Well, this is certainly something. As you may know, Intel has been focused like mad on mobile computing for the past few years, attempting to insert itself into a growing market against established rivals like ARM. Desktop computing has kind of been on the backburner as a result. But a funny thing happened on the way to the death of the PC: yet another revival at the high end of the market. PC gaming is more alive and vibrant than ever, and folks are pioneering new applications like virtual reality on the PC, as well.
Intel has decided to acknowledge the thriving PC gaming market by throwing us a big, juicy bone. The first-ever version of the Skylake, its next-generation CPU architecture, is making its debut today in a pair of socketed processors for desktop PCs. The Core i7-6700K and Core i5-6600K are the first ever Skylake parts available to the public, and they’re arriving alongside an armada of motherboards based on the new Z170 chipset.
It’s a lake in the sky!
2015 has been a busy year in PC hardware, but it’s been full of strange product introductions. We’ve covered a number of product unveilings that have involved big architecture reveals and great fanfare but very little actual hardware to review. Heck, Intel announced the desktop version of its Broadwell CPUs back in June, but you still can’t buy them in North America. Skylake is the opposite situation. We have a Core i7-6700K chip in our grubby hands, but we don’t yet know the details of this new CPU microarchitecture. Intel says it’s planning to reveal those in a couple of weeks, at its Intel Developer Forum event in San Francisco. So we can show you how Skylake performs, but we can’t yet tell you exactly why.
We also don’t yet know the exact shape of the entire lineup of Skylake-based products. Intel says it will be releasing the rest of the family later in the third quarter of this year, after IDF. For now, these two desktop CPUs will stand alone.
Truth be told, though, we really do know quite a bit about Skylake already. The desktop CPU we’re reviewing is a quad-core, eight-threaded processor with 8MB of L3 cache, much like its predecessors. Skylake parts are manufactured on Intel’s 14-nm fabrication process with tri-gate transistors, like the Broadwell parts of the prior generation. However, Skylake is a “tock” in Intel’s familiar “tick-tock” development cadence; it brings with it a new CPU microarchitecture.
As usual, this architectural refresh is meant to improve clock-for-clock performance through various clever enhancements to the processor’s efficiency and throughput. Intel claims the 6700K is “up to 10% faster” than its predecessor, the Haswell-based Core i7-4790K.
In fact, the Core i7-6700K replaces the 4790K at the exact same price, just as the 6600K replaces the 4690K. The 6700K’s peak clock speed is down a little bit, at 4.2GHz instead of the 4.4GHz Turbo peak on the 4790K, but the 6700K can run all four cores at 4.2GHz under load. Since the 6600K and 6700K are both K-series parts, they have unlocked multipliers for easy overclocking, too.
Intel has pushed a little on the power front in order to deliver higher performance in recent generations. The top Ivy Bridge quad core topped out at 77W, while Haswell moved up to 88W. Now, Skylake nudges the limit up to 91W.
That added CPU power draw could be offset at a platform level by the transition to a new memory type, DDR4. Skylake supports both DDR4 and DDR3L type memories, for lower voltage operation and power savings. Bog-standard DDR3 isn’t officially supported at its usual 1.5V, although we may see some motherboard makers hack their way around that limitation. Most of the market will likely embrace DDR4 as the new standard, since it promises higher throughput and more headroom for transfer rates going forward. Intel’s high-end Haswell-E platform made the transition to DDR4 last year, as did the dual-socket Haswell-EP Xeons.
Like those platforms, Skylake desktop parts are getting a conservative start with DDR4. The first products only officially support 2133 MT/s operating speeds. Running your memory any faster is technically overclocking. That said, our test rigs are outfitted on day one with Corsair Vengeance LPX DIMMs rated for 2666 MT/s operation, and much higher speeds are possible.
The LGA 1151 socket
As you might expect given the new memory type, Skylake processors are not socket-compatible with prior generations. They require a motherboard with a new socket type known as LGA1151. Although the pinouts and plastic retention tabs are different, this socket is virtually the same size and shape as the one that came before it. As a result, any CPU coolers meant for Haswell CPUs ought to be compatible with LGA1151-based motherboards.
