Snapdragon 8 Gen 3 vs Tensor G4

We compared two 8-core processors: Qualcomm Snapdragon 8 Gen 3 (with Adreno 750 graphics) and Google Tensor G4 (Mali-G715 MP7). Here you will find the pros and cons of each chip, technical specs, and comprehensive tests in benchmarks, like AnTuTu and Geekbench.

Review

General comparison of performance, power consumption, and other indicators
CPU Performance
Single and multi-core processor tests
Gaming Performance
GPU performance in games and OpenCL/Vulkan
Battery life
Efficiency of battery consumption
NanoReview Score
Overall chip score

Key Differences

Main differences and advantages of each chip
Pros of Qualcomm Snapdragon 8 Gen 3
  • Performs 65% better in floating-point computations
  • Shows significantly better (up to 54%) AnTuTu 11 score – 2345K vs 1519K
  • 6% higher CPU clock speed (3300 vs 3100 MHz)
  • Higher GPU frequency (~6%)
  • Supports 13% higher memory bandwidth (76.8 against 68.2 GB/s)

Benchmarks

Performance tests in popular benchmarks

AnTuTu 11

The AnTuTu Benchmark measures CPU, GPU, RAM, and I/O performance in different scenarios
2345563
1519937
CPU 700525 536742
GPU 805478 414930
Memory 296043 215572
UX 543517 352693
Total score 2345563 1519937
Samples [443 tests] [25 tests]

GeekBench 6

The GeekBench test shows raw single-threaded and multithreaded CPU performance
Single-Core Score
2203
1707
Multi-Core Score
6511
4431
Asset compression 256.8 MB/sec 177.3 MB/sec
HTML 5 Browser 190.9 pages/sec 134 pages/sec
PDF Renderer 209.1 Mpixels/sec 157 Mpixels/sec
Image detection 168.6 images/sec 122 images/sec
HDR 208.6 Mpixels/sec 128.3 Mpixels/sec
Background blur 24 images/sec 15.8 images/sec
Photo processing 63.6 images/sec 39.2 images/sec
Ray tracing 7.27 Mpixels/sec 4.65 Mpixels/sec
Samples 139 tests 11 tests
Compute Score (GPU)
15378
8744
Background blur 26.9 images/sec 13.4 images/sec
Face detection 13.2 images/sec 8.51 images/sec
Horizon detection 0.7 Gpixels/sec 0.32 Gpixels/sec
Edge detection 0.61 Gpixels/sec 0.3 Gpixels/sec
Gaussian blur 0.98 Gpixels/sec 0.36 Gpixels/sec
Feature matching 0.18 Gpixels/sec 0.11 Gpixels/sec
Stereo matching 43.4 Gpixels/sec 26.4 Gpixels/sec
Particle physics 3083.6 FPS 1273.7 FPS
API OpenCL OpenCL
Samples 90 tests 9 tests

3DMark

A cross-platform benchmark that assesses graphics performance in Vulkan (Metal)
3DMark Steel Nomad Light
1652
1049
Score 1652 1049
Stability 79% 84%
Graphics test 12 FPS 8 FPS
3DMark Solar Bay
Score 8190 -
Stability 72% -
Graphics test 31 FPS -
Section 1 36 FPS -
Section 2 31 FPS -
Section 3 26 FPS -
3DMark Wild Life Extreme
Score 4676 -
Graphics test 27 FPS -

PCMark 3.0

Benchmark for evaluating overall system performance in everyday workloads
PCMark 3.0
16318
13174
Web score 16536 10521
Video editing score 7905 7569
Data manipulation score 14260 14718
Photo editing score 31248 21074
Writing score 19168 15880
Phones tested 41 models 4 models

Recent User Tests

The latest benchmark tests that have been submitted by users
Qualcomm Snapdragon 8 Gen 3
DateBenchmarkResult
πŸ“˜ 2026-09-30 (EA Fran)AnTuTu 122452758
πŸ“˜ 2026-09-30 (Nyvronis)AnTuTu 122649328
πŸ“˜ 2026-09-29 -> Alejandro SorinAnTuTu 122597186
πŸ“˜ 2026-09-29 (Gadji TΓ¦mΓ¦v)AnTuTu 122438906
πŸ“˜ 2026-09-29 (Ratiii)Geekbench 7 CPU1995 / 6875
Google Tensor G4
DateBenchmarkResult
πŸ“˜ 2026-09-18 -> Oleg 16 cmAnTuTu 111423579
πŸ“˜ 2026-08-27 (feraroof43)AnTuTu 111626655
πŸ“˜ 2026-08-21 (Leif)AnTuTu 111590049
πŸ“˜ 2026-08-21 (Kazin)AnTuTu 111723964
πŸ“˜ 2026-07-14 (123)Geekbench 6 GPU8744

