Google's long-awaited Pixel 8 and Pixel 8 Pro are now "official," after being unofficially and then officially "leaked" in recent months. In terms of physical appearance, there hasn't been much change since Google introduced this design language with the launch of the Pixel 6 series around three years ago. Over that time, Google has polished and developed the design slightly yet successfully. However, how Google's new Tensor G3 performs, particularly in the eyes of tech fans, is of great interest.
Although Google loves to take credit for the Tensor chips,
they are the result of a tight technical collaboration with Samsung. This
involves utilizing the Exynos chip design architectures from Samsung LSI as
well as the silicon technologies from Samsung Foundry. As many people are
aware, Samsung's own Exynos chips have underperformed partly due to Samsung
Foundry's difficulties optimizing its fabrication method. As a result, chips
have been somewhat inefficient, and they can't match the outright performance
of the competition because of this inefficiency, which restricts clock speeds.
Current leakage (inefficiency) generates heat, which causes instability under
long-term loads.
Unsurprisingly, the same flaws that afflicted Samsung's
Exynos processors have also affected Google's Tensor-branded chips. Despite
providing enough performance for normal use and daily work, the lack of
outright performance hasn't been as worrisome as the lack of efficiency. As a
result, Pixel battery life lags behind the competitors. This hasn't seemed to
be as big of an issue for Pixel users as it has for Samsung Galaxy owners in
the recent past, where Exynos CPUs were fitted to flagships in certain countries
while higher performing Snapdragon chips were placed in others.
As is common these days, some lucky consumers have already
received the Pixel 8 series and have begun releasing their benchmark results.
The Tensor G3 is built on Samsung's 4LPP (4 nm, Low Power Plus) node, rather
than the company's newer 4LPP+ node, and has a 9-core configuration with 1x
Cortex-X3 (3.00 GHz), 4x Cortex-A715 (2.45 GHz), and 4x Cortex-A510 (2.15 GHz).
The Qualcomm Snapdragon 8 Gen 2 has an 8-core CPU design that looks like 1x
Cortex-X3 (3.2 GHz), 2x Cortex-A715 (2.8 GHz), 2x Cortex-A710 (2.8 GHz), and 3x
Cortex-A510 (2.00 GHz). It's built on TSMC's far more efficient and powerful N4
(4 nm) node.
Despite having one less core, the initial Tensor G3
Geekbench results are underwhelming. Instead of being comparable to the
Snapdragon 8 Gen 2, it has ended up being more comparable to the performance of
last year's Snapdragon 8+ Gen 1. While Google has been able to tweak the Tensor
by cooperating with Samsung, such as integrating it with a Google-designed
Tensor Processing Unit (TPU), Samsung's node technology plainly continues to
impede its performance. This includes the first GPU test, courtesy of
@Tech_Reve, which shows the new Pixel 8 models running the 3D Mark Wild Life
Stress Test.
The Tensor G3 includes an Arm-designed Mali-G715 GPU, which,
like its CPU, is a respectable design on paper. It, like the new Apple A17 Pro,
has hardware-accelerated ray-tracing, making it highly sophisticated. However,
Samsung's 4LPP node has repeatedly failed to deliver on this promise. The Wild
Life Stress Test loops the 3D graphics scene from the normal Wild Life Test for
20 minutes. It evaluates the GPU's long-term performance and stability. The
GPU, CPU, node, and device cooling system are all factors to consider here. The
Pixel 8 has a greatest loop score of 8,216 and a lowest loop score of 4,316,
with a very low stability of 52.5%, but the Pixel 8 Pro has a maximum loop
score of 8,572 and a lowest loop score of 5,029, with a somewhat higher stability
of 58.7%. These are not ideal outcomes.
Despite the fact that both versions use the Tensor G3, the
ordinary Pixel 8 lacks a vapor chamber, which helps to explain the majority of
the difference - the other being its somewhat more thermally limited design.
The sustained performance is inferior to that of the Apple A17 Pro or the
Snapdragon 8 Gen 2. The A17 Pro returned a stability value of 78.9% in our
iPhone 15 Pro Max review, whereas the Snapdragon 8 Gen 2 returned a stability
result of 69% in our Galaxy S23 Ultra review. The iPhone 15 Pro Max's stability
lowers to 65.4% in the somewhat more difficult Wild Life Extreme Stress test,
while the Galaxy S23 Ultra slips to 58.7%.
If the Tensor G3 had been manufactured on a TSMC 5nm or 4nm
node, there is no doubt it would have performed far higher in both the
Geekbench and 3D Mark benchmarks. While it will still provide enough
performance for day-to-day usage, it is doubtful that it will outperform its
competitors in terms of performance or efficiency, even if on theory it should.
It's difficult to envision this being anything other than a sore point in
Google's relationship with Samsung; Google intended to distinguish with Tensor,
but not in this way.
