XPG MARS 980 BLADE PCIe Gen5 x4 M.2 SSD Unboxing Review

The AVerMedia XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD uses the SMI SM2508 Solution, which adopts TSMC's 6nm process and effectively reduces heat generation compared to the Qunlian E26. It offers a choice of 1TB/2TB/4TB capacities, with sequential read/write speeds of up to 14000 MB/s and 13000 MB/s, and comes with a five-year warranty and 1480 TBW of total bits written. 13000 MB/s, with 5-year warranty and 1480 TBW (Total Bit Writes). The actual unboxing solution is equipped with Micron 232L 3D TLC chips and samsung DDR4 SDRAM-2666 2GB physical DRAM cache, and we will install the included thin profile heat sink for performance testing. In this article, we will install the supplied thin heatsink to test the performance of this Gen5 SSD, which is probably the cheapest SMI 2508 host SSD on the market.

XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD 2TB Specifications:

Interface: PCI-Express 5.0 x4
NVMe: 2.0
Interface format: M.2 2280
Continuous read speed: 14000 MB/s
Continuous Write Speed: 13000 MB/s
Capacity options: 1TB / 2TB / 4TB
Controller: SMI SM2508
NAND Flash: Micron Micron 232L 3D TLC chips
Cache: Samsung samsung DDR4 SDRAM-2666 2GB (two 8Gb cells on both sides) with Pseudo Single-Level Cell (pSLC) technology.
Size: 80 x 22 x 4.5 mm
Warranty: 5-year limited warranty
TBW Durability: 1480 TB

Huirong SM2508 Host Control XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD Unboxing

I just opened the box last time.AVerMedia ADATA LEGEND 970 PRO PCIe Gen5 x4 M.2 2280 Solid State DriveRecently, Vigor has launched the MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD under its e-Sports sub-brand XPG, which utilizes the Hui-Jung SMI SM2508 host controller.

The MARS 980 STORM and MARS 980 PRO have been launched simultaneously. The MARS 980 STORM uses liquid and air cooling, as well as a vertical symmetrical dual-fan cooling module, while the MARS 980 PRO has thick aluminum cooling fins and an active fan, and the MARS 980 BLADE unboxed today has a thin copper radiator with passive cooling. The three models are identical in terms of specifications and main controls, with the only difference being the heat sinks.

Because of its low footprint, the MARS 980 BLADE measures 80 x 22 x 4.5 mm with a thin heatsink, a total thickness of only 4.5 mm, and uses a motherboard mounting slot size of M.2 2280.

∆ XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD 2TB capacity specification.

∆ Product Features.

∆ The contents are only the SSD body and the heatsink.

∆ Copper heat sink, thickness about 0.8 mm.

The back of the ∆Copper heat sink is secured by a 3M thermal paste.

 

The XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD uses the PCI-Express 5.0 x4 transmission bandwidth interface, also known as PCIe Gen5 x4, and comes in a choice of capacities of 1TB/2TB/4TB, with sequential read and write performances of up to 14 GB/s (14,000 MB/s) and 13 GB/s (13,000 MB/s) for the 4TB and 2TB capacities, and only 10 GB/s (10,000 MB/s) for the 1TB capacity. The 4TB and 2TB have sequential read/write performance of up to 14GB/s (14,000 MB/s) and 13GB/s (13,000 MB/s), while the 1TB capacity has a write performance of only 10GB/s (10,000 MB/s).

The MARS 980 BLADE is powered by TSMC's 6nm Wisepac - Silicon Motion (SMI) SM2508, which is labeled as supporting Pseudo Single-Level Cell (pSLC) caching algorithm and LDPC (Low Density Parity Check Code) error correction. Parity Check Code (LDPC) error correction mechanism.

The 2TB capacity is a double-sided die layout with SM2508 controller, Samsung DDR4 SDRAM-2666 1GB (K4A8G165WC-BCTD) DRAM Cache chips, and two Micron 232L 3D TLC cells on the front side, while the back side has the same number of DRAM Cache chips and 3D TLC cells, each with 8Gb (Gigabyte) of DRAM Cache chips, which translates to 1GB (Gigabyte). On the back side, there are the same number of DRAM Cache chips and 3D TLC cells. The single DRAM Cache chip is 8Gb (Gigabit), which translates to 1GB (Gigabyte), and the total capacity of the two DRAM Cache chips on the front and back sides is 16Gb, which translates to 2GB.