Specifications

Full list of technical specifications of Snapdragon 8 Gen 3 and Tensor G4

CPU

Architecture 1x 3.3 GHz – Cortex-X4
3x 3.15 GHz – Cortex-A720
2x 2.96 GHz – Cortex-A720
2x 2.27 GHz – Cortex-A520
1x 3.1 GHz – Cortex-X4
3x 2.6 GHz – Cortex-A720
4x 1.92 GHz – Cortex-A520
Cores 8 8
Frequency 3300 MHz 3100 MHz
Instruction set ARMv9.2-A ARMv9.2-A
L2 cache 1 MB -
L3 cache 12 MB -
Process 4 nanometers 4 nanometers
TDP (Sustained Power Limit) 8 W -
Manufacturing TSMC Samsung

Graphics

GPU name Adreno 750 Mali-G715 MP7
Architecture Adreno 700 Valhall 4th gen
GPU frequency 903-1000 MHz 940 MHz
Pipelines 6 7
Shading units 256 128
Total shaders 1536 896
FLOPS 2774 Gigaflops 1684.4 Gigaflops
Vulkan version 1.3 1.3
OpenCL version 2.0 2.0
DirectX version 12.1 -

AI Accelerator

Neural processor (NPU) Hexagon Yes

Memory

Memory type LPDDR5X LPDDR5X
Memory frequency 4800 MHz 4266 MHz
Bus 4x 16 Bit 4x 16 Bit
Max bandwidth 76.8 Gb/s 68.2 Gb/s
Max size 24 GB 16 GB

Multimedia (ISP)

Storage type UFS 4.0, UFS 4.1 UFS 3.1, UFS 4.0
Max display resolution 3840 x 2160 3840 x 2400
Max camera resolution 1x 200MP -
Video capture 8K at 30FPS, 4K at 120FPS 8K at 30FPS, 4K at 120FPS
Video playback 8K at 60FPS, 4K at 120FPS 8K at 30FPS, 4K at 120FPS
Video codecs - H.264
- H.265
- AV1
- VP8
- VP9
- H.264
- H.265
- AV1
- VP9
Audio codecs - AAC
- AIFF
- CAF
- MP3
- MP4
- WAV
- AAC
- AIFF
- CAF
- MP3
- MP4
- WAV

Connectivity

Modem Snapdragon X75 Exynos 5400c
4G support LTE Cat. 24 -
5G support Yes Yes
Download speed Up to 10000 Mbps -
Upload speed Up to 3500 Mbps -
Wi-Fi 7 7
Bluetooth 5.4 6.0
Navigation GPS, GLONASS, Beidou, Galileo, QZSS, NAVIC GPS, GLONASS, Beidou, Galileo, QZSS

Info

Announced October 2023 August 2024
Class Flagship Flagship
Model number SM8650-AB, SM8650-AC GS401 (S5P9875)
Official page Qualcomm Snapdragon 8 Gen 3 official site -
Successor - Qualcomm Snapdragon 8 Elite Gen 5 - Google Tensor G5
Further details
Notes on Snapdragon 8 Gen 3:
    - Snapdragon 8 Gen 3 for Galaxy has a 100 MHz higher CPU and GPU clock speed. It has a higher GPU performance rate of up to 3072 Gigaflops.

