SN750's poor read and write speed, when can it be solved?

I bought the ACASIS Thunderbolt 3 external hard drive enclosure for SN750 2T. I saw many people commenting that SN750 is very fast, and basically can read and write more than 2000MB/S. However, I bought SN750 2T with firmware number 112000WD. The writing speed of the hard drive box is only about 1100, and the reading speed is up to 2700MB/S, which is very embarrassing data. Through constant search, it is found that SN850 also has this problem, but it is solved by firmware update, then whether SN750 has updated firmware plan of?


Hi @Kentz,

Please refer to the link to check new firmware updates:

Thank you for your reply. My embarrassing problem is that I use macbookpro M1 MAX, I use a USB4.0 hard disk box to connect to the computer, and the virtual machine win10 or 11, the Dashboard prompts that the current disk cannot be detected, and I cannot view the SSD on my computer. I can’t detect whether the firmware has been updated or not. I can see that the firmware number is 112000WD from another computer using the USB interface. This version 750 has speed limitations, which makes the SSD unable to perform. I hope it can be resolved as soon as possible.

It seems that my brand new SN750 is even worse :slightly_frowning_face:.
Can those SSDs be returned?

The shop said that I have powered on the hard drive and have written it, so I am not allowed to return it. My firmware version is the same as yours. I don’t know if WD has a firmware upgrade plan for SN750.

So we’re stuck with an overpriced/underperforming SSD? :angry:

If you use a PCIe slot that has direct connection to the CPU and not using a signal path that runs through the chipset, then you get the full performance.

If you connect it to something that runs through the chipset, then performance takes a large hit due to the payload size bug in the current SN750 firmware.

WD seems to be refusing to make a statement on any progress towards a firmware update to fix the bug like they did on the SN850 which has the same bug.

@Razor512 so if I buy a PCIe card to « house » the SSD, I would get full performance? :thinking:

It is hard to tell, especially if it is an Apple system since they tend to be tight lipped on any block diagrams.

Thunderbolt macbooks share bandwidth with many other components, and in many cases, a device with multiple thunderbolt 3 ports can end up with a back end allocation of effectively 2 PCIe lanes per port, of which you will have overhead and further bandwidth sharing, thus while the interface is technically capable of 40gigabit/s the actual speeds will be lower. For example, many older gen Intel and AMD platforms (from the PCIe 3.0 days) will have 4 PCie lanes connecting the chipset to the CPU, and from there the chipset will offer around 16-20 PCIe lanes. They will all be listed as a fully implemented and full speed PCIe lanes, but they are still all sharing the 4 actual lanes to the CPU, this is why on systems like that, the chipset m.2 slot will run at x4 speeds, but actual read/ write speeds will be much lower than what is advertised for the SSD.

Beyond that, the enclosure also seems to specify similar limitations. https://www.amazon.com/Enclosure-External-Station-Tool-Free-Portable/dp/B08P7L6SDD

Thanks for the advice @Razor512, much appreciated.
I have an 24 lanes Asus-based PC and I could use this card:
https://www.funkykit.com/reviews/accessory/asus-hyper-m-2-x16-card-review/
I must say though that a fully functional M.2 WD SSD would be better or at least less hassle :tipping_hand_man:

A device like that will not be a good choice if you want to use your GPU. To give a more exact answer, the detained system specs would be needed. Furthermore a device like that will not offer 16 lanes to a single card, instead it will effectively split the X16 slot into 4, 4 lane m.2 slots.

Any lanes that are connected to the DMI bus will be bottlenecked by that bus regardless of the lanes that the chipset offers.

PS, if when passing devices through to a virtual machine, there is a performance hit that takes place. Even low bandwidth devices such as USB Bluetooth dongles incur a latency penalty when passed through to a VM.

Depending on the specs of the PC, if you have one where the top most m.2 slot is connected directly to the CPU, then that would be the ideal slot as that will be 4 lanes that are not shared with anything else, and an SSD in such a slot will only have the basic PCIe NVMe overhead, and various other overheads from drivers used. For example, the SN850’s reads and writes are both well in excess of the PCIe 3.0 x4 throughput, but if that drive is on a PCIe 3.0 interface, it will still top out at around 3.6GB/s.

I agree with the lane split @Razor512 but if I only have one SSD (or even 2) on that PCIe card, performance should be fine?
And I do have a four lane M.2 slot but my ■■■■■■ WD Black SSD is “faulty”. :tired_face:

If in a direct to CPU x16 slot then that 4 m.2 card can handle 4 m.2 ssds without any slowdown, but such a setup would mean losing your X16 video card.

As for the SSD, is using an m.2 slot on the motherboard, look in hwinfo64 Free Download HWiNFO Sofware | Installer & Portable for Windows, DOS
(best to get the portable version which can be had via the green download button over the white background.

While the issue primarily impacts AMD, it can also impact Intel based platforms.
Check if on your platform if the SN750 is experiencing the payload size bug. If it is, then it might be cheaper for you to get a different SSD, such as an SN850 or probably a different brand such as Samsung that has managed to avoid these issues in their firmware.

Getting another brand would have been nice but those things are not returnable as you know. I am on Intel though and ran new benchmarks using the well-known CrystalDiskMark8_0_4 and the results were quite different:
CrystalDiskMark_20220109153524_SN750_apres
Even my older SN700 is doing good once installed on the PCIe extension card:
CrystalDiskMark_20220109152849_SN700_surPCI
My video card is, indeed, no longer in an X16 slot though but the performance hit is pretty small.
Was my worry all cause by the userbenchmark.com and a more “aggressive” measuring? I don’t know…

For videocards, when on a lower speed bus, the main areas that will have a performance hit are games that make heavy use of texture and level streaming where they want a game world with no load screens. In those cases, the system memory is more heavily used and data is swapped between the VRAM and system RAM more heavily, though the performance hit varies with the card, for example if the game is already heavily bottlenecked by the compute aspect of the GPU then the bus limitation is less visible. This is why you most often see slower PCIe slots having very little impact on mid to upper mid range cards, and larger impacts as you go higher end. In those cases, the majority of games will have no noticeable performance hit.

The exception to this is an edge case. When a game needs to actively use more VRAM than the card has available. e.g., a card card with 6GB of VRAM and a game needs to actively use close to 8GB of VRAM. In any case, you will notice a large performance hit, but if you are on a slower bus, then the performance hit becomes significantly larger. as a GPU that is designed around being fed data at 250+GB/s is now bottlenecked to 16GB/s on an X16 3.0 slot, or 8GB/s or less if if on a slot with fewer lanes. In all cases, the experience is poor. For example, with a game like X4 foundations, if using a card with 4GB VRAM and max settings, the game can go from offering 60+FPS to offering 25FPS when the game needs to actively use more than the available VRAM. The PCIe bus usage will max out, and the game slows down. If on a slower slot like experienced by people who were using older SLI setups and the cards were limited, to slower bus speeds, they would end up dropping well below 20FPS with lots of hitching.

Thanks for the detailed answer @Razor512 :slightly_smiling_face:
I am not a gamer and don’t care about frame rate and stuff like that. That computer is used for photo editing (and maybe some video editing at times).
It seems like my current setup gives me top performance on SSDs and good enough video.
That’s probably fine for me then. :tipping_hand_man: