Table of Contents
Modern system memory can be flashy and eye-catching but not always up to par when it comes to overall system performance. Showing a bit of flash without RGB LEDs is a set of DDR4 Mushkin from the Redline Frostbyte series. In this case, a 2×16 GB 3600 MHz CL18 kit. Mushkin is a name we’ve come to know and trust offering high-performance memory with an emphasis on XMP 2.0 compliance and quality IC’s. We’ll explore this memory further and put it to the test with overclocking and benchmark programs.
Specifications and Features
Mushkin’s Redline product line spans a wide range of module densities, rated frequencies, kit configurations, and even a few different heat sink choices. Starting out with humble 2666 MHz modules and ranging up to the speedy 3600 Mhz ones, the Redline product series has a good range to fit modern DDR4 needs. You can get individual stick densities of 4 GB, 8 GB, and 16 GB. With kit combinations of up to 4 sticks. The options are seemingly endless, and that’s a good thing for consumers. If that weren’t enough options, they also offer every memory module in either the Frostbyte G3-R or the Ridgeback G2-R heat sink.
At the top of the product line is the 3600 MHz kit composed of two 16 GB modules priced at $154.46. At the lower end of the retail spectrum is the 2666 Mhz kit composed of two 8 GB modules with a retail price of $68.15. With an attractive price tag of just $78.99, the 3200 Mhz, 16 GB kit would make a nice addition to any budget-conscious gaming PC. Today, we will be evaluating the top end memory within this product series. In the table below we examine the particular details of our test memory.
|Mushkin Redline Frostbyte 3600 CL18
|32GB (16GB x 2)
The XMP rated speed of 3600 MHz CL18-22-22 might be difficult for the AMD platform to handle. Below is a screenshot of Thaiphoon Burner, which is a free tool that allows one to read the Serial Presence Detect (SPD) firmware of the DRAM. The SPD information is critical in determining how the stick will perform and how the computer will recognize it out of the box.
For the first time in recent history, running Thaiphoon Burner did not tell us what memory ICs this memory is using. Later on in the review, we will remove the heat sink and see if we can identify the IC from a visual inspection.
Thaiphoon Burner does tell us that this memory utilizes the modern B1 style PCB layout. That translates to the modern A2 style of PCB. This refers to the orientation of the 8 ICs on the actual memory PCB itself. While it may seem like a useless tidbit of information, this is relevant, because it can play a very substantial role in the relative overclocking ability on some motherboards.
Packaging and Product Tour
With an exceptionally low price tag of just $154.46 for a 32 GB kit, Mushkin needed to cut costs somewhere. When opening up the shipping box and unearthing your new Redline Frostbyte, don’t expect high-end packaging with felt-lined, hinged boxes. What we have here is the standard low-cost memory packaging option that is commonplace in the computer industry.
The plastic packaging has not been heat-sealed shut, so there is no need to cut into this package.
Now that the packaging out of the way, we get an up-close look.
With the shocking red coloring and sawtooth-like design on top, this memory makes a big first impression. The heat sinks are constructed of stamped aluminum which has been anodized and finished leaving a matte look behind. The bottom of the aluminum has a brushed look.
The aluminum is thin and lightweight, but it doesn’t have a low-quality feel in hand. In fact, it’s quite the opposite.
Overall the memory presents itself nicely with an attractive design for a budget-conscious build. The look is simplistic, but with a few added elements to make it stand out.
One stick has a total weight of 37 grams. The maximum dimensions, in millimeters, are 133 x 38 x 8. In terms of clearance height for some air coolers, we don’t feel this memory is tall enough to become a problem.
While the ASRock X570 Taichi doesn’t have any complimentary, red-colored parts, it does have many different RGB LED,s which can be easily configured to match this memory. The Frostbyte heat sink looks right at home on our test system.
Because Thaiphoon Burner was not able to tell us the IC type, we decided to remove the heat spreader and take a closer look. Unfortunately, in this instance looking at the IC does not help us solve the mystery.
Based on the price point and XMP rating, we can make an educated guess that this memory is composed of Hynix ICs.
Testing and Overclocking
AMD introduced serious changes to the memory capabilities and overclocking of the new X570 series chipset. Our first goal is to check and see if this memory even works with the XMP profile straight out of the box. We will attempt to apply the XMP profile and run benchmarking software to assess the relative stability.
