Update: Linus Tech Tips just released a new video that addresses this issue pretty directly. In their limited blind tests, the participants couldn’t accurately tell the difference between different types of SSD storage.
Last week, I was engaged in a discussion on Twitter about the relative importance of different gaming PC component choices. I stated my opinion that storage performance was less important than graphics performance or CPU performance for most games. Here is what I said exactly:
Storage performance is not the main bottleneck for most games. Graphics is #1, CPU #2, and storage is a very distant #3. As long as you at least have a decent SATA 3 SSD, you won’t notice much difference.
The person I was talking to strongly disagreed with me. One of his responses was:
Given the size of modern assets in games, storage performance is anything but a distant third. Middle of the road graphics cards can do 1080p at 120Hz without a problem. So your “#1” bottleneck being graphics cards is bunk.
I won’t go through a blow by blow recap of the entire discussion, but I will recap the main point I was trying to make…
Not All Gaming Workloads Are The Same
This should be obvious, but I think some discussion would be useful. There are a number of common variables when it comes to gaming performance and how your component choices will affect that. For any particular game, you should consider these items:
- What video resolution are you playing your game at?
- 1080P or lower is pretty easy for modern graphics cards to handle
- 2K or 4K is much harder for a graphics card
- Do you have a high refresh rate “gaming” monitor?
- If not, then high FPS rates are not as useful
- What video quality settings are you using in the game?
- This has a huge effect on the workload for your graphics subsystem
- How old is the game you are playing?
- Older games are usually much less graphics intensive
- Many older games won’t use very many CPU cores, often no more than 4
- Some newer games will struggle when there are fewer than 6-8 cores
- What gaming engine does the game use?
- This has a big effect whether the game is graphics intensive or CPU intensive
- Is it an online game or not?
- If it is a real time online game, then your internet latency can be a big factor
- What kind of game is it?
- Is it a FPS or RTS, or is it a turn-based strategy game
- Does your game have to periodically load new levels or maps as you are playing?
- If this is the case, then storage performance will matter more than normal
- I don’t think this is a common scenario, but maybe I am wrong
How Is Gaming Performance Tested?
If you are playing games at 1080P or lower (especially with lower video quality settings), then any mid-range graphics card can give you high frames per second (FPS) performance. This is true as long as you have at least a “decent” CPU. At 1080P or lower, the CPU becomes the bottleneck once you have a certain, relatively low level of graphics performance.
When new desktop CPUs are released, it is very common for reviewers to perform detailed gaming benchmark tests of that CPU at 1080P resolution. These results are compared to results for other desktop CPUs to see how they compare. They typically have at least two main testing rigs/environments. One will be a high-end Intel Z390 motherboard and the other will be a high-end AMD X570 motherboard. Each rig will use as many identical components as possible. So for example, they will often each have a top of the line NVidia GeForce RTX 2080Ti. The reason for this is to remove the graphics card as any possible bottleneck at 1080P.
They will also use the same brand and model memory sticks so that their detailed specifications are identical. As much as possible, everything but the CPU and motherboard will be identical to eliminate as many other variables as possible. When gaming benchmarks are done in this manner, you will usually see a pretty strong correlation between the single-threaded CPU performance of the CPU and the FPS it gets for a particular game. Depending on the game, the number of cores in the CPU will also have an effect on the FPS scores.
Once you repeat these same benchmarks on the same games at 1440P (2K) or higher resolution, you see a much tighter clustering of the FPS scores between low-end and high-end CPUs. This happens as you move from being CPU-bound at low resolution to being GPU-bound at high resolution. I’ve listed a few YouTube videos that have good examples of this.
- AMD Ryzen 9 3900X & Ryzen 7 3700X Review, Zen 2 Has Arrived!
- Intel i7-4790K & i5-4690K in 2020: Benchmarks vs. Ryzen 3600, 9700K
- Intel i7-6700K & i5-6600K in 2019: Benchmarks vs. Ryzen, 9900K, 9700K, 3600
- Intel Core i3-9100F vs. AMD Ryzen 5 1600 AF, Battle of the sub $100 CPUs
What About Storage Performance for Gaming?
Personally, I think any new desktop PC build (whether it is for gaming or not) should have at least a 500GB SSD as a boot drive. Ideally, this would be an M.2 NVMe PCIe 3.0 x4 SSD drive or better. If your budget is tight, you can get by with a conventional 500GB AHCI SATA SSD, and you will probably not notice that much difference in daily usage (compared to an NVMe SSD).
There is a huge difference in the specifications between the two, and for many use cases, an NVMe SSD is far superior. But for many daily tasks (including most gaming), you just won’t notice much difference between a SATA SSD and an NVMe SSD. You can measure a difference in a benchmark, but you won’t notice it as a human. Going from a magnetic drive to any type of SSD is where people do notice a big difference.
One big difference between a conventional SATA SSD and a M.2 NVMe PCIe 3.0 x4 SSD is sequential performance. This is limited by the interface, so 6 Gbps SATA 3 SSDs are limited to about 550 MB/s, while PCIe 3.0 x4 SSDs can get up to about 3,500 MB/s. If you have a game that regularly loads a large amount of data from disk, and you find yourself waiting for this a lot, then that faster disk is going to help. Most games that I am aware of do an initial load when the game is first started, and then much smaller loads as needed during game play.
Modern PCIe NVMe SSDs also have other advantages over older SATA SSDs. For example, they will have better random I/O performance and lower latency. Intel Optane storage cards do even better in those areas.
Luckily, the prices of good quality M.2 NVMe PCIe 3.0 x4 SSDs have come down to the point where they are not that much more expensive than a similar size and quality SATA SSD from the same brand. The price difference is typically in the $20-30 range for a 500GB SSD. If you can afford it, I would recommend you get a good M.2 NVMe SSD as your boot drive. Just don’t think that it is going to really improve your gaming experience.
A standard rule of thumb for allocating your budget for a dedicated gaming rig is to spend roughly 50% of your budget on the graphics card, then 20-25% on the CPU, and the rest on all of the other components. This reminds me of the old advice about spending 50% of your budget on the speakers when you were buying a component stereo system. If you are playing modern FPS/RTS games and want the best FPS performance (and you have a high refresh rate monitor), then this is pretty good advice. This is especially true if you are gaming at higher resolutions.
Since there is a pretty small difference in cost between a SATA SSD and a PCIe NVMe SSD, you should favor the faster type if it is in your budget. But, if you are on a strict budget, I would save money on storage in favor of the graphics card and CPU. I just would not use a mechanical hard drive as the boot drive to save money, because that is too much of a compromise.
Think about your gaming PC component choices from a workload perspective, because not all games have the same demands. I would also think about your intended budget, to make sure you are making wise choices that will actually improve the performance area that you care about.