CPU benchmarking can be handled a lot of different ways for game tests. We’ve got the obvious option of singling-out applications which are fully CPU dependent – like the Total War games, where GPU has less impact – or we can throw-in mixed workloads and GPU workloads. The former makes the most sense from a strict methodological standpoint, while the latter provides a look at limitations of CPU scaling when considering the more realistic user scenario of mixed workload / GPU workload games.
We test using both approaches, with CPU-heavy titles providing the most ‘stretched-out’ look at CPU limitations and mixed workload titles (Watch Dogs 2) providing a realistic look at CPU/GPU limitations. Test methodology is defined on the first page, if you’d like more information on the setup.
Watch Dogs 2 Benchmark Performance with i5-2500K vs. 7600K, 6600K, i5 vs. i7
Starting with Watch Dogs 2, we can now reveal why we added this game to our CPU benchmarking workload. With the high-end i7 CPUs, we undoubtedly begin to battle with other system resources – like the GPU, even though it’s a GTX 1080 FTW ($590) – and that means the difference between the Skylake and Kaby Lake i7 CPUs is harder to detect. As we scale down, though, the CPU choke becomes clear.
The i5-2500K is limiting our GTX 1080 FTW to 59FPS AVG when at 1080p with High settings, a card which is capable of achieving nearly 2x that performance when using the latest i7 CPUs. If you’re running a stock i5-2500K, this is about the maximum performance you can expect on our particular benchmark course. Frametime performance isn’t exactly bad, it just scales linearly with the average.
The i5-3570K shows a reasonable improvement, but even that gain is outdone by an overclock on the 2500K to 4.5GHz. This isn’t hard to do and can be held at around 1.3v, though your mileage may vary, and produces higher averages and 0.1% lows than the 3570K.
More interestingly, we see the same-gen i7-2600K really stretch its legs in Watch Dogs 2, producing an additional 15 or so frames per second than the 2500K. This is where you get your value in those i7 purchases – better longevity – and is probably one of the most interesting stories here. We’ll soon see if that carries over to other games, though.
With Watch Dogs 2, we’re generally seeing that multi-core CPUs are advantaged. The i5-4690K and 4790K have a sizeable gap between them as well, something we’ll talk about in another content piece.
But as for the 2500K versus a modern CPU purchase, linear upgrades would land you on a 7600K or 6600K, both of which produce an additional 20FPS or greater average framerate throughput. Frametimes improve in step with this.
Watch Dogs is a game where you could get by at 40-45FPS if you weren’t a stickler about hitting 60, but the point is that an upgrade would better allow for higher graphics settings.
These results are genuinely pretty exciting to study, since the differences are so measurably apparent between all the devices tested. That’s not always the case. Let’s move on to another game and see if that continues.
Intel i5-2500K in Battlefield 1 FPS Benchmark vs. 7600K, 6600K, i7 CPUs
Battlefield 1 doesn’t show much change at the very high-end, where our 1080p/Ultra settings are landing the Kaby Lake, Skylake, and Devil’s Canyon i7 CPUs all roughly in the same performance range. Where we do see the change, though, is dropping down to the 2500K CPU. The 2500K still performs reasonably well in BF1, despite becoming a bottleneck to the 1080 FTW, and is operating at 115FPS AVG with the overclocked version at 124FPS AVG.
The 3570K performs about where the overclocked 2500K performs, and the 4690K starts pulling away from the 2500K in a more dramatic fashion. The CPU is certainly slower, but still keeps up pretty well in Battlefield 1. Only if you’re pushing for higher refresh gameplay or upgrading into 1070 and 1080 class hardware will a bottleneck be noticeable.
Unlike Watch Dogs 2, the difference between the i7-2600K and i5-2500K is not as substantial.
Total War: Warhammer Benchmark – i5-2500K vs. 7600K, 6600K, 4690K
Total War is new to our CPU bench, and thus far only features the i5 series CPUs. Before diving in, note that Total War does output framerates with a good amount of variance in the worst frametimes. This means we won’t really be using the 0.1% metric as much, since it fluctuates pass-to-pass and seems inconsistent overall. We’re also only testing with DirectX 11, because the DirectX 12 beta version of the game just isn’t as performant as the Dx11 version.
We’re seeing the i5-2500K perform at about 92FPS with 1080p and High settings for this benchmark. Overclocking provides substantial gains, pushing us up to 114FPS AVG and with bolstered frametimes. The performance improvement is about 24% -- absolutely worth the overclock.
Looking to modern CPUs, the 6600K is capable of operating at 156FPS AVG, with the 7600K at 165FPS AVG. From the 2500K to the 7600K, we’re seeing a stock clock-rate difference totaling 73FPS AVG, or a percentage increase of about 80%.
