Technology content trusted in North America and globally since 1999
7,701 Reviews & Articles | 57,110 News Posts

Intel Optane in RAID 0 - World's Fastest System Disk

By: Jon Coulter | m.2 SSDs in Storage | Posted: Jun 28, 2017 12:50 am
TweakTown Rating: 100%Manufacturer: Intel

Moderate Workload Model

 

We categorize these tests as indicative of a moderate workload environment.

 

PCMark Vantage - Hard Disk Tests

 

Version and / or Patch Used: 1.2.0.0

 

The reason we like PCMark Vantage is because the recorded traces are played back without system stops. What we see is the raw performance of the drive. This allows us to see a marked difference between scoring that other trace-based benchmarks do not exhibit. An example of a marked difference in scoring on the same drive would be empty vs. filled vs. steady state.

 

Typically, we run Vantage three ways. The first run is with the OS drive 75% full to simulate a lightly used OS volume filled with data to an amount we feel is common for most users. The second run is with the OS volume written into a "Steady State" utilizing SNIA's consumer guidelines. Steady state testing simulates a drive's performance similar to that of a drive that been subjected to consumer workloads for extensive amounts of time. The third run is a Vantage HDD test with the test drive attached as an empty, lightly used secondary device.

 

With Optane, it doesn't matter what state the drive is in, full, empty or steady-state it runs the same regardless. With this in mind, we are only showing performance at 75% full for comparison.

 

 

OS Volume 75% Full - Lightly Used

 

intel-optane-raid-worlds-fastest-system-disk_50

 

intel-optane-raid-worlds-fastest-system-disk_51

 

intel-optane-raid-worlds-fastest-system-disk_52

 

As we've been telling you, workload performance is primarily driven by low queue-depth random performance. This chart really puts that into perspective. Flash-based workload performance isn't even in the same universe as what we are getting from our Optane Memory arrays. Keep in mind that this is performance with data on the disk.

 

 

PCMark 7 - System Storage

 

Version and / or Patch Used: 1.4.0

 

We will look to Raw System Storage scoring for evaluation because it's done without system stops and, therefore, allows us to see significant scoring differences between drives. When testing NVMe SSDs on PCMark 7, we are looking for a minimum score of 10,000.

 

OS Volume 75% Full - Lightly Used

 

intel-optane-raid-worlds-fastest-system-disk_53

 

intel-optane-raid-worlds-fastest-system-disk_54

 

intel-optane-raid-worlds-fastest-system-disk_55

 

Again, flash offers no competition at all. Keep in mind that Optane Memory is just giving us a fraction of what 3D XPoint can deliver, and even that is overwhelmingly more than flash can do at its fullest potential. I know what you are thinking, what if I had $2600 burning a hole in my pocket and bought 2-2TB 960 Pro's and RAIDed them? Would that beat an Optane Memory array? No, it would not. That will only get you a score of 25K at best.

 

 

PCMark 8 - Storage Bandwidth

 

Version and / or Patch Used: 2.4.304

 

We use PCMark 8 Storage benchmark to test the performance of SSDs, HDDs, and hybrid drives with traces recorded from Adobe Creative Suite, Microsoft Office, and a selection of popular games. You can test the system drive or any other recognized storage device, including local external drives. Unlike synthetic storage tests, the PCMark 8 Storage benchmark highlights real-world performance differences between storage devices.

 

OS Volume 75% Full - Lightly Used

 

intel-optane-raid-worlds-fastest-system-disk_56

 

intel-optane-raid-worlds-fastest-system-disk_57

 

intel-optane-raid-worlds-fastest-system-disk_58

 

PCMark 8 is the most intensive moderate workload simulation we run. With respect to moderate consumer type workloads, this test is what we consider the best indicator of a drive's performance. For us, this really seals the deal. Flash-based SSDs are no longer in the discussion for bleeding-edge enthusiast-level storage performance. These results also serve to illustrate, once again, that in an OS environment low queue depth random read performance is king. We have to look no further than our 950 Pro array to show us why. Notice how the 950 Pro array scores LESS than a single 950 Pro? This is because when you RAID drives to create a boot volume, random read performance at QD1 actually goes down.

 

intel-optane-raid-worlds-fastest-system-disk_59

 

Again, we want to illustrate how meaningless sequential performance is in an OS environment. This is four 950 Pro 256GB SSDs in RAID 0 delivering 9GB/s of sequential read performance. This is 16-PCIe lanes direct to CPU not routed through a chipset. Now look at the storage bandwidth:

 

intel-optane-raid-worlds-fastest-system-disk_60

 

Only 883 MB/s storage bandwidth when running consumer workloads. Compare that with the 1,453 MB/s storage bandwidth we are getting from our 3-drive Optane Memory array when running consumer workloads with only 4-PCIe lanes.

 

 

BAPCo SYSmark 2014 SE System Performance

 

Version and / or Patch Used: 2.0.0.70

 

SYSmark 2014 SE is considered the gold standard for testing system performance because it is an application based benchmark. This test gives us the ultimate in real-world results because it utilizes actual applications running on the system, instead of playing back recorded traces. If you want to know what kind of impact a particular SSD will have on your system's overall performance; this test will show you.

 

intel-optane-raid-worlds-fastest-system-disk_61

 

intel-optane-raid-worlds-fastest-system-disk_62

 

intel-optane-raid-worlds-fastest-system-disk_63

 

intel-optane-raid-worlds-fastest-system-disk_64

 

Disk performance has the greatest impact on the Responsiveness Score, so that is what we will focus on.

 

intel-optane-raid-worlds-fastest-system-disk_65

 

Our system is much more powerful than the calibration system (1000-point baseline) used by BAPCo, so we ran an OCZ TL100 120GB SATA III SSD to establish a comparison point relative to our test system. We will be running this test going forward, and we will add drives to our chart as we test them. We decided to chart our results with Server 2012 as well as our results with Windows 10.

 

It is important to keep in mind that with SYSmark 2014 SE a few points are a big deal when comparing one drive to another on the same platform when running the same operating system. Additionally, real-world application performance is driven primarily by low-queue depth random read performance which is why we are seeing our 2-drive array slightly outperforming our 3-drive array. The 2-drive array has slightly better 4K random read performance at QD1.

 

The main thing we want point out is that our Optane arrays are scoring well over 100 points more than any flash-based NVMe drive/array is capable of, which is a massive difference in the real-world. Next, we would point out Server 2012 R2 delivers far better real-world disk performance than Windows 10 does.

 

This caps off our assertion that you can build the world's fastest OS disk for $240 using Intel's Optane Memory modules. BAPCo seems to agree with our findings:

 

intel-optane-raid-worlds-fastest-system-disk_66

 

TweakTown's score with our 2-drive Optane Memory array is the best BAPCo has ever seen.

    We at TweakTown openly invite the companies who provide us with review samples / who are mentioned or discussed to express their opinion of our content. If any company representative wishes to respond, we will publish the response here.

Related Tags

Got an opinion on this content? Post a comment below!
loading