The Intel Optane SSD 900P Review
As a Solid State Drive buyer in 2018 it is hard to not take a lot for granted. Over the past decade this burgeoning industry has gone from niche products that were known for being (at best) ‘temperamental’ to mainstream darlings that are now the de rigueur choice for even entry level systems. Every inch of this amazing progress has been dotted with what trail-blazers like to call ‘challenges’ but us less forward-thinking mortals like to call problems. Intel’s new 900P series was created to solve many of today’s most glaring storage challenges but it may also create a few of its own.
The problems I’m referring to are countless in number and date back to the dawn of the SSD age. There was a lack of storage capacity, short lifespans, even performance bottlenecking, all of which started out as issues that had to be solved for the market to survive its infancy. Most of these were indeed solved, and sometimes seemingly overnight. When this happens, it creates a paradigm shift in the market with the first model offering the answer becoming the new gold standard upon which all following drives are judged.
The Intel X25 series, OCZ Vertex or even Samsung 850 Pro are all examples of these disruptive forces at play, and each one is an example of good engineers solving hard problems. The $389 to $600 (USD) Intel Optane SSD 900P series hopes to be another instant classic series that shows the rest of the industry how things should be done when it comes to performance stability and longevity.
While there are still numerous roadblocks on the path to a “perfect SSD”, the one hiccup that has proven to have a rather elusive solution is long term performance consistency. This is best described as the difference between a new solid state drive’s performance and one that has been either hammered constantly for a while or simply filled to capacity. Put another way, the lack of consistency is why a drive does indeed seem to get slower the older it is and the less free space it has.
Intel, Micron, Samsung, Toshiba, and all the other stakeholders in the Open NAND Flash Interface Working Group have known for a long time now that NAND based storage -especially newer TLC or QLC NAND- is not overly fast to begin with and only gets slower with time. To help solve this issue some companies took the route of simply including a write acceleration buffer, where a portion of the NAND acts like SLC NAND. However, once this buffer is exhausted performance drops like the proverbial lead balloon.
Other companies are trying to use a different logic gate that is indeed faster than the more common Floating Gate design, but as we saw in the Toshiba OCZ TR200 (http://www.hardwarecanucks.com/foru…toshiba-ocz-tr200-960gb-480gb-ssd-review.html) review Charge Trap designs still have a way to go. Their performance does indeed suffer over time and there are other challenges they bring to the table as well.
With the 900P series, Intel has taken a different path and created an entirely new storage technology that promises to be faster, more robust, and just plain better than NAND storage. This new technology is of course 3D XPoint non-volatile storage and is considered by nearly anyone as a potential game changer or ‘disruptor’ of the industry.
Ever since the Intel Optane Memory and its groundbreaking 3D XPoint non-volatile storage was released several months ago enthusiasts and workstation users alike have been waiting with nearly breathless anticipation for these new technologies to trickle down into more practical models. Intel’s more recent release of the rather niche and expensive DC P4800X did little more than fan the flames of desire. With many a storage enthusiast dreaming of the day when near RAM speeds in a non-volatile package would finally come.
With the release of the Intel Optane SSD 900p series it is pretty obvious that more mainstream folk will have to wait a little while longer for this to be a reality, but this is indeed the foundation upon which future solid state drives will be built.
In order to show exactly what level of performance the $389 Optane SSD 900P 280GB can offer we have included a wide variety of solid state drives – with all the major classes covered. For the entry level we have opted for a thirty cent per Gigabyte Toshiba TR200. For more mainstream both the BX300 and MX300 (also in the 27- 30 cent range) covers this class nicely, and of course for NVMe we have included results from a 1.2TB Intel 750 (which still costs seventy cents per GB). With all these data points potential buyers of the Intel Optane SSD 900P can make an informed decision on if the blazingly fast – and it is fast – Optane SSD 900P does indeed justify its asking price.
The Cost of Disruption
On the surface of things, an all new high-performance NVMe solid state drive that uses 3D XPoint technology and the same controller as its expensive and difficult to find DC P4800X sibling seems like a match made in heaven. Who wouldn’t want near DC P4800X performance in a package that is not only tailored more to home user / gaming scenarios but also a heck of a lot cheaper? This is the winning formula that Intel used with their DC P3700 to 750 series generation and a major reason why the Intel 750 has latest so long in an industry filled with mayfly lifespans.
