The Intel Optane 905P is the Titan of SSDs: incredibly fast, enormously expensive and basically the best SSD on the market, and a piece of PC hardware we just had to test out. It's our contention that a good SSD is essential at this point for your gaming PC - just in terms of the extra user-friendliness you get from Windows, plus the reduced loading times in games - but just how much do you need to invest to get a decent experience? How do mid-range and more value-orientated SSDs compare to the absolute state of the art? And where does this leave the traditional mechanical hard drive? Technology has continued to evolve there so in assessing solid-state performance for gaming, we decided to include a fast modern drive too.

To get a good overall view of the market and the kind of capabilities that solid state storage delivers across a range of price-points, we used five drives for our tests overall. The Optane 905P is staggeringly expensive at around $600 or 550 and its performance is indeed unparalleled, but we've also covered a range of more cost-effective alternatives. The Western Digital Black 500GB SSD uses the PCI Express-based NVMe standard and is about one third of the price of the 905P. Cheaper still are SATA-based drives like Crucial's well-received MX500 at around $90/90. Up against that we have a 1TB Western Digital Blue SSD. Again, that's SATA in nature but its extra capacity commands a 150/$160 price-point. Finally, representing the classic mechanical HDD, WD supplied us with the Black 6TB, priced at around 240/$240.

So, why is the 905P so expensive? Well, it's a different kind of product, more intended more for workstations or deep-pocketed enthusiasts than the average gamer. The drive's 480GB of 3D XPoint memory, developed jointly by Micron and Intel, theoretically provides a comprehensive advantage in terms of latency and endurance. For the purposes of our testing here it represents the current state-of-the-art, the standard by which all of our other test drives should be judged.

We began our test by installing a fresh copy of Windows 10 Pro to the Intel 905P, then the latest OS updates and our testing software. Next, we cloned the disk image to other drives using Macrium Reflect to ensure that each drive was set up identically. It's worth noting that our test rig includes an Intel Core i5 6600K processor, 16GB of Crucial DDR4 RAM and a Nvidia GeForce GTX 1080 graphics card, all running on a Asus Z170i Pro Gaming motherboard. This motherboard doesn't have a U.2 connector, so we're using the included U.2 to M.2 PCI-e adapter to hook up the Intel 905P. This shouldn't affect performance.

Next, we ran a synthetic benchmark, CrystalDiskMark 6.0.1, to give you a better idea of how these drives differ in sequential and random read performance. Different games read data in a different ways, but we expect the majority of read requests in games to be random reads between 4KB and 32KB in size with a low queue depth (say, eight or less). That suggests that the third column below should be the best predictor of load time performance in most games.

group_shot_tweak
CrystalDiskMark 6.0.1 (MB/s) Sequential Read (Q32T1) Random 4K Read (Q8T8) Random 4K Read (Q32T1) Random 4K Read (Q1T1)
Intel NVMe 2689 1793 607 254
WD NVMe 3441 1314 615 47
Crucial SATA 562 373 318 45
WD SATA 559 367 367 39
WD HDD 118 1.7 1.7 0.6

Note: Sequential data is stored together; random data is stored disparately. The terms in parentheses following the name of each read test show the number of queues and threads used.

As you can see, the Western Digital Black NVMe drive claimed the highest sequential read speeds, but the Intel 905P drive showed much better random read performance, particularly at lower queue depths. This should make the Intel drive the favourite in our game load tests. The two SATA SSDs followed behind, with Crucial's MX500 slightly outperforming the Western Digital Blue in most tests. Our mechanical HDD from Western Digital brought up the rear, showing the significant gap in sequential read and particularly random read performance between HDDs and SSDs.

Now that we have a rough hierarchy for how these drives compare, let's move onto our game load time tests. These were conducted by starting our stopwatch once we pressed the 'load' button in each game, and stopping the timer once the game was playable. We then turned off the game and restarted the computer, then tested again until we had three load time results. We also tested three times without restarting the computer in between, to see how Windows RAM caching affected loading speeds. We tested five games, including strategy titles, shooters and an RPG. All games used their default settings at 1440p resolution with v-sync turned off.

