Samsung Magician Benchmark - Reliable?

[yngndrw]

I've just received a new 990 Pro 4TB and went to test it against my existing 970 Pro 1TB.

I'm using Samsung Magician 8.0 with the latest firmware for both drives. Both drives are in a PCIe Gen 3 x4 slot.

I'm not expecting ground-breaking sequential throughput as I'm using PCIe Gen 3, but Samsung Magician is reporting oddly low IOPS for the 990 Pro.

I also tested using CrystalDiskMark (C: is the 970 Pro, Z: is the 990 Pro) - This shows the drives as being comparable on my system.

Is Samsung Magician incorrect or have I received a bad 990 Pro?

[Hudz]

You need to make sure the 990 is your main boot drive.

If your 970 will be your main boot drive then the 990 will only match it's speed.

[yngndrw]

Oh I didn't realise that, thank you - I'll have to get my operating system migrated and re-try it then!

[yngndrw]

Just to give an update to this, I'm now using this new 990 drive as my main OS drive - Sequential speeds are better than the 970 but the IOPS are still far lower than expected. (Half of the IOPS performance that the 970 achieved) It says power saving but that's just because I have it configured to use 10% over-provisioning.

The latest benchmark is below:

[PhoenixTank]

Responding to the other commenter, not OP - I thought this forum had a threaded view.

This is complete nonsense. I have a PCIE3 drive as my boot drive, and now a separate 990 Pro 4TB. The 990 doubles the performance of the boot drive with PCIE4. I have ensured that the 990 is connected to the CPU & not the chipset but I struggle to give you the benefit of the doubt that this is what you meant.

[PhoenixTank]

Hi,
Appreciate that this is a month after your post but I was searching online. Hopefully you've resolved it but your figures do not look right to me.

This is the same model as yours in a PCIE4 x4 slot but I can't see why your IOPS wouldn't be in the region of half of mine since you're running at gen 3.

You didn't mention a motherboard platform but on the consumer/mainstream gear only the first m.2 slot is connected to CPU directly. Possible that'd dip IOPS if you installed to your second slot that is connected to the chipset instead... but that is a really big dip. I'd investigate further if it were me and bring it up with support.

Best of luck!

[yngndrw]

Yes the numbers are far below what I expected as well. At the very least, I'd have expected to see the same as my old 970.

My motherboard is an Asus Strix Z390-F Gaming - It has two slots but I actually swapped the old 970 into the other slot when I installed the 990 (The top slot doesn't have a cooler, so I put the 990 into the one with the cooler) and I didn't notice a drop in performance on the 970. Sadly the manual doesn't give any details on their topology. I could certainly try swapping them over though.

[PhoenixTank]

Thanks for listing your board. The top slot will have the lanes directly to the CPU.

Didn't check the manual but the tech spec page for the motherboard gives just enough info, and the lower & physically larger slot will be using the chipset here.

1 x M.2 Socket 3, with M key, type 2242/2260/2280 storage devices support (both SATA & PCIE mode)*¹

Intel^® Z390 Chipset :

1 x M.2 Socket 3, with M key, type 2242/2260/2280/22110 storage devices support (PCIE 3.0 x 4 mode)

I know that PCIE4 drives do tend to run hotter and benefit from a heatsink much more than Gen 3. Whether you'd be at any risk of thermally throttling a PCIE4 drive at PCIE3 speeds... I'm unsure. Worth a go to swap & test if you have ample time but I'd still be opening some form of ticket with support. These are not cheap drives.

The only other datapoint I can offer is that my 1TB PCIE3 drive gets double the IOPS you do for your new drive in Magician: 295K Rand Read, 227K Rand Write. Not dissimilar from your 970 really.

[yngndrw]

Ah I completely missed that line when I read over it. I wouldn't mind but I read it about three times! I'll give it a go and see if that changes anything and if not I'll raise a support ticket. Will report back!

I appreciate the help.

[yngndrw]

Well I swapped them over and now I'm even more confused. When I booted up Windows the machine fest faster, noticeably faster. Considering that I already had a fast NVME drive, a noticeable difference is significant.

That said, the benchmarks are all over the place. They vary a lot, sometimes dropping by an order of magnitude between runs. I tried a few different tools, all of them were variable but I think that they generally show the drives to be performing about the same.

You can see the difference is sequential speed, but the IOPS are coming out very weird in the benchmarks. Either way it feels faster as I say so I'm happy with that - I think it's just a benchmark issue now.