As we will show you later in this review the Intel Optane SSD 900-series is not really meant to replace the Intel 750 series. Rather it is meant to satisfy the needs of an entirely different group of consumers, those who live at the top of the proverbial food chain. They have a need for speed that is never satisfied and have a bank account that can support their addiction.
The model name itself should make this clear as it is the first new Intel 9 series SSD to be releases since the Intel SSD 910 series way back in 2012. Yes it has been a long, long time since Intel showed this group of buyers any love. Sadly 7-series buyers needed to wait a bit longer before their drives were replaced with the new 760p drives….but more on that in another review.
Long time coming or not, the Intel Optane SSD 900p series is going to find it difficult to compete in this ever-evolving marketplace. Unlike when the 910 was released, the Intel Optane 900p is entering a market that is filled with competitors who have their own vision for the future and a proven track record for delivering on their goals. Also, unlike when the 910 was released, the Intel brand name can no longer be relied upon to carry a model across the finish line and win any hearts and minds of consumers.
Intel may have indeed created some of the most iconic solid state drives in this industry’s short history – such as the Intel X25-m and Intel 750- but their more recent forays have been less than stellar. Models like the Intel 540S, 600P, or even Intel Optane Memory have shown that Intel is not synonymous with perfection. So our pessimistic side fears the same will hold true for the Intel Optane SSD 900P series.
My inner performance junkie on the other hand sees promises of RAM-like performance and salivates at the mere idea of obtaining such performance levels from a single drive. Meanwhile my inner conservative sees durability ratings in the Petabyte range, a five-year warranty, both U.2 and Add In Card form-factor specifications, and really wants to give in to sheer temptation of not waiting for the eventual 7 series refresh. We are sure many enthusiasts will be battling with their inner selves for these exact same reasons.
Making things a tad more difficult to give into temptation is the MSRP and capacity options Intel is making available at launch. At this time the Optane 900p comes in two flavors: an anorexically small 280GB capacity version and an overly svelte 480GB version. Even compared against the five-year-old Intel 910 series these capacities are small.
More concerning still, in the intervening years storage requirements have ballooned for both the enthusiast and workstation user alike. Games now routinely weigh in at 50GB or more. 4K UHD, 6K ‘RED’ UHD, and 8K UHD video easily consume 40GB to hundreds of Gigabytes of storage space per hour of video recording. Even a modern Operating System like Windows 10 can gobble up anywhere from 5 to 10 percent of the entire capacity of a Optane SSD 900P. That certainly will get many enthusiasts to pause for thought before taking the plunge into the world the new Intel 9-series Optane SSD has created.
Arguably the smaller capacities are certainly not the end of the world. These are uber-high performance models meant to dominate the competition in those all important – to some – performance metrics. Unfortunately, as mentioned before the competition is now highly competent. Even excluding unreleased models and technologies like Samsung’s ZNAND and Micron QuantX based devices, models like the Samsung 960 Pro offer not only excellent performance and larger storage capacity but also a much better cost per gigabyte. Intel will even be competing against itself as the Intel 750 has proven to a lifespan rarely seen in this corner of the market and costs about half as much per Gigabyte as the Optane SSD 900P series.
Put simply, with asking prices of $389 and $599 – or $1.25 to $1.39 per Gigabyte – the Optane 9-series is not cheap. It is so far away from inexpensive territory that it shockingly costs twice the price per Gigabyte of the Samsung 960 Pro series – and the Samsung 960 Pro is not an affordable drive to be begin with.
Put another way the Intel Optane 900p has to not only be fast but a lot faster than the competition to justify both its asking price and smaller capacity. That is indeed a hard task the Intel engineers have set out to accomplish but they do have a better than average track record of seeing possibilities where others see problems.
A Closer Look at the Optane 900P
Unlike the majority of NVMe drives recently released the Optane SSD 900P bucks the less than optimal trend of using the small M.2 form-factor. Instead it relies upon either a half-height, half-length Add In Card ‘HHHL PCIe board’ or U.2 (aka ‘12.5mm high 2.5-inch drive’) design. It is my firm opinion that Intel made the right call here by sticking to their tried and true form factors instead of following the heard.
