Skip to main content

WD Black SN850 SSD Loses Over 40% of Performance When Connected To X570 Chipset

WD Black SN850
(Image credit: Western Digital)

One of Western Digital's Black drives and one of the best SSDs on the market, the SN850, is significantly underperforming in writes speeds on X570 chipset motherboards when connecting the SSD to a chipset-connected M.2 slot, according to a report from ComputerBase. The company told the site it's looking into the matter.

This strange situation started out when multiple people started complaining about performance results on tech forums, but it wasn't until ComputerBase tested the SSD for themselves that the issue really became apparent.

Modern motherboards are equipped with a generous number of M.2 slots so consumers can use multiple SSDs. Due to the limited PCIe lanes on processors, not every M.2 port communicates directly with the chip. Some M.2 slots are connected to the chipset instead. In the case of AMD's X570 motherboards, it's public knowledge that there is a performance disparity between chipset-and processor-based M.2 slots. This is due to the higher latency on the former.

The difference in performance is typically less than 10% between a M.2 slot connected to the Ryzen processor and a M.2 slot wired to the X570 chipset. PCIe 3.0 SSDs are unaffected. The WD Black SN850, however, appears to suffer a major performance hit when accommodated in a M.2 slot that isn't linked to the Ryzen processor.

For some reason, this handicap only seems to be affecting this particular SSD, and on the chipset lanes. If you installed this SSD directly to the CPU-controlled M.2 slot, there would be no performance penalty.

CrystalDiskMark 8.0.1WD Black SN850 1TB via CPUWD Black SN850 1TB via X570Performance Loss
SEQ1M Q8T15,254.83,247.838.2%
SEQ1M Q1T15,2552,97243.4%
RDN4K Q32T1652.8660.4N/A
RDN4K Q1T1250217.812.9%

But when connected to the chipset M.2 slots, the performance is cut in half for write speeds. ComputerBase measured write speeds of 5,254 MBps for the SN850 when connected to the CPU-based M.2 slot.  But when switching over to the chipset-controlled slot, performance was nearly cut in half to 3,247 MBps.

ComputerBase's tests showed that the WD Black SN850 1TB lost up to 43.4% in write performance when the drive was installed on a M.2 slot hooked up to the X570 chipset. The drive achieved Western Digital's advertised speeds when inserted into a M.2 slot coupled to the processor. 

Apparently, sequential workloads were the most affected. The delta between random workloads, on the other hand, was less than 13%.

CrystalDiskMark 8.0.1WD Black SN850 1TB via CPUWD Black SN850 1TB via X570Performance Loss
SEQ1M Q8T17,067.56,304.210%
SEQ1M Q1T14,375.64,008.58.4%
RDN4K Q32T1705.8712.8N/A
RDN4K Q1T183.281.12.5%

The nature of the M.2 slot didn't have a significant impact on the WD Black SN850's read performance. There was still a performance loss, but it was only around 10%, which is within the expected margin. It would seem that only the SN850's write performance suffered a major drop.

When ComputerBase tested other SSDs, performance only dropped by 10% when using the chipset lanes. Having a minor drop like this is normal since the chipset is very far away from the CPU, and there could be latency penalties that will lower performance.

But a near 50% drop in speeds is not normal at all for the SN850. Western Digital is actively looking into the matter right now, so we should know more about this situation sooner than later. We suspect the chipset is downgrading the SN850 to Gen 3.0 support, as 3,200 MBps is a very common speed among Gen 3.0 SSDs.

  • nostrocrompt
    Ha, tfw a very random performance issue affects your rig. Well geez, what to do now? Can I swap my boot drive with my 850? I'm not wild about the idea... any idea if this is an issue that might get fixed at some point?
    Reply
  • King_V
    In the case of AMD's X570 motherboards, it's public knowledge that there is a performance disparity between chipset-and processor-based M.2 slots. This is due to the higher latency on the latter.

    Is this supposed to say "on the former," since it mentions the chipset first?
    Reply
  • littlechipsbigchips
    If The problem is about write speed then you should check out where the write test huge files are stored , it seems they share the chipset lanes om another SSD ? , that is are installed on a drive connected to the chipset as well ?

    If the benchmark does not load the file into RAM before writing it to the SSD then the Benchmark software is the problem.
    Reply
  • Sleepy_Hollowed
    I wonder if this is with auto-PCIE detection. Sometimes the x570 on some motherboards, places the entire board in PCIE 3.0 when some devices are connected.

    I've had to explicitly set it to 4.0 when those devices are present.
    Reply
  • satai
    Is this Windows specific? Or does it happen at Linux too?
    Reply
  • jpe1701
    This was an issue that popped up when x570 first came out with certain pcie 3.0 drives too. My hp ex950 has never been as fast as it was supposed to be. Some reviewers found it too but hp gave them firmware to fix the issue and then after the good reviews they never bothered to release the firmware for anyone else.
    Reply
  • CerianK
    King_V said:
    Is this supposed to say "on the former," since it mentions the chipset first?
    It appears to have been fixed.
    Reply
  • AndyTomlin
    If this is just write, I would be concerned its a SoC issue. Write is sensitive to PCIe latency, the SoC maybe doesnt have enough outstanding PCIe DMA's. If the latency is low enough it won't matter, but if SSD is connected to non CPU lanes there could be an issue.
    Reply