After nearly six years and countless posts about how my i7-2600K was still good enough, I decided that I’d had enough of good enough when we published our Core i7-7700K review. It was time to upgrade my PC, and I recently completed my new build. I can hear the palms contacting faces already. “Fish, you idiot, Ryzen is almost here! You should have waited.” That could be, but I won’t be buffaloed into second-guessing my decision. As it happens, I’m quite pleased with the results and I’m pretty confident that Ryzen couldn’t do any better.
Now, I don’t have any insider info about AMD’s upcoming chips. Pulling the trigger on my upgrade was influenced by the same speculation and rumors that anyone reading TR would consider. What I did know was that that the games I want to play need every last bit of single-threaded CPU power they can get. Simply put, my guess is that when it comes to clock speed and IPC, Kaby Lake is going to beat Ryzen handily.
To succeed in DayZ you need the right gear: a lot of apples and a lot of single-threaded CPU performance.
First off, it’s good to know where I’m coming from. My old rig (now paired with a Radeon RX 470 for dedicated The Sims 4 duty on the TV) was an i7-2600K mildly overclocked to 4.2 GHz and sporting 16GB of DDR3-2133 RAM. That Sandy Bridge build served as my main PC for probably twice as long as any of my previous rigs. It’s still pretty respectable, but seeing these results in our i7-7700K review pushed me over the edge. The only thing I kept from my old PC was my EVGA GTX 980 Ti Hybrid which, if I behave myself, still has over a year of service ahead of it to reach my personal every-three-year GPU upgrade threshold.
Here’s the complete list of everything that went into my upgrade:
- Intel Core i7-7700K CPU
- Asus Prime Z270-A motherboard
- G.Skill TridentZ 16GB DDR4-3866 memory
- Corsair Hydro H115i cooler
- Corsair Obsidian 450D case
- Samsung 960 EVO 500GB
- EVGA SuperNOVA 850 G3 PSU
Asus Prime Z270-A
Much of my parts list should be self-explanatory. Many of the items therein are TR favorites, while others are at least generally recognized as quality stuff. Of course, there’s some personal preference involved in my choices, but I doubt many gerbils would complain if their own PCs suddenly contained identical hardware to the stuff I bought. One unknown for me was the Asus Prime Z270-A. I purchased it the day it went up for sale with nary a review to be found online. I usually buy motherboards in the $150 range, so it fit the bill there. It also seemed to have a solid combination of features for the price, so I took a chance and ordered one up.
That was over a month ago, and so far I have no major complaints with the Z270-A. I mean, it would be nice if it had a few more USB ports on the rear panel, but I knew what I was getting into when I bought it, so it’s hard to judge it for that. It does have RGB LEDs, but they behave themselves. By default, the LEDs light up white for a while when you power on the PC, then they respectfully turn off after the boot process is complete. The only time you’ll otherwise see them is if you futz with them in Asus’ software. Maybe I’ll mess with the colors on some rainy day in the future, but the lighting doesn’t matter to me in the slightest. My computer is behind my monitor on a corner desk, so I’ll never see inside it anyway despite the fact that the Corsair Obsidian 450D case I chose does have a window.
One minor bummer about this board is that I can’t run my fancy G.Skill Trident Z 3866 MT/s memory at full speed in it yet. G.Skill doesn’t list the board on its official compatibility list, but I was hopeful the pairing would work out anyway. I am able to run the memory at 3733 MT/s with the same XMP-dictated timings as the slightly-faster speed would have, so I’m not exactly starved for memory bandwidth. Of course, there’s more than one factor that could be to blame for this issue, and it might not be the board’s fault. Regardless, I’m hoping for a BIOS update that will let me run the RAM at full speed.
The only other nitpick I have with the Z270-A is that the teeny-tiny standoffs for the pair of M.2 slots aren’t pre-installed. I haven’t built a lot of systems with M.2 drives in them yet, but I know the NUCs I’ve dabbled with had standoffs built-in. When I first plugged in and tightened down my 960 EVO, I realized that something didn’t look quite right, and then I found the bag with the nearly invisible standoffs in it. Thankfully, nothing was damaged, but it’s something to watch for.
The small stuff aside, the Z270-A is a great board: simple but comprehensive, in the vein of the Z97-A and Z170-A. It’s got an Intel NIC, SLI support, USB 3.1, and great fan controls, all at a reasonable price. In fact, we recently recommended this board in our February system guide for those very reasons. If you’ve used or read a recent review of other Asus boards, you more or less already know what to expect from the UEFI. It’s standard Asus goodness all around. The AVX offset feature that arrived with Kaby Lake, which automatically reduces the CPU multiplier for AVX workloads so you can run a swift overclock most of the time and still maintain stability when AVX comes into play, is especially handy to have.
i7-7700K vs. i7-2600K
Surely you’ve already read our i7-7700K review, but if not, it would be a good idea to catch up before reading further. The details below aren’t anywhere near as comprehensive as our standard suite of testing. I’ll be talking about the results of just one benchmark and my own personal pair of chips. I’m going for real-world results, as one might compare their own two builds post-upgrade. We’re talking about completely different systems and even slightly different driver versions here. The only things in common are the video card, the operating system (64-bit Windows 10, of course), and the benchmark itself.