M.2 cards are indeed easy to use, and can be added to laptops, but their smaller footprint does come at a rather large cost, a cost that Intel has neatly sidestepped. This major concern is of course heat, or more precisely thermal limits. All drives -even solid state- create waste heat and the M.2 size is just too small to efficiently transfer this heat to the air. Instead larger M.2 drives can and do throttle when stressed unless they have a heatsink added. When this happens the drive’s performance tanks and life span of the storage cells is decreased.
The relatively massive heatsink of the Optane SSD 900P series may not be actively cooled but is more than up to the task of keeping storage IC’s temperatures in check. As a nice bonus the increased mass of this heatsinks allowed Intel to use an ultra-high performance NVMe controller that was originally designed for the enterprise market. This is the same controller as the one used in the server grade Intel DC P4800X – albeit with different firmware tuned for better low queue depth performance.
As an aside Intel did not just recycle the DC P3700 or Intel 750’s heatsink. Instead this new all black heatsink boasts much improved cooling fins and an all-black exterior. This combined with the plastic top fascia, does noticeably impact cooling performance since the TDP is just 14W. It also allows it to become a centerpiece of any high end build and considering the asking price of these rather small NVMe drives that is indeed a good thing.
For those interested in using the 900P in smaller half-height cases Intel has once again included a half-height bracket. Removing and replacing the factory mounted bracket with the included adapter only requires unscrewing the two backmost screws. This will not void your warranty. Instead all screws are easily accessible and free of such annoyances.
That metal backplate on the back of the Optane 900P was the one issue with this cooling configuration that did concern us. When held up to the light there appeared to be significant air gap between the PCB and this heatspreader. Once removed our fears were quickly realized.
It appears that even though Intel did spend significant resources on the top of the PCB and its components the back side received precious little attention. Instead, this heatspreader is for protection and aesthetic purposes only as the seven 3D XPoint chips are not connected to it with heat pads like the fourteen top side ICs.
I’m unsure why Intel did not feel the need to cool these seven (suspected) phase-change ICs as they do create heat and will get warmer than the ones on the top side. Equally puzzling is most of the even more expensive 480GB’s additional storage capacity will all be located on here – as you can see there are an additional seven empty IC mounting points. This is a significant misstep on any device that costs as much as this model does. I was truly shocked that a company with so much experience would have made such an obvious mistake on their consumer grade flagship model.
Counterbalancing this initial impression is 3D XPoint is a most likely a phase-change based storage medium that does run a touch cooler than NAND based storage. The IC size is larger than the typical 3D NAND IC (206.5 mm2 vs 168.2 mm2 IMFTs 3D NAND) and has more surface area for passive cooling. As such, Intel could simply have not felt additional cooling was needed on the back. Instead the main heatsink on the top of the card is there to cool the high performance controller while cooling of the ICs on that side is simply a fringe benefit.
The reason for this also could be that the 900P series does not make use of high capacity ICs. Instead they are low density chips which do not run as hot even under sustained use. In the 280GB’s case all 21 ICs contain two layers of 3DXPoint memory. In the case of the 480GB versions there is a combination of dual and tri-layer ICs used.
With all that said, thermal pads cost at most pennies each, and this cost cutting on such an expensive drive is nearly intolerable. Required or not Intel should have spent the couple cents to ensure proper cooling of all the 3D XPoint chips. This is not an inexpensive drive and corner cutting like this does not belong on such a costly series.
On first glance the lack of onboard capacitors that graced previous enterprise and high-end consumer Intel solid state drives also smacks of cost cutting where none should exist. After all, those capacitors allow for enterprise grade ‘flush in-flight’ data loss protection. But the reality is they are just not needed. 3D XPoint storage is not NAND storage so it does not suffer from this critical issue. Instead any state in which the 3D XPoint cells are in when power is lost is the state that they will be in when power is restored. All that is needed is aggressive table mapping so that the data is not corrupted upon the next power cycle. Intel does indeed include this feature and as such this ‘issue’ is in fact a nice upgrade.
