5, SM8850 - expected)
     */
    private fun isAdreno800Series(): Boolean {
        // 1. Check strict SOC model names (Android 12+)
        if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.S) {
            val socModel = Build.SOC_MODEL ?: ""
            // SM8750 = Snapdragon 8 Elite (Adreno 830)
            // SM8850 = Snapdragon 8 Gen 5 (Adreno 840 - likely)
            if (socModel.contains("SM8750") || socModel.contains("SM8850")) {
                Log.d(TAG, "Detected Adreno 800 series via SOC_MODEL: $socModel")
                return true
            }
        }

        // 2. Check hardware codenames (Fallback for older Android versions or missing SOC_MODEL)
        val hardware = Build.HARDWARE ?: ""
        // "sun" is the codename for Snapdragon 8 Elite
        // "pakala" is another known codename
        if (hardware.equals("sun", ignoreCase = true) || hardware.equals("pakala", ignoreCase = true)) {
            Log.d(TAG, "Detected Adreno 800 series via HARDWARE: $hardware")
            return true
        }

        return false
    }

Thanks for the update.

Labeling this for further investigation

face same issue after enabling impeller on Adreno 710

Device data:

Field	Value
Device	Motorola Moto G96 5G
Chipset	Qualcomm Snapdragon 7s Gen 2 (4nm)
GPU	Adreno 710
RAM	8GB LPDDR4X
Display	6.67" pOLED, 144Hz, 10-bit color
Android version	15
Flutter version	3.41.6 (stable)
Engine hash	5cdd32777948fa7a648fac915f8da7120ac7e97a
Impeller enabled	Yes (opt-in via manifest)

This issue is marked P1 but has had no recent status updates.

The P1 label indicates high-priority issues that are at the top of the work list. This is the highest priority level a bug can have if it isn't affecting a top-tier customer or breaking the build. Bugs marked P1 are generally actively being worked on unless the assignee is dealing with a P0 bug (or another P1 bug). Issues at this level should be resolved in a matter of months and should have monthly updates on GitHub.

Please consider where this bug really falls in our current priorities, and label it or assign it accordingly. This allows people to have a clearer picture of what work is actually planned. Thanks!

The first thing to check would be:

Is that device running with the opengles backend or vulkan backend? (seen in the log)
If it is running vulkan, does manually dropping it to opengles fix the performance? [Impeller] Severe performance degradation on Adreno 840 #180958
There have been a lot of changes since this has be filed, is it reproduced on main?

That GPU is recent enough that it isn't something we'd want to opt out of vulkan as a long term solution.

Another thing that would help would be capturing an accompanying perfetto trace: https://perfetto.dev/

Navigation Menu

Repository navigation

[Impeller] Severe performance degradation on Adreno 840 #180958

Steps to reproduce

Cause

CONFIRMED FINDINGS:

Additional context

Expected results

Actual results

Code sample

Screenshots or Video

Logs

Flutter Doctor output

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[Impeller] Severe performance degradation on Adreno 840 #180958

Description

Steps to reproduce

Cause

CONFIRMED FINDINGS:

Additional context

Expected results

Actual results

Code sample

Screenshots or Video

Logs

Flutter Doctor output

Activity

danagbemava-nc commented on Jan 14, 2026

edTheGuy00 commented on Jan 14, 2026

danagbemava-nc commented on Jan 15, 2026

edTheGuy00 commented on Jan 15, 2026

edTheGuy00 commented on Jan 15, 2026

danagbemava-nc commented on Jan 16, 2026

ain-py commented on Apr 22, 2026

flutter-triage-bot commented on Jun 2, 2026

gaaclarke commented on Sep 24, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions