Performance

So we know the Trion 100 has a controller that is very similar to the S10 but beyond that we don’t know too much about it. We also know that for NAND they went with new Toshiba TLC NAND. As mentioned earlier TLC is similar to MLC but is able to fit 3 bits per cell due to having third state/level. So in order to put the new NAND and controller to the test I ran the Trion 100 through our standard benchmark suite. This consists of a range of tests that touch on most situations to help us get a better feel for how the drive will perform. To start off the testing I jumped right in with CrystalDiskMark to see sequential write and read speeds. For read speeds the Trion 100 performed extremely well pulling 548.8 MB/s putting it nearly at the top of our charts. The write speeds were a little slower with a few of the Phison S10 drives out performing but the 518.1 MB/s write speeds of the Trion 100 were still very good.

graph1

graph2

Next I tested the Trion 100 in AS SSD, to avoid too much test overlap I use AS SSD’s file copy benchmark that uses three different file types. The results are the timed copy results with all three combined to help average out the performance differences the different file types. The end result is the Trion being faster than all of the older drives but struggling against the higher end OCZ drives as well as the S10 based drives with MLC.

graph3

The next set of tests are similar to the previous test in that I test using four different settings. Only this time each setting simulates a different usage scenario in a business environment using Passmarks advanced storage benchmark. The results are the average transfer speed so the higher the result the better. For the graph I combine all four together to get a peek at what drives might be the best overall drive no matter the situation. So how does the Trion 100 perform? Well It performed better than the MLC based S10 drives in the File Server test helping it pull ahead in the overall numbers. That said the higher end OCZ drives are still faster overall with the exception of the Vector 180 in the Web Server test.

graph4

The next benchmark I run is in my opinion the best representation of a real world benchmark that you will find. This is the PCMark 8 Storage Benchmark. I avoid using the scores that PC Mark likes to give out because they don’t really show the difference between drives. I use the overall average bandwidth result that shows the average transfer speed of the drive over the entire course of the benchmarks. This includes the gaps between each test so the numbers look a little slower than what you might expect the drives to do over the full 1-3 hour test, but they do compare very well to each other. So how did the Trion 100 do? Well sadly it fell at the complete bottom of the chart with both the MLC S10 drives and higher end OCZ drives out performing by a big gap.

graph5

What about total IOPS? Well for IOPS I test using Anvils Storage Benchmark with a 4k file size and a Queue Depth of 16 on both the read and write tests. Again I combine the results in the graph because I prefer to look at the overall drives performance, great read speeds aren’t anything unless the write performance is good for example. Here The Trion 100 performed very well on the write speeds but fell on its face on the write speeds. I think this is because the benchmark wasn’t able to take advantage of the SLC Caching on the write test, but I will get into that in the next test.

graph6

For my last test I stuck with Anvils Storage Benchmark but this time I focused on how well the drive would handle various queue depths across both read and write performance. To do this I doubled the queue depth for each test starting at 1 and moving up to 64 on write testing and 128 on read testing. On the read testing the Trion 100 performed well accelerating faster than the other drives as the queue depths went up until a queue depth of 32 where the OCZ Vector 180 pulled ahead and stayed ahead for the rest of the tests. That said the Trion 100 performed really well here, being the second fastest drive tested in the high queue depths and the fastest drive in the lower queue depths.

graph7

For the write performance on the queue depth testing things got a little more interesting. Specifically, I ran into an issue on my first round of testing that I wanted to highlight. The second graph below was my first tests and I was surprised to find that the Trion 100 zig zagged its way through the chart. After looking into it it was determined that as I was doing the tests I was actually testing every other test while the drive was clearing out the SLC Cache. Basically the time between my benchmarks was enough to prompt the drive to clear the catch but not enough time for it to finish. So every other benchmark was actually showing the true write speed of the Trion 100 without its SLC Cache. I was able to confirm this by running the same test over and over and even with the exact same queue depth setting I would see a good performance every other time. I took a little additional time between each benchmark and was able to get our proper graph (the top one) to show that the Trion 100 when using its SLC Cache is very quick on the write speed tests. Of course this is only testing the cache really so in real world testing you will sometimes run into times where the write speed is a little slower as well. That is why I wanted to include both results as I think pretending the retested result is what you will see every time would be a little disingenuous.

graph8

SLC

 

Log in to comment

garfi3ld's Avatar
garfi3ld replied the topic: #37315 28 Oct 2015 17:34
Happy hump day everyone, today I take a look at the OCZ Trion 100 SSD

We have 1908 guests and no members online

supportus