MEASURING PERFORMANCE
As with all of our tests, the following tests were performed after a secure erase of the drive. The drive was also conditioned with a predefined workload until it reached steady state. We also test across the entire span of the drive. The only drive that we have reviewed that come anywhere close to price and performance is the P420m. Even though it is not NVMe, its controller is PCIe based and they have spun their own customer driver, so it should be able to hold its own again the P3700. We also threw in some 12Gbps HGST SAS SSDs. While not a direct comparison, in a form-factor sense, we include them because they are the best that the traditional SATA/SAS ecosystem has to offer.
The P3700 really shines with sequential write performance. Its single queue depth write performance is more than double the maximum rate of the Micron P420m. You might start to notice a theme in that the P3700 does really well in low queue depth write performance.
In terms of raw sequential read performance, the P3700 could quite match the P420m, which is the current gold standard.
We need a little more information in order to interpret the random 4KB write results. On the surface, the P3700 holds its own at a queue depth of 1, but quickly falls behinds as queue depths increase. Unfortunately, we are comparing different price points and different capacities. As much as we would like to test every capacity for a given product, we sometimes are limited in our samples. Because of this, we are showing the 800GB P3700 and the 1.4TB P420m. At face value, the P420m wins, but when you compare the 800GB P3700 against the 700GB P420m, or the 1.6TB P3700 against the 1.4TB P420m, the Intel wins every time. Surprisingly, the dual HGST drives, totaling 800GB, delivers the best maximum write performance, albeit at a steep price.
Thankfully, read performance doesn’t take nearly as much thought. At queue depths of 64 or fewer, the Intel delivers the best performance, in some cases by nearly 50%. Once you get to high queue depths, the P420m pulls away and pulls away fast. It eventually posts a 250K IOPS edge. You can expect the same results no matter how you compare the Intel NVMe drives and the P420m, as they all keep the same read characteristics across their capacities.
SNIA IOPS TESTING
The Storage Networking Industry Association has an entire industry accepted performance test specification for solid state storage devices. Some of the tests are complicated to perform, but they allow us to look at some important performance metrics in a standard, objective way.
SNIA’s Performance Test Specification (PTS) includes IOPS testing, but it is much more comprehensive than just running 4KB writes with IOMeter. SNIA testing is more like a marathon than a sprint. In total, there are 25 rounds of tests, each lasting 56 minutes. Each round consists of 8 different block sizes (512 bytes through 1MB) and 7 different access patterns (100% reads to 100% writes). After 25 rounds are finished (just a bit longer than 23 hours), we record the average performance of 4 rounds after we enter steady state.
- Preconditioning: 3x capacity fill with 128K sequential writes
- Each round is composed of .5K, 4K, 8K, 16K, 32K, 64K, 128K, and 1MB accesses
- Each access size is run at 100%, 95%, 65%, 50%, 35%, 5%, and 0% Read/Write Mixes, each for one minute.
- The test is composed of 25 rounds (one round takes 56 minutes, 25 rounds = 1,400 minutes)
Our SNIA tests show how well an SSD transitions between read/write workloads. The P3700 does a really good job of handling these transitions. Many times, we see more of a saddle in the graphs, where mixed workloads dip between the extremes. For the P3700, performance only increases between writes and reads, with no visible dips.
Since you say that you will test later on with consumer workloads,it would be nice to see an ANVIL and AS-SSD to see what is this device able to do compared to our normal everyday 2.5″ SSD’s
This is just a taste posted on our Forums yesterday…but yes the testing will follow our typical consumer testing…and then some…
https://www.thessdreview.com/Forums/ssd-discussion/4082-hardware-nvme-running.html#post36277
Looking great!!!! Thanx!!!!
very impressive thnx
Try and sweet talk Intel into giving you a P3500 to review.
Why do you change the meaning of colors on graphs (iops against other drives)? It’s rather annoying.
It’s looking like PCI-E SSDs may hit mainstream in a few years, especially if the price goes down.
The question is, how big is the market for this type of SSD versus say, an M.2 drive? Hopefully we’ll see more entrants come in.
Well really…. how long before we see M.2 NVMe?
Nice Review! looking forward to the consumer workloads tests
maybe one of these in my Music /workstation computer and a 3500 for the gaming machine
Is it bootable?
Sure is…with Windows 8. I have tested it on the Asus H871 Plus, ASRock Z97 Extreme 6 and Z87 Extreme 11ac. My thoughts are that it is bootable with ANY newer motherboard that allows one to switch to URFI boot rather than legacy. I will cover this in the consumer review.
Is Intel actually doing RAID 0 on this board or is RAID now obsolete for the PCIe SSDs?
Negative….straight controller…single PCB…no RAID solution…it is pretty sweet.
You guys are taking away the meat and gravy of my consumer side of the report!
So in a VMWare host, say, would you rely on Intel’s “End-to-End” data protection & backups and not worry about raid / drive failures?
Absolutely.
Sorry if I missed this in the article, but what is the USA/North America release date for this new NVMe drive from Intel? I am building a new enthusiast pc rig and have contemplated splashing out for one. Also, I am going to be using a mix of M.2, sata ssds (Samsung 850 Pros) and maybe one of these Intel P series NVMe drives. I wanted to get an LSI RAID card (12gb/s) and run RAID 5. What setting RAID setting would you recommend?
Hey, quick question: what software was used to run the SNIA tests?