Cast your vote

So, which SoC would you choose?
1066 (86.4%)
168 (13.6%)
Total votes: 1234

Comments

Please give your opinion on the comparison of Tensor G4 and Snapdragon 8 Gen 3, or ask any questions
🌐 Register your profile and become part of NanoReview community!
Avatar
Tay February 20, 2025 at 04:49 PM
To be fair the snapdragon is clocked higher on every CPU core if you look at that alone. They have the same X4, 720 and 520 CPU cores so if Google wanted to match 8 gen 3 in CPU core speed alone they could've, but they were seeking more power efficiency and heat management over over clocked over heated battery drain processing power. Y'all put to much thought into it being "slow" or not able to catch up with actually looking at the core layout yet alone snapdragon just started using Av1 decoding meanwhile tensor has it since G1 LMFAO some ppl are idiots and will never understand soc design is more than just performance
+19 Reply
Avatar
Phil November 16, 2024 at 10:44 AM
People are mad about the performance of the chip but ultimately, someone holding a pixel 9 pro is likely holding a better phone than yours. He can play the same games you play, his phone's battery life is now comparable to yours, his photos look better, his videos have much higher potential with video boost, his phone's display is better surpassing the s24 ultra's technically, his smartphone is generally smarter, being able to better detect spam calls, separate/isolate any audio, utilise an AI assistant on device etc. So sure, if you buy a pixel, you're getting the technically less powerful chip, but you are also (very likely) getting the better phone. At least for now until Samsung comes out with the s25 line up. Apple's 16 pro is whatever at this point so no-one should care. Their "apple AI" is freaking ChatGPT lmao.
+13 Reply
Avatar
Pixelproispoop November 19, 2024 at 08:41 PM
Yet his chip is not "technically" as good as my s24 ultra at running all day to day tasks 🀣 It's not technically the less powerful chip, it is the less powerful chip.If you're claiming technicalities you've already lost. This chip is absolutely bodied by 1 year old technology and on a flagship that costs £1000 that's not good enough
+32 Reply
Avatar
Bro August 26, 2024 at 07:18 PM
The tensor G4 is meant for photography no for gaming,I think they learn about the previous experience of the tensor g1,2,3 who produce lot of heat and lack of stability
+33 Reply
Avatar
Garrett August 25, 2024 at 05:02 PM
My understanding was that this years Pixels don't even use LPDDR5X ram, it's just LPDDR5.
+9 Reply
Avatar
igor August 18, 2024 at 02:20 PM
No matter what you call it, it will still be a failed Exynos product.
+39 Reply
Avatar
igor August 18, 2024 at 02:19 PM
Big mistake to use Samsung processors. No matter what you call it, it will still be a failed Exynos product.
+21 Reply
Avatar
Yous September 19, 2024 at 10:30 AM
Actually exyons 2400 is like doubling the performance even old exyons 2200 is far better than it Google was need to make the Samsung 2400 but with touched it
+1 Reply
Avatar
Trex September 24, 2024 at 06:13 PM
You're right, Imagine how good pixels would have been if they used the same chip with the s24 ultra
+1 Reply
Avatar
Pixel user March 21, 2025 at 01:42 PM
Yous, I have a s22+ and a pixel 8 pro, the tensor G3 is slightly better then the S22+ and it is way more stable plus it has much better battery life, my s22+ lasted 4 and a half hours of sot, my pixel 8 pro lasts 6 and a half hours sometimes even 7 hours of sot.
0 Reply
Avatar
Krzycho August 17, 2024 at 08:34 AM
Tensor G4 Class Flagship??? This is some kind of mistake. This is a mid-range processor, please fix it. Very poor processor for 2024.
+26 Reply
Serhii August 19, 2024 at 01:39 PM
Yeah but it is used in flagships smartphones.
+3 Reply
Avatar
Pixel user March 21, 2025 at 01:40 PM
It's a upper midrange processor, it kicks the snapdragon 7 gen 3's ass in a lot of tasks however compared to the flagships it's 3-4 years behind.
+1 Reply
Avatar
Eeaaoo April 16, 2025 at 02:52 AM
Only flagship chips uses X4 cores. The mid-range only have A cores
+1 Reply
Avatar
Dilaw Nac August 15, 2024 at 01:01 AM
it's crazy how the 8 gen 3 is 1 year behind and still double the performance of tensor g4! i think google should focus on performance and hardware not just software and AI, for longterm usage..
+74 Reply
Avatar
Pixel User August 17, 2024 at 09:06 AM
Tensor is for AI purposes (fast processing of tensors like Google Coral), so these tests are not complete. If we run LLMs I doubt Snapdragon will fail and ranked under 10% of Tensor.
+90 Reply
Avatar
EvilAliK October 20, 2024 at 04:35 PM
Pixel User, Trying to justify your purchase!
+12 Reply
Avatar
Dayum bitch! November 10, 2024 at 04:26 AM
Pixel User, Bruh have you even checked the neural engine on sdgn and mdtk, then can run 10B parameters model locally at reading speed and tensor will shit at 2B. It's absolutely a fucking garbage!
+8 Reply
Avatar
Tay February 20, 2025 at 04:51 PM
Dayum bitch!, That's bs. Pixel have industry leading results with 45 tokens per second compared to 8 gen 3 20 token per second. Learn what tokens mean and you'll understand why tensor is actually better and the important Ai related stuff
+5 Reply