Google Tensor G6 vs Tensor G5: Benchmarks and Specs

The Tensor G6 has seen the biggest jump in the CPU department, efficiency and connectivity front, but Google still can't sort out the GPU

Arjun Sha profile pictureby Arjun Sha
Link Copied
copy link iconcopy link icon
tensor g6 vs tensor g5 side by side benchmark comparison

Image Credit: Beebom Gadgets

Summary

  • The Tensor G6 scored 1,687,489 points in our AnTuTu test, around 31% more than the 1,291,252 points we recorded on the Tensor G5.
  • Geekbench 6 puts the Tensor G6 at 2,691 in single-core and 7,203 in multi-core, compared with 2,285 and 6,191 on the Tensor G5.
  • In the CPU Throttling Test, the Tensor G6 throttled to 47% of its peak performance, while the Tensor G5 held on to 61%.

Notes

We benchmarked the Tensor G6 on the Pixel 11 Pro XL and the Tensor G5 on the Pixel 10 Pro.

Google is powering the Pixel 11 series with the new Tensor G6 chipset and we are seeing something new after a long while. It's fabricated on a 2nm process node, directly jumps to Arm's C1-series cores and packs a modem that is not made by Samsung. So, we have benchmarked the latest Tensor G6 chipset against last year's Tensor G5 to measure the performance difference. On that note, let's begin.

Tensor G6 vs Tensor G5: AnTuTu Score

AnTuTu BenchmarkTensor G6Tensor G5
AnTuTu Score1,687,4891,291,252
CPU739,209457,073
GPU251,126382,578
Memory340,262242,613
UX356,892208,988

Let us start with the AnTuTu benchmark. The Pixel 11 Pro XL, running on the Tensor G6, achieved 1,687,489 points, whereas the Pixel 10 Pro with the Tensor G5 scored 1,291,252 points. It means the Tensor G6 chip has a lead of around 31%.

In the CPU department, the Tensor G6 scored 739,209 points, which is about 62% higher than the Tensor G5. The GPU score, however, has surprisingly gone down this year instead of going up. The Tensor G6's PowerVR CXTP-48-1536 GPU managed just 251,126 points, which is around 34% lower than the DXT-48-1536 GPU in the Tensor G5.

Tensor G6 vs Tensor G5: Geekbench Score

Geekbench 6 CPUTensor G6Tensor G5
Single-core2,6912,285
Multi-core7,2036,191

Next, we ran Geekbench 6 to measure raw CPU performance. Here, the Tensor G6 put up 2,691 points in single-core and 7,203 points in multi-core, compared to 2,285 and 6,191 points on the Tensor G5. It means that the Tensor G6 is around 18% faster in single-core and about 16% faster in multi-core performance.

The single-core improvement comes from the new Arm C1-Ultra prime core clocked at 4.11GHz, which replaces the Cortex-X4 core on the Tensor G5. In effect, Google has skipped the Cortex-X925 core and moved straight to Arm's C1-series CPU. By the way, the same C1-series cores also power the Dimensity 9500 processor and the next-gen C2-series CPU cores are available on the Dimensity 9600 Pro.

Note that the multi-core gain is smaller because Google has dropped one core this year. The Tensor G6 uses a 7-core CPU (1+4+2), while the Tensor G5 came with an 8-core CPU (1+5+2).

Related Articles

Tensor G6 vs Tensor G5: CPU Throttling Test

CPU Throttling TestTensor G6Tensor G5
Maximum Score343,713 GIPS408,383 GIPS
Average Score248,160 GIPS289,135 GIPS
Minimum Score133,360 GIPS239,312 GIPS
Throttling Percentage (Higher is better)47%61%

Now, coming to the 15-minute CPU Throttling Test, which keeps all CPU cores under heavy load to check sustained performance. Interestingly, the older Tensor G5 posted a peak score of 408,383 GIPS, which is around 19% higher than the Tensor G6. In addition, the Tensor G5 throttled to 61% of its peak performance, while the Tensor G6 dropped to 47%.

That said, the Pixel 11 Pro XL did not get abnormally hot during our test. The phone has a larger body than the Pixel 10 Pro and both come with a vapour chamber, so heat is not the reason behind the drop.

In our assessment, Google is power-limiting the Tensor G6 and capping clock speeds under sustained load to keep the phone cool, which is why it scores lower here despite the newer cores. In contrast, Qualcomm is pushing its prime cores to 5GHz on the latest Snapdragon 8 Elite Extreme Gen 6 chip to deliver strong peak performance.