For further details on why this is the case we recommend reading our article on 3D XPoint and its benefits here: (http://www.hardwarecanucks.com/foru…/74989-intels-optane-dc-p4800x-deep-dive.html). For those who only want a brief overview, 3D XPoint is incredibly fast, can be accessed at a much lower level, does not need to be erased before a cell can be used again, is more durable, more data corruption resistant, and simply better than any NAND created to date.
As it is brand new technology it sadly is also bloody expensive. The cost of producing this new storage technology plays a major role in why this drive costs twice the price of its more contemporary counterparts. 3D XPoint simply does not have efficiency of scale working in its favor like traditional NAND does. This was proven true as a 3D XPoint based replacement for the Intel 750 did occur once Intel actually had the technology up to proper production levels. However, the release of the 900P proves Intel is making great strides in this area so the wait for even more drives should not be too long.
The one issue that is mildly concerning –at least on first glance- is Intel has not included an external RAM buffer for the controller. As history has shown RAM-less controllers are simply not as optimal a solution compared to those that do have a RAM buffer.
The reason Intel does not feel the need to include one is because this is not a NAND-based drive. 3D XPoint supports direct read and incredibly low level write access that is simply impossible with existing NAND technology. This fine grain control in conjunction with latency levels that simply aren’t possible with existing NAND technology is why a buffer is not really needed. Instead the over-provisioning portion of the drive can act as a large buffer for those times when the IO requests saturate the incredibly high abilities of this new storage technology.
This may sound familiar to some enthusiasts as SandForce controller-based designs made similar claims that never really worked as intended outside laboratory conditions. However, this time it actually does make sense and is not just marketing department dreams.
Remember this new technology has latency that is nearly in RAM territory and can be re-written to without first being erased. As such the performance penalty from not using a RAM buffer is minor at best. It may in fact be lower than the multiple controller cycles required to first write to the RAM, then read it from the RAM, then write to the non-volatile storage. Compare this with writing to any free cells and updating the internal mapping table on which cells are being used for storage and which are actually part of the over-provisioning section.
It is indeed a brave new world where most pre-conceptions about non-volatile memory have to be thrown out. For those interested in such things the actual capacity of this drive is 336GB (or 21 ICs with 16GB of capacity each) with the extra 56GB being set aside as over-provisioning. Of course, as this is a ‘280GB’ drive the OS reported capacity is just a hair over 260.6GiB of useable space. Hopefully one day both software and hardware governing bodies will agree on a single unified standard for measurement.
The 3D XPoint used in the Optane SSD 900P is also why enthusiasts do not need to purchase the largest capacity in order to get the highest performance possible. This storage medium is so fast that interleaving (the number storage cell levels per controller channel) really is of little concern beyond deep queue depth scenarios. It is only here that the controller will routinely call on the over-provisioning section to act as a buffer, and it is only here that the difference in layers per channel will create a potential performance difference between the 900P’s 280GB and 480GB versions. Even then it is more a case of having access to more idle storage cells than layers on the controller.
This is also why the new Intel controller is a seven-channel design instead of eighteen like the first generation controller used in the Intel 750 series. If such a scenario will be a common occurrence neither the 280 nor 480GB capacity version will be an optimal solution. Instead the only real option is the DC P4800X series which offers even larger capacity options.
With that being said, the offshoot issue of no RAM buffer is the write speed is limited to that of 3D XPoint itself. In a NAND-based drive the controller can push write requests into the RAM and allow for peak write performance well in excess of what the NAND is capable of. This is not possible with the Intel Optane SSD 900P, but would have been if Intel has included a RAM buffer.
Instead the 900P’s maximum speed is the true speed of existing 3D XPoint storage technology. This is why the sequential write specifications of the 900P do not seem all that impressive when compared to say a Samsung 960 Pro. However, these specifications are more realistic as the 900P can sustain this level of performance well beyond what any RAM buffer could provide in a NAND based drive. As such the marketing optics may not be overly impressive but the real world performance should more than make up for this shortcoming.
