Big leak details specs for Google Tensor G5 and G6
- E
- EsseLowNitro
- L3@
- 03 Dec 2024
Listic, 06 Nov 2024Did I miss it and do we have CPU Virtualization and virtual... moreARM processors have been capable of handling hardware-accelerated virtualization for a while now. The choice of using an Imagination GPU in the Tensor G5 is to have GPU virtualization as well, for security and performance reasons inside VMs.
We don't have virtual machines in Android yet, but Google supposedly has plans to merge ChromeOS and Android, making Android a convergent OS for both mobile and desktop. That implies a lot of fun stuff is coming to Android in the coming years: see the rumors of an upcoming Pixel Laptop, an upcoming Linux terminal in Android, etc etc...
- ?
- Anonymous
- x@f
- 28 Nov 2024
I Just Read, 26 Oct 2024This Distraction statement look there and guess what NO SPO... moreThe trick is to buy the base Pixel 3-4 months after release at mid-range price :) Then you have a very good phone with great camera priced reasonably, Only think you'll miss is the telephoto and lack of LTPO but then again you are paying mid-range price and get 7 years support.
- Listic
- dWe
- 06 Nov 2024
Did I miss it and do we have CPU Virtualization and virtual machines on Android already? Since when? How are they used?
- Kangal
- Hqj
- 30 Oct 2024
Anonymous, 27 Oct 2024They initially used imagination ip, then licensed renewed a... moreThere is 0% chance that Google will use Imagination Technologies.
Not willingly. There would have to be some collusion and corruption involved.
To put it bluntly, I-T were innovators of the best GPU Technology actually ahead of Nvidia, AMD, Qualcomm, and ARM. But this British company got gutted by Apple, who managed to secure a very favourable license, and then took their chief engineers. It was at that point this company basically floundered.
But several years back, under the nose of the UK Monarchy, there was a hostile takeover from China. They basically took this company, it's intellectual property, all under the radar. This was around the time of the ARM China fiasco, which still is unresolved.
In short, around 2017 China was meant to lose access to this new technology due to Trade War by the USA. But somehow they not only secured the latest ARM IP, but one of the most lucrative GPU Companies, and also access to trade secrets from TSMC. So in principle, China could have access to Android 9, 20nm, Cortex-A76, and Mali-G76 levels of technology. But in practice, they have independently have copied 16nm silicon, use Android 11, have access to Cortex-X1, and much more advanced iGPUs. The application of these for personal electronics is one thing, but being adapted into smart weapons is the real dilemma. Like missiles which have self-defence capabilities are now a reality.
- Kangal
- Hqj
- 30 Oct 2024
Anonymous, 24 Oct 2024As one of the biggest players in the mobile industry, why i... moreIt comes down to which company has access to better equipment and the better engineers.
TSMC basically has a monopoly (80%) of the high-end stuff, with Nvidia owning the lion's share of the chips and becoming the major source. It is primarily server stuff and Machine Learning chipsets. Believe it or not, this would have been worse if it was not for the pandemic causing TSMC to falter their timeline by +2 years, whilst the state sponsored SMIC have managed to get their hands on a billion dollar machine (EUV).
Right now TSMC (+3nm) is a solid 1.5 generations ahead of Samsung (4nm), who themselves are a solid 1.5 generations ahead of Intel (-5nm), who then are a decent 3.0 generations ahead of SMIC (-9nm) from Huawei/China Military, which is ahead of the floundering ambitions of (+16nm) Global Foundries. There are other fabs out there, mostly small scale stuff, based around Europe and Japan to build chips for low-complexity stuff like Volite Flash, Solid Storage, Connectivity, Automotive Relays, and Appliances.
- ?
- Anonymous
- rK8
- 27 Oct 2024
Anonymous, 27 Oct 2024didn't apple use imagination(powervr) gpu's on th... moreThey initially used imagination ip, then licensed renewed a deal to use parts of imagination IP from 2020 later it became troublesome after Apple swindled imagation staff and ended in divorce
Now Apple uses own GPU with knowledge gained from handling imagination IP and having former imagination staff
So imagination doesn't have big contract for mobile IP, a new deal from Google would be great for both and the ecosystem
- ?
- Anonymous
- uSV
- 27 Oct 2024
Anonymous, 26 Oct 2024I support Google using Imagination GPU, immortalis GPU is v... moredidn't apple use imagination(powervr) gpu's on their phones before? then later on offshoot their own after getting some info tech from them?
- ?
