News

Nearly a quarter of the Pixel 10’s RAM is reserved for AI

On-device AI models eat up a lot of memory

Google Pixel 10

Google’s Pixel 10 sports 12GB of RAM, just like last year’s Pixel 9. However, less of that RAM is available for apps and games to use because on the Pixel 10, about 3GB are reserved for AI.

Some digging by Android Authority spotted that roughly 3GB of RAM are reserved for the AICore service, meaning it can’t be used even if the other roughly 9GB get used up for other tasks. The reason for this? Ensuring that on-device AI models launch nearly instantaneously.

The issue comes down to launch speeds. When people launch something on their phone (or other electronics like computers), the device must fetch the necessary pieces from the device’s storage and load it into the RAM. Most apps don’t use a ton of RAM, and therefore launch quite quickly. More intensive software, like games, takes longer to load because there’s more the device needs to put into RAM. But AI models occupy a space in between, where people expect near-instant launches, but the software needs are more akin to a game.

By reserving a chunk of memory for AI, it ensures fast launch and access to the on-device AI models that power so much of the Pixel experience — things like Magic Cue, computational photography, on-device Gemini, and more.

Last year, Google reserved memory for AI on the Pro Pixel models, but not on the Pixel 9, which would load in AI models as needed. However, the Pro Pixel models have 16GB of RAM, so there’s more room to set aside for AI models.

It’s not clear why Google decided to reserve RAM for AI on the Pixel 10 this year when it chose not to with the Pixel 9, but my money is on Magic Cue. One of the Pixel 10 series’ flagship features, Magic Cue leverages on-device AI to provide contextual suggestions to users. For example, if your spouse texts you asking what time your date is, Magic Cue can use data such as a calendar event or a reservation in your email to suggest the time, saving you from having to check it yourself.

However, only time will tell if sacrificing potential app and game performance to lock down space for AI models will be worth it. I’ve only had the Pixel 10 in hand for a few days and haven’t noticed any performance issues, while AI tools have felt snappy. I don’t have a Pixel 9 for side-by-side comparisons, but I don’t recall the Pixel 9’s AI ever feeling slow from back when I tested it.

Regardless, it’s worth knowing that not all 12GB of RAM are available. Heavy multitaskers might want to go for the Pro models to avoid any issues.

Source: Android Authority

MobileSyrup may earn a commission from purchases made via our links, which helps fund the journalism we provide free on our website. These links do not influence our editorial content. Support us here.

Related Articles

Fatal error: Uncaught Aws\S3\Exception\PermanentRedirectException: Encountered a permanent redirect while requesting https://ms-staging-baselayer-static-assets.s3.ca-central-1.amazonaws.com/?list-type=2&delimiter=%2F&prefix=uploads%2Fwpcf7_uploads%2F. Are you sure you are using the correct region for this bucket? in /var/www/html/vendor/aws/aws-sdk-php/src/S3/PermanentRedirectMiddleware.php:49 Stack trace: #0 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(209): Aws\S3\PermanentRedirectMiddleware->Aws\S3\{closure}(Object(Aws\Result)) #1 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(158): GuzzleHttp\Promise\Promise::callHandler(1, Object(Aws\Result), NULL) #2 /var/www/html/vendor/guzzlehttp/promises/src/TaskQueue.php(52): GuzzleHttp\Promise\Promise::GuzzleHttp\Promise\{closure}() #3 /var/www/html/vendor/guzzlehttp/guzzle/src/Handler/CurlMultiHandler.php(163): GuzzleHttp\Promise\TaskQueue->run() #4 /var/www/html/vendor/guzzlehttp/guzzle/src/Handler/CurlMultiHandler.php(189): GuzzleHttp\Handler\CurlMultiHandler->tick() #5 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(251): GuzzleHttp\Handler\CurlMultiHandler->execute(true) #6 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(227): GuzzleHttp\Promise\Promise->invokeWaitFn() #7 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(272): GuzzleHttp\Promise\Promise->waitIfPending() #8 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(229): GuzzleHttp\Promise\Promise->invokeWaitList() #9 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(272): GuzzleHttp\Promise\Promise->waitIfPending() #10 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(229): GuzzleHttp\Promise\Promise->invokeWaitList() #11 /var/www/html/vendor/guzzlehttp/promises/src/Promise.php(69): GuzzleHttp\Promise\Promise->waitIfPending() #12 /var/www/html/vendor/aws/aws-sdk-php/src/AwsClientTrait.php(58): GuzzleHttp\Promise\Promise->wait() #13 /var/www/html/vendor/aws/aws-sdk-php/src/ResultPaginator.php(151): Aws\AwsClient->execute(Object(Aws\Command)) #14 /var/www/html/vendor/aws/aws-sdk-php/src/functions.php(52): Aws\ResultPaginator->valid() #15 /var/www/html/vendor/aws/aws-sdk-php/src/functions.php(69): Aws\map(Object(Aws\ResultPaginator), Object(Closure)) #16 [internal function]: Aws\flatmap(Object(Aws\ResultPaginator), Object(Closure)) #17 /var/www/html/wp-content/plugins/s3-uploads/inc/class-stream-wrapper.php(695): Generator->valid() #18 [internal function]: S3_Uploads\Stream_Wrapper->dir_readdir() #19 /var/www/html/wp-content/plugins/contact-form-7/includes/file.php(362): readdir(Resource id #734) #20 /var/www/html/wp-includes/class-wp-hook.php(322): wpcf7_cleanup_upload_files() #21 /var/www/html/wp-includes/class-wp-hook.php(348): WP_Hook->apply_filters(NULL, Array) #22 /var/www/html/wp-includes/plugin.php(517): WP_Hook->do_action(Array) #23 /var/www/html/wp-includes/load.php(1280): do_action('shutdown') #24 [internal function]: shutdown_action_hook() #25 {main} thrown in /var/www/html/vendor/aws/aws-sdk-php/src/S3/PermanentRedirectMiddleware.php on line 49