Test Configuration, Results, And Final Analysis
The Tumbler PC makes its way back to the bench to receive this ASRock E3V5 Gaming/OC sample with only a few modifications. The Corsair AX860 continues to supply this system with power, with its 80 Plus Platinum certification delivering efficient power throughout a wide range of load scenarios. We rely on the DeepCool GAMMAXX 400 recommended by fellow Tom's Hardware members to cool the Xeon processor.
Test System Configuration
|RAM||Kingston KVR24N17S8/8 4x8GB DDR4-2400 CL17 UDIMM|
|Graphics||Gigabyte GeForce GTX 970 G1 Gaming 4GB|
|Sound||Integrated HD Audio|
|Network||Integrated Gigabit Networking|
Gigabyte donated a GTX 970 G1 Gaming GPU for use in today's tests. The main difference between this card and our previous Windforce sample (procured for the Gigabyte-X170 review because of time constraints) is a few MHz on the core and memory clocks and a fancy backplate. It will be interesting to see if the extra hertz provide any discernible performance differences.
Toshiba has sent the Tom's Hardware motherboard team samples of its 250GB OCZ RD400 NVMe M.2 offering. The package comes with both a PCIe break-out board and the actual drive. The lack of M.2 connectors on this motherboard board has forced us to deploy a PCIe M.2 riser card to properly compare ASRock's C232 motherboard to Gigabyte's C236.
When you talk about server or workstation PC guts, most of the time they are boring, green, and have no pizazz. For this review, Kingston supplied four sticks of its DDR4-2400MHz 8GB ValueRAM to replace the previous configuration's uninspiring green DIMMs. With the ECC vs non-ECC performance debate settled, we opted to only include the UDIMM results from Kingston here.
Gigabyte GA-X170 Extreme ECC
Synthetics and Applications
There are two different ways to look at the data this time around. We're comparing the ASRock E3V5 Gaming/OC board to the previous C236 platform (Gigabyte-X170) and to our last AMD article results (Gigabyte 970 Gaming-SLI), of course, using the same Windows 10 image. From a synthetic perspective, both platform analyses are clear. The (C232) E3V5 Performance Gaming/OC comes within points of the (C236) X170-Extreme ECC, and the only PCMark application where it doesn’t smoke the AMD system is the storage test. Sandra continues the overall trend, and the only significant change is the memory bandwidth test, where the ASRock inches out a 0.2 GB/s lead compared to the C236 platform.
There are mixed results with DiskSpd, where sequential reads were down 200 MB/s but writes were up by 50 MB/s. This could be due to the change in interface with the PCIe riser card, but the 3% difference is well within the margin of error for such a high-speed connection. Given the nature of the application tests, these results nearly put us to sleep. Handbrake clearly prefers the Intel platforms, and file compression is again so close that the scale of the chart is misleading. However, Blender shows a 20 second increase in render time on the ASRock E3V5 Gaming/OC motherboard compared to the Gigabyte X170 Extreme-ECC. Are there any other programs that we should run against this hardware? Let us know in the comments!
The inclusion of the G1 Gaming in place of the standard Windforce 970 will account for much of the differences between the Xeon-based systems. Judging by the 3DMark synthetics, both graphics solutions behave as expected and stomp the AMD 7970 card’s performance. The F1 2015 results have us perplexed, given the synthetic results for both the CPU/memory and 3DMark scores. If you’ve seen similar results in this application, let us know in the comments. Regardless, this GPU is overkill for this resolution and monitor.
Continuing with Ashes, when using High settings both chipsets are within a few frames of each other, but cranking up the settings to Crazy clearly suits the G1 Gaming solution’s additional core and memory frequencies. Metro gives a slight edge to the ASRock E3V5 Gaming/OC configuration by a few frames across all tested settings. The Talos Principle is still uninspiring as always, showing very consistent results across settings, though the G1 Gaming does gain a frame here and there.
Power And Efficiency
With our trusty Kill-A-Watt in place, we can start observing true differences between the C236 and C232. Though the ASRock E3V5 platform draws five more watts from the wall at idle, the power distribution of this C232 sample is 30W more efficient at full CPU load and 11W more efficient at full system utilization. Temperatures are higher despite the win in power draw for the ASRock E3V5. The 7° difference on the voltage regulators are clearly the result of the design of ASRock’s heatsink. The Gigabyte X170 has a behemoth of a solution compared to the ASRock E3V5, and surface area is important. Even with sizably smaller heatsinks, these components can effectively provide stable power to these voltage rails at modest temperatures.
From an efficiency perspective, AMD is clearly struggling and if the leaked Zen charts give me any hope, I have something to look forward to in the next few months. Back to Xeon, given the better power draw and the almost-up-to-snuff performance, efficiency goes to the ASRock E3V5 Gaming/OC sample.
Overclocking, Performance, and Value
Having already spilled the beans with performance, the ASRock E3V5 Gaming/OC performs nearly identically in Apps and Games. Let's not forget to mention that AMD is dragging by nearly 30%.
The wildcard for the E3V5 is overclocking. Despite Intel mandating that multipliers be locked and that chipsets not enabling overclocking, manufacturers have opted to enable overclocking by adjusting the BCLK frequency fed into the CPU. Though increasing BCLK does directly impact frequency applied by stock multiplier settings to the CPU, BCLK also gets fed into memory operating frequencies and other processor and chipset reference clocks.
Over the course of overclocking this Xeon E3-1230, we wrestled with the memory settings of this configuration and nearly threw in the towel. With each adjustment of BCLK, we had to dance around the DDR4-2133 mandated specification of the platform and eventually settled on 2200 MHz at CL17 speeds. With BCLK set to 120 MHz (up 20% from spec) and adding 50 mV to the processor, we successfully set a stable 4078 MHz clock with all power saving options disabled through the UEFI. Though nowhere near what Thomas achieved, we feel like we proved that this board can apply the juice and options to those willing to risk their CPUs and wallets.
Back to the original question: is there value in a budget gaming workstation? Judging by performance alone, the C232 can deliver near identical levels of performance to the C236 in the configuration tested. If we planned on deploying any kind of high performance attachments, such as a SAS RAID card, high performance gigabit Ethernet network, or an SLI graphics configuration, C232 is not the droid you're looking for. Ditching the NVMe riser card, we could gain four additional PCIe lanes, but compared to the wealth of options available to C236, we may just stick with what we have in deployed in this C232.
What about consumer grade options? Yes, you can get most of the same bells and whistles on Z170, if not H170, but if your livelihood relies on uptime and reliability, spend the extra cash on ECC and buy at least a motherboard with the C232 chipset.
As far as C232 is concerned, this ASRock E3V5 Performance Gaming/OC delivers a similar experience to other more expensive solutions while providing more stability and features, such as larger VRM heatsinks and that Fatal1ty flair. Outside of complaints about C232, we can’t find anything wrong with this product. However, unless ECC is necessary, we wouldn't blame anyone for using a standard consumer board instead.
MORE: Best Motherboards
MORE: All Motherboard Content