So we stumbled upon something we really didn't expect to find while benchmarking for this article - Hyperthreading takes a serious toll on Battlefield 4 performance, and yet core unparking seems to put Hyperthreading nearly on par with a non-Hyperthreaded CPU. As we noted previously, our initial benchmarks were conducted using Windows 7, which is still the most commonly-used OS. There's been plenty of speculation regarding the improvements made under the hood in Windows 8, including the more effective use of virtual cores, so we decided to conduct a few extra tests to help get to the bottom of this issue.
Rather than upgrade our Windows 7 system to Windows 8.1, however, we decided to simply build a second benchmarking rig, which allows us to conduct back-to-back comparisons. This machine, using an i7-4770K at 4.2GHz, is speed-matched as closely as possible to the original machine, which uses a 3770K at 4.5GHz. We've carried over the results from Page 2 for our 3770K machine, and run some new benches on the 4770K. The point here is not actually to see which is faster, but rather to see how Hyperthreading performs on each machine versus the same machine without HT. We think you'll agree that our results do a pretty good job of putting this issue to rest.
Here are two test systems:
- Test System #1: Intel Core i7-3770K@4.5GHz, Asus Maximus V Gene, EVGA GeForce GTX 780 3GB Superclocked, 16GB Samsung DDR3@1866, Windows 7 x64, GeForce Driver Version 331.65
- Test System #2: Intel Core i7-4770K@4.2GHz, Asus Gryphon, EVGA GeForce GTX 780 3GB Superclocked, 8GB Corsair DDR3@1866, Windows 8.1 x64, GeForce Driver Version 331.65
3DMark Fire Strike
We start with 3DMark Fire Strike for two reasons: first, to show that these two systems are about as close to each other in theoretical performance as they could be, and second, to show that yes, Hyperthreading does actually work equally well under both Windows 7 and Windows 8.1 when an application is designed to take advantage of it. The Win7 system picks up 38 percent in its Physics Score when using Hyperthreading, while the Win8.1 system picks up a nearly-identical 39 percent. Throughout these tests, keep in mind that the real goal here is not to compare the Win7 system directly to the Win8.1 system, but rather to note the delta in Hyperthreaded versus non-Hyperthreaded performance to determine if Win8.1 uses Hyperthreading more effectively. Here Win7 and Win8.1 use it to equal effect.
While we considered just posting Battlefield 4 results, given the surprisingly poor results using HT in that game, that wouldn't do justice to the equally-shocking results in Crysis 3, which demonstrated that Hyperthreading can have tremendous benefits. The results were more pronounced at lower core speeds, but even at the highest overclock, HT provided a significant boost.
Well, under Windows 8.1, the HT system still bests the non-HT system, but the delta isn't quite as significant. Under Win7, HT provided a 20 percent boost at the maximum overclock, while in Win8.1, the boost is only 7.5 percent. The benefits to minimum framerates are similar: 21 percent on Win7 versus 11 percent on Win8.1. One possible explanation for this is that with the HT systems, we may be approaching a potential GPU bottleneck as the systems reach 54fps. Alternatively, it could be that Win8.1 uses HT lesseffectively than Win7. But there's something else going on here - why is the 4770K system without HT so much faster than the 3770K system without HT, despite their near identical performance in 3DMark? Perhaps this is a performance gain resulting from the use of Win8.1 that we weren't actually looking for...let's look at a few more tests.
Battlefield 4 Single-Player
The single-player mode of BF4 is generally quite GPU-bound. Under Windows 7, the results with and without HT were about the same at the 4.5GHz clock speed, although the HT system didn't perform as well at lower clock speeds, as shown on page 2. Under Win8.1, we see HT again stumbling ever so slightly, as the 4770K with HT just barely underperforms a 4770K without HT. The delta of 2.5 percent in favor of disabling HT under Win8.1, versus the 0.5 percent delta in favor of using HT on the 3770K. Given that both of these results are basically within the margin of error, we won't draw any definitive conclusions based on them. But the next results, well, that's a different story...
Battlefield 4 Multi-Player
Case closed. HT simply doesn't provide any benefit in Battlefield 4 multi-player, regardless of whether you're using Win7 or Win8.1. We performed these tests on multiple maps and on multiple servers, and every time, the systems with HT enabled were slower. And yet, the use of Win8.1 definitely tightens up the measurements somewhat. The delta for average frames per second under Win7 was 9.5 percent, while it's only 3 percent under Win8.1.
But perhaps what's more interesting here is that the trend we saw in Crysis 3 has reappeared: our 4770K/Win8.1 system seriously outperforms the 3770K/Win7 system, despite the fact that performance was nearly identical in 3DMark. So while the HT system definitely performs better under Win8.1, so does the system without HT. If we'd just looked at the two systems with HT enabled, we'd probably have concluded that Win8.1 solves the HT problem. Well, it doesn't, but it does something even better - it boosts CPU performance across the board in CPU-limited situations. We don't consider this preliminary testing conclusive, so it's definitely something we'll be looking at again. [Update: For more on this topic, see our follow-up article Does the OS Matter: Windows 7 vs. Windows 8.1 in Games.]
How does one summarize results that are clearly all over the map? In one game, Crysis 3, you simply cannot have enough CPU power, and Hyperthreading looks like the greatest thing since sliced bread, especially under Windows 7. In Battlefield 4, the much-anticipated follow-up to the extremely popular Battlefield 3, we're let down by Hyperthreading, just as we were in BF3. Windows 8.1 helps alleviate the problem, but it does not cure it. In every test we ran, both of our systems were slower with HT enabled than with it disabled.
And that's not where the confusion ends. In two games, Crysis 3 and Battlefield 4 multi-player, overclocking provides a big boost in performance, whereas in other games, in particular Tomb Raider, it has almost no effect. So, how much CPU do you really need, and should you pay extra for HT, for that matter? Unfortunately, it absolutely depends on the game, but we do have some general guidelines.
We've previously recommended that folks in the midrange market strive to build balanced systems, and that typically means investing a bit more in the video card than the CPU. Nothing in our findings here changes that recommendation. We think a newer i5 quad-core is almost always the ideal foundation for a gaming system, and they start at around $180. For folks building higher-end systems ($1000 and above), we'd recommend a -K series chip that's unlocked for overclocking - today that means the i5-4670K, although both the 3570K and 2500K have plenty of life left in them. If every game were coded like Crysis 3, we'd probably all need hexa-cores, or at the minimum an i7 with Hyperthreading, but until then, it seems a ~$225 i5 processor is just the ticket to gaming bliss. As for Windows 8.1, it doesn't appear to use Hyperthreading much more effectively than Windows 7, but it did seem to provide a bit of a speed boost overall in CPU-limited scenarios. Therefore, we recommend it to anyone building a new system today.
If what you've read here has you thinking about your next upgrade, have a look at our System Builder's Guides. We particularly recommend our $750 Budget Gaming PC Builder's Guide and our $1,500 High-End Gaming PC Builder's Guide, which we believe provide an ideal balance of CPU and GPU power at their respective prices.