Besides making Exynos 2400 for their Galaxy S24 and S24 Plus, Samsung also makes the Tensor G4 chip for the Pixel 9 series. Considering both chips are made by Samsung Foundry, you may be curious to know how they hold up against each other. Let’s pit the Galaxy S24 (with Exynos 2400) and Pixel 9 (with Tensor G4) against each other to compare not only the processors but also to see how the two compact Android flagships perform against each other.
Note: Since the Pixel 9 lacks the vapor cooling chamber found in its elder siblings, I’ve also included scores from the Pixel 9 Pro XL to provide a clearer picture of the Tensor G4’s potential against the Exynos 2400.
Table of Contents
Tensor G4 vs Exynos 2400: Specifications
| Tensor G4 | Exynos 2400 | |
|---|---|---|
| Release Date | August 2024 | January 2024 |
| Manufacturer | Samsung | Samsung |
| Fabrication Process | 4nm | 4nm LPP+ |
| CPU | Eight-core 1x Cortex-X4 (3.1 GHz) 3x Cortex-A720 (2.6 GHz) 4x Cortex-A520 (1.95GHz) | Ten-core 1x 3.21 GHz – Cortex-X4 2x 2.9 GHz – Cortex-A720 3x 2.6 GHz – Cortex-A720 4x 1.95 GHz – Cortex-A520 |
| GPU | Arm Mali-G715 MP7 | Samsung Xclipse 940 |
| RAM | LPDDR5X | LPDDR5X |
| Storage Type | UFS 3.1, UFS 4.0 (Pixel 9-series uses UFS 3.1) | UFS 3.1, UFS 4.0 |
| Machine Learning and AI | Third-gen Tensor Processing Unit, Gemini Nano chip | 17K MAC NPU (2-GNPU+2-SNPU) |
| Camera and Video Capture | Max resolution up to 320MP 8K at 30FPS, 4K at 120FPS | – 8K at 30FPS, 4K at 120FPS |
| Modem | Samsung Exynos 5400 Downlink speeds of up to 14.79Gbps | Integrated 5G Modem Downlink speeds of up to 12.1Gbps |
| Connectivity | 4G LTE 5G sub-6GHz and mmWave Wi-Fi 7 Bluetooth 5.3 Satellite connectivity | 4G LTE 5G sub-6GHz and mmWave Wi-Fi 6E Bluetooth 5.3 |
| Codecs | AV1, H.264, H.265, VP9 | AV1, H.264, H.265, VP9 |
| Security | Titan M2 Security Chip | Samsung Knox Security |
Tensor G4 vs Exynos 2400: CPU
The Tensor G4 is an octa-core processor, down from nine cores on the Tensor G3, as we talked about earlier. It has one Cortex-X4 core clocked at 3.1GHz, three Cortex-A720 performance cores clocked at 2.6GHz, and four Cortex-A520 efficiency cores clocked at 1.95GHz.
On the other hand, the Exynos 2400 is a deca-core (10) processor. It has one Cortex-X4 core clocked at 3.21GHz, two Cortex-A720 cores clocked at 2.9GHz, two Cortex-A720 cores slightly lower at 2.6GHz, and then four efficiency Cortex-A520 cores clocked at 1.95GHz.
The Exynos 2400 has two additional cores. Its prime core is clocked 110MHz higher, and two of its Cortex-A720 cores are clocked 300MHz higher.
While both chips are based on Samsung’s 4nm fabrication, the Exynos 2400 uses enhanced 4nm LPP+ technology and Fan-Out Wafer-Level Packaging (FOWLP) tech for improved thermal management. It’s unclear whether the Tensor has these features.
AnTuTu Benchmark
In AnTuTu benchmarks, the Exynos 2400 clearly outperforms the Tensor G4, boasting a roughly 60% higher AnTuTu score. Notice the difference in CPU scores specifically; even though our Galaxy S24 got hot during the test, it scored almost three times the Pixel 9’s score.
| AnTuTu | Tensor G4 (Pixel 9) | Tensor G4 (Pixel 9 Pro XL) | Exynos 2400 (Galaxy S24) |
|---|---|---|---|
| CPU | 125,828 | 140,345 | 387,695 |
| GPU | 425,262 | 441,349 | 588,344 |
| Memory | 191,635 | 190,784 | 313,965 |
| UX | 163,690 | 197,774 | 283,946 |
| Overall | 906,415 | 970,252 | 1,573,950 |
Geekbench 6 CPU
In Geekbench single-core tests, the Tensor G4 closely matches the Exynos 2400. However, the one on the regular Pixel 9 lags slightly behind. The Exynos’s additional cores and higher clock speeds give it a clear advantage in multi-core performance, scoring about 45% higher at 6,224 points.
| Geekbench CPU | Tensor G4 (Pixel 9) | Tensor G4 (Pixel 9 Pro XL) | Exynos 2400 (Galaxy S24) |
|---|---|---|---|
| Single-Core | 1512 | 1918 | 1922 |
| Multi-Core | 3350 | 4319 | 6224 |
CPU Throttling Test
In a 15-minute throttling test, the Tensor G4 throttled early on, dropping to nearly half its performance. The Exynos 2400 fared slightly better, maintaining 61% of its peak on average. It even recovered slightly towards the end and consistently clocked higher GIPS than the Tensor.
Tensor G4 vs Exynos 2400: GPU
The Tensor G4 features an Arm Mali-G715 MP7 GPU clocked at 940 MHz. In contrast, the Exynos 2400 boasts a Samsung Xclipse 940 GPU, clocked at a higher 1109 MHz and based on AMD’s powerful RDNA 3 architecture. To test graphic capabilities, I ran a couple of GPU benchmarks on the two phones.