Before I get all excited about the actual benchmark details, I’m going to bust out a dirty ten-letter word: subjective. That’s not generally how we roll at TR, so I’ll avoid claiming anything definitively without the data to back it up. That said, the “seems faster” factor is real, even though I wasn’t expecting it to be noticeable. It’s there, even just in normal day-to-day use.
I’m not alone in noticing this, either. Our resident code monkey, Bruno Ferreira, jumped from an i5-2500K to an i7-6700K last year and spammed the TR Slack channel about how awesome it was. That happened even without the hot-clocked memory and NVMe SSD in my new build. Perhaps the most surprising boost is how the lowly gem Rimworld benefits from my new hardware. My colonists can now raise substantially larger Emu armies before frame rates suffer. Excellent.
Bohemia Interactive’s DayZ is where I spend most of my gaming time, however, and its performance can be approximated well with Yet Another Arma Benchmark, an AI-heavy scenario for Arma III made by community member Greenfist. This is the same benchmark that Jeff used to tease performance differences from various memory speeds in his i7-7700K review. That brings us back to my feelings about Ryzen. I suspect Kaby Lake will have a significant leg up on AMD’s latest when it comes to the games that I’m looking for better performance in. Let’s see what my upgrade bought me in Arma III and DayZ.
After a couple hours of 100% load w/ Prime95 (non-AVX)
For this test, all of Arma‘s graphics settings were cranked up to the max, except for FSAA. My i7-2600K was running at 4.2 GHz and my i7-7700K was running at 4.8 GHz, which is as far as I’ve pushed it so far. (Side note: I think I got a pretty good Kaby chip, and I may have won the TIM lottery, as well.) I’m using 16GB of DDR3-2133 with my i7-2600K and 16GB of DDR4-3733 with my i7-7700K. My GTX 980 Ti is the same in both cases but the SSDs, motherboards, PSUs, cases, and cooling hardware are all different. However, I don’t think that any of those are major contributors to the performance delta between the systems. On to the results.
i7-2600K @ 4.2 GHz w/ 16GB DDR3-2133
Ouch. This notoriously poorly-performing title from 2013 puts the hurt on my old system. Even at 1920×1080, it doesn’t get over 45 FPS and averages just over 30 FPS. Jeff’s unofficial testing for the i7-7700K review placed a stock i7-3770K (3.5 GHz) with DDR3-1866 and similar settings at about 35 FPS when paired with a GTX 1080, so we at least know we’re in the ballpark of what to expect. Let’s see what six years of ticking and tocking gets us.
i7-7700K @ 4.8 GHz w/ 16GB DDR4-3733
Yowza! Nearly double the average FPS from “just” a CPU upgrade? That’s not normal. Only a game engine as goofy as Arma III‘s could produce that result. I’ll take it, though. This jump in performance is a huge quality of life improvement and I can confirm that, subjectively, it carries over to DayZ, as well. What’s that, you say? None of that matters because 1920×1080 is for chumps and if I was playing at a proper resolution the CPU wouldn’t matter nearly as much? Well, let’s check out one more result, then: this time at 3440×1440.
This is so wrong…
You’re looking at just a 2.5-FPS drop after an approximately 2.5x increase in the number of pixels being pushed around. There’s a saying for this sort of thing: That’s Arma! I guess my GTX 980 Ti isn’t out of a job yet. In DayZ, the results have always been similar. In the game’s worst-performing areas, it didn’t matter if you ran 1080p with everything on low or at 4K with AA on high, your frame rates were going to be in the low-teens no matter what, and your GPU wouldn’t even break a sweat.
These results are relevant to my upgrade story because as recently as early last summer, before the DX11 renderer and version 0.60 of the game were released, my trusty i7-2600K and GTX 980 Ti combo was nigh-unplayable in many areas of the map. The 0.60 update would ultimately double or triple the frame rate in the most troublesome areas, and my new PC is frequently doubling those numbers yet again. All with the same graphics card, and all while Bohemia Interactive is improving the look of the game with dynamic shadows and lighting. Cumulatively, it’s completely transformative. I didn’t know if I would ever see the day when DayZ ran buttery smooth. Such a thing was practically unachievable just eight months ago, no matter what hardware you had or how much money you spent.
Make no mistake: I understand that my experience is clearly a niche case. Most games aren’t as CPU-limited as DayZ and Arma III. However, the lesson I learned after chasing perfect DayZ performance for years is that sometimes we just need CPUs to get things done now. In those cases, there’s no substitute for clock speed, IPC, and, apparently, even memory bandwidth and latency. I was guilty of underestimating the importance of the CPU in the overall system, especially since I’ve always gamed at above-average resolutions where conventional wisdom suggests the CPU matters even less.
All told, I stretched my Sandy Bridge system too far. I could have experienced most of the improvements I’ve seen in the last month earlier if I had jumped on Skylake when it came out. Then again, Z270 seems to be better-suited to super-fast memory than Z170, so maybe waiting was the right choice. What about Ryzen? Well, we’ll just have to see how that pans out. I don’t think I’ll regret my choice, but I geared my upgrade toward a specific task. You win some benchmarks, you lose some. Other builds will be better suited for other needs. All told, though, I’m happy with my decision, and if you’re wondering whether it’s time to upgrade your own Sandy system, the answer is probably “yes.”