[ad_1]
The phone hasn’t even launched yet, but the prospect of a Pixel 6 series powered by a custom Google Tensor SoC already raises some big questions. Can the chip catch Apple? Will it really use the latest and greatest technology?
Google could have bought chipsets from its longtime partner Qualcomm or even bought an Exynos model from its new friends at Samsung. But it wouldn’t have been so fun. Instead, the company worked with Samsung to develop its own chipset using a combination of standard components and some of its internal machine learning (ML) silicon.
According to a solid report, the Pixel 6’s Google Tensor SoC will be a little different from other flagship chipsets on the market. Of course, we’ll keep the benchmarking and all performance and battery verdicts when we have the device in hand. But we already have a lot of information to dive into a paper comparison between Qualcomm’s latest chipsets (and Samsung too while we’re at it). How is the confrontation between the Google Tensor chipset and Snapdragon 888 going? Let’s take a look at the beginning.
Google Tensor vs. Snapdragon 888 vs. Exynos 2100
While Qualcomm and Samsung’s next-gen SoCs aren’t too far removed, the Google Tensor chip is designed to compete with the current-gen Qualcomm Snapdragon 888 and Samsung Exynos 2100 flagship chipsets. We will therefore use them as the basis for our comparison.
Google tensor | Snapdragon 888 | Exynos 2100 | |
---|---|---|---|
CPU |
Google tensor: 2x Cortex-X1 Arm (2.80 GHz) |
Snapdragon 888: 1x Cortex-X1 Arm (2.84 GHz) |
Exynos 2100: 1x Cortex-X1 Arm (2.90 GHz) |
GPU |
Google tensor: Mali-G78 arm (854MHz) |
Snapdragon 888: Adreno 660 |
Exynos 2100: Mali-G78 MP14 arm (854MHz) |
RAM |
Google tensor: LPDDR5 |
Snapdragon 888: LPDDR5 |
Exynos 2100: LPDDR5 |
ML |
Google tensor: Tensor processing unit |
Snapdragon 888: Hexagon 780 DSP |
Exynos 2100: Triple NPU + DSP |
Multimedia decoding |
Google tensor: H.264, H.265, VP9, AV1 |
Snapdragon 888: H.264, H.265, VP9 |
Exynos 2100: H.264, H.265, VP9, AV1 |
Modem |
Google tensor: 4G LTE |
Snapdragon 888: 4G LTE |
Exynos 2100: 4G LTE |
To treat |
Google tensor: 5 nm |
Snapdragon 888: 5 nm |
Exynos 2100: 5 nm |
As might be expected given the nature of their relationship, Google’s Tensor SoC relies heavily on Samsung’s technology present in its latest Exynos processor. According to the report, the modem and GPU configuration is borrowed directly from the Exynos 2100, and the similarities extend to support for similar AV1 media decoding hardware.
If the GPU configuration does indeed match Samsung’s Exynos 2100, then the Pixel 6 will be a decent gaming phone as well, although there are still a few images behind the graphics capabilities of the Snapdragon 888. Nonetheless, it will be a relief for those. who are hoping for proper flagship-level performance from the Pixel 6. However, we expect the chip’s tensor processing unit (TPU) to deliver even more competitive machine learning and AI capabilities.
Read more: Snapdragon 888 vs Exynos 2100 tested
The Google Tensor SoC appears to be competitive on CPU, GPU, modem, and other technologies.
Google’s 2 + 2 + 4 processor setup is a weirder design choice. It’s worth exploring in more detail, which we’ll come back to, but the important point is that two powerful Cortex-X1 processors should give the Google Tensor SoC more grunt for single threads, but older Cortex-A76 cores. can make the chip a weaker multitasking. It’s an interesting combination reminiscent of the unfortunate configurations of Samsung’s Mongoose processor. However, there are big questions to be answered about the power and thermal efficiency of this design.
On paper, the Google Tensor processor and Pixel 6 series look very competitive with the Exynos 2100 and Snapdragon 888 found on some of the best smartphones of 2021.
Understanding the design of the Google Tensor processor
Let’s move on to the big question on every tech enthusiast’s lips: why would Google choose the 2018 Arm Cortex-A76 processor for a cutting-edge SoC? The answer lies in a compromise between surface, power and heat.
I dug up a slide (see below) from a previous Arm announcement that helps visualize the important arguments. Admittedly, the scale of the graph is not particularly precise, but what should be remembered is that the Cortex-A76 is both smaller and less powerful than the new Cortex-A77 and A78 given the same clock speed and same manufacturing process (ISO-Comparison). This example is on 7nm but Samsung has been working with Arm on a 5nm Cortex-A76 for some time. If you want numbers, the Cortex-A77 is 17% larger than the A76, while the A78 is only 5% smaller than the A77. Likewise, Arm only managed to reduce power consumption by 4% between the A77 and A78, leaving the A76 as the smaller and less power-hungry choice.
