There are a few ways to boost your gaming PC’s performance, but upgrading the components, especially the graphics card or GPU, is one of the key ways to unlock faster frame rates, higher-quality visuals and higher resolutions. However, you can’t always be certain that a GPU upgrade alone will give you the performance lift you’re after. For gaming, other factors can hold your system back, and one of the big culprits is actually the central processor. Understanding CPU bottlenecks, how you can run into them and how you can work around them can help you unlock greater performance gains.
The recipe for “CPU-bound” scenarios
With gaming, your CPU and GPU both have important tasks on their own and work together to complete the overall job of smooth gaming. Put simply, your CPU figures out what needs to happen in the game. It calculates things like physics and enemy AI, then tells the GPU to get to work. The GPU figures out where everything is going to go within the virtual space of the game and what it’s all going to look like. It then sends that out to your TV or monitor.
You can think of it like a restaurant. A waiter (the CPU) seats guests and takes orders, then passes those orders to the kitchen. The cooks in the kitchen (the GPU) whip up the food for the guests. Ideally, the cooks finish the food in the right amount of time so the waiter can bring it to the guests (your monitor, in this case). Ideally, the waiter isn’t waiting for the cooks, and the cooks aren’t waiting for the waiter. When it’s all timed correctly, your monitor gets a steady stream of fully rendered frames to create smooth on-screen motion. All the guests get their food on time, in other words.
It’s fairly easy to get this sequence out of balance if your gear isn’t capable of the performance required for your chosen resolutions, game detail, screen resolution and so on. If the cooks (your GPU) are too slow, the result is often a slow or stuttering frame rate. This is usually pretty obvious. When the bottleneck is the CPU, slow-moving waiters in our analogy, the cooks will be waiting, underperforming what they’re capable of. When the CPU is what’s limiting your performance, that’s called being CPU-bound.
When you drop in a new graphics card to your PC, it’s like replacing the cooks you originally had in the kitchen with even faster cooks. However, if you keep the wait staff the same, the orders might not get to the kitchen staff any faster, so overall, a similar amount of food is getting to the customers. The orders will just sit under a heat lamp waiting for pickup.
Or to put it more plainly, if you just get a new graphics card and your CPU is already at its limit, you won’t improve your overall performance that much. You’ll likely see some improvement, but not nearly as much as you would have gotten had you also upgraded your CPU.
A concrete example: Upgrading my rig
I’ve been running a gaming rig with an Intel Core i7-12700K (or in our analogy, a good “waiter”) and an AMD Radeon 7900XT (an OK “cook”) for a while. Then I saw an opportunity to upgrade to an RTX 5080 (a way better “cook”). That GPU is considerably newer than the Core i7 CPU, though, and I knew there was some potential for the system to be held back because of being CPU-bound.
Curious to see how much the older i7 could hamper performance, I tested my GPU-upgraded system against another PC running a much more potent gaming CPU (an AMD Ryzen 7 9800X3D). I put the exact same graphics card into each system so there’d be no card-to-card variance. The results raised a lot of valuable findings regarding gaming performance, benchmarks and balancing CPU/GPU load to get the most out of your hardware.
Synthetic benchmarks don’t tell the whole story
I ran CNET’s usual suite of 3DMark benchmarks on both systems. Although my PC improved its scores from before the upgrade, I was surprised by how marginal the differences were, especially compared to the Ryzen-powered system. The Ryzen 7 9800X3D is widely considered the best gaming CPU, and yet here was my aging 12700K somehow going neck and neck with just a simple graphics swap. In the Wildlife Extreme benchmark, the Ryzen system performed just 0.12% better. Time Spy and Steel Nomad also failed to show the Ryzen system offering even a 1% improvement in performance.
Neither system was getting a heat advantage, either, as both ran 3DMark’s Steel Nomad Stress test and passed with matching 99.3% stability, confirming consistent performance and adequate cooling for both. In Steel Nomad Lite, Fire Strike Ultra and Port Royal benchmarks, the Ryzen system won by as much as 6% and as little as 2%.
Those aren’t exactly gains I would want to buy a whole new CPU and platform for. All of this happened in spite of 3DMark’s CPU Profiler showing a 15.4% lead for the Ryzen system in single-threaded processing (albeit with a 0.8% deficit in max-threaded processing, thanks to the 12700K’s 20 threads to the 9800X3D’s 16 threads). So what’s going on? Is being CPU-bound a myth? Not so fast. Time to dig deeper.
1080p gaming shines a light on the weaknesses
Turning to actual game benchmarks, the tiny crevice between the two systems widened to a chasm. Running Shadow of the Tomb Raider at 1080p and the highest graphical preset, my PC hit 202 frames per second on average. A respectable number — but actually slower than with my old video card! With the older Radeon, it was able to get an average of 208fps.
Meanwhile, the Ryzen system reached an impressive 360fps. My PC with the RTX 5080, running Guardians of the Galaxy at 1080p and high settings, managed 162fps, while the Ryzen system leaped ahead to 267fps. Remember, that’s with the same graphics card. This is the kind of improvement I’d hope to see from an expensive graphics card upgrade. This isn’t just a graphics card upgrade, but it’s also an entire platform swap with a CPU, motherboard and new memory (so another $1,000 to build). It does tell us a few things, however.
