Performance

Before jumping into testing I always like to take a look at CrystalDiskInfo to document the firmware revision that I am testing with and to confirm that I am testing in the correct transfer mode. In this case that is PCIe 3.0 at x4 and everything is good to go.

image 6

My first tests are all in Cristal Disk Mark and I tested with both the older CDM 6 and the new version 7. For CDM 6 I focused on the sequential read and write performance and the P2 had a read speed of 2498 which was a nice jump over the original P1. The write speeds were a little lower at 1840 but again higher than the P1, especially if you compare the 500GB version of the P1 which did very badly in overall writes.

graph1
graph2

While in Crystal Disk Mark I also took a look at the Random IOPS performance at a queue depth of 8 and with 8 threads on both read and writes. I stacked the two together to get an overall idea of the drive's performance Which doesn’t hold up when compared with the fastest drives but was faster than the older WD Blue but slower than the new WD Blue SN550. The Lexar NM610 is also a good comparison and another budget-friendly drive and the P2 was way out ahead of it, especially on the write IOPS.

graph3

As for the newer Crystal Disk Mark 7, the P2’s read performance was a little lower here than in the previous version but still solidly ahead of the Lexar and not far behind the WD Blue SN550. It was even closer to the SN550 on the writes and again faster than the Lexar.

cdm7read
cdm7write

Now PCMark 10 is a great test because it includes a lot of real-world usage examples into its test. Sadly the P2 fell off here on the full system drive benchmark but it did do well on the data drive benchmark where it came in ahead of the SN550.

graph4

For AS SSD I skipped the standard tests which hit the same points that Crystal Disk Mark tests and I went for their timed file transfer test that tests the transfer time for three different file types. I stacked them all together because overall performance is important unless you only plan on using the drive for one specific type of file. In this case for the ISO transfer, the P2 didn’t do so well but it did make up a lot with the program files and wasn’t bad on the game files as well. Compared to the original P1 the P2 looks even better, especially when in the combined results it outperformed the SN550 and the NM610 but also the MP600 which is a monster of a drive.

graph5

In Passmark Performance Test 9 I ran a few different tests. I did their advanced disk test with all four of the default tests which look at a few different enterprise workloads and I also tested using the standard Performance Test with the Disk Mark which combines multiple tests to create a score. For the advanced tests, the P2 did well on the Web Server and File Server tests being three times as fast as the original P1 and similar compared to the NM610. Its workstation performance is good, not at the top of the class but right in line with everything else. The same could be said for the database results as well. As for the disk mark score, it was significantly faster than the SN500 but way behind the SN550 and a big jump up from the P1. It is right in between all of the higher-end drives and last year's budget drives. A happy medium and about what you would expect with a new budget-focused drive.

graph6
graph7

I went back to checking out IOPS performance. This time using Anvils Storage Utilities with a Random IOPS at a queue depth of 16. The P2 did well here, really well with 346390 in write IOPS which was way out ahead of the SN550 and the NM610 with just the high end drives ahead of that. Then for the read IOPS, it was right with the SN550 and WAY ahead of the original P1.

graph8

Sticking with Anvil’s Storage Utilities I wanted to see how the drive would handle different queue depths so I ran through the same test with both read and writes while ramping the queue depth up to see how the Phison E13 controller would handle things. For the reads, it ramped up well and only stopped gaining when I reached the queue depth of 128. It wasn’t the highest in the charts at any point during that but the end result at those higher queue depths was above average with just the five higher-end drives up at the top. With the write test, it was a little different. Specifically, the results were making sense until it hit a queue depth of 16 and it spiked up then dropped back down. I was completely puzzled about these and spent a LOT of time testing and retesting thinking it was a cache or thermal issue. I’ll talk a little more about this below.

graph9
graph10

So like I said that queue depth of 16 had me puzzled so I wanted to show some of my results. Normally if a cache issue is popping up you will run a test once and then for the second time it will speed up and stay fast after that because it is cached. The four tests below are all in order. The first one was 1340, the second was 746, the third was 720. At this point, I thought well clearly it is throttling down maybe from heat. But then the fourth test was back up to 1366 MB/s.

image 1
image 2
image 3
image 4

I still always take a look at thermals and for this, I run AIDA64 read stress test and take a thermal image. In AIDA64 I am looking for any dips or spikes and boy did I get those. The test ran fine for 15 minutes then the results spikes up to 2752 and then back down again over and over then it went away and went back to the lower speed. Was our drive possessed? I’m not completely sure, but I can say that it didn’t get too hot during the testing. The controller was the hottest area of course which reached 119F but the rest of the drive was running at 97F. Both aren’t enough to be a concern.

image 5
image 7

 

Log in to comment

We have 888 guests and one member online

supportus