- Anonymous
- U@6
- 26 Oct 2024
bukeyolacan, 26 Oct 2024No one else is using Imagination GPUs since years, Google w... moreI support Google using Imagination GPU, immortalis GPU is very well known at this point and also since ARM doesn't support open source GPU drivers
Imagination has been making open source GPU drivers since 2023
And with Google adopting imagination GPU it may give imagination more motivation to make better GPU IP
- b
- bukeyolacan
- mJ8
- 26 Oct 2024
No one else is using Imagination GPUs since years, Google will be the first it seems. According to website, it has 1.5GFlops FP32 power so equal to Adreno 730 in 8gen1. Not a good flagship level for 2025
- ?
- Anonymous
- 99s
- 26 Oct 2024
I Just Read, 26 Oct 2024This Distraction statement look there and guess what NO SPO... more1 year* but keep yapping
- I Just Read
- 8mp
- 26 Oct 2024
Just my opinion, 24 Oct 2024why not just use d9400This Distraction statement look there and guess what NO SPOILER ALERT, it's just too good to spoil
"Google Tensor G5 (codename “laguna”), likely to be the chip in next year’s Pixel 10 series, will be manufactured on TSMC’s 3 nm-class N3E — the same exact process node Apple uses for the iPhone 16 Pro’s A18 Pro and its M4 chips."
Disingenuous at best
"Long story short, Samsung’s recent process nodes typically underperform those of its rival TSMC."...
"Fortunately, Google will soon rectify that mistake by dropping Samsung and designing its upcoming chips internally."****
It been reported in articled here that Samsung and Google didn't get along, the Tensor G3 nine core was the unique and has / had, better potential for improvement (look at what the 10 Core Exynos has done is doing for Samsung.
Afaik there was little or no Samsung personal involvement in the lack lustre Tensor G4 (reverting back to 8 Core) The G4 is ranked below Exynos 2400E and the 2022 D9200 & Snapdragon 8 Gen 2.
****Design internally
SoC ain't easy , fortunately Google can use the Snapdragon 8 Gen 3 as a template
1x 3.3 GHz – Cortex-X4
3x 3.15 GHz – Cortex-A720
2x 2.96 GHz – Cortex-A720
2x 2.27 GHz – Cortex-A520
Still 2 years behind.
Pixels are OVERPRICED, so whilst Apple / Samsung (Qualcomm) Vivo (Mediatek) have flagship phone, Google is charging similar premium price for mid range SoC
- ?
- Anonymous
- LfV
- 25 Oct 2024
Anonymous, 25 Oct 2024It's bad for emulation; that part is true.Not just that though, look it up
- ?
- Anonymous
- 0Uc
- 25 Oct 2024
Anonymous, 25 Oct 2024Because there are other chipsets with powervr gpu like d930... moreIt's bad for emulation; that part is true.
- y
- yungfishstick
- Ib5
- 25 Oct 2024
Support for GPU virtualization definitely means Google is definitely working on a solution to run Linux apps on Android. It just depends how good G5's GPU actually performs and how efficient it is
- ?
- Anonymous
- 4D3
- 25 Oct 2024
Anonymous, 25 Oct 2024DM9400 has big core as Cortex X925 which is the latest one.... moreSo again, it uses old cores and nobody complains
- ?
- Anonymous
- 4D3
- 25 Oct 2024
SShock, 25 Oct 2024That's not true. You can't just take a chip "... moreThat is true. Please educate yourself on chip fabrication before talking nonsense
- ?
- Anonymous
- XRx
- 25 Oct 2024
Anonymous, 25 Oct 2024also forgot to mention that this is based on the assumption... moreprocess node size*
- ?
- Anonymous
- XRx
- 25 Oct 2024
Anonymous, 25 Oct 2024I highly doubt that, either foundry can make the chip, ther... morealso forgot to mention that this is based on the assumption that both foundries support the same node process but obviously, in reality, tsmc is always a step ahead
- ?
- Anonymous
- XRx
- 25 Oct 2024
SShock, 25 Oct 2024That's not true. You can't just take a chip "... moreI highly doubt that, either foundry can make the chip, there's no code or physical limitation that only allows a particular foundry to make it. The only other variable is letting the foundry know in advance so they can make sure they are prepared to produce it.
- S
- SShock
- AKj
- 25 Oct 2024
Anonymous, 25 Oct 2024The problem is that Google isn't designing any chips d... moreThat's not true. You can't just take a chip "blueprint" and take it to whatever foundry and they'll just make it. It doesn't work that way. You need to be designing the chip with specific foundry from the beginning. You can't just say take chip that was designed with Samsung fabs, take it to TSMC and they'll just make it in no time.