If and when the XMP profile has been established to be stable, we will evaluate the memory from an overclocking perspective. We want to see what this memory can do, but without hurting it. Therefore, we will stick to what could be classified as 24/7 stable daily memory voltages. According to the XMP 2.0 certifications, the absolute maximum allowable voltage is 1.50 V VDDR. Thus, all overclocking endeavors will be conducted with no more than 1.50 V.
Below are the test system and resulting memory speeds that will be used to evaluate the memory and run the benchmarks.
|AMD RYZEN 7 3900X
|NZXT Kraken X62 280mm
|ASRock X570 TAICHI AM4
|PowerColor RED DEVIL Radeon RX 580
|Solid State Drive
|Team Group L5 LITE 3D SSD
|Enermax RevoBron 700W 80+ Bronze
|Windows 10 x64 v1909
|Memory Speeds Compared
|XMP ~ 3600 MHz CL18-22-22 + XMP Sub Timings @ 1.35 V
|Test Case 1 ~ 4000 MHz CL18-22-22 + XMP Sub Timings @ 1.35 V
|Test Case 2 ~ 3200 MHz CL16-18-18 + Tight Sub Timings @ 1.35 V
|Test Case 3 ~ 3600 MHz CL18-20-20 + Tight Sub Timings @ 1.35 V
As is the case with all overclocking adventures, your results may vary, so proceed only if you assume all risk. To view and examine all of the various memory profiles, we use two primary tools which include AIDA64 and AMD Ryzen Master utility. AIDA64 is a powerful system diagnostic and benchmarking tool that can be purchased for a reasonable price.
The modern DDR4 market offers a vast range of XMP memory. Shopping on Newegg, we find everything from DDR4 2400 MHz, all the way up to DDR4 5000 MHz. Improvements in XMP profile speeds are greatly attributed to modern manufacturing processes and memory PCB layouts; however, it might not be attainable on all motherboards. XMP profiles are intended to be a one-click overclock, but it likely only applies to enthusiast-grade motherboards.
As you can see below, we had no problem running our memory using the one-click XMP memory profile. In this case, the FCLK = 1800 MHz and UCLK = 1800 MHz. The ASRock X570 Taichi effortlessly applied the one-click XMP profile and we never had any stability issues.
Within the new architecture, AMD decoupled the infinity data fabric clock (FCLK) and the unified memory controller clock (UCLK). In the previous generation, the clocks were inherently linked with memory frequency and dividers. The process of decoupling meant that AMD could push the frequency to substantially higher limits, from about 4000 MHz to over 5000 MHz.
When we go to test the memory and overclocking capabilities we need to take into consideration two new variables: FCLK and UCLK. Even though these variables are somewhat out of our control, they play a huge role in the overall performance and benchmark scoring. That being said, it’s not always fair to compare one frequency against another, because the FCLK and UCLK might be running at different speeds and will, thus, skew the test results. While it’s not within the scope of this memory review to dive deeper into the Ryzen 3000 memory structure, just keep in mind that those variables are there.
For the first test case only, the operating frequency was increased with no other settings adjusted. Unlike our effortless experience with the XMP profile, attempting to overclock this memory proved very difficult. Our initial plan of setting 1.50 V failed instantly, and we learned this memory will stop posting over 1.35 V. Overclocking mileage and headroom will vary depending on your motherboard, but we were only able to increase the frequency by 400 MHz.
Below is the resulting overclock of 4000 MHz CL18-22-22-42 with FCLK = 1800 MHz and UCLK = 1000 MHz.
For the next overclock example, we wanted to see if we could lower the primary and secondary timings at all. This provides an alternative example of what types of overclocks are possible.
It’s worth noting that this memory had difficulties running low primary and secondary timings. This is one area in which a direct competitor, Samsung memory, has a clear advantage. The now infamous Samsung B-Die memory modules are able to run at substantially lower timings all around.
Running the XMP frequency of 3600 MHz, we were not able to decrease the primary timings much. Furthermore, after several attempts to lower the secondary timings, it became apparent that there was not much headroom for overclocking. In order to provide a good example of timing overclocks, we needed to also decrease the DRAM frequency to 3200 MHz.
Below is the resulting overclock with FCLK = 1800 MHz and UCLK = 1150 MHz.
For the third and final overclocking test case, the goal was to achieve a well-rounded overclocking profile which includes both tighter timings and higher frequency.