Test System & Testing Methodology
Properly testing a modern Solid State Drive to fully understand its abilities is not a simple undertaking. It takes time and it takes experience as relying upon tried and true applications is no longer good enough. Modern solid state drives come with a whole arsenal of tricks to ensure that the end-user never sees the true capabilities of a drive long enough to form a negative opinion. They have gotten so good at coming up with workarounds that minimize any underlying issues that even less experienced reviewers can be fooled.
This certainly is a laudable goal as at the end of the day a SSD is not meant for reviewers it is meant for users. As such anything that can make the overall experience a more positive one has to be considered a good thing. It does however make it difficult to make an informed decision a drive is never truly pushed past its performance envelope – as only then can you the potential buyer know if a given model is right for you.
This new testing methodology is the distillation of a decade’s worth of Solid State Drive reviewing. In these years we have seen all the tricks, all the workarounds and have spent a lot of time and effort on creating an improved methodology that is designed to strip away them all. Only then can we show you our readers exactly what a drive is made of. To do this we have blended in new with the old. Long term readers will notice that many of our tests are similar to the way we used to do things, but even here things have changed greatly. The size, the scope, and even the underlying methodology has been improved.
In the past we, like other review sites, would test a drive when empty of all other data. This is unrealistic and while we did do some limited partial and full drive performance it was based on an unrealistically optimistic scenario. As such from now on all solid state drives will be tested only when they are first filled to 50% capacity. The only exceptions are testing applications that require an empty drive to work. For example, HD Tune requires not only an empty drive but a drive that is also unpartitioned in order to run. These are now the exceptions not the rule.
Long term readers will also notice a few new additions to our testing suite. These custom tests are worst case scenarios that we have come up with that are still in the realm of possibility – as all tests are focused in on showing overall performance in as realistic a manner as possible.
For all of the benchmarks, appropriate lengths are taken to ensure an equal comparison through methodical setup, installation, and testing. The following outlines our testing methodology setup:
A) Windows is installed using a full format.
B) Chipset drivers and accessory hardware drivers (audio, network, GPU) are installed.
C) To ensure consistent results, a few tweaks are applied to Windows 10 Pro and the NVIDIA control panel:
• UAC – Disabled
• Indexing – Disabled
• Superfetch – Disabled
• System Protection/Restore – Disabled
• Problem & Error Reporting – Disabled
• Remote Desktop/Assistance – Disabled
• Windows Security Center Alerts – Disabled
• Windows Defender – Disabled
• Screensaver – Disabled
• Power Plan – High Performance
• V-Sync – Off
D) All available Windows updates are then installed.
E) All programs are installed and then updated, followed by a defragment.
F) All networking is disabled so as to eliminate this variability in overhead
G) Benchmarks are each run four to ten times, and unless otherwise stated the results are then averaged.
The full system specs are as follows:
Case: Lian-Li PC-T70W
Motherboard Chipset: Intel X299
CPU: Intel 7940x
RAM: DDR4-3200 16-16-16-18
OS: 64-Bit Windows 10 RS2 Pro
OS Drive: 1x 1TB Corsair MX300 SSD
Graphics card: EVGA GeForce GTX 1070 SC Gaming
Power Supply: Seasonic Focus Gold 850FX
For this benchmark, HDTune was used. It shows the potential read speed which you are likely to experience with these hard drives. While this application will provide numerous results the most important number is the Average Speed number. This number will tell you what to expect from a given drive in normal, day to day operations.
For this benchmark HD Tune Pro was used. To run the write benchmark on a drive, you must first remove all partitions from that drive and then and only then will it allow you to run this test. Unlike some other benchmarking utilities the HD Tune Pro writes across the full area of the drive, thus it easily shows any weakness a drive may have.
Due to the unique nature of 3D XPoint and Optane technology, buyers expecting to see PCIe x4 saturating read or write speeds will be disappointed. Even when compared to the last generation’s Intel 750 SSD the differences are not all that great in sequential r/w benchmarks.
This is because Optane technology is all about performance consistency and not short-term benchmarking metrics. By that same token, a RAM buffer could have indeed alleviated some of this and potentially made this new series even faster.
