Test Results, Overclocking And Conclusion
Test System Configuration
We continue to use the test configuration from our initial Z170 round-up in the evaluation of concurrent samples.
Traditional "new on top" chart order gives way to "descending order of price" on this occasion, as the Z170 Extreme4 most closely matches the features and design of ASRock's own Z170 Extreme6. The search for value puts a target on the higher-priced board.
Synthetic Benchmark Results
The Z170 Extreme4 takes a few small wins not because it's better, but because it comes factory-overclocked by just under 1 percent. That same slight overclock is likely to hurt the board a little in power readings, and we'll keep it in mind during future performance evaluations.
3D Game Benchmarks
That 0.9 percent overclock doesn't help the Z170 Extreme4 at all in games, even though the lower half of our game settings are designed to stress the CPU and/or memory.
The Z170 Extreme4's tiny default overclock defines tiny leads in Premiere, Blender and 7-Zip, but we don't see any true breakaway moments. A look at the combined scores might just show this board slower than its 0.9 percent error.
Power, Heat And Efficiency
The Z170 Extreme4's 0.9 percent overclock appears to cost buyers a few percent in power consumption, while producing an artificial performance advantage of only 1 percent above-average. It comes out 1.2 percent above-average in performance-per-watt, but that's mostly because another board skewed the average so far downward.
Heat was likely another factor in the Z170 Extreme4's efficiency reduction when compared to the Extreme6, as ASRock boasts higher-quality voltage regulator components for the Extreme6.
|ASRock Z170 Extreme4 Frequency and Voltage settings|
|Base Clock||90-600 MHz (62.5 kHz)|
|CPU Multiplier||8x-120x (1x)|
|DRAM Data Rates||800-4133 (100/133.3 MHz)|
|CPU Vcore||0.90-1.52V (5 mV)|
|System Agent||0.95-1.35V (10 mV)|
|CPU I/O||0.85-1.25V (5 mV)|
|PCH Voltage||0.90-1.30V (5 mV)|
|DRAM Voltage||1.00-1.80V (5 mV)|
|CAS Latency||4-31 Cycles|
Equipped with one-less SATA controller than its Extreme6 sibling, the Z170 Extreme4's drastically lower price raises a few value questions. Some of those questions might be addressable by overclocking the unit, since both ASRock boards are designed for this purpose.
Both boards would get the CPU to 4.6 GHz, but the Z170 Extreme4 wouldn't hold a 1.30V CPU core under changing loads regardless of settings. Sticking to the "all boards get the same voltage limit" rule for the sake of fairness, the lower setting with a 1.30V peak resulted in a second-place maximum clock rate.
DRAM overclocking was far worse, with even the low-cost Z170-HD3 producing better results. Of course some of that could be due to the Extreme4 using more performance-oriented, rather than stability-oriented, timings.
A quick test at DDR4-2933 shows at least part of the reason why the Z170 Extreme4 came up shy in DRAM clock, as its DRAM performance at this relatively-moderate clock is excellent. That can only come from performance-oriented automatic adjustment, since I haven't set any of the timings manually. And, just between us reviewers and readers, 34GB/s is a higher bandwidth rating than many of the Z170 Extreme4's competitors have produced at DDR4-3200.
The overclocking results leave this editor in somewhat of a jam, since I was hoping to get Z170 Extreme4 results at least as good as the Z170 Extreme6. After all, the cheaper board has over a month's worth of additional firmware development. It looks like the hardware just isn't there, as ASRock really does promote the hidden value of the Extreme6's higher-priced voltage regulator.
If the comparison between the $146 Z170 Extreme4 and the $180 Z170 Extreme6 proves too difficult, perhaps it would be better to compare Gigabyte's Z170-HD3. It's a solid board for the low price of $115, but it also lacks any USB 3.1 controller, comes up a little short in CPU overclocking, and doesn't even have those little PCIe pathway switches needed to get eight of the CPU's lanes to a second graphics card. That combination of features is certainly worth $30, no?