On the front of the ∆XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD is the SM2508 controller chip, samsung DDR4 SDRAM-2666 1GB (K4A8G165WC-BCTD) DRAM Cache chip, and two Micron Micron 232L 3D TLC chips.

∆The specification sticker on the back of the unit, if torn, it is considered as human damage and will lose the factory warranty.

∆ I'm going to open the box, so rip! Back layout at a glance.

∆ Thin thermal pads are shown in real life. The difference in height between the main controller chip and the DRAM Cache chip is not that big, so the thermal pads and the heat sinks can be flatly attached to the main controller chip.

 

The 3D NAND Flash die has ADATA 600799DG 5305857653 printed on it, and when viewed through the SSD utils software, you can see that it's a Micron 232L (232-layer) 3D TLC die, and the XPG MARS 980 BLADE 2TB comes with a 5-year warranty and a 1480 TB Tera Bytes Written per Day (TBW), using 1.86 TB = 1904 GB. Tera Bytes Written (TBW), DWPD_Drive Writes Per Day (DWPD_Drive Writes Per Day) is converted to 1.86 TB=1904 GB, which means that you would have to write 810.959 GB to the SSD every single day for 5 years before exceeding the nominal TBW value.

The ∆ 2TB capacity has 1480 TB of Tera Bytes Written (TBW) available with a five-year warranty.

∆ The XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 2TB SSD that I have in my hands uses Micron Micron 232L (232-layer) 3D TLC chips.

 

XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD Performance Tests

Testing Platform Usage AMD Ryzen 9 9950X3D The XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD 2TB is installed in the first M.2 slot of the motherboard, which provides full PCIE Gen5 x4 bandwidth from the processor's direct connection channel for performance testing. Also, please note that test performance may vary depending on SSD firmware version, system hardware configuration, and other factors, so these results are for reference only.

Testing Platform

Processor:AMD Ryzen 9 9950X3D
Cooler: ASUS Ryuo IV SLC 360 ARGB (Full Speed)
Motherboard: CROSSHAIR X870E EXTREME ( BIOS Version: 1504 )
Memory:XPG LANCER BLADE DDR5 6000 MT/s 2x 16 GB CL 28-36-36-96 1.4V
Graphics: ASUS Prime Radeon RX 9060 XT 16GB GDDR6 OC
Operating System: Windows 11 Professional 24H2
Graphics driver: AMD 25.10.09.01

 

The first step is to use CrystalDiskInfo software to view the basic information of the M.2 SSD, which uses PCIe 5.0 x4 transfer mode and NVM Express 2.0 standard, and supports features such as S.M.A.R.T. (Self-Monitoring, Analysis, and Reporting Technology), TRIM, and VolatileWriteCache, and the firmware version tested by the author is Y0218A. The firmware version tested by the author is Y0218A.

∆ formatted and converted to different bit system units can use 1.86 TB.

∆ CrystalDiskInfo software view information.

 

Use with empty disk CrystalDiskMark The software was used to test the read/write performance of the M.2 SSD. The test file size was set to 1GiB in NVMe SSD configuration mode, and the sequential read/write speeds measured in the default profile were 14404 MB/s and 12836 MB/s. The software was used to test the performance of the M.2 SSD.

The Q8T1 test result in the first column represents a thread with eight queue depths, which means that there are eight 1MB-sized accesses in the worklist waiting to be accessed. This corresponds to a real situation, such as reading or writing eight 1MB files from or to a hard disk at the same time, which is generally less likely to be done in general.

The random read/write performance of the RND4K Q32T16 item is 4097 MB/s and 3493 MB/s.

The RND4K Q32T16 operates with 16 queue depths out of 32 threads, and the file types are random access 4 KB size files written or read into the SSD.

∆ CrystalDiskMark Settings:NVMe SSD / Profile:Default, Sequential Read/Write and Random Read/Write test scores.

 

For daily use or for gamers, the QD1 to QD4 range is a good reference. If we switch the CrystalDiskMark profile to real-world performance, the first column will be Q1T1. Although the result will be a bit lower than the Q8T1, it will be more in line with the actual performance of the daily use experience.

The reason for this is that most of the daily operating systems use the Q1T1, which is a queue-depth mode with one thread, so the Q1T1 is more suitable for daily use than the Q8T1. The read and write speeds of the Q1T1 are 6100 MB/s and 10640 MB/s with the test file size set to 1GiB in the Q1T1 mode.