Overall though it really does not matter as even at a ‘mere’ two gigabytes per second this 280GB drive will only be able to sustain its performance for a bit over two minutes before running out of capacity. Needless to say, the 900P is not going to be an optimal solution if you need to move mega amounts of data from point A to point B – as it just does not have the capacity to do so. This is actually a larger issue than only hitting 2GB/s sequential speeds.
ATTO Disk Benchmark
The ATTO disk benchmark tests the drives read and write speeds using gradually larger size files. For these tests, the ATTO program was set to run from its smallest to largest value (.5KB to 8192KB) and the total length was set to 256MB with a queue depth left at its default of 4. The test program then spits out an extrapolated performance figure in megabytes per second. Of all the results there are four that we consider the most important. 0.5KB, 2KB, 4KB, and 8192KB. The first three show how a given drive can handle those critical small files, while the largest shows what the drive can do under optimal scenarios.
In ATTO testing owners of an existing Intel 750 1.2TB drive can expect to see at best a 50% to 100% performance boost by purchasing the new Intel Optane SSD 900P 280GB. That is indeed a nice boost from one generation to the next, and we would love to see that from Intel’s CPU division, but don’t forget there is a significant decrease in capacity and expense for mostly double-digit improvements.
This is going to make justifying the 900Ps price a touch more difficult than the promises Intel made about Optane before it was released. Keep this in mind as this is going to be a reoccurring theme throughout this review.
Crystal DiskMark is designed to quickly test the performance of your drives. Currently, the program allows to measure sequential and random read/write speeds; and allows you to set the number of tests iterations to run. We left the number of tests at 5 and size at 100MB.
While there are numerous suites of tests that make up PCMark 7, only one is pertinent: the Storage 2.0 test. The Storage 2.0 consists of numerous tests that try and replicate real world drive usage. Everything from how long a simulated virus scan takes to complete, to MS Vista start up time to game load time is tested in these core tests; however, we do not consider this anything other than just another suite of synthetic tests. For this reason, while each test is scored individually we have opted to include only the overall score.
Now that we are turning our attention to slightly more realistic synthetic testing the performance difference between the last generation and the new generation solid state drives does start to come into stark contrast. As you can see the critical 4K single queue depth performance results of this new series is bloody fantastic – and we have not even gotten to the amazingly low latency it also offers. In other words, this is a pricey drive but it does have the performance to back up this price tag. It just does not have the capacity.
AS-SSD is designed to quickly test the performance of your drives. Currently, the program allows to measure sequential and small 4K read/write speeds as well as 4K file speed at a queue depth of 6. While its primary goal is to accurately test Solid State Drives, it does equally well on all storage mediums it just takes longer to run each test as each test reads or writes 1GB of data.
Anvil Storage Utilities Pro
Much like AS-SSD, Anvil Pro was created to quickly and easily – yet accurately – test your drives. While it is still in the Beta stages it is a versatile and powerful little program. Currently it can test numerous read / write scenarios but two in particular stand out for us: 4K queue depth of 4 and 4K queue depth of 16. A queue depth of four along with 4K sectors can be equated to what most users will experience in an OS scenario while 16 depth will be encountered only by power users and the like. We have also included the 4k queue depth 1 results to help put these two other numbers in their proper perspective. All settings were left in their default states and the test size was set to 1GB.
Once again there is a rather significant difference in performance potential from one Intel generation to the next. The Optane SSD 900P really is setting the bar awfully damn high for future solid state drives and what it will take to be considered ‘fast’. If Intel only had also raised the bar on what it meant to be ‘big’ and we would be truly impressed.
IOMeter Latency Torture Test
In a perfect world the response time of a storage device should be as close as instantaneous as possible. This of course is impossible, instead any delay that is under 0.100 of a second (100miliseconds) is considered the gold standard of storage responsiveness. This is because 100ms is generally considered the smallest perceptible interval of time humans can perceive. Anything above this will result in the occasional perceptible ‘stutter’. However, a single solitary 200ms pause is better than a significant cluster of 150ms pauses. As such any and all results must be considered in their totality and not just based on a single data point. This is why we have included four charts instead of just two. The first two charts represent the total results of an IOMeter 10 minute read test and a 10 minute write test. The last two just show the average read/write results as well as the maximum read/write response rate that occurred during these tests.