Geekbench 6 GPU
In Geekbench GPU tests, the Tensor G4 scored 6,467 points in OpenCL and 6,973 points in Vulkan API. On the other hand, the Exynos 2400 scored 17,459 points in OpenCL and 17,792 points in Vulkan, which is nearly three times the scores of the Google Tensor.
The Exynos’s GPU is the more powerful one here, meaning it would better handle graphic-intensive games and tasks like video editing.
| Geekbench GPU | Tensor G4 (Pixel 9 Pro XL) | Exynos 2400 (Galaxy S24) |
|---|---|---|
| OpenCL | 6,467 | 17,459 |
| Vulkan | 6,973 | 17,792 |
GFXBench
Here, we’ve run GFXBench on the Tensor G4 and Exynos 2400, which tests the processors’ graphic performance. Notice the frame rate in each test: the higher the number, the smoother the visuals and gameplay. Although Exynos is stronger on paper, the Tensor can also handle any game you throw at it at respectable graphic settings.
| GFXBench | Tensor G4 (Pixel 9 Pro XL) | Exynos 2400 (Galaxy S24) |
|---|---|---|
| Aztec Ruins OpenGL (High Tier) | 4,625 frames (72 fps) | 4,272 frames (66 fps) |
| Aztec Ruins Vulkan (High Tier) | 4,541 frames (71 fps) | 4,470 frames (70 fps) |
| Manhattan 3.1 (High Tier) | 6,776 frames (109 fps) | 7,208 frames (116 fps) |
3D Mark WildLife Stress Test
In 3DMark WildLife Stress Test, the Tensor G4 yielded a stability score of 68%, while the S24’s Exynos 2400 scored a mere 58.4%. Surprisingly, the Exynos 2400 also ran hotter than the Tensor G4.
While both processors aren’t as stable as their Qualcomm counterparts and experience performance dips under sustained loads, Tensor actually does a better job here.
| WildLife Stress Test | Tensor G4 (Pixel 9 Pro XL) | Exynos 2400 (Galaxy S24) |
|---|---|---|
| Best loop score | 8960 | 12501 |
| Lowest loop score | 6116 | 7297 |
| Stability | 68.3% | 58.4% |
Tensor G4 vs Exynos 2400: NPU
The Tensor G4 uses third-gen Tensor Processing Unit for neural processing which is the same as last year’s Tensor G3 on Pixel 8. But it does have a newer Gemini Nano model for Pixel AI features. On the other hand, Exynos 2400 has a 17K MAC NPU (2-GNPU+2-SNPU).
Geekbench AI
Geekbench AI tests the AI performance of your device, specifically focusing on how well it handles machine learning, deep learning, and other AI-centric workloads. It measures the capabilities of your device’s CPU, GPU, and NPU in executing AI-related computations across different precision levels and a variety of AI tasks.
Here’s how both the Tensor G4 and Exynos 2400 fare in AI-related tasks:
| Tensor G4 (Pixel 9 Pro XL) | Exynos 2400 (Galaxy S24) | |
|---|---|---|
| Single Precision Score | ||
| CPU | 2030 | 2077 |
| GPU | 720 | 1435 |
| NNAPI | 249 | 243 |
| Half Precision Score | ||
| CPU | 1974 | 2089 |
| GPU | 821 | 1775 |
| NNAPI | 4068 | 226 |
| Quantized Score | ||
| CPU | 3169 | 2958 |
| GPU | 784 | 1700 |
| NNAPI | 6076 | 671 |
The Exynos 2400’s powerful GPU gives it an edge in AI tasks that leverage the GPU, particularly for tasks requiring lower precision (half and quantized). The Tensor G4, on the other hand, shows its strength in NNAPI-accelerated tasks at half-precision, highlighting the effectiveness of Google’s optimizations for its TPU.
Overall, the Exynos 2400 seems to be the more capable AI processor, especially for tasks that can utilize its GPU. However, the Tensor G4 holds its ground in specific scenarios, demonstrating the potential of Google’s custom silicon.
Tensor G4 vs Exynos 2400: Modem & Connectivity
The Tensor G4 uses the Samsung Exynos 5400 modem, which can deliver a maximum speed of 14.79 Gbps with its NR dual connectivity of both FR1 (sub6 GHz) and FR2 (mmWave) bandwidth. It further supports Wi-Fi 7, Bluetooth 5.3, and satellite connectivity.
The Exynos 2400 has an integrated 5G NR modem, which also supports both frequency ranges. It can achieve maximum downlink speeds of up to 9.6 Gbps in FR1 and 12.12 Gbps in FR2. It supports Wi-Fi 6E and Bluetooth 5.3 — no satellite connectivity here.
In my testing, both phones have had stable connectivity. The Pixel 9 has improved a lot compared to its predecessors, and that’s something I’m happy about.
So Who Is Winning the Race?
So, in this Tensor G4 vs Exynos 2400 face-off, it’s kinda surprising to see the older Exynos coming out on top! Even though it’s been around longer, it still boasts newer technology, higher clock speeds, and additional cores that deliver superior performance in benchmarks.
Both chips grapple with thermal issues and can get hot and slow down when you push them hard, but Exynos handles that a bit better. Tensor G4 does score some points with its fancy Wi-Fi 7 and satellite connectivity, but overall, Exynos is just the stronger chip with better performance and likely a longer lifespan. It’s not like we can just pick which chip we want in our phones, but at least we know what we’re getting into before we buy.