∆ CrystalDiskMark Settings:NVMe SSD / Profile:Real World Mode 1GiB profile test results.

 

follow ATTO Disk Benchmark To test the continuous read/write performance and stability of different file sizes, the maximum values are 12.53 GB/s for writing and 13.88 GB/s for reading.

∆ ATTO Disk Benchmark.

 

Professional video multimedia file write and read performance testing using the AJA System Test To simulate the scenario test, the test settings were set at 5120×2700 5K RED format for 64GB and 1GB files, and the Codec Type was 10bit YUV by default, the final results were 11146 MB/s for writing and 10171 MB/s for reading in 1GB mode, and 10761 MB/s for writing and 9242 MB/s for reading in 64GB mode. The final results were 11146 MB/s written and 10171 MB/s read in 1GB mode and 10761 MB/s written and 9242 MB/s read in 64GB mode.

∆ AJA System Test 1GB.

∆ AJA System Test 64GB.

 

Next, through TxBENCH The test was conducted in the default test file 512MB mode and achieved 11780 MB/s for Read and 14127 MB/s for Write.

∆ TxBENCH Default test file 512MB mode results.

 

3DMark Storage BenchmarkingTested with game startup loading, copying game files, game archiving, installing games, and OBS game recording, the scenario mainly uses a number of games for actual testing, so that gamers can clearly refer to the performance of the hard disk in the use of the game, the above test process is recorded in units of time, but the final results are calculated using the bandwidth and the average access time, and the higher the final storage benchmark test score, the better.

∆ 3DMark Storage Benchmark Test.

 

3DMark DirectStorage Feature TestIt can help gamers understand the performance enhancement of PC game loading with Microsoft's DirectStorage technology, a technology introduced by Microsoft for Windows PCs equipped with PCIe SSDs, which reduces the amount of resources consumed when loading game data.

DirectStorage can bring more benefits through BypassIO, which reduces the CPU utilization of the game by minimizing the CPU load when transferring data. When DirectStorage is used in conjunction with GDeflate data compression technology, it can further reduce the game loading time, and GDeflate eventually decompresses game data to VRAM through the GPU, resulting in more transfer bandwidth for better game data transfer efficiency. GDeflate ultimately decompresses game data to VRAM through the GPU, bringing more bandwidth for better game data transfer efficiency.

At the end of the test, three scores are generated. The first is the performance difference between DirectStorage on and off, which is expressed as a percentage. Next is the average bandwidth of game data transfer from SSD to GPU memory (VRAM) with DirectStorage enabled, and finally the average bandwidth of game data transfer from SSD to GPU memory (VRAM) with DirectStorage disabled. The Disabled item score also represents the average bandwidth of standard Windows APIs in transferring game data. The deactivation score also represents the average bandwidth of the standard Windows API for transferring game data.

∆ 3DMark DirectStorage feature test on and off scores improved.

The ∆ 3DMark DirectStorage feature tests the average bandwidth comparison between on and off.

 

Finally, AIDA64 was used to conduct the Read Test Suite, Linear Read, and Linear Write tests. The Read Test Suite mainly uses Linear Read, Random Read, and Buffered Read to test the read performance of the hard disk through these three read methods.

∆ AIDA64 Read Test Suite reads the performance combination test.

 

AIDA64's Linear Read and Linear Write full disk read/write tests will read/write the hard disk's MAX capacity large files for users to observe its read/write stability. In the Linear Read test, the SSD mostly read around 6399 MB/s on average until the end of the test, with a significant drop in read speed at one point, but recovered immediately.

In the Linear Write test, the write speed started to drop when it reached about 18 %, with a minimum of about 389 MB/s, and the overall write speed averaged at 3165 MB/s.

∆ AIDA64 Linear Read.

∆ AIDA64 Linear Write.

 

XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD Temperature Testing

The next part of the test was the cooling performance, where the platform was mounted on a bare bracket, so the actual cooling effect would be a bit better than in a closed chassis. During the test, all the slots were set to run at full speed in the motherboard's BIOS, and the test scenario was a sealed room at 24 °C. This test is for reference only, as it's difficult to control the ambient temperature in a normal room.

M.2 SSDs are used with the thin heatsink included in the box, and are not accompanied by a motherboard heatsink or an additional fan to help dissipate the heat.