To obtain these results we configured IOMeter to use a 10 second ramp up followed by a 10 minute run for each test using the entire drive’s capacity. We also configured IOMeter to record the results in one second increments (the smallest time slice allowable). The first test was using 4K aligned data chunks that were 100% random, 100% write only using its Full Random pattern. The second used 4K aligned data chunks that were 100% random, 100% read. This is done to show how the controller handles emergency housecleaning even when inundated with read I/O request.
In this test we are not focusing in on steady-state results or other Enterprise orientated determining factors. We are simply looking for overall latency under what can be considered a realistic worst-case scenario for home users via a method that can still reliably strip away the various protection mechanism the controller has in its arsenal to keep up appearances. This is not the bad old days where ‘SSD Stutter’ is still truly a thing. Instead this test is designed to solely highlight how good or bad a controller and NAND combination really is.
All tests were run four times and the most common result was used.
These results are why any enthusiast should even be thinking about the Optane SSD 900P – as there is fast and then there are speeds that would give a RAM drive a run for its money. Not only is the latency incredibly low -but not as low as expected- the sheer consistency of this new device is amazing. Remember the Intel 750 is an eighteen channel, huge RAM cache backed solid state drive and yet the Intel Optane SSD 900P simply blows the doors off it.
By that same token the combination of a lack of capacity and high asking price is keeping the Optane SSD 900P from being the revolution that was promised. Remember, it is not like the Intel 750 1.2TB capacity model is slow, it is just slow in comparison to what this new technology offers. Keep reading to see how this improvement translates to more real-world type scenarios and if this new level of performance is worth the capacity trade-off.
Firefox Portable Offline Performance
Firefox is notorious for being slow on loading tabs in offline mode once the number of pages to be opened grows larger than a dozen or so. We can think of fewer worse case scenarios than having 120 tabs set to reload in offline mode upon Firefox startup, but this is exactly what we have done here.
By having 120 pages open in Firefox portable, setting Firefox to reload the last session upon next session start and then setting it to offline mode, we are able to easily recreate a worst case scenario. Since we are using Firefox portable all files are easily positioned in one location, making it simple to repeat the test as necessary. In order to ensure repetition, before touching the Firefox portable files, we have backed them up into a .rar file and only extracted a copy of it to the test device.
Data Transfer Torture Test
New to our testbed suite is a simultaneous read and write test using real world data. Unlike almost all other tests in our arsenal this is a test that literally pits the controller and NAND against itself. The faster the controller reads data from the NAND the more pressure it puts on itself to write to the NAND – and vice versa. This is truly a no win scenario for the controller. Rather it has to find the optimal balance between read and writes in real-time, while also juggling house-keeping and other behind the scene tasks that allow the controller to be able to write to ready to use NAND.
By doing this we not only strip away all cache boosting performance, as well as short term performance boosting algorithms, we also see exactly how good the firmware, the controller, and even the NAND is at handling high stress environments. Further helping to show what a controller & NAND’s true abilities are we have opted for 60GB single file for the large file test and 20GB for the small file test. This way even the largest pseudo-SLC buffer will be unable to mask any underlying weakness.
On the surface the idea of the average home user running into a scenario that requires simultaneous read and write performance seems minimal at best. The reality however is it is a very common occurrence. Most PC users do not have multiple solid state drives and instead rely upon a single storage device to handle all their needs. As such when they download a steam game and then install it, this is the type of scenario they will run into. Albeit to a more limited extent.
In order to allow for consistency from run to run we have chosen RichCopy to carry out this arduous task. Also, in order to replicate as close as possible a home user environment we have limited RichCopy to a single thread / queue depth.
Once again this drive is bloody fast, but how often is the difference between the Optane SSD 900P and Intel 750 going to be noticed, and when noticed is the difference going to be enough – and often enough – to justify the 900P’s asking price? For most we doubt the differences will be large enough to matter. This is not criticism of the Intel 900P per se, rather it is a statement on how far manufactures have come in solving the long list of issues that NAND had in the early days.
