PC gaming has always been an expensive hobby, even if you stick to entry-level hardware. So much so that the majority of enthusiasts only upgrade their systems piecemeal, one component every so often, or get a new rig after many long years. But if you were hoping to build a new gaming PC or update the CPU and motherboard this year, the shocking cost of DRAM kits right now is making the process considerably challenging.
Back in July 2025, you could pick up a 32 GB kit of DDR5-6000, with a CAS latency of 30 cycles, for under $90 (albeit during the sales). For AMD Ryzen processors, you honestly wouldn’t need anything better, and although Intel’s chips can handle faster DRAM, they’ll still be just as happy with a kit like that.
Article continues below
The DRAM test setup
PC Gamer test rig: AMD Ryzen 9 9900X | Asus ROG Strix LC III 360 AIO cooler | Nvidia GeForce RTX 5090 FE | 2 TB Samsung 990 Evo Plus SSD | Thermaltake ToughPower PF3 1050 W PSU | Thermal Grizzly Der8enchtable chassis | Acer XB280KH monitor | Windows 11 25H2
To test how much of an impact DRAM speed has on today’s games (and to a lesser extent, timings as well), I’ve used a set of Lexar Thor OC DDR5-6000 CL32. Without enabling EXPO in the motherboard BIOS, it will run at 4800 MT/s, with a CAS latency of 40 cycles. While it’s not quite a worst-case vs best-case scenario, as there is even slower DDR5 RAM out there, it’s still relevant because that’s the kind of ‘cheap’ memory around at the moment.
I’ve picked seven games in total: three that I didn’t expect to see any difference in performance, and four that I suspected would run slower with the DRAM set to 4800 MT/s. I’ll explain the reasons behind my forecasts as I break down the results for each game, so without further ado, here’s what I discovered.
With each game, I carried out five minutes of gameplay (no built-in benchmarks), using Nvidia’s Frameview to record the average and 1% low frame rates for a run at 1080p, 1440p, and 4K. As I used a GeForce RTX 5090, the graphics quality in the games was set at the maximum value, though I didn’t use path tracing if it was available. Ray tracing was used in all but one of the games that support it.
I didn’t use DLSS upscaling and frame generation, either, as I wanted the hardware in the test rig to be loaded up relatively evenly. Applying DLSS Performance, for example, would just have the GPU doing very little behind the scenes and artificially pushing the frame rate constraints onto the CPU.
Games that aren’t DRAM-sensitive
Gaming performance at 1080p
AMD Ryzen 9 9900X, Nvidia GeForce RTX 5090
| Product | Value |
|---|---|
| Lexar Thor OC DDR5-6000 CL32 | 112 Avg FPS, 67 1% Low FPS |
| Lexar Thor OC DDR5-4800 CL40 | 114 Avg FPS, 71 1% Low FPS |
The games I didn’t expect to see any performance difference with the two DRAM speeds are Cyberpunk 2077, Black Myth: Wukong, and Stalker 2. And that’s exactly what happened, as you can see in the results. Even though Stalker 2 was marginally slower with DDR5-4800 CL40 than DDR5-6000 CL32, it’s best to ignore any differences of 5% or less with the frame rates (either higher or lower).
That’s because there will always be a reasonable margin of error when capturing gameplay rather than benchmark performance. A game’s built-in graphics test will run the same code pretty much every time, whereas it will often alter quite a bit when playing it as intended. This is especially true for open-world games that have a ‘real-time life’ aspect to them, such as NPCs seemingly wandering off and doing their own thing.
As to why I didn’t expect to see any performance differences, well, that’s because these games are heavily dependent on the GPU for performance. As powerful as the RTX 5090 is, it’s not able to handle the graphics load with the settings I used to such an extent that the CPU and/or system RAM becomes the limiting factor behind the frame rate.
To have that kind of scenario, you either need a game that has an enormous CPU workload or one that has a very light graphics load. Enter stage left, Microsoft Flight Simulator 2024 (MSFS 24) for the former, and Counter-Strike 2 (CS2) for the latter.
Games that are a little bit DRAM-sensitive
Gaming performance at 1080p
AMD Ryzen 9 9900X, Nvidia GeForce RTX 5090
| Product | Value |
|---|---|
| Lexar Thor OC DDR5-6000 CL32 | 50 Avg FPS, 27 1% Low FPS |
| Lexar Thor OC DDR5-4800 CL40 | 48 Avg FPS, 29 1% Low FPS |
If you’re looking at the MSFS 24 results and thinking, “They look a bit all over the place”, then you’d be thinking right. There’s so much going on, even with a simple flight around Heathrow airport, that the performance is rarely consistent. That said, there is at least some evidence that DRAM speed does affect how well the game runs, even at 4K Ultra.