Tensor G6 vs Tensor G5: On-Device AI Performance

On-Device AI (Gemma-4-E2B-it, CPU)Tensor G6Tensor G5
Prefill speed271.48 tokens/sec165.97 tokens/sec
Decode speed24.67 tokens/sec10.42 tokens/sec
Time to first token0.98 sec1.65 sec

Since Tensor chips are built around on-device AI, we ran the Gemma-4-E2B-it model on both phones through the AI Edge Gallery app. Google does not allow third-party apps to access the TPU yet, so the model ran on the CPU. In our on-device AI test, the prefill speed on the Tensor G6 reached 271.48 tokens/sec, which is 64% faster than the Tensor G5.

The decode speed, which decides how quickly the model generates a response, has seen the biggest jump. The Tensor G6 generated a response at 24.67 tokens/sec, while the Tensor G5 could only manage 10.42 tokens/sec, giving the Tensor G6 a lead of 137%. The time to first token also dropped from 1.65 seconds to 0.98 seconds.

Keep in mind that these results do not engage the TPU at all. According to Google, the 5th-gen TPU in the Tensor G6 has 50% more compute and more memory bandwidth than the Tensor G5. Google also claims a 3.5x gain in both speed and energy efficiency for on-device AI tasks.

Related Articles

Tensor G6 vs Tensor G5: Benchmark Gallery

antutu comparison between Tensor G6 vs Tensor G5
antutu comparison between Tensor G6 vs Tensor G5
geekbench comparison between Tensor G6 vs Tensor G5
geekbench comparison between Tensor G6 vs Tensor G5
cpu throttling test comparison between Tensor G6 vs Tensor G5
cpu throttling test comparison between Tensor G6 vs Tensor G5
on-device ai performance comparison between Tensor G6 vs Tensor G5
on-device ai performance comparison between Tensor G6 vs Tensor G5

Tensor G6 vs Tensor G5: Specs Comparison

SpecsTensor G6Tensor G5
Process NodeTSMC 3nm (N3P)TSMC 3nm (N3E)
CPU7-core Arm C1-series CPU8-core Arm Cortex CPU
CPU Cores1x Arm C1-Ultra at 4.11GHz, 4x Arm C1-Pro at 3.38GHz, 2x Arm C1-Pro at 2.65GHz1x Cortex-X4 at 3.78GHz, 5x Cortex-A725 at 3.05GHz, 2x Cortex-A520 at 2.25GHz
GPUImagination PowerVR CXTP-48-1536 (C-Series), 1.29GHzImagination PowerVR DXT-48-1536 (D-Series), 1.1GHz
NPUFifth-gen Google TPU, 50% more compute and higher memory bandwidthFourth-gen Google TPU
ISPUpgraded ISP with a dedicated video portrait accelerator, 4K Portrait Video, Instant Night Sight (Pro models)New in-house ISP, 4K video at 60fps
SecurityTitan M3 security chipTitan M2 security chip
Storage and MemoryLPDDR5X, UFS 4.0 (Zoned UFS 4.0 on 512GB and above)LPDDR5X, UFS 3.1 (128GB), UFS 4.0 (256GB), Zoned UFS 4.0 (512GB and above)
ModemMediaTek M90 5G modem, up to 12Gbps peak downloadSamsung Exynos 5400 5G modem, up to 14.79Gbps peak download
ConnectivityWi-Fi 7 and UWB (Pro models), Wi-Fi 6E (Pixel 11), Bluetooth 6.0Wi-Fi 7 and UWB (Pro models), Wi-Fi 6E (Pixel 10), Bluetooth 6.0

Verdict

To conclude, the Tensor G6 is a clear upgrade over the Tensor G5. The new Arm C1-series CPU cores deliver 18% faster single-core performance, on-device AI responses are generated more than twice as fast on the CPU and the Pixel 11 Pro XL stays cool and smooth in daily use. The MediaTek modem should also fix radio connectivity and battery drain on the Pixel 11 series.

Google has chosen to prioritise thermals and efficiency over peak scores with the Tensor G6. However, in gaming , the chip is far behind the rivals.

All in all, if you are upgrading from a Pixel 10, the Tensor G6 will feel quicker and smarter, but don't expect a GPU  upgrade. Reports suggest that Google might move Tensor to a 2nm node next year so expect the upcoming chip to finally reduce the gap with rivals. If you want to see where Google's Tensor G6 stands against MediaTek's older flagship, go through our Tensor G6 vs Dimensity 9500 comparison.

#Tags

pixel

Recommended For You

Popular Mobile List