GTA V Benchmark with i5-2500K, 7600K, 6600K, 6700K, 3570K
The final benchmark is GTA V. Big note for this one: We had some confusing issues with the 6600K and 7600K that didn’t seem to affect other CPUs, and even took to Twitter to ask about this. Jay reported something similar, as did a few of our readers, and we’ve been unable to fully figure it out yet.
So far, it seems like the 6600K and 7600K have worse stuttering with our GTA V deployment than other CPUs, and this seems to be less of an ordeal when dropping XMP on the memory. Still, there’s a lot of research to do.
We haven’t got it fully figured out yet, re: 6600K & 7600K. We’ve reinstalled Windows, the game, and done all manner of testing to ensure it’s not on our end. Just wanted to point that out before we get to the results. We’ve done due diligence to figure it out, but just haven’t yet. If anyone’s got thoughts, post them below! We’ve tried XMP settings, overclocking and downclocking, new memory, motherboards, PSUs. Some of our followers and colleagues in the industry reported that they’ve seen similar issues when we asked on twitter.
As for the rest of the results, we know those are all accurate as ever – it’s just the 6600K and 7600K that we’re having trouble with.
OK, so the charts. We’re looking at an average FPS of 101 with the i5-2500K, with overclocking pushing us to an impressive 124FPS AVG. Compared to the i7-2600K, there’s not a lot of advantage in GTA V from the extra threads. We’re seeing frequency seems to provide the biggest impact to performance, for the most part. The i5-3570K operates about 11FPS faster than the stock i5-2500K.
Ashes of the Singularity – i5-2500K vs. 7600K, 2600K, 3570K, 6600K, etc.
Ashes of the Singularity with Dx12 and the CPU-focused benchmark positions the CPUs as we’d expect, based on Battlefield 1, Watch Dogs 2, and synthetic tests. We’re seeing the i5-2500K outperformed about two-fold by the modern i7 CPUs (6700K, 7700K), with the i5-7600K about 10FPS over the 2500K stock (~55% difference).
Overclocking the 2500K nets us a performance boost that plants the CPU between the 3570K and 4690K stock CPUs, showing again that an OC may actually be worthwhile on SNB platforms.
AOTS clearly favors multithreading, as Watch Dogs 2 did. The i7 CPU stack-up is interrupted only by the 2600K’s lower placement, clearly a result of the aged architecture and slower clock-rate of the SNB chip.
Metro: Last Light – i5-2500K vs. 2600K, 3570K, 4690K, 7600K
As has historically been the case for Metro: Last Light, we are seeing the i5 CPUs (and i7s with hyperthreading disabled) struggle in the frametime department, failing to achieve even half of what the i7 CPUs can output for 0.1% low FPS.
The i5-2500K stock operates an average of nearly 90FPS – really not bad, but there’s a big bottleneck here placed on the GTX 1080 FTW. The bottleneck isn’t overcome until we start hitting the high-end i7 devices, like the i7-7700K (stock or overclocked). Overclocking the i5-2500K to 4.5GHz gives us a significant FPS gain (~+20%), planting the i5-2500K ahead of the i5-3570K (though low FPS does have some trouble as we approach stability limits), and landing below the i5-4690K.
Conclusion: Is It Worth Upgrading from an i5-2500K to 7600K or Zen?
The i5-2500K has held up relatively well over the past five years, but it’s starting to show some serious age in a few specific games. Watch Dogs 2, for one, posts sizeable differences between modern CPUs and the 2500K – though multi-core seems to matter more for that particular title.
We’re also seeing big gains from overclocking in a lot of games, especially because the 2500K was so easy to overclock. If you’ve got one, it might be worth throwing it under a good cooler and pushing it to 4.5GHz for the remainder of its life.
Blender and rendering tasks are particularly abusive to the i5-2500K, which is being outpaced nearly two-fold by modern successors.
As for upgrades, if you can grab an Intel i5-6600K ($239) for a good bit cheaper than a Kaby Lake i5-7600K ($242), it might be a worthwhile upgrade. The 7600K is also worth looking into, potentially, but wait for our forthcoming review on that before making any purchases.
It’s a good time to upgrade from the 2500K, though it is still hanging in there pretty well. We don’t know what Zen’s / Ryzen’s performance will look like just yet, but we do know that it should be shipping in late February. Normally, we don’t advise waiting for upgrades since you could wait forever with hardware. Zen is very close at this point, though, and seems like it’d be good to wait on. Give it another month and we’ll revisit this topic. At that point, it’s either on to the 7600K, a used 4690K or 6600K, or some Zen equivalent.
Editorial, Test Lead: Steve “Lelldorianx” Burke
Video Production: Andrew “ColossalCake” Coleman
Test Technician: Patrick “Germ King” Lathan