Performance Consistency

Performance consistency tells us a lot about the architecture of these SSDs and how they handle internal defragmentation. The reason we don’t have consistent IO latency with SSD is because inevitably all controllers have to do some amount of defragmentation or garbage collection in order to continue operating at high speeds. When and how an SSD decides to run its defrag or cleanup routines directly impacts the user experience as inconsistent performance results in application slowdowns.

To test IO consistency, we fill a secure erased SSD with sequential data to ensure that all user accessible LBAs have data associated with them. Next we kick off a 4KB random write workload across all LBAs at a queue depth of 32 using incompressible data. The test is run for just over half an hour and we record instantaneous IOPS every second.

We are also testing drives with added over-provisioning by limiting the LBA range. This gives us a look into the drive’s behavior with varying levels of empty space, which is frankly a more realistic approach for client workloads.

Each of the three graphs has its own purpose. The first one is of the whole duration of the test in log scale. The second and third one zoom into the beginning of steady-state operation (t=1400s) but on different scales: the second one uses log scale for easy comparison whereas the third one uses linear scale for better visualization of differences between drives. Click the buttons below each graph to switch the source data.

For more detailed description of the test and why performance consistency matters, read our original Intel SSD DC S3700 article.

  Intel SSD 730 480GB Intel DC S3500 480GB Intel SSD 530 240GB SanDisk Extreme II 480GB Seagate 600 480GB
Default
25% Spare Area - -

Thanks to the enterprise DNA in the SSD 730, IO consistency is outstanding. We are looking at S3500 level consistency here, which isn't surprising given the similarity between the two. The faster controller and NAND interface mainly help with peak performance but IO consistency is built deep into the architecture of the drive. The only drive that can really challenge the SSD 730 is OCZ's Vector 150 while even the SanDisk Extreme II falls short once it reaches steady-state. Also of note is that Increasing the OP yields a healthy boost in performance and the SSD 730 actually manages more IOPS than the S3700 even though it has slightly less OP (25% vs 28%).

  Intel SSD 730 480GB Intel DC S3500 480GB Intel SSD 530 240GB SanDisk Extreme II 480GB Seagate 600 480GB
Default
25% Spare Area - -

Here you can see the differences a bit better with the linear scale. The SSD 730 manages around 15K IOPS compared to a slighly lower 10K IOPS on the SanDisk Extreme II. With the increased overprovisioning, the SSD 730 is in a class of its own, maintaining a minimum 30K IOPS.

  Intel SSD 730 480GB Intel DC S3500 480GB Intel SSD 530 240GB SanDisk Extreme II 480GB Seagate 600 480GB
Default
25% Spare Area - -

TRIM Validation

To test TRIM, I filled the drive with incompressible sequential data and proceeded with 120 minutes of incompressible 4KB random writes at queue depth of 32. I measured performance with Iometer after issuing a single TRIM pass to the drive.

Intel SSD 730 Resiliency - Iometer Sequential Write
  Clean After TRIM
Intel SSD 530 240GB 351.3MB/s 402.9MB/s

TRIM definitely works as performance is actually higher than after a secure erase.

Introduction, The Drive & The Test AnandTech Storage Bench 2013
Comments Locked

96 Comments

View All Comments

  • Namey - Friday, February 28, 2014 - link

    Agreeded
  • Death666Angel - Friday, February 28, 2014 - link

    Hm, so the Oxford Dictionary recognized word "monies" sounds uneducated to you? And it seems that documentated and adaptated are perfectly fine and acceptable words as well. Just because you don't use them in every day language or they sound weird does not mean they are not correct. There are many different types of spoken and written English language. And making fun of one of the (correct) ways to write that language does seem, what's the word, uneducated. :P
  • Mipmap - Thursday, February 27, 2014 - link

    "build a drive significantly that was better than the other offerings."

    build a drive that was significantly better than the other offerings.
  • JarredWalton - Thursday, February 27, 2014 - link

    Man, I seriously flubbed that sentence/paragraph on the editing pass. I think I tried to copy/paste to rearrange words but somehow missed. LOL
  • Guspaz - Thursday, February 27, 2014 - link

    The endurance figures must be wrong. They're quoting that the drive supports only 267 cycles before it dies (128TB / 480GB = ~267), which is absurdly low. This is supposed to be high endurance MLC, but even Samsung's TLC has multiple times higher endurance than that...
  • KAlmquist - Thursday, February 27, 2014 - link

    Perhaps Intel is assuming 3000 cycles and a worst case write amplification of 12. (If you aren't familiar with write amplification, see Anand's early SSD articles on this site.)
  • Guspaz - Thursday, February 27, 2014 - link

    Intel previously claimed a write amplification of 1.1 way back in the G1 days. Are you saying that this drive has an order of magnitude worse write amplification than their oldest drives?
  • futrtrubl - Friday, February 28, 2014 - link

    Remember that they are using enterprise metrics and not client metrics on this. It could be the you can be more flexible on the WA you can assume on client drives.
  • Guspaz - Friday, February 28, 2014 - link

    Perhaps, but don't you think that the order of magnitude reduction in claimed write endurance deserves a call out in the article, if only to explain why that's the case?
  • KAlmquist - Friday, February 28, 2014 - link

    The 1.1 write amplification was presumably typical rather than worst case. I wouldn't expect the write amplification to be that much different from the G1. The large block and page sizes probably make write amplification a bit worse, but the algorithms may have improved a bit.

    I agree with your comment below that it would have been nice if the reviewer had queried Intel about the endurance spec, rather than leaving us to speculate.

Log in

Don't have an account? Sign up now