Our Mushkin Redline Frostbyte memory proved to be difficult to overclock. We were not able to overclock the frequency as well as lowering the timings, it just proved too difficult and the motherboard would not boot. We slightly lowered the primary timings and kept the frequency at the XMP rating of 3600 MHz. However, we were able to tighten up the sub-timings. This has proven to be extremely beneficial for performance in the past. It comes down to just a few timings that make the biggest difference.
Adding voltage simply didn’t help the situation and we ended up sticking to 1.35 V because our system would not post with more. Below is the resulting first overclock test case with FCLK = 1800 MHz and UCLK = 1800 MHz.
First up, we used AIDA64 Cache and Memory Benchmark. In the graph below, it’s clearly visible that each of the four different memory speeds compared had a noticeable change in the benchmark result. Looking at the read performance, we found it interesting that overclocking the frequency resulted in a lower score than stock XMP. This can be attributed to the substantially lower UCLK frequency in that case.
How did our memory overclocking impact the overall performance? In total, our overclocking endeavors didn’t have a huge impact on performance. It comes down to how we interpret the results and which tests we choose to look at. For the read performance, increasing the frequency had a clearly negative effect. Conversely, in copy performance, high frequency was the clear winner.
While we observe a small interesting relationship between the frequency and timing overclocking relationship in AIDA64, there is one big trend that’s hard to ignore. Our attempt at tightening up the timings–and sacrificing frequency to do so–has failed. It proves that this X570 platform employs performance penalties when lowering the frequency too far. Due to the inverse relationship of overclocking performance, we feel the XMP profile is the best option.
Next up is Geekbench 4, and it has proven itself to be an excellent tool for determining the real-world performance of the system being tested. This type of benchmark is purely 2D calculation-based and there is no graphical processing element, so it’s a great analytical tool to evaluate memory performance.
Focusing on the overall score, we see that this ‘real world’ test benchmark favors the tighter sub-timings. Here, all of our overclocking endeavors showed an increase in performance. The performance increase was small across the board, which is understandable in that this has low overclocking headroom. Adjusting the sub-timings and frequency had an overall beneficial effect on the performance.
Next, we will examine the performance using a few of the memory benchmark tests offered within the SiSoftware Sandra suite of benchmarks. The flagship product, known as Sandra, is a powerful suite of many different benchmarks used to evaluate computer performance of all major components, including the processor, graphics, memory, and disk.
As with the AIDA64 memory test results, interpreting the SISandra memory results is tricky because different tests represent different findings for overclocking. The high-frequency profiles are clearly better for memory bandwidth. In the Cache & Latency test, we saw the biggest improvement by keeping the frequency at 3600 MHz and lowering the timings as much as possible.
In this next section, we will examine the XMP profile performance of a few kits of memory we have in the lab. All of the memory compared here has been tested on the same exact X570 test bench and with the same exact operating system. Due to this fact, we are able to make some direct performance comparisons. For the purposes of continuity, we will examine only the XMP profile and exclude any overclocked comparisons.
First up is the AIDA64 Memory Benchmark. The graph below gives us a nice representation of the various XMP memory profiles and how they stack up against each other. As we have commented on previously, this benchmark favors frequency over timings, and it’s quite clear in this example.
Next up we look at results using Geekbench 4. We would like to point out that the graph below is not a fair comparison. The Mushkin memory has an unfair advantage with a total of 32 GB, while all others tested are only 16 GB. If nothing else, the comparison shows the advantage of more system memory for certain tasks.
There’s a lot to like about the Mushkin Redline Frostbyte memory. With red, anodized aluminum heat sinks and a saw-like pattern on top, this memory will surely make a statement in your next build without getting in the way. The XMP profile worked effortlessly on our X570 test platform and the rating of 3600 MHz CL18 proved to be quite efficient.
When it came to overclocking, our test sample left a lot to be desired. In terms of frequency and timing adjustments, there just wasn’t much headroom. Our best attempt at overclocking resulted in modest performance increases for all the benchmarks we evaluated. This is not a sad story, however, as some memory just doesn’t need to be overclocked. Mushkin delivers excellent performance right out of the package with no overclocking required.
With a price tag of just $154.46 for 32 GB of speedy DDR4, we feel that this memory has wide appeal. Other competitors in the non-RGB arena are offering the same XMP rating in a price range of $135 to $170, so Mushkin lands about midway in the field. Click the stamp for an explanation of what this means.
David Miller – mllrkllr88