Bad performance NVMe SSD Kingston OM8PDP3512B - Swift 3 SF314-43-R2LX

I just purchased an Acer Swift 3 (model: SF314-43-R2LX). It suffers from terrible NVMe SSD performance.

I got the following test results from the built-in NVMe SSD Kingston OM8PDP3512B:

It was measured whilst plugged-in with full power settings and a switched off virus scanner. All drivers have been updated with Windows Update and Acer Care center.

When I lookup the drive's benchmarks at PassMark, then it should be able to perform like this:

Disk Test Suite Average Results for KINGSTON OM8PDP3512B-A01

Sequential Read	2,265 MBytes/Sec
Sequential Write	1,028 MBytes/Sec
Random Seek Read Write (IOPS 32KQD20)	1,045 MBytes/Sec
IOPS 4KQD1	64 MBytes/Sec

As you can see my results are nowhere near those values.

The drive is Kingston OM8PDP3512B (EP-SSM280512TKS by AAEON UP) with specs: SSD M.2 SSD 2280 512GB,PCIe Gen3 x 4, M Key,3D TLC.

Who knows the SF314-43 internal M.2 port specs?

What is its PCIe version and the number of PCIe lanes available to the SSD?

Who knows how I can improve the performance of the internal NVMe SSD or do I have a faulty unit and need to return it ASAP?

[//Edited the content to add model name]

Accepted answers

All comments

StevenGen

Try to take the M.2 drive out and re-plugg it then try again, also enable TRIM within windows and see the results. The specs of CDM should not be so slow with an SSD M.2 PCIe Gen3 x 4 drive. Otherwise this NVMe SSD Kingston OM8PDP3512B is faulty (which is very unusual) or your M.2 mobo plug has some pins broken, so check all that out also.

Marty11

StevenGen When I open the chassis, then my warranty is lost.

Acer used to build nice access hatches to service memory, SSD and cooling fans in their laptops. The Swift 3 is one big clunker, that you have to dismantle completely in order to access the M.2 NVMe slot. There's nothing in the user manual about upgrading it. And the service manual isn't available for consumers. Above that all, there is a warranty void sticker over the screws.

When the NVMe SSD is not working properly. I'll have to return the laptop A.S.A.P.

Is there another way to validate the assumption that the NVMe drive is badly seated or there are damaged PCIe leads?

Marty11

I read out the NVMe SSD's SMART data:

Image: https://us.v-cdn.net/6029997/uploads/editor/x4/qnmlvaey6s5e.jpg

I can't make sense of the Tranfer Mode duality (PCIe 1.0 | PCIe 3.0)? Is there anything peculiar about this data?

Marty11

I downloaded Kingston's SSD Manager Software and ran it in order to check the SSD and if there are firmware updates available:

Image: https://us.v-cdn.net/6029997/uploads/editor/0y/1dbp4f6v7llq.png

The drive is not recognized by Kingston's own the software. I can do nothing with it.

Where can I find management software for the drive?

billsey

I believe they only support their retail line of drives with that software, not the OEM models. The SF314-43 models come with one of six different SSDs, from four different manufacturers, but they should all be NVMe x4 on a PCIe 3.0 buss. Yours is acting like it's x2 instead of x4 or something else is accessing it during your benchmark run. Are you still running the stock antivirus or have you removed that to run Defender instead? Maybe a boot to safe mode and then run the benchmark will show you better numbers? If so we can point the finger at a third party app or driver using up the bandwidth.

Marty11

@billsey Thanks for your reply.

Great to know that the interface on the laptop side is an NVMe x4 on a PCIe 3.0 bus. Where did you find that information? I wasn't able to find it.

It's still running stock anti-virus, but it was switched-off until reboot.

In the mean time also tested with a live Linux CD:

/dev/nvme0n1p5:
Timing buffered disk reads: 2158 MB in 3.00 seconds = 719.19 MB/sec

That's in line with CrytalDiskMark's measurement. I'll try the safe mode too.

kirmgrauza

Hello,

You're expecting results same as from PassMark benchmark, but you're benchmarking with CrystalDiskMark application. Applications are using different algorithms to test your hardware. Different algorithms = different results.

Check out other benchmarks results with type of SSD of yours made with CrystalDiskMark, here. They don't differ much from what I see in your screenshot.

Commodore_1995#

This issue may be related to the sata controller driver! Check if the performance is still lower by changing the sata mode to ahci:
a) first enable safeboot: press windows+r, type msconfig, go to system boot and enable safeboot!
b) then go to bios main tab, press ctrl+s and change sata mode to ahci!
c) then run crystaldismark again.

Marty11

kirmgrauza said:

Hello,

You're expecting results same as from PassMark benchmark, but you're benchmarking with CrystalDiskMark application. Applications are using different algorithms to test your hardware. Different algorithms = different results.

Check out other benchmarks results with type of SSD of yours made with CrystalDiskMark, here. They don't differ much from what I see in your screenshot.

kirmgrauza thanks for the link. There's a lot of information there. I'll need some time to make sense of it all. It's a bit confusing. Is there a way to benchmark your own machine in order to compare apples with apples over there?

Marty11

egydiocoelho said:

This issue may be related to the sata controller driver! Check if the performance is still lower by changing the sata mode to ahci:
a) first enable safeboot: press windows+r, type msconfig, go to system boot and enable safeboot!
b) then go to bios main tab, press ctrl+s and change sata mode to ahci!
c) then run crystaldismark again.

Thanks for your reply.

I did switch off safeboot in order to be able to boot Linux.

Is safeboot really necessary in order to run ahci? (As far as I know ahci was still on.)

Commodore_1995#

This model supports sata ahci and raid mode, see:

Image: https://us.v-cdn.net/6029997/uploads/editor/ac/zw7rptqkbinq.jpeg

Yes, safeboot is required first before changing sata mode to ahci so that no problems occur during windows boot.

kirmgrauza

Marty11 said:

kirmgrauza said:

Hello,

You're expecting results same as from PassMark benchmark, but you're benchmarking with CrystalDiskMark application. Applications are using different algorithms to test your hardware. Different algorithms = different results.

Check out other benchmarks results with type of SSD of yours made with CrystalDiskMark, here. They don't differ much from what I see in your screenshot.

kirmgrauza thanks for the link. There's a lot of information there. I'll need some time to make sense of it all. It's a bit confusing. Is there a way to benchmark your own machine in order to compare apples with apples over there?