Discussion Google Tensor SoC thread

[hemedans]

You have LineageOS?
It used to be that some banks did not work on phones with custom ROMs and/or rooted.

It can work but not for average consumer, you need magisk and lot of modules for workaround.

 

[ToTTenTranz]

He's talking about phones that have had generic Android installed on them over the vendor install. The Pixel does it fine because its Android is the vendor Android. If you replaced it with Lineage or any other generic Android install you'd find Android Pay no longer works, as well as a number of banking/financial apps.

Who the hell buys a Pixel to not use Google's Android?

I get doing that with a Xiaomi or a Honor for the excellent hardware plus custom software. But a Pixel?!

 

[MarkizSchnitzel]

Most others dropped support and google remains.
Pixelos is nothing special anyway.
Limited launcher, font issues..
I know I've read recently an article claiming the underground has switched to Pixels for it 🙂

 

[misuspita]

What are you talking about?

Was talking about going to lineageOS and nfc payments are not available. 10 years ago I was content buying cheap xiaomi phones (Mi8 was great and only 200€) but at the time, payments were mostly cash and card readers sparse... Now everything is changed, my whole banking is done by phone.

 

[DZero]

I need to say this but...
There are the SoC leaks of the Kirin 9030 on the Mate 80 Pro Max

Huawei Mate 80 Pro Max: specs, benchmarks, and user reviews

Huawei Mate 80 Pro Max – full specifications, battery, camera, display, and performance tests. User and expert reviews.

nanoreview.net

How Google had to screw up very hard with the Tensor G5 to the point that a"7nm" Kirin 9030 matches them?

That might be a clear lesson on how to NOT to make bad desicions when making a processor.

 

[mvprod123]

Finally lol

https://twitter.com/x/status/2001493963174187237

 

[Geddagod]

Why is the tensor's X4, which is clocked 15% faster than the one in Qualcomm's 8gen3, 5% slower?

 

[511]

View attachment 139768
Why is the tensor's X4, which is clocked 15% faster than the one in Qualcomm's 8gen3, 5% slower?

Google Skill issue

 

[gdansk]

Why is the tensor's X4, which is clocked 15% faster than the one in Qualcomm's 8gen3, 5% slower?

You know the first question should be if it is even running at that clock rate for the duration of the test

 

[Geddagod]

You know the first question should be if it is even running at that clock rate for the duration of the test

This reviewer does check that, tho idk if he did specifically in this video.
In the exynos 2500 video, he did note that the X925 was not holding it's boost, and noted it in the graph.

 

[gdansk]

This reviewer does check that, tho idk if he did specifically in this video.
In the exynos 2500 video, he did note that the X925 was not holding it's boost, and noted it in the graph.
View attachment 139769

No plot of frequency vs time?
It's not worth speculating further without it. Because that's the #1 most likely culprit

 

[Geddagod]

No plot of frequency vs time?
It's not worth speculating further without it. Because that's the #1 most likely culprit

Well, as I mentioned in my previous comment, the reviewer does keep track of the frequency of the cores. In the exynos 2500 video, not only did he point out thermal throttling, but even showed a snippet of the clip where you can see the frequency of each cores during testing as it was being tested.
Given he did not mention thermal throttling in this video though, at least from what I parsed of it (I don't understand mandarin lol), I think it's safe to assume that isn't the culprit.
Plus, you see a ~10% perf gap for the mid core too, despite the 8gen3 mid core only clocking ~3% faster. Do you think it's thermal throttling here too?

 

[gdansk]

Do you think it's thermal throttling here too?

Not exactly. But I suspect it's not a peak frequency as often as it should be. Voltage and thermal constraints abound. But I don't know at which time scale Tensor adjusts clocks.

 

[Geddagod]

Not exactly. But I suspect it's not a peak frequency as often as it should be. Voltage and thermal constraints abound. But I don't know at which time scale Tensor adjusts clocks.

If only I had a bilibili account, I can ask the dude directly.