The test software still uses CrystalDiskMark Settings:NVMe SSD / Profile:Default, but manually adjust the number of times to 9; test file size setting to 64 GiB, some foreign media will set their own scripts to stress test, but the author personally think that Taiwanese gamers are more commonly used. CrystalDiskMark The M.2 SSD software is used to test your own M.2 SSDs, so it would be better to continue using the same software to reproduce the test on your own, in the same way as in the previous M.2 SSDs.Pressurize with All-in-One Water Cooling! What kind of cooler do I need for M.2 PCIE Gen5 SSDs?The article is the same, so you can check it out.

Temperature logging is done by using HWinfo64 software to record the maximum temperature of the hard disk, and manually adjusting the polling cycle in the HWinfo64 software to record the temperature of the M.2 SSD itself in a more real-time manner by adjusting the following settings. Why don't we use CrystalDiskInfo software to record the temperature? Because when there are multiple temperature sensors on an M.2 SSD, CrystalDiskInfo only displays the first sensor in order; sometimes the drive manufacturer will adjust the sensor temperature order so that the master chip is not always first, resulting in viewing temperatures from other locations. Furthermore, CrystalDiskInfo is very slow in updating the temperature sensors, and the difference between the displayed temperature and the actual temperature can be as much as 5~10 °C due to the lack of real-time updating of the temperature sensors.

• Global:20 ms
• Disk SMART every 1 cycle
• Embedded Controller every 1 cycle

∆ CrystalDiskMark Settings:NVMe SSD / Profile:Default, but manually adjusted the number of tests to 9; the test file size setting was changed to 64 GiB, and the maximum temperature is 98 °C.

 

Conclusion

We tested AVerMedia XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 SSD with 2TB capacity, and ran the score test with its thin heatsink throughout the whole process, and only with the thin heatsink can reach the level of results in CrystalDiskMark software through the test, which is partly visible SMI SM2508 Solution has a lower heat output than the Phison PS5026-E26-52.

However, in longer-term stress tests, we can still see that the cooling performance of this thin heatsink is still limited, for example, in the AIDA64 linear write and read performance test, and the CrystalDiskMark temperature test planned by the author, there will be high temperatures or "may be due to" overheating resulting in less than expected performance, it is not true that the thin heatsink can't be used in emergencies, but if you often have a long time to read and write data, the author still recommends using it together with the heavy heatsink included with the motherboard. However, if you are going to be reading and writing data for a long period of time, I would recommend using the thicker heatsink that comes with the motherboard, but the thinner heatsink can be left in place for now.

In the actual test, the continuous read/write speeds measured in CrystalDiskMark are 14404 MB/s and 12836 MB/s, which is in line with the specification level indicated on the official website, and the after-sales part provides a five-year warranty as well as 1480TB of TBW (Tera Bytes Written), and the amount of data written per day (DWPD_ Drive Writes Per Day) is converted to 1.86 TB=1904 GB, which means that 810.959 GB would have to be written to the SSD every day for five years to exceed the standard. Drive Writes Per Day (DWPD_) uses 1.86 TB=1904 GB to convert, which means that you have to write 810.959 GB to the SSD every day for five years to exceed the standard TBW value, which is basically impossible to exceed within the scope of normal use.

 

At the time of the author's writing, the price of each model of 14000 MB/s Gen5 2TB SSD in the scalper's price increase house is ranked, XPG MARS 980 BLADE is the cheapest model, in addition to being a big section cheaper than the Qunlian E26 solution, it is also cheaper than other models that also use the SMI SM2508 Solution (Kingston G5, Acer GM9000, China Brand X570 PRO), Acer GM9000, China X570 PRO), want cheaper 14000 MB/s Gen5 2TB SSD then InnoGrit solution ADATA LEGEND 970 PRO, this part depends on how to choose, at least the XPG MARS 980 BLADE PCIe Gen5 x4 M.2 2280 2TB. Probably the cheapest SMI 2508 host Gen5 SSD on the market at this stage.

 

The X670E Taichi Carrara motherboard was used for additional testing, and an additional cooling test was conducted with the aluminum marble passive cooling armor on the motherboard. The SSD was tested with bare strips and the motherboard thermal pads in direct contact with the motherboard without the use of thin copper heatsinks, and the maximum temperature was 77 degrees, which would have been even lower with the M.2 Fan Sink active heatsink that is included in the motherboard kit. If used with the M.2 Fan Sink active heatsink included in the motherboard kit, the temperature should be even lower. All in all, it's just a good idea to use it directly with a PC motherboard heatsink, and if the motherboard heatsink doesn't perform as well as it should, then just buy an additional Luminous or Josper's type to use it.