Today we will be covering enthusiast NVME SSDs and more specifically one from Kioxia. To the general consumer this may sound like a non-name brand but that could not be further from the truth. Kioxia is formerly known as Toshiba Memory who is actually the inventor of modern day flash memory and is a rather large player in the league, with estimated global market share of almost 19%. With that being said, let’s check out their latest drive, the EXCERIA PLUS G2.
The drive we have here is a 1TB variant, it also comes in a larger capacity of 2TB and a smaller 500GB. Do note that the smaller drive will be slightly slower. While it is based on the older PCI-E Gen 3 interface, I believe for most people that would be good enough.
We will compare this Kioxia drive to another PCI-E gen 3 drive from Sabrent as well as PCI-E Gen 4 drive from Samsung. Both of these drives are actually 2TB in size.
On paper when placing the specs side by side, the Samsung drive has a clear lead (and spoiler alert - it is really damn fast), but does it actually matter? We are going to cover a few benchmark results and then do real life examples to check this from different angles. First let’s analyse the difference between the drives.
When we put Sabrent and Kioxia in head-to-head battle, there is no difference in sequential reads but we see a nice bump in sequential writes. However Kioxia drive has a considerably higher performance results in random read and write speeds. And as expected - Samsung drive shows much higher results on all of those parameters.
Then we have the warranty - all three drives boast a 5 year term. So it’s safe to say they are pretty even here.
Looking at TBW, also known as Terabytes Written Endurance - we see Sabrent taking a clear lead, with Samsung and Kioxia being rather far behind. To be fair if you’re planning to use one of these as your primary boot drive then even 400 TB is plenty, however if you use it in more of a production environment - then that may not be enough. Examples of this could be cache drive for video editing or more extreme case - storage server drive.
With all of this said, the details on paper always show the best case scenario so let's check out some benchmarks to see how they really stack up.
First test is ATTO Disk Benchmark and the results are validating the speeds that Kioxia advertises, these are within a few hundred MB per second. Interestingly enough both Sabrent and Samsung deliver slightly lower speeds when compared to advertised. Sabrent comes about 600 MB short on the read but surpasses on the write speeds and Samsung is about 700 MB short on the read speed.
Next we delve deeper in Crystal Disk Mark benchmark, where we are running “real” profile with a test size of 1GB, and we find Kioxia drive is a little faster as compared to Sabrent drive but when we look at Samsung drive, we find that both of gen 3 drives are blown out of the water.
After changing the test to 8GB size we find basically the same results on Sabrent and Samsung drives but sequential write speeds on the Kioxia drive plummets down.
There is a good reason for this. Most modern TLC SSDs run in a dynamic mode where a portion of the drive is allocated for cashing. This portion of the drive works at a much higher speed and for normal day-to-day tasks can significantly improve performance, especially for small files or applications and while you are not using that data - the drive then in the background offloads it to the slower part making it ready to receive more.
In this scenario we pushed a larger load which saturated the cache and forced the drive to run at a lower speed. You might be asking - why did other drives not have this problem in this test? The answer here is actually pretty simple, our testbench here simulates a pretty standard system - it has few applications, a few games and other files. The total size is just over 450GB. As the other drives are twice the size, they have more space for the dynamic cache. This is why most SSDs slow down considerably when they get full. The real question should be - does this slow down in benchmarks also translate to real life examples?
For this we ran a few quick tests to see the real situation. We set up our test bench and cloned the operating system to ensure everything is exactly the same. The first test was booting into Windows and here Samsung drive came in at 15 seconds and both Kioxia and Sabrent took 18 seconds.
Next test is opening up a game, in this case we used Shadow of The Tomb Raider, and found Samsung in the lead but only by 0.5 to 1 seconds from the other two drives.
While in the game we also started the benchmark which loads up all the data for the level and found all three drives were within margin of error of each other.
As you can see, just because benchmarks show particular performance - it does not always translate exactly to the real world situations. In some cases an average user may not even notice the difference between one of the fastest NVME drives on the market and something that is fast enough. There will always be specific workflows where speed will make a difference, but this particular drive is aimed at a general public and here it will do just fine. Important to note - size does matter. Do try to get a drive larger than you need today to both maximise that dynamic cache and also ensure you have enough space for the files you intend to hold on it in the future.
Affiliate disclosure: as an Amazon Associate, we may earn commissions from qualifying purchases from Amazon.