CS2 was just as surprising. Being a competitive, multiplayer shooter, it doesn’t boast GPU-crushing graphics loads, as indicated by the high frame rates. However, once above 1080p, there’s practically no difference between the two DRAM speeds. It would be different if I had used a lower quality setting and/or resolution, of course, because that’s what professional gamers do, to get the best possible frame rates.
At 720p low quality, most GPUs will be doing very little, and CS2 will almost certainly be impacted by more than you can see above.
However, it’s not always about the intensity of the graphics or what GPU you’re using. There are some games that are always sensitive to DRAM performance, often because they’re constantly streaming data from the system RAM. The two examples I’ve chosen to demonstrate this are Spider-Man Remastered and HogRAM Legacy. I mean Hogwarts. Sorry about that.
Games that are very DRAM-sensitive
Gaming performance at 1080p
AMD Ryzen 9 9900X, Nvidia GeForce RTX 5090
| Product | Value |
|---|---|
| Lexar Thor OC DDR5-6000 CL32 | 117 Avg FPS, 72 1% Low FPS |
| Lexar Thor OC DDR5-4800 CL40 | 106 Avg FPS, 66 1% Low FPS |
With an RTX 5090, Spider-Man is very CPU-limited, as indicated by the fact that increasing the resolution doesn’t reduce the frame rate by any appreciable amount. But unlike CS2, which isn’t affected by DRAM speeds unless at 1080p, this game does run a little slower with DDR5-4800. Not by much on average, but the drops in the 1% low figures are quite marked. Spider-Man is still thoroughly playable with the slower memory, of course.
As is Hogwarts Legacy, even though the 1% lows in this game dropped considerably with DDR5-4800. However, just like with MSFS 24, it’s pretty inconsistent in terms of performance when run on top-end hardware, even when you disable ray tracing, as I did for my tests. Depending on where you are in the game’s world and what’s going on around you, the limiting factor for the frame rate bounces between the CPU and GPU.
So I pulled the RTX 5090 out of the test rig and swapped it for an RTX 3060 Ti, and retested Hogwarts Legacy across all of its graphics presets, but just at 1080p.
|
FPSAvg. FPS / 1% Low |
DDR5-6000 CL32 |
DDR5-4800 CL38 |
% difference |
|
Low preset |
160 / 80 |
149 / 70 |
-7% / -13% |
|
Medium preset |
138 / 71 |
130 / 64 |
-6% / -10% |
|
High preset |
111 / 65 |
110 / 64 |
-1% / -2% |
|
Ultra preset |
94 / 54 |
93 / 53 |
-1% / -2% |
Here you can clearly see what happens when the game becomes more CPU-limited than GPU-limited. With the low and medium presets, the slower DRAM causes a reduction in the minimum performance, whereas the rendering-heavy high and ultra settings push the limit entirely onto the GPU.
Not that any of this should come as a surprise. If your gaming PC’s graphics card is working flat out to produce all the graphics, with not even a hint of spare capacity anywhere, then you’ll still get the same performance with a slower CPU and/or DRAM kit.
On the other hand, if you only play competitive shooters at 1080p and on low settings, then you are going to see a reduction in performance with something like DDR5-4800. However, even though such DRAM has 20% less bandwidth than DDR5-6000, only one of the games tested above displayed the same level of fps reduction, and it wasn’t an e-sports shooter.
So in the current RAMpocalypse, where memory kits now cost more than some mainstream graphics cards, you’ll be fine with picking up a set of slow DDR5, as long as you’re not interested in achieving the maximum possible frame rates for competitive gameplay.
A loss of 7% in performance in CS2 at 1080p might not seem like much, but that’s with very high settings. It’ll only get worse at lower resolutions and quality settings.
And then there’s the likes of Spider-Man and Hogwarts Legacy, which both stream a lot of data, and are therefore affected by DRAM speeds. The latter lost up to 20% in its minimum performance using DDR5-4800, compared to DDR5-6000, but if you tweak the settings to put the frame rate limit back onto the GPU, then you can reduce the impact a fair bit.
For most other scenarios, though, especially those that will have your GPU busy as a bee, then grabbing a set of slow DDR5 to save money won’t seriously ruin your gaming experience. It’ll still be horrible for your bank account, but there will be no escaping that problem any time soon.
Best PC build 2026
Memory,Hardware#memory #crisis #fast #memory #gaming1775895152
