Notebookcheck Logo

Pixel 8 Gemini fail spotlights Google's Pixel AI failings

The Pixel 8 fails to deliver on Google's AI promises. (Image: Notebookcheck)
The Pixel 8 fails to deliver on Google's AI promises. (Image: Notebookcheck)
Google has billed its Pixel smartphones as being all about their AI capabilities. With revelations that the Pixel 8 can’t support Gemini Nano, Google’s first on-device LLM for Android phones, that reputation is in tatters.
Views, thoughts, and opinions expressed in the text belong solely to the author.

Google has to be applauded for what it has accomplished with its AI endeavors on its Pixel line of smartphones. Right from the start, the company defined the Pixel line by its AI capabilities with a clear vision of where smartphones were headed and how AI could enhance the user experience. Whether it was the introduction and subsequent popularization of computational photography or clever features like call screening, it has led the way on how AI can be used to make smartphones even smarter.

However, it is now abundantly clear that it has well and truly tripped up with news emerging this past week that the Pixel 8 launched just last October is unable to support either one of the two Gemini generative AI-based models. To be clear, while OpenAI may have wrested AI leadership from Google, its AI software is still very competitive. The issue here lies with what Google itself has described as “hardware limitations.

The Pixel 8 is fitted with the same Tensor G3 chip fitted to the Pixel 8 Pro, but where the latter has 12 GB of RAM, the Pixel 8 is stuck with just 8 GB of RAM, which appears to be the cause of the system bottleneck on this occasion. This is still somewhat surprising as Gemini Nano comes in two model sizes; one running at just 1.8 billion parameters and the other running at 3.6 billion parameters. It is unclear which of the two models is running on the Pixel 8 Pro, but at least it can support Google's first on-device mobile LLM. 

As far as on-device AI models go both are relatively modest, however. Qualcomm, which has just launched a new AI Hub with over 75 AI models compatible with its Snapdragon chips, has highlighted that its Snapdragon 8 Gen 3 can support AI models of up to 10 billion parameters. Even its Snapdragon 8 Gen 2 can support AI modes running at up to 7 billion parameters. 

This is significant for two reasons. Firstly, the greater the parameters, potentially the more sophisticated and accurate the model. Secondly, it highlights that it is not just system RAM where the Pixel 8 suffers, but the Tensor G3 is, as we have previously examined, far from being the AI champ that Google’s marketing would lead us to believe. Google has pitched the Pixel series as being all about the AI and defended the Tensor for its lack of outright performance one the pretense that outright performance is less important than AI capability. 

Of course, AI capability and chip performance go hand-in-hand, especially when it comes to the size of large language models (LLMs) and large multimodal models (LMMs). Google has been able to get away with the Tensor’s performance shortcomings because it has been using smaller machine learning models until now. This has allowed it to hoodwink some Pixel fans into thinking its chips were somehow specially “tuned” for AI and that their performance in benchmarks does not matter – newsflash - they do.

LLMs and LMMs require the full processing muscle a chip has to offer, with the neural core acting as an accelerator with both the CPU and GPU fully utilized to process machine learning and deep learning models on-device. The more powerful the chip across the board, the more able these multibillion parameter AI models. It is no coincidence that the Snapdragon 8 Gen 3 trounces the Tensor G3 in benchmarks and can also handle much larger AI models on-device.

This is why, as we have also highlighted, Google has been forced to off-load a number of the new generative AI features available to Pixel 8 Pro users to the Google Cloud for off-device processing. The Tensor G3 is simply not “up to snuff” nor is the 8 GB of RAM fitted to the Pixel 8. 

In dropping Qualcomm’s Snapdragon chips to make its own Tensor chips, the aim Google said, was to create chips that could “keep up” with its AI software endeavors. Clearly, that goal is currently in tatters, especially with this latest Pixel 8 Gemini fiasco. A mere five months after its launch, it cannot keep up with Gemini Nano – even though it also comes in a comparatively tiny 1.8 billion parameter model. On the other hand, the Pixel 8 Pro, which while capable of running Gemini Nano on device, is still forced to off-load many new generative AI features off to the cloud. 

It also makes a mockery of Google’s announcement at the launch of the Pixel 8 series that the phones would come with 7 years of software updates. We are not even half way through their first full year of release and the Pixel 8 is unable to keep up with its stable mate. While it’s a commendable goal, the reality is that it won’t be too many years before the Pixel 8 and Pixel 8 Pro OS “updates” will be largely limited to security patches and other system tweaks. With AI the raison d'etre for the very existence of the Pixel series, the two devices aren't properly able to handle Google's latest AI features even now.

Google has partnered with Samsung LSI and Samsung Foundry for its design, development and fabrication. Until now, this has been their Achilles heel as these chips have suffered from inefficiency and overheating due to current leakage and packaging concerns. This has limited the performance potential of their underlying Arm-based architecture in terms of peak and sustained performance.

However, the apparent success of the Exynos 2400 in the Galaxy S24 and Galaxy S24+ models thanks to node refinements and the utilization of a new fan-out wafer-level packaging (FOWLP), bodes well for the Tensor G4. This chip will find its way into the forthcoming Pixel 9 series later this year, while Google will deliver its first completely custom chip, the Tensor G5, which will be fabricated by the more fancied TSMC.

This of course, is little comfort for Pixel 8 owners. Not only do they miss out on the generative AI features reserved for Pixel 8 Pro owners, they can’t even get a taste of Gemini Nano. The Pixel 8 still be a solid mid-range phone, but it really does spotlight Google’s current - and self-inflicted - Pixel AI failings.

Pixel 8 "hardware limitations" to blame for Gemini Nano fail. (Image: PBKReviews)
Pixel 8 "hardware limitations" to blame for Gemini Nano fail. (Image: PBKReviews)



Read all 1 comments / answer
static version load dynamic
Loading Comments
Comment on this article
Please share our article, every link counts!
> Expert Reviews and News on Laptops, Smartphones and Tech Innovations > News > News Archive > Newsarchive 2024 03 > Pixel 8 Gemini fail spotlights Google's Pixel AI failings
Sanjiv Sathiah, 2024-03-10 (Update: 2024-03-10)