Introduction and Test Hardware
How much does a processor matter for gaming?
The PC gaming world has become divided by two distinct types of games: those that were designed and programmed specifically for the PC, and console ports. Unfortunately for PC gamers it seems that far too many titles are simply ported over (or at least optimized for consoles first) these days, and while PC users can usually enjoy higher detail levels and unlocked frame rates there is now the issue of processor core-count to consider. This may seem artificial, but in recent months quite a few games have been released that require at least a quad-core CPU to even run (without modifying the game).
One possible explanation for this is current console hardware: PS4 and Xbox One systems are based on multi-core AMD APUs (the 8-core AMD "Jaguar"). While a quad-core (or higher) processor might not be techincally required to run current games on PCs, the fact that these exist on consoles might help to explain quad-core CPU as a minimum spec. This trend could simply be the result of current x86 console hardware, as developement of console versions of games is often prioritized (and porting has become common for development of PC versions of games). So it is that popular dual-core processors like the $69 Intel Pentium Anniversary Edition (G3258) are suddenly less viable for a future-proofed gaming build. While hacking these games might make dual-core CPUs work, and might be the only way to get such a game to even load as the CPU is checked at launch, this is obviously far from ideal.
Is this much CPU really necessary?
Rather than rail against this quad-core trend and question its necessity, I decided instead to see just how much of a difference the processor alone might make with some game benchmarks. This quickly escalated into more and more system configurations as I accumulated parts, eventually arriving at 36 different configurations at various price points. Yeah, I said 36. (Remember that Budget Gaming Shootout article from last year? It's bigger than that!) Some of the charts that follow are really long (you've been warned), and there’s a lot of information to parse here. I wanted this to be as fair as possible, so there is a theme to the component selection. I started with three processors each (low, mid, and high price) from AMD and Intel, and then three graphics cards (again, low, mid, and high price) from AMD and NVIDIA.
Here’s the component rundown with current pricing*:
- AMD Athlon X4 860K – $74.99
- AMD FX 8350 – $165.93
- AMD FX 9590 (with AIO cooler) – $259.99
- Intel Core i3-4130 – $118
- Intel Core i5-4440 – $184.29
- Intel Core i7-4790K – $338.99
Graphics cards tested:
- AMD Radeon R7 260X (ASUS 2GB OC) – $137.24
- AMD Radeon R9 280 (Sapphire Dual-X) – $169.99
- AMD Radeon R9 290X (MSI Lightning) – $399
- NVIDIA GeForce GTX 750 Ti (OEM) – $149.99
- NVIDIA GeForce GTX 770 (OEM) – $235
- NVIDIA GeForce GTX 980 (ASUS STRIX) – $519
*These prices were current as of 6/29/15, and of course fluctuate.
Looking over the list above I’m immediately drawn to the R9 280 at just $169.99 on Amazon, a great price for a mid-range card (we'll see how well it performed). On the NVIDIA side I'll note that my choice of mid-range card might be questioned, with the GTX 960 the new $200-$220 option; I picked the 770 over the 960 simply because I already had one on hand to test. Finally, all of the AMD cards tested were overclocked retail models, such as the MSI R9 290X Lightning with a 1080 MHz core. Stock performance will be a bit lower, but these are all off-the-shelf cards and nothing was run beyond retail spec. If it was overclocked by the manufacturer, I ran it that way. Each platform was configured using default settings, with 8GB of dual-channel 1600 MHz DDR3 used for each testbench.
The six games used to benchmark this hardware
For game testing I made the decision to use only automated benchmarks for the sake of consistency. I wanted to eliminate the possibility of variance with the test results, as there will often be minute differences in the results with this much hardware. As a result of this there were some excellent candidates that I simply couldn't use without an automated benchmark tool. I tried to vary the mix of games to provide an uncolored look at the true potential performance of the hardware, and of the 6 games selected half use AMD's Mantle API (these were tested with DX11 as well) and at least one (Civilization: Beyond Earth) is known to be highly CPU-bound.
All tests were run at both 1920×1080 and 2560×1440 resolution, with three identical runs at each resolution for each hardware component. Drivers were current when testing began in January, and therefore out of date by current standards. This was necessary to provide a true comparison between hardware results.
AMD cards were tested using Catalyst Omega 14.12
NVIDIA cards were tested using GeForce Game Ready Driver 347.25
Windows 8.1 64-bit was used for all game testing, and games were loaded using identical Steam backups for each title.
Without further preamble let’s get to the test results!