Conclusion – Unbelievably Fast, Completely Limited
The Intel 900P SSD is easily the most difficult to judge and assess non-volatile storage device I have seen in many, many years. Make no mistake about it, this is a golden god of performance that all others will have hard time measuring up to. In a single generation Intel has gone from 2D Planar NAND to 3D XPoint and this change makes the Optane SSD 900P series the fastest, most powerful, most consistent SSD in the consumer market today. The word “impressive” doesn’t even begin to cover that feat of engineering. However, to obtain all this power buyers will have to sacrifice many, many things on this hungry god’s altar.
The first sacrifice is of course the asking price. On first blush spending $389 to $600 for a solid state drive of this caliber should not cause any experienced SSD user to even bat an eye. In this regard the Optane SSD 900P is actually an improvement. For example, the flagship capacity version of the Intel 750 still costs more than the largest Optane SSD 900P. When compared to the only drive in its performance class – the DC P4800X – this series is a veritable bargain.
For first time buyers who could never imagine why spending this much for a drive, the Optane SSD 900P is much like owning your own Leer jet. Like a luxury jet it will allow you to soar to new heights that few will ever have seen outside of multi-drive RAID configurations. More importantly it will do so with such low latency response times that it has to be experienced to be believed. So of course it is going to cost more than the competition. Performance costs. If this still does not make sense to you then this is just not the series, nor price bracket for you.
If raw pricing was the only sacrifice that was necessary, Intel would have a true cult classic on their hands. Possibly one even bigger and better than the Intel 750 it now supersedes at the top of Intel’s lineup. When properly harnessed the 900P will make even the most jaded enthusiast giggle at how absurdly fast it churns through workloads – workloads that made even an Intel DC P3700 blanche.
Enthusiast grade SSDs may not even have the word ‘value’ in their lexicon but enthusiasts do expect a certain amount of luxury to go along with all that performance. Creature comforts like having an entire Steam library on one expensive SSD or being able to store and edit more than a couple hours of video on one drive.
Instead of following this well-established pattern of using performance and capacity to make up for the asking price, Intel has significantly handicapped this new series with rather anemic capacity options. Even the largest 480GB capacity version of the 900P is relatively small when compared against Samsung, Toshiba, and a whole host of competitors offering workstation/enthusiast orientated models that come in the 1+ Terabyte range and do so for about the same overall price as the 480GB 900P.
Yes, the Optane SSD 900P certainly is going to be faster than a RD400, 960 Pro, or any other NVMe drive in the consumer market you care to mention. Yes, it certainly is also going to offer more consistent performance, and be less sensitive to overheating issues thanks to its more sensible form-factor. This is all true but it just does not have the capacity that both enthusiasts and professionals have come to expect.
This is why the real question any enthusiast considering the Optane SSD 900P is simple: are you willing to trade capacity and price for even more performance? That is not a question I can answer for anyone else. For us the performance is simply too high to ignore as two 900P’s combined with Intel’s VROC (Virtual RAID on CPU) will be pure computing nirvana for LGA 2066-based systems.
Put concisely the Intel 900P is not like owning a Leer jet. It is like owning a one-seater fighter jet. Yes, it is incredibly fast. Yes, it is capable of feats of performance consistency that no NAND based model can match right now, but at the end of the day it still is a one-seat jet without space for any luggage. The Optane SSD 900P’s lack of real world usability will interest few buyers beyond those who truly want, need or just desire the best of the best, but do not want to pay the DC P4800X’s even higher asking price.
As such, in its current form, the Optane SSD 900P is a halo product meant to show the world that Intel is still the reigning king of performance and its 3D XPoint technology is indeed ready for prime time. It however is not meant to be sold in quantities that would overstress a production line that may have reached its capacity with the Intel DC P4800X series. This is why Intel resurrected the 9-series moniker for this new drive in the first place and it is why Intel is not expecting a large demand for it. Instead it was a teaser. One meant to keep enthusiasts interested in Optane memory while they scrambled to ramp up 3D XPoint production for the 760-series.
Only you can know what sacrifices you are willing to make and only you can decide if this is the right series for you. Rest assured however that if you do decide to live with these concessions the Intel Optane SSD 900P is sure to impress. It is the new gold standard upon which all NVMe drives will be judged.