Table of Contents
Specifications and Features
Back in July of 2020, we reviewed the Silicon Power US70 NVMe drive. We concluded that it was one of the best value drives on the market at that time. So, when they approached us about reviewing a new DDR4 kit they are launching soon, we were incredibly excited. They’ve made a name for themselves in the memory industry by always pushing the price per performance barrier.
The new XPOWER Zenith RGB release offers pretty much everything you’d want in a memory module. The sides are clad in a stamped and anodized aluminum heat sink to keep them from overheating. In terms of the lighting, it’s a fairly straightforward RGB design with a clean and elegant look. The entire top of the module is an unobstructed light diffuser. With a tight budget in mind, the Zenith RGB module does not have the premium build quality we’ve seen with some other modules, but it does check all the boxes for RGB memory targeted to gamers.
They launched it with a fairly sizable product lineup consisting of 3200 MHz and 3600 MHz variants. The website does allude to a 4133 MHz rating coming down the line, but for now, the initial product launch is as follows:
|XPOWER Zenith RGB Ordering Information|
|Speed Rating||Capacity||CAS Latency, Voltage||Part Number|
|3600 MHz||8 GB||CL18, 1.35 V||SP008GXLZU360BSD|
|3600 MHz||16 GB||CL18, 1.35 V||SP016GXLZU360FSD|
|3600 MHz||16 GB||CL18, 1.35 V||SP016GXLZU360BSD|
|3600 MHz||8 GB x 2||CL18, 1.35 V||SP016GXLZU360BDD|
|3600 MHz||16 GB x 2||CL18, 1.35 V||SP032GXLZU360BDD|
|3600 MHz||16 GB x 2||CL18, 1.35 V||SP032GXLZU360FDD|
|3200 MHz||8 GB||CL16, 1.35 V||SP008GXLZU320BSD|
|3200 MHz||16 GB||CL16, 1.35 V||SP016GXLZU320FSD|
|3200 MHz||16 GB||CL16, 1.35 V||SP016GXLZU320BSD|
|3200 MHz||8 GB x 2||CL16, 1.35 V||SP016GXLZU320BDD|
|3200 MHz||16 GB x 2||CL16, 1.35 V||SP032GXLZU320FDD|
|3200 MHz||16 GB x 2||CL16, 1.35 V||SP032GXLZU320BDD|
Features at a glance
- High-quality gaming memory module built for hardcore gamers
- Equipped with a colorful band of RGB lights to electrify a gaming PC rig
- Blazing speeds ranging from 3200 MHz to 4133 MHz with almost zero lag time allow for maximum DDR4 performance
- 100% tested for stability, durability, and compatibility on major motherboard brands
- Durable aluminum heat spreader provides maximum heat dissipation and thermal management
- Supports Intel® Extreme Memory Profile (Intel® XMP) Standards
- Up to 10-layer PCBs to ensure stability and reliability
- Low voltage of 1.35 V – 1.4 V for less power consumption
Silicon Power supplied us with what is effectively the highest-rated kit of their initial launch, the 3600 MHz 16 GB x 2 variant. In the table below are specific details of our test kit:
|XPOWER Zenith RGB|
|Dimensions (mm)||133.4 x 38.5 x 8.0|
|Capacity||32 GB (2 x 16 GB)|
|Type||288-Pin DDR4 SDRAM|
|Rated Frequency||DDR4 3600 (PC4 28800)|
|Chipset||Intel XMP 2.0|
|Multi-channel Kit||Dual Channel Kit|
|Certification||CE, FCC, Green dot, WEEE, RoHS|
Below is a screenshot of Thaiphoon Burner, a free tool that allows users to read the Serial Presence Detect (SPD) firmware of the DRAM. The SPD information is critical in determining how the motherboard will recognize it out of the box.
As the screenshot shows, this specific memory kit is composed of Hynix ICs arranged in a dual-rank configuration. From the SPD read, we are not able to determine the exact die used on the modules. Based on the price point and specification, we would guestimate that it’s the C-Die, which could be considered the second-tier IC within the Hynix family. This refers to the fact that there are ICs located on both sides of the PCB. In the case of our 16 GB module, there are 8 on each side for a total of 16 x 1 GB ICs. The SPD read also shows us that this PCB runs the B1 layout, which we confirmed with a physical inspection.
DDR4 PCBs are broken down into three major layout designs. The older design, called A0, has the ICs spaced out evenly on the PCB and can limit the maximum frequency. The newer A1 and A2 designs (effectively the same) place the ICs closer together and closer to the PCB connection edge. The A2-style PCB has become the new unofficial industry standard because it allows for higher frequencies and generally better compatibility. As a result, motherboard manufacturers are now routing memory traces to coincide with the IC placement on A2-style PCBs.
