Western Digital Readies WD Black SN850 Firmware Update Restoring AMD X570 Performance

[btarunr]

Western Digital is reportedly preparing a firmware update for its WD Black SN850 M.2 NVMe SSD that restores the drive's write performance levels on PCs based on the AMD X570 platform. This problem is localized to X570, specifically to when the drive is installed on an M.2 NVMe slot that is wired to the X570 chipset. Drives that are installed on the slot that's directly wired to the Ryzen processor perform as expected (Ryzen 3000 "Matisse" and Ryzen 5000 "Vermeer").

The drive performs as intended on AMD B550, as well as Intel platforms that support PCIe Gen 4, as the only Gen 4-capable M.2 slots are the ones directly wired to the processor. Western Digital localized the problem to certain X570 motherboards that have their PCIe maximum payload size (MPS) value set at 128 bytes. This dictates the maximum transaction layer packet (TLP) that goes through the PCIe controller, and a low MPS value cripples performance. The firmware update by Western Digital possibly works around this limitation. The company is expected to release the firmware update by 12 July, 2021.



View at TechPowerUp Main Site

 

[VulkanBros]

Grrr.....

 

[Guwapo77]

Bring on the speed thank you very much!

 

[londiste]

Western Digital localized the problem to certain X570 motherboards that have their PCIe maximum payload size (MPS) value set at 128 bytes. This dictates the maximum transaction layer packet (TLP) that goes through the PCIe controller, and a low MPS value cripples performance. The firmware update by Western Digital possibly works around this limitation.

I would really love to see a good technical description of what is going on there. Low MPS value affecting - especially maximum sequential - performance makes perfect sense but why are other SSDs not affected by the same thing and what exactly is the workaround?

 

[TheLostSwede]

I would really love to see a good technical description of what is going on there. Low MPS value affecting - especially maximum sequential - performance makes perfect sense but why are other SSDs not affected by the same thing and what exactly is the workaround?

Yeah, this is rather bizarre and I have to own up that I was wrong in the thread where someone was having this issue, as I really thought it was OS related, but apparently it was something entirely different.

 

[Tomorrow]

This does not appear to affect Samsung. I have 3 PM9A1's (980 Pro OEM version) connected to all 3 M.2 slots on my X570 Aorus Master. No difference between using the top or bottom two slots.

 

[Maxx]

I would really love to see a good technical description of what is going on there. Low MPS value affecting - especially maximum sequential - performance makes perfect sense but why are other SSDs not affected by the same thing and what exactly is the workaround?

I actually did notice other drives having issues way back in late 2019 (but not nearly as drastically as this) and documented it on Reddit (google: sm2262en x570). I was largely ignored and it was chalked up to "just being over the PCH" (which isn't true as I had many drives that operated just fine). With the SN850, though, it's much more significant. So I guess "stay tuned."

 

[Foobario]

As I understand it, using the chipset slot instead of the slot connected to the CPU always results in about a 10% degradation in performance but with this WD card the loss in performance was upwards of 40%.

Glad to hear that Western Digital is finally addressing the issue. Even though they are spinning it as something wrong with certain X570 boards when all the other SSD manufacturers somehow made their cards work just fine from day one.

 

[monkeyboy46800]

WD have been aware of this for quite a while but kept there mouths shut on the subject even after pages of posts here, reddit and on WD Forum.

@TheLostSwede next time don't be so insistent about a subject that you have no idea about, we could have brought attention to this a lot sooner if you and others had not been so insistent that we were newbies and had no idea how to set up are operating systems to correctly use a NVME drive.

 

[born2rock4life]

No update yet with less than 48 hours on their provided timeline. We'll find out soon if it was just trying to help fend off losing sales due to it becoming a more widely-known issue given it's their latest flagship NVMe SSD to launch, or if they actually have a means to resolve the issue. Given their silence of the issue other than the initial news, I hope they're diligently working on the firmware fix