The Guardians of the Galaxy test helps highlight which component is responsible for performance shortcomings. It gives us specifics about the average frame rate, but diving deeper into the results, it also gives us frametimes — in other words, the amount of time required by the CPU and GPU to produce a frame.
Looking at that same 1080p run, the Ryzen CPU needed an average of 3.2 milliseconds per frame while the Core i7-12700K needed 5.9ms. That seemingly small difference adds up considerably over dozens or hundreds of frames. After all, the CPU and GPU need to spit out a frame every 16ms to deliver just 60fps. In that same benchmark, the GPU only needed an average of 2.2ms per frame on the Ryzen system versus 2.4ms on the Intel system. Going back to the restaurant analogy, the cooks were clearly getting their job done so quickly that they had to wait around for new orders from the waiter.
The Ryzen system’s lead over the older Intel system wasn’t always so substantial. Running a more modern, far more demanding game changed things. For Assassin’s Creed: Shadows at 1080p with high settings, the Ryzen system managed 91fps while the Intel system trailed at 87fps. That gave the Ryzen system a marginal 5% lead. Now, why is that?
While 1080p gaming can show a gulf between CPU performance, it may only do so when a game demands a lot out of the CPU and not as much from the GPU. The GPU still has a heavy load at 1080p with Assassin’s Creed: Shadows, so things look quite different. To show just how that looks in practice, I also ran benchmarks at 1440p and 4K (and higher graphical settings in a couple cases for good measure). The results of those tests highlight a key way to get past the shortcomings of an older, slower CPU by shifting the load.
Shifting the load
Your waiter can only work so fast, and you don’t want your cooks sitting around doing nothing. So what do you do? Give the cooks more to do. Perhaps more food per order or something fancier for each table. Assassin’s Creed: Shadows leveled the playing field considerably between the Ryzen 7 9800X3D and Intel Core i7-12700K. Make no mistake: It’s pretty demanding on a CPU, but even at 1080p, it can slam an RTX 5080 hard with all its ray-traced graphics. So a graphically demanding game can “hide” a CPU shortcoming. By adjusting settings, you can often make a game more or less graphically demanding to fine-tune your system based on its specific CPU and GPU strengths (or weaknesses).
The role of resolution
Let’s go back to Shadow of the Tomb Raider. At 1080p, the Ryzen system crushed the Intel system. The RTX 5080 was hardly breaking a sweat on the Intel system, even at the highest graphics setting. Even dialing up the resolution to 1440p, the average frame rate didn’t drop. It still hit 202fps and was now delivering 744 million pixels per second (resolution by average frame rate) instead of 418 million pixels per second like it was at 1080p.
That bump in resolution gives you sharper visuals, and in this case, it didn’t cost any frames per second. It’s as if the cooks were making an extra-large serving and still serving it up in the same amount of time. The Ryzen system was still faster at 282fps when jumping to 1440p (that’s 1 billion pixels per second), but it did see its frame rate decline compared to 1080p.
The jump to 4K with Shadow of the Tomb Raider takes some frames off the Intel system’s average, bringing it down to 153fps (or 1.27 billion pixels per second). That shift implies that the CPU is no longer the major limiting factor. Success! We’re no longer CPU-bound. In fact, back on the Ryzen system, the 4K results are also 153fps. That’s no coincidence since, remember, these are using the same video card.
The role of other graphics settings
Guardians of the Galaxy highlights a different case. The Intel system delivered little change whether the resolution was 1080p (162fps average), 1440p (162fps average) or 4K (161fps average). In each case, the GPU had to work harder, but the CPU remained the limiting factor. Some games are just CPU-heavy, including those with complex physics, elaborate nonplayable character AI and world destruction, as explained in this Intel interview with game devs.
My tests used high graphics settings. If we changed those settings instead of the resolution, we can also shift the load to the GPU. Running at 1080p but upping the settings to Ultra, the Intel system ran at 157fps on average, barely losing any speed while getting a big graphical upgrade. In this case, the cooks made filet mignon instead of a hamburger, and it barely took any additional time.
This gives you a second option to break free of being CPU-bound. You can increase the in-game graphics settings while staying at the same resolution. If you’re already playing at your monitor’s or TV’s native resolution, this may be the more useful option.
Another element to consider is dynamic or subsampling resolution. If your games are using DLSS, FSR or XeSS, it’s akin to running at a lower resolution and can, therefore, present a similar opportunity to be CPU-bound. In these cases, you can try upping the setting (for example, upping Quality or DLAA instead of Performance) or you can disable them altogether and render at your display’s native resolution.
Game graphics and resolutions can only go so high (practically speaking), so there will still come a time when the only reasonable answer will be to upgrade your CPU. Armed with this information on CPU-bound scenarios and how to shift the load, hopefully, you can make sure that time doesn’t come too soon.
Reverse! Reverse!
One last thing: Understanding CPU-bound scenarios and how to work around them also arms you with information about how to work around GPU-bound scenarios.
Just like how you can find ways to shift the balance of the load to be greater on the GPU than it is on the CPU, you can shift the load back to the CPU by performing the opposite adjustments. For example, want higher frame rates in a competitive or fast-paced game? Lower the resolution and dial back graphics settings. You may even find that changing just a few specific settings can help without a major drop in visual quality.