The trade-off is that the Cortex-A76 offers much less peak performance. By sifting through Arm’s numbers, the company achieved a micro-architectural gain of 20% between the A77 and A76, and an additional 7% on a process comparable to the switch to the A78. As a result, multithreaded tasks can run slower on the Pixel 6 than its Snapdragon 888 rivals, although that of course depends a lot on the exact workload. With two Cortex-X1 cores for heavy loads, Google can be sure its chip has the right mix of peak power and efficiency.
That’s the crux – the choice of the older Cortex-A76s is inextricably linked to Google’s desire for two high-performance Cortex-X1 processor cores. There is only a small amount of floor space, power, and heat that can be spent on a mobile processor processor design, and two Cortex-X1s push those limits.
Opting for smaller, less powerful cores frees up the chip’s silicon, energy, and thermal budget for those larger components. Alternatively, you could argue that the choice of two Cortex-X1 processor cores forces Google to adopt two smaller, less powerful mid-tier cores. But why would Google want two Cortex-X1s when Qualcomm and Samsung are happy and work great with just one?
Read more: Why the Pixel 6’s Tensor chip is actually a big deal (and why it’s not)
Besides the raw single-thread performance improvement, the kernel is 23% faster than the A78, the Cortex-X1 is an ML workhorse. Machine learning, as we know, is a big part of Google’s design goals for this custom silicon. The Cortex-X1 offers 2x more machine learning number computation capabilities than the Cortex-A78 thanks to the use of a larger cache and doubles the bandwidth of the SIMD floating point instructions. In other words, Google is swapping some general multicore performance in exchange for two Cortex-X1s that boost its TPU ML capabilities. Especially in cases where it might not be worth running the dedicated machine learning accelerator. Although we don’t yet know how much cache Google intends to associate with its processor cores, which will also make a difference to their performance.
Two powerful Cortex-X1 cores deviate from Qualcomm’s successful formula which has its own pros and cons.
Despite the use of Cortex-A76 cores, there is still a potential trade-off between power and warmth. Tests suggest that a single Cortex-X1 core is quite power hungry and can struggle to maintain peak frequencies in today’s flagship phones. Some phones even skip performing tasks on the X1 to improve power consumption. Two onboard cores double the heat and power issue, so we have to be careful suggesting that the Pixel 6 will beat the competition simply because it has two powerful cores. Sustained performance and energy consumption will be essential. Remember that Samsung’s Exynos chipsets powered by its powerful Mongoose cores suffered from this issue.
Google’s TPU differentiator
One of the few unknowns remaining about the Google Tensor SoC is its Tensor processing unit. We know that it is primarily responsible for performing Google’s various machine learning tasks, such as speech recognition, image processing, and even video decoding. This suggests a reasonably general-purpose inference and media component that connects to the chip’s media pipeline.
Related: How on-device machine learning has changed the way we use our phones
Qualcomm and Samsung also have their own dedicated ML silicon parts, but what’s particularly interesting about the Snapdragon 888 is how diffuse these processing parts are. Qualcomm’s AI engine is spread across its CPU, GPU, Hexagon DSP, Spectra ISP, and Sensing Hub. While this is good for efficiency, you won’t find a use case that runs all of these components at once. Thus, Qualcomm’s 26TOPS of system-wide AI performance is not used often, if ever. Instead, you’re more likely to see one or two components running at a time, such as the ISP and DSP for computer vision tasks.
Google says its prowess in TPU and ML will be the key differentiator.
Google’s TPU will undoubtedly include various sub-blocks, especially if it also performs video encoding and decoding, but it looks like the TPU will house most, if not all of the Pixel 6’s ML capabilities. can leverage most of its TPU power at the same time, it might just be able to outperform its competition for some really interesting use cases. But we’ll just have to wait and see.
Google Tensor vs Snapdragon 888: the early verdict
With Huawei’s Kirin on the back burner, the Google Tensor SoC has poured much-needed fresh blood into the mobile chipset’s colosseum. Of course, we’ll wait until we have the phone in our hands before drawing any conclusions. But on paper, the Google Tensor looks just as convincing as the flagships Snapdragon 888 and Exynos 2100.
As we anticipated from the start, the Google Tensor is not designed to outperform current generation processors. However, he is pursuing his own innovative approach to the mobile treatment problem. With two high-performance processor cores and its in-house TPU machine learning solution, Google’s SoC is shaping up to be a little different from its rivals. Although the real game changer may be Google offering five years of OS updates to its own silicon.
What do you think of the Google Tensor vs Snapdragon 888 and Exynos 2100? Is the Pixel 6’s processor shaping up to be a real flagship competitor?
[ad_2]
Source link