Latest Samsung NVMe Driver Released

Wobitancrawfodi said:

Either way, my speeds are slower than those of other people's tests here. The drives aren't very old, even the ones on M2_2 and M2_3. Quite disappointing.

reinstalled Windows with v3.2 drivers. Still Slow.

Contacted Samsung Technical support, they fobbed me off with "Different tools give different results. Background activity will affect results". My system is very lean, with 56 processes total. Terrible response from Samsung.

I'm going to start testing other brands and potentially dump Samsung. The v3.3 drivers made no difference to the poor performance of the drives, either better or worse.

CountMike said:

Samsung Magician can test other drives too.

I know. As can Seagate's Seatools. There are lots of tools around. And Samsung's statement might even be reasonable, but I used the same tool as others, and compared my results with theirs. I am skeptical as to whether Magician is above board. They want their drives to look good. Perhaps Magician gives more "favorable" results than other tools for a reason. Any way, the driver versions don't appear to be the cause of the problems, and I think this could start to get off topic now.

Mod: can you please move this to a more appropriate sub forum.

Yes I do have 3 connected. But these speeds are typical across all PCs. Of the 10 PCs, 7 have only one M.2 drive and the other 3 have 3 each.

The CPU has 48 lanes. (10940X), and of course the motherboard supports 48 lane CPUs

GPU + M.2 + M.2 = 16 + 4 + 4 = 24 from CPU

1 x M.2 on chip set = 4. Rest of 20 chip set lanes assigned to motherboard features

With only 1 M.2 drive installed I get the same results, whether it be chip set or CPU. It also doesn't matter what port I try (M2_1, M2_2 or M2_3). I Also tried Pro vs Evo on each of the three ports. Same Deal.

I also tried placing one then two drives in to the MSI Expander card that comes with the motherboard (not the boot drive - I haven;t tested if I can boot from that). That means that all M.2 drives were connected to the CPU and none via the chip set, so 28 CPU PCIe lanes used.

I also have a USB 3.1 Gen 2 2 port card which uses 4 lanes, but I took that out while doing tests. But even with that connected, I'm still only at 32 lanes.

What irked me about Samsung's response was that they didn't ask a single question. They don't know if any system has more than one drive. They didn't ask how old the drives were, what their TBW was or how full they are. SSDs definitely slow as they fill, but I have no SSD more than 30% full.

They don't know if they are connected to CPU or chip set and they didn't ask me what tests I'd done, or ask me to do any tests. They just went with the fob off. And that came from "Samsung HQ in the states" according to Australian Level 2 tech support.

I'm going to grab a Corsair MP600 PCIe4 and possibly a Gigabyte Aorus drive as well as see how they fare in comparison.

You've done your homework then, well done.

Wobitancrawfodi said:

I'm going to grab a Corsair MP600 PCIe4 and possibly a Gigabyte Aorus drive as well as see how they fare in comparison.

I would do actually opposite: find a greenfield system and check these NVMe SSDs - your result are below the par across the board and to me it indicates a limitation or a bottleneck of some sort in your system.
Also, do you have this chappy installed in one of the PCIe slots? Aquantia 10G NIC

Did you try removing it?

I installed Samsung NVMe Driver v3.3.0.2003 instead of using Microsoft Driver v10.0.19041.1.
Is this OK? By the way. It changed my Disk order around.

Wobitancrawfodi said:

What do you mean by "Greenfield system"?

I have a test PC which I restored to a raw Windows base install. No Apps only motherboard driver, graphics and NVMe drivers. Same results. Then did the same on my PC. The results were even worse than before.

Uninstalling the AQN107 drivers did not make any difference.

But is this card present in all PCs where/when you run these tests (regardless of its drivers being installed, or not)?