Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
Office & Productivity Applications
Blender 2.78b
The multi-threaded Blender test scales well when you back it with multiple physical cores. But it also responds well to higher frequencies and Hyper-Threading.
Notice the pronounced jump in render times when we switch to Pentium processors? Blender uses AVX instructions on compatible CPUs, and the Pentium's lack of AVX support shows through. They're instead forced to utilize the SSE instruction set.
Some rendering and encode/decode programs are optimized for AVX2 extensions. And while most software can fall back to older instructions, performance is typically lost in the process.
Cinebench R15
Maxon's Cinebench single-core benchmark yields predictable results, given what we know about each architecture. The Core i5-7400 uses its top 3.5 GHz Turbo Boost bin to provide a slight advantage over the 3.5 GHz Pentium G4560. Intel's Haswell-based Pentium G3258 employs a lower 3.2 GHz frequency, so it naturally lands below the G4560. The Athlon suffers from an architecturally-imposed IPC deficit and falls to the bottom of our test pool.
The multi-core test scales well. Intel's Pentium G4560 and G4620 enjoy a healthy lead over the G3258 due to their higher frequencies and Hyper-Threading support. But this test prefers the Core i3's more aggressive clock rates and increased cache, along with the Core i5's additional physical cores.
Although Cinebench's OpenGL workload is mostly a test of graphics performance, our Radeon RX 470 behaves differently depending on the host processor it's paired with.
HandBrake
Our HandBrake workload involves converting a 4.19GB movie file into an MP4, so it's a long and repeatable threaded benchmark. Like our Blender results, we notice a big drop in performance as we transition to the AVX-deprived Pentium series. Hyper-Threading helps the benchmark along compared to Intel's older Pentium G3258, but the company's Core i3s definitely enjoy an advantage due to their ISA enhancements.
LAME
Our suite favors future-looking threaded applications, but the nuances of Intel's Turbo Boost technology are most perceptible in lightly-threaded tests. We use LAME to characterize single-threaded performance in a simple .WAV to .MP3 conversion.
Although we're accustomed to seeing the highest available Turbo Boost frequencies in this metric, Intel's Core i5-7500 only hit 3.7 GHz (instead of its 3.8 GHz ceiling), allowing the 3.7 GHz Pentium G4620 to keep pace. We also noticed the Core i5-7600K behaving similarly; it jumped to 4.2 GHz briefly, but fell to 4.1 GHz for the remainder of our test.
7-Zip 16.02
7-Zip utilizes all available execution resources, rewarding processors with more cores and higher clock rates. The Pentium G4560 lags behind its more powerful relative by a mere 4%.
Notably, the Pentium G4560 beats the G3258 by 40%, which is a result of both its faster clock rate and the additional logical cores.
Adobe After Effects CC
After Effects is one of the few Adobe applications that launches several concurrent threads to spread work across multiple cores. As a result, physical cores are favored, though Hyper-Threading does facilitate greater utilization of the dual-core models. Unsurprisingly, there is a slim margin between the Pentium G4560 and G4620.
Adobe Illustrator CC (64-bit)
A responsive storage subsystem and high clock rates come into play during the Illustrator workload.
Adobe InDesign CC (64-bit)
AMD's Athlon X4 750K stands out due to its extremely low performance, while the Core i5-7400 falls between Intel's Pentium G4620 and G4560.
Adobe Photoshop CC (64-bit)
The Photoshop Light test performs simple tasks like changing the color balance and auto-leveling, while the Heavy sequence employs more stressful manipulations. The CPU-centric benchmarks perform to our expectations, and the results of both tests align similarly.
Advanced Photo Editing & 4K Video Editing
Photo and video editing is a universal task that benefits greatly from OpenCL acceleration. The conventional test relies on host processing, while the accelerated results quantify the benefit of heterogeneous computing with a Radeon RX 470.
There's a one-second difference between the Pentium G4620 and G4560 in our photo manipulation workload, which grows to seven seconds under a more demanding 4K video editing benchmark. OpenCL acceleration has a potentially tremendous impact on performance, in this case whittling the difference between Pentiums down to an imperceptible level.
The Athlon X4 750K is unbearably slow on its own as we edit 4K video. Fold in the power of heterogeneous compute, though, and it becomes competitive.
Microsoft Excel 2016 - Word, Excel, & PowerPoint
Mundane Excel tasks don't require a lot of computational horsepower, so our test pool offers generally acceptable performance across the board.
Power & Thermals
Unfortunately, BCLK-based overclocking is mostly fruitless. Intel encouraged motherboard vendors to disable the feature, even via third-party clock generators, with recent BIOS updates.
We didn't see any thermal bottlenecks with our Corsair H100 v2 cooler and these mainstream CPUs, so there's frankly not much to say about heat. We did record temperatures with AIDA, but a changing ambient environment likely exaggerates any reported delta between our test runs.
The Pentium G4620 averaged just 29W during the stress test, and the G4560 dipped to 24W. In comparison, the Core i5-7600K averaged 46W, and the Core i3-7350K consumed 29W. The low-power Pentiums perform fine with Intel's stock air cooler, helping you save money for other components.
The number of configurations we tested was so large that our custom graphics charts couldn't accommodate another sample and remain readable, so we dialed in a relatively simple 4.3 GHz overclock on the aging Athlon X4 750K at 1.43V rather than bore you with stock results. The tuned settings push AMD's power consumption to over 100W, which has an understandably detrimental effect on performance per watt comparisons.
