Intel SSD 750 PCIe SSD Review: NVMe for the Client
by Kristian Vättö on April 2, 2015 12:00 PM ESTPerformance Consistency
We've been looking at performance consistency since the Intel SSD DC S3700 review in late 2012 and it has become one of the cornerstones of our SSD reviews. Back in the days many SSD vendors were only focusing on high peak performance, which unfortunately came at the cost of sustained performance. In other words, the drives would push high IOPS in certain synthetic scenarios to provide nice marketing numbers, but as soon as you pushed the drive for more than a few minutes you could easily run into hiccups caused by poor performance consistency.
Once we started exploring IO consistency, nearly all SSD manufacturers made a move to improve consistency and for the 2015 suite, I haven't made any significant changes to the methodology we use to test IO consistency. The biggest change is the move from VDBench to Iometer 1.1.0 as the benchmarking software and I've also extended the test from 2000 seconds to a full hour to ensure that all drives hit steady-state during the test.
For better readability, I now provide bar graphs with the first one being an average IOPS of the last 400 seconds and the second graph displaying the standard deviation during the same period. Average IOPS provides a quick look into overall performance, but it can easily hide bad consistency, so looking at standard deviation is necessary for a complete look into consistency.
I'm still providing the same scatter graphs too, of course. However, I decided to dump the logarithmic graphs and go linear-only since logarithmic graphs aren't as accurate and can be hard to interpret for those who aren't familiar with them. I provide two graphs: one that includes the whole duration of the test and another that focuses on the last 400 seconds of the test to get a better scope into steady-state performance.
Given the higher over-provisioning and an enterprise-oriented controller, it's no surprise that the SSD 750 has excellent steady-state random write performance.
The consistency is also very good, although the SSD 750 can't beat the 850 Pro if just focusing on consistency. When considering that the SSD 750 provides nearly three times the performance, it's clear that the SSD 750 is better out of the two.
 |
|||||||||
| Default | |||||||||
At the initial cliff the performance drops to around 15K IOPS, but it quickly rises and seems to even out at about 22-23K IOPS. It actually takes nearly an hour for the SSD 750 to reach steady-state, which isn't uncommon for such a large drive but it's still notable.
I couldn't run tests with added over-provisioning because NVMe drives don't support the usual ATA commands that I use to limit the LBA of the drive. There is similar command set for NVMe as well, but I'm still trying to figure out how to use them as there's isn't too much public info about NVMe tools.
 |
|||||||||
| Default | |||||||||
Facebook
Twitter
RSS
132 Comments
View All Comments
Kristian Vättö - Friday, April 3, 2015 - link
As I explained in the article, I see no point in testing such high queue depths in a client-oriented review because the portion of such IOs is marginal. We are talking about a fraction of a percent, so while it would show big numbers it has no relevance to the end-user.voicequal - Saturday, April 4, 2015 - link
Since you feel strongly enough to levy a personal attack, could you also explain why you think QD128 is important? Anandtech's storage benchmarks are likely a much better indication of user experience unless you have a very specific workload in mind.d2mw - Friday, April 3, 2015 - link
Guys why are you cutpasting the same old specs table and formulaic article? For a review of the first consumer NVMe I'm sorely disappointed you didn't touch on latency metrics: one of the most important improvements with the NVMe busKristian Vättö - Friday, April 3, 2015 - link
There are several latency graphs in the article and I also suggest that you read the following article to better understand what latency and other storage metrics actually mean (hint: latency isn't really different from IOPS and throughput).http://www.anandtech.com/show/8319/samsung-ssd-845...
Per Hansson - Friday, April 3, 2015 - link
Hi Kristian, what evidence do you have that the firmware in the SSD 750 is any different from that found in the DC P3600 / P3700?According to leaked reports released before they have the same firmware: http://www.tweaktown.com/news/43331/new-consumer-i...
And if you read the Intel changelog you see in firmware 8DV10130: "Drive sub-4KB sequential write performance may be below 1MB/sec"
http://downloadmirror.intel.com/23931/eng/Intel_SS...
Which was exactly what you found in the original review of the P3700:
http://www.anandtech.com/show/8147/the-intel-ssd-d...
http://www.anandtech.com/bench/product/1239
Care to retest with the new firmware?
I suspect you will get identical performance.
Per Hansson - Saturday, April 4, 2015 - link
I should be more clear: I mean that you retest the P3700.And obviously the performance of the 750 wont match that, as it is based of the P3500.
But I think you get what I mean anyway ;)
djsvetljo - Friday, April 3, 2015 - link
I am unclear of which connector will this use. Does it use the video card PCI-E port?I have MSI Z97 MATE board that has one PCI-E gen3 x16 and one PCI-E gen2 x 4. Will I be able to use it and will I be limited somehow?
DanNeely - Friday, April 3, 2015 - link
if you use the 2.0 x4 slot your maximum throughput will top out at 2gb/sec. For client workloads this probably won't matter much since only some server workloads can hit situations where the drive can exceed that rate.djsvetljo - Friday, April 3, 2015 - link
So it uses the GPU express port although the card pins are visually shorter ?eSyr - Friday, April 3, 2015 - link
> although in real world the maximum bandwidth is about 3.2GB/s due to PCIe inefficiencyWhat does this phrase mean? If you're referring to 8b10b encoding, this is plainly false, since PCIe gen 3 utilized 128b130b coding. If you're referring to the overheds related to TLP and DLLP headers, this is depends on device's and PCIe RC's maximum transaction size. But, even with (minimal) 128 byte limit it would be 3.36 GB/s. In fact, modern PCIe RCs support much larger TLPs, thus eliminating header-related overheads.