The B1-style PCB layout is similar to the A1/A2 style but carries the ‘B’ to signify a dual-rank format. We will use the ASRock Z590 OC Formula for testing today, which has been specifically optimized for A2-style memory PCBs, so we should have the optimal configuration.
Packaging and Product Tour
The packaging could not get any more basic, and we think it’s a perfect choice. It’s a typical single-use style blister pack with a color-printed insert. Once the memory is in your hand, the packaging for budget memory has fulfilled its usefulness. The printed insert is not specific to the memory kit and contains little to no relevant information on the back. Furthermore, we found it very interesting that the Silicon Power brand name was completely absent on the front. It’s just going to get tossed in the trash as soon as you open up the memory, but still, we feel it’s a slightly missed opportunity for brand placement.
Inside the package, there are no accessories, manuals, or case stickers. Each memory module has a high-quality printed sticker on the side facing the 24-pin power connector. The sticker does actually provide a fair bit of information, such as the part number, frequency, CAS latency, and kit density.
Meet the XPOWER Zenith Memory
As RGB memory goes, this is by far one of the lowest height models we’ve seen. With AIO CPU coolers being the first choice for most enthusiasts, there’s no longer a concern for memory height. As such, we’ve seen memory modules, especially RGB ones, grow in height and complexity.
The XPOWER Zenith has a total height of 38.5 mm – that’s excellent news for air coolers as there shouldn’t be any interference problems.
The heat sink material is aluminum and was stamped and anodized to develop the shape, size, and color. Each side is one piece that has been bent and shaped to form the signature design. While it doesn’t provide much in the surface area for cooling, it does help with heat dissipation and makes the memory look nicer.
Recalling above that the overall height is under 39 mm, it doesn’t leave much room for a large or thick light diffuser. As you can see in the pictures below, the result is that there are indeed some visible hot spots. The light diffuser is quite thin and results in light and dark areas, called hot spots. Some may see the visible hot spots as a deterrent, but not everyone does. Looking at the combined effect of the heatsink and RGB element, it’s a beautiful combination.
Almost the entire top of the stick is an unobstructed RGB light diffuser. Looking at the top-down view, we can see that it’s not just a straight rectangular piece, they’ve given it some shapes and curves. Printed in black ink is the ‘XPOWER’ logo and the text ‘DDR4 RGB’.
Silicon Power does not offer any standalone lighting control software so we tested the RGB functionality with ASRock Polychrome Sync. The bundled motherboard software easily handled the control of the lighting. Although it’s not as advanced as other control software suites we’ve used in the past, it does the job and gives you full access to the lighting.
Software download: Polychrome RGB
Testing and Overclocking
If and when the XMP profile has been established to be stable, we evaluate the memory from an overclocking perspective and 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, except in rare circumstances.
Below is the test system and resulting memory speeds used to evaluate the memory and run the various benchmarks.
|CPU||Intel Core i9-11900K Rocket Lake 8-Core|
|Cooler||Corsair H115i RGB PRO XT|
|Motherboard||ASRock Z590 OC Formula|
|Graphics Card||EVGA RTX 2080 Ti Kingpin Edition|
|Solid State Drive||T-Foce CARDEA Liquid 1 TB|
|Power Supply||EVGA Supernova 1600 W P2 80 Plus Platinum|
|Operating System||Windows 10 x64 V1909|
|Memory Speeds Compared|
|XMP Profile ~ 3600 MHz CL19 + Auto Sub Timings @ 1.35 V|
|Overclock 1 ~ 3600 MHz CL16-18-18 + Improved Sub Timings @ 1.45 V|
|Overclock 2 ~ 4000 MHz CL18-20-20 + Improved Sub Timings @ 1.50 V|
|Overclock 3 ~ 4533 MHz CL20-24-24+ Auto Sub Timings @ 1.55 V|
XMP Profile ~ 3600 MHz CL18-22-22 + Auto Sub Timings @ 1.35 V
It’s not a foregone conclusion that the XMP profile will work, in fact, it’s somewhat common for the XMP profile not to work or have stability issues. However, we’re happy to report that we had absolutely no trouble with the XMP profile with all BIOS settings on auto.
The XMP profile proved to be incredibly stable too. We never once crashed from the OS, even after a few hours of hard testing. Below are the resulting AIDA64 test bench result and the XMP profile timings.
Overclock 1 ~ 3600 MHz CL16-18-18 + Improved Sub Timings @ 1.45 V
The objective of the first overclock test was to create the absolute best profile we could with the Gear 1 setting. The Gear 1 setting can be a huge performance increase for some programs, and it can also be slightly slower in others. We know that single-threaded applications can benefit from Gear 1, while multi-threaded ones might be hindered. Regardless of the potential efficiency difference, we wanted to show the best that Gear 1 has to offer.
