Exclusive: Geekbench founder responds to Pixel 8 benchmark blocking controversy

[Redaktion]

Geekbench 6 founder John Poole has responded to the Pixel 8 and Pixel 8 Pro benchmarking controversy. In exclusive comments obtained by Notebookcheck, Poole says Primate Labs is unaware of any issues with compatibility that might have otherwise resulted in the app from being made unavailable to reviewers.

https://www.notebookcheck.net/Exclusive-Geekbench-founder-responds-to-Pixel-8-benchmark-blocking-controversy.761277.0.html

[QertD]

The Google has been open and there are several articles over the last week about Google directly stating that they use cloud processing for some of their new AI features on the Pro 8. Does not however take from the fact that there are dozens of AI features that are on most pixels that are run on device only and they work perfectly well and faster than on previous models.
No CPU or GPU today. Could in under 10 seconds render a full 4K video with HDR+, It's kind of obvious why they would need cloud computing to offer such an awesome feature.

Or maybe you can talk about the several others and other AI features that work on phone only that aren't available on any other device.

[Vince789]

IMO Google did a good job trying to explain the different features that are on-device vs off-device, but with so many features there's inevitably going to be some confusion
AFAIK here's some on-device vs off-device features using generative AI:

• The new updated Magic Eraser is on-device and features generative AI, but note that Magic Editor is different and off-device
• And Recorder will be able to provide summaries using generative AI done on-device, but note that Google Assistant with Bard's summary feature will be off-device
• Zoom Enhance will be on-device and feature generative AI, but note that Video Boost with Night Sight Video will be off-device
• Gboard's updated Smart Reply will be on-device and feature generative AI, but note that Gboard's new updated proofread is off-device

[enlightenment]

Last two articles trashing Google on this site used a YouTuber and now a synthetic benchmark owner Geek Bench as sources.  Neither of these individuals know anything about AI or chip design.
Questions that the author needs to research
1. How are AI models trained
2. Why some AI task are done on device and others are not.

[Erick]

There were reviewers with the requisite knowledge to by-pass Google's Play Store block by simply downloading the app package for Geekbench 6 and installing it manually using a process known as side-loading.

If you don't know how to do this, you have no business reviewing Android phones. It takes a few seconds of searching to find out.

[Modernman]

Did anyone buy a $1000 pixel 8 pro because they thought the Ai wallpaper and magic eraser were processed on the device?

Theres plenty of good critical reviews of the pixel and how it performs but the author fails to engage in any meaningful critique. Resting on the authority of a lone YT channel not widely regarded by anyone and some attempts at "gotchas"

Ai features I'm interested in are those like transcription/caption, translation,call screening, compose which work offline. The camera quality tends to be a major selling point. Criticizing the delay on 50mp images would have been better than every point here.


This is the second article now and it's just as lazy as the first one I saw. I can't imagine anyone who writes on the industry is actually this in sophisticated.

So while it pains me to give the clickbait it's engagement, I just don't want anyone to actually get mislead by such a misleading writer.

[Sanket]

Googles software is on their way down but still okay. However, It's laughable that you Google fans defend its hardware. I guess after iSheep, gSheep is finally a thing now, too.
Tensor G3 is laggy, underperforming and constantly heating up. That is just known fact. In general, google's devices are just Overpriced. If they cut 200 dollars off their phone product lines. They may still have a chance in the market. $1000 for a Pixel pro? No, thank you, no.