  1. Grand Theft Auto 5 - single-player, early save
  2. Total War: Warhammer 2 - The Eye of the Vortex campaign, early save
  3. The Witcher 3: Wild Hunt - single-player, late game save
  4. Sid Meier's Civilization VI - small map, 180 turns in
  5. Counter-Strike: Global Offensive - play with bots, Dust 2
Avg Load Time (s) GTA 5 Warhammer 2 Witcher 3 Civ 6 CSGO Average
Intel NVMe 36.97 29.41 8.21 54.86 24.12 30.71
WD NVMe 37.59 35.43 8.89 56.8 24.89 32.72
Crucial SATA 38.23 40.59 8.88 67.85 27.02 36.51
WD SATA 38.69 43.20 9.14 67.93 27.36 37.26
WD HDD 63.99 106.22 17.16 114.54 36.55 67.69

In general, we saw significantly faster loading speeds for SSDs compared to mechanical HDDs in all of the games we tested. Game load times dropped between 25 to 60 per cent by switching to an SSD; on average we saw a time savings of 45 per cent for the Western Digital SATA drive and 46 per cent for the Crucial SATA drive. We also saw noticeably better results for NVMe SSDs over SATA SSDs in three of the five games we tested; in The Witcher and Grand Theft Auto the differences were within our margin of error. On average, our fastest NVMe drive reduced loading times by 54 per cent compared to HDDs. We also noticed that if we didn't restart our computer between tests, we saw vastly improved load times after the initial load was completed, with minimal differences between any of our storage media. This is because the game data loaded from the drive was kept in RAM even after the game was closed, allowing it to be re-used to save time later on.

Cached Load Time (s) GTA 5 Warhammer 2 Witcher 3 Civ 6 CSGO Average
Intel NVMe 36.74 25.49 8.20 54.55 22.57 29.51
WD NVMe 37.10 25.63 8.17 54.11 21.82 29.37
Crucial SATA 36.65 25.74 8.19 53.20 22.64 29.28
WD SATA 36.52 25.62 8.26 50.68 23.11 28.84
WD HDD 37.19 25.64 8.14 54.19 22.25 29.48

Of course, what these tests don't measure is in-game stutter caused by background streaming systems where a mechanical drive can't keep up with a game's demands. We put the WD Black 6TB through its paces here in titles that have traditionally caused issues for mechanical drives - and found no issues whatsoever. To put this into perspective, this is a top-end performance drive (it can write at 200MB/s!) and it's also based on a clean install with no fragmentation issues. We suspect the test results would be very different were they run on an older, fragmented drive or a laptop HDD. What we can say is that an SSD's primary advantage is near instant seek times - it can locate data anywhere on the drive almost immediately, while the mechanical drive needs to physically move the head to another location on the platter. That incurs a lot of latency, which rises massively if the drive is also doing something else at the same time.

Returning to the benchmarks we have, we decided to put our hardware under a more sustained, heavier load. In order to examine how the different drives perform over time, we performed a storage consistency test using Futuremark's PCMark 8. This marathon of a benchmark tests drive performance with drive activity that mimics real-world usage of applications like After Effects or games like Battlefield 3 (this shows the age of the bench, but the results still hold up). We focused on the most demanding trace, which mimics applying complex effects to a high resolution image in Photoshop.

The benchmark is complex, but it essentially involves a degrading series of random writes interspersed with performance tests. In the first two phases, degradation and steady, the drive isn't given idle time to run its usual cleanup duties, but this is allowed in the third phase, recovery. This allows us to see how the drive performs when it is under heavy load and how quickly it recovers. The consistency test takes a long time to run - about 42 hours per drive - so we have only tested three drives: the Intel 905P, the Western Digital Black NVMe and the Crucial MX500 SATA SSD.

consistency_test_v2

The consistency test makes it easy to spot the difference between the two NVMe drives. The WD Black performs as expected in these gruelling tests, with heavily reduced performance during the degradation and steady phases as the drive is bombarded with commands. However, it does return to near peak performance after just five minutes of recovery time. The Crucial MX500 loses a similar proportion of performance, but recovers more slowly.