We had hoped to achieve stability with 3733 MHz, but unfortunately, it failed about half of the benchmarks we tested, even with relatively loose XMP timings. Therefore, we backed the frequency down to 3600 MHz with Gear 1 and worked on the timings instead.
Overclock 2 ~ 4000 MHz CL18-20-20 + Improved Sub Timings @ 1.50 V
For the second overclock attempt, we tried to create a balanced profile between frequency and timings. Unfortunately, as we’ve seen before, Hynix ICs don’t offer much flexibility regarding timings. We ended up finding stability at 4000 MHz CL18, but it needed 1.50 V.
Overclock 3 ~ 4533 MHz CL20-24-24+ Auto Sub Timings @ 1.55 V
Lastly, we wanted to know where the frequency wall is. We started with very loose primary timings of 22-28-28-48 and soon found out that we couldn’t train anything beyond 4533 MHz. Once we had that frequency dialed in as our relative max, we worked to enhance all of the timings to be the best they could be.
This frequency proved to be very difficult but possible. To achieve stability, we needed to bump up the voltage to 1.55 V, which is technically against our own rules. Furthermore, the primary timings needed to be a fairly loose CL20-24-24, and the sub-timings needed to be set on auto for the trick to work.
Benchmark download: AIDA64
AIDA64 is perhaps the most important tool we have for assessing memory performance. It’s quite old but still very relevant because it scales beautifully with changes to both memory speeds and timings. It’s worth noting, however, that with modern DDR4, we see the biggest effect in the scoring from frequency overclocking. In terms of our results today, the scores came out very linear because our frequency overclocking had a linear progression. Keeping the frequency at 3600 and only changing the timings did show a noteworthy gain. However, the percentage of gain is much smaller when compared to overclocking the frequency.
Geekbench 4 Multi-Core
Benchmark download: Geekbench 4
Next, we used Geekbench 4 to test our memory profiles. It’s one of the best synthetic benchmarks for predicting ‘real-world’ performance. This benchmark utilizes 25 unique sub-benchmarks of varying complexity and then collates them to create an overall picture of performance.
Focusing on the total score, we see that the ‘real world’ benchmark results look similar to AIDA64. Typically we see Geekbench 4 scores trend with timing efficiency more than frequency overclocking, but our scores came out linear. Although the percent of change is small across the board, our Gear 1 test showed no performance degradation compared to the Gear 2 ones.
Si Software Sandra
Benchmark download: SiSoftware Sandra Lite
Lastly, we examined the performance using a few memory benchmark tests offered within the SiSoftware Sandra suite. The flagship product, known as Sandra, is a powerful suite of many different benchmarks used to evaluate the computer performance of all major components, including the processor, graphics, memory, and disk.
So we’ve seen how the memory compares against itself when overclocked, but how does it compare against other XMP profiles on the market today? In this section, we exclude all overclocking results and use only XMP profiles for comparison against different brands. All of the test results below were newly generated on the Z590 OC Formula motherboard.
Dual-rank modules appear to have no extra advantage with scoring potential in AIDA64. The benchmark is too limited by frequency to make use of improved interleaving. Therefore, we find the 3600 MHz a big limiting factor in this benchmark compared to speedy XMP profiles such as the Crucial Ballistix MAX 4400 19-19-19.
Our Geekbench 4 results show off the power of dual-rank modules nicely. Looking at the total score, we find that all three of our dual-rank kits came out ahead of the pack. The XPOWER Zenith came in third within the other dual-rank kits we examined, but it still beat even the fastest single-rank memory modules.
Silicon Power put together an attractive and solid-performing memory module that’s right at home in any gamer or enthusiast’s computer. Overall, they delivered a budget-conscious kit that offers more than its price would indicate. Style is always subjective, but the dark grey heat sink is attractive and blends right in with most motherboards out there. The memory just about qualifies as low-profile, with a total height of just under 39 mm. The downside to this is that the light diffuser didn’t blend the colors as well as we would have liked and it showed hot spots.
The Hynix ICs held true to their stereotype in terms of overclocking and gave us the experience we expected. There was plenty of overclocking headroom with frequency, but not much with regards to the timings. We managed to make 4533 MHz stable, which is a huge feat considering we are working with 32 GB of DDR4. There’s no doubt about it: the memory is incredibly fast out of the box and provides plenty of overclocking room for those who want to step outside of the XMP rating.
The MSRP of $203 is at the lower end of the price spectrum for this speed and capacity. Retail pricing usually comes in under the MSRP for DDR4 modules, so we’d expect the actual retail price to be extremely aggressive. We know that Silicon Power DDR4 memory promises to give you a big bang for the buck, but the XPOWER Zenith even surprised us with the overall value they deliver.
Click the stamp for an explanation of what this means.
David Miller – mllrkllr88