Jul 28, 2022
Intel's new Haswell-E CPU supports quad-channel DDR4, but my testing reveals it may not matter. As if memory channels were shotgun barrels, you may think of them as such. There are two barrels better than one since you've seen it in video games. When it comes to computers, quad-channel RAM is like a four-barrel shotgun: There is greater memory bandwidth provided to the CPU if there are more memory ports. A separate stick of RAM is required for every channel in a contemporary PC. Of course, the CPU plays a factor here as well. Up to four channels may be supported by consumer CPUs like the Core i7-5960X, the latest Core i7-6700K, and the Core i7-4790K, respectively. In most cases, this is irrelevant. When purchasing a high-end motherboard and high-end processor, you wouldn't purposefully lower the functionality of the X99's quad-channel memory system by adding just two RAM modules instead of four.
What exactly is the issue? It's impossible to use a Mini-ITX motherboard with Intel's Haswell-E processor and also the four memory ports it requires. They are physically incompatible with full-sized memory modules. As a result, microprocessors will be limited to quad-core processors at best. Asrock came up with the absurd idea of entirely omitting two slots from their X99E-ITX/ac motherboard. However, if you want to construct a computer like the Falcon Northwest or the CyberPower Trinity Extreme and operate greater than four CPU cores, you'll have to decrease your bandwidth in half. You must decide how big of a financial blow you can bear.
For the sake of research, I wanted to see how much real-world performance you lose by just using half of your system's available bandwidth. Only a comprehensive X99 motherboard could be used to test this since the Asrock X99E-ITX/ac is limited to dual-channel RAM. I used a MicroExpress B20 machine that we had previously tested. One full-size ASUS motherboard with six cores of CPU powering the system, a GeForce GTX 970 graphics card & 16GB of DDR4/2666 Memory space in quad-channel capacity utilizing four 4GB modules are included.
I performed numerous benchmarks in quad-channel settings and switched to two 8GB DDR4/2666 units. Instead of removing two of the computer's four memory sticks, I felt some might be worried that the 16GB versus 8GB of overall RAM would influence the findings. Of course not, but I'll play along anyhow. We are comparing the performance of 16 GB of DDR4/2666 memory operating in dual-channel mode to that of 16 GB of DDR4/2666 memory operating in quad-channel mode.
Memory bandwidth was the first test I ran. This all-in-one benchmark suite tests and measures almost everything about your PC. Measuring a computer's available memory bandwidth has long been common practice. The outcomes were exactly what we had predicted. Quad-channel DDR4/2666 substantially increases memory bandwidth from dual-channel. Apologies, but quad-channel RAM is better than dual-channel RAM.
Another matter is where theoretical performance shows up in real-world tasks when measured by synthetic testing. Handbrake was the following experiment I ran on the system to see whether it could handle it. It's a CPU-intensive test using a well-known and free video encoder. Regarding video compression, I felt multiplying the memory bandwidth might be an excellent investment. Unfortunately, as you can see from the graph below, there was nothing. Surprised by this since I've always held that memory bandwidth improves encoding speed. I've seen it before on earlier hardware systems, and this was a surprise. Dual & quad-channel RAM made no distinction in our encoding test.
My next job was to complete the Creative Standard test in PCMark 8. Photo editing, video compression and light gaming are all included in this test's workload simulation. To limit the burden to the CPU alone, I use the traditional version instead of the GPU version. The outcome was, once again, shocking and disheartening. Additionally, Creative Standard tests in PCMark 8 reveal a small effect.
I also performed PCMark 8's Home & Work Standard tasks to mix things up a little. Again, almost doubling the bandwidth of the system RAM had no effect. No need to waste bandwidth on the PCMark 8 Work results graphic since they are the same. As with other tests, the conventional PCMark 8 Home Standard test reveals a split between dual & quad-channel RAM.
File compression, such as video encoding, is among the jobs that normally benefit from a large amount of memory bandwidth. To figure it out, I utilized WinRAR 5.21 and the compression test that came with it. Although it wasn't much, the extra memory bandwidth definitely paid off. Despite the fact that memory bandwidth is often an advantage to file compression, the difference with WinRAR is rather small.
There isn't much difference in gaming efficiency between quad-channel and dual-channel RAM setups. If you want faster app loading speeds, you won't be able to "double your RAM speed." Quite the contrary. Raising the number of processor cores might be beneficial. Furthermore, the experiments described above confirm this conclusion.