Meanwhile, the Intel 905P isn't phased by the degradation whatsoever. The drive's bandwidth remains remarkably consistent throughout the two days of continuous testing, which speaks to how well Intel's Optane storage controller and 3D XPoint memory combination works. It goes some way to justifying the 905P's price premium and suggests that the 905P should be able to deliver consistent performance even in worst-case scenarios - like when Windows decides to start preparing for an update in the background while you're playing an important game.

So what general statements and recommendations can we make based on our results? Well, it's definitely worth upgrading to an SSD if you're still on a mechanical hard drive, as games load and install much faster on this flash-based storage medium. SSDs also make other programs on your computer start quicker and feel more responsive, saving you time and potential frustration. And we have to say, returning to a mechanical drive running Windows after years using an SSD proved to be a torturous experience - even with a high-end HDD. In our view, mechanical drives should only be considered for bulk storage at this point.

However, when it comes to gaming, although an SSD clearly makes a big difference, there's a law of diminishing returns when it comes to investing more money and buying into later, more performant technology. Choosing a faster NVMe drive like the Intel 905P only makes sense for high-end systems as you don't see as much of an improvement in game load performance moving from SATA to NVMe as you do moving from a mechanical HDD to a SATA SSD. Of course, NVMe drives are much more worthwhile in non-gaming use cases, such as 4K video editing, where their lower latencies and higher speeds can be better put to use, but for gaming a good-but-cheap SATA drive delivers most of the benefits of solid-state storage.

For many gamers, we'd recommend upgrading your graphics card or processor instead of picking up an NVMe SSD, as these components will have a much bigger impact on in-game performance. Upgrading to at least 8GB and preferably 16GB of RAM should also ensure that load times after the initial load are faster, although measuring the precise impact of different RAM capacities and speeds falls outside the scope of this article. However, for gamers that are already sporting high-end graphics cards and processors, or content creators, moving to the Intel 905P or another high-end NVMe drive should provide noticeable performance benefits - we've been using Samsung Evo NVMe drives for 4K editing recently and the upgrade here over SATA alternatives is substantial.

The bottom line:

  • Moving from an HDD to an SSD can reduce initial game load times by up to 62 per cent
  • NVMe drives show a smaller boost over SSDs, in the region of 10 to 20 per cent
  • SATA SSDs offer better value for money; spend your savings on a better graphics card
  • Ensure you have a decent amount of RAM (at least 4GB) to reduce subsequent load times

Sometimes we include links to online retail stores. If you click on one and make a purchase we may receive a small commission. For more information, go here.

Jump to comments (94)

About the author

Will Judd

Will Judd

Senior Staff Writer, Digital Foundry

A bizarre British-American hybrid, Will turns caffeine into technology articles through a little-known process called 'writing'. His favourite games are Counter-Strike, StarCraft and Fallout 2.

More articles by Will Judd

Comments (94)

Comments for this article are now closed, but please feel free to continue chatting on the forum!

Hide low-scoring comments
Order
Threading

Related

Like what we do at Digital Foundry? Support us!

Subscribe for only $5 and get access to our entire library of 4K videos.

Digital Foundry

Digital FoundryDF Retro: we play every single Doom console conversion

Doom is 25 today - and this is every console port the PC game inspired.

Digital FoundryThis RX 580 8GB at 170 is a bargain for 1080p PC gaming

Get the best value graphics card plus two free games at 30 off.

Digital FoundryJust Cause 4 performance analysis: every console tested

The game's massively more ambitious than JC3 - but do frame-rates hold up?

Advertisement