Okay for testing I took most of our tests directly from our normal SSD benchmarking suite. But how do I really get a look at the performance of StoreMI? Well, I decided to benchmark both the hard drive and the SSD independently before getting into any other testing to get a look at their normal performance. Then from there I set up the StoreMI drive and benchmarked that. Then lastly I tested a fourth time with the 2GB RAMCache turned on as well.
Setting everything up wasn’t that difficult. You have to download the program from AMD to start off. Once that is downloaded and installed you can run it and it will walk you through everything. You can do it automatically where it decides which drive is slow and fast or you can do a manual setup where you pick the fast and slow.
Once set up, because AMD sent a 256GB drive and StoreMI is limited to 128GB it actually sets up two drives. One is the Tiered drive with the 2TB capacity combined with the 128GB and the other is the remainder of our SSD allowing normal use of that.
Remember once you set this up you still need to go into Computer Management and activate the drives and then also format them so you can use them.
Once setup you can spot the DRAM cache option. This gave me trouble turning it on. At one point it actually was boosting that secondary non-tier half of the SSD. I ended up having to delete everything and start the setup over. It seemed to work better when I didn’t manually set everything up.
With all of the setup out of the way, I can finally get into the overall results. I started with Passmark’s Performance Test 9 where I ran the full Disk Mark for an overall score. It's nothing like benchmarking a spinning disk, even though this is a relatively fast one, against RAM Cache to put speeds into perspective. Running just the StoreMI configuration actually showed an improvement over the SSD itself and as you can see adding the cache on top of that really kicked things into high gear as well.
For my next test, I used AS SSD’s copy benchmark. This is a timed benchmark so the lower the score the better the result. It copies files that are similar to game usage, program usage, and an ISO for the tests. Not surprisingly the hard drive was much slower. But what might surprise you is that here the RAM Cache was the second slowest and the StoreMI Tiered setup was fastest.
Next in Anvils Storage Utilities I tested both read and write speeds in IOPS at 4k file size 16 queue depth. Here the RAM Cache dominated on the read side of things but for write speeds, the cache dropped the ball. It defaulted back to the regular StoreMI and SSD performance. As you can see StoreMI did, however, tower above the hard drive speeds both on write and reads.
Next, I went with Crystal Disk Mark for the sequential read and write speed results. These results looked a lot like the IOPS results only here the RAM Cache was even faster on read speeds with a crazy 4440.7 MB/s read speed. The hard drive doesn’t look as slow with these numbers, in fact, 150+ on both read and write would have been amazing to have 5-7 years ago but even the limited speeds of the SATA based M.2 drive offer a big improvement when paired up together for StoreMI. Notice the SSD and StoreMI results are basically the exact same, as are the write speeds on the cached version.
For my last set of tests, I wanted to look at how all of the drives responded as you increase the queue depth. On the read results it's not a surprise that the Cache was well above everything else, but it is interesting to see that it actually slows down after a queue depth of 64. The StoreMI configuration was a little slower than the SSD independently, but it was also more consistent, not really improving from 16 on up. The SSD results on the write speeds are so close to the StoreMI results that the line is basically hidden behind it, the same goes for QD8 and above with the RAM Cache as well. The RAM Cache was actually slower on the writes in the early tests and just even on the rest shows that you only see an improvement on the reads, but man it’s a big improvement.