The new socket brings another change of note: the death of FIVR, the fully integrated voltage regulator first introduced and hailed as progress on Haswell processors. I expect we’ll hear more about the reasons behind the decision to spike the integrated VR at IDF, but we already know FIVR caused some complications for the mobile Broadwell chips. The inductors required by FIVR increased the Z-height of the processor, and Intel had to cut a hole in the motherboard in order to shoehorn Broadwell CPUs into extra-thin devices. Furthermore, FIVR wasn’t optimally efficient at low voltages, and that fact prompted Intel to build a bypass mechanism into Broadwell called LVR. As we wrote back then, “The need for 3DL and LVR makes one wonder whether the level of VR integration in Broadwell makes sense for future generations of Intel SoCs.” I suppose we don’t have to wonder anymore.
The question now is what exact arrangement has replaced FIVR. Presumably, Intel hasn’t taken a step back on things like independent supply rails for the CPU cores and the chip’s internal I/O ring.
Whatever the case there, Skylake does bring progress on other fronts, including tweakability. Intel says this CPU is its first ever to have features included expressly for overclocking. For one thing, the CPU’s base clock or BCLK has been liberated. The Asus Z170 Deluxe motherboard in my test rig offers options from 40 to 500MHz for BCLK speeds in 1MHz increments. Folks tweaking the BCLK don’t need to worry about the DMI and PCI Express ratios, either; a new PEG/DMI domain has its own isolated 100MHz clock. That should take a lot of the pain out of BCLK-based overclocking. Finally, Skylake’s memory controller supports a ton of different DDR4 speeds, up to 4133 MT/s in increments of 100 and 133MHz.
As a guy who just likes to overclock his stuff and not a dude with a professional overclocking career, I’m not sure what to make of these changes. I mean, I’m quite happy with the unlocked multipliers on K-series processors, which will let you squeeze a little extra out of a processor using conventional cooling. And I’m down with the memory clock flexibility. But those BCLK-oriented tuning knobs are probably meant for the folks with liquid nitrogen pots. I mean, I’m happy that they get these features, but I think it’s pretty obvious these capabilities aren’t for everyone. I seriously doubt Intel will allow lots of BCLK tuning leeway on the non-K variants of Skylake, for instance. That prospect seems very unlikely.
A new chipset: the Z170
One of the biggest changes from Skylake to Haswell happens at the platform level with the introduction of the new 100-series chipsets. By “chipset,” I mean “companion I/O chip,” since Intel’s platforms have consolidated things into a single chip for several generations.
Skylake’s companion chip is built using Intel’s 22-nm fab process, and the enthusiast-class variant of it is called the Z170. This chip supports 26 high-speed I/O ports, each one offering bandwidth equivalent to a single lane of PCI Express Gen3 connectivity. That’s a huge upgrade from the eight PCIe Gen2 lanes in the Z97 chipset. The Z170’s high-speed ports can be deployed by motherboard makers in various configurations. The possibilities include up to 20 PCIe Gen3 lanes, up to 10 USB 3.0 connections, and up to six SATA 6Gbps ports—though not all at the same time, since there are 26 high-speed ports in total.
Our Asus Z170 Deluxe is packed with slots and ports
The tremendous bandwidth available via the Z170 prompted Intel to upgrade the DMI link between the CPU and the chipset, as well. The new DMI 3 link offers bandwidth equivalent to four lanes of PCIe Gen3. That’s not nearly enough to sustain simultaneous I/O operations across all 26 of the Z170’s high-speed ports. Heck, it’s not even close, which is kind of a big deal since the system’s memory sits beyond that DMI link, hanging off of the CPU’s integrated memory controller. If this were a server architecture, I’d be worried about that fact. For typical desktop PC use, though, I suspect DMI’s new four-lane arrangement should generally be sufficient.
Although the Z170 doesn’t natively support the higher-bandwidth USB 3.1 standard, Intel offers a chip code-named Alpine Ridge that supplies both USB 3.1 and Thunderbolt capabilities, and several motherboard makers are adopting it. As a result, you can expect to see quite a few Skylake boards with USB 3.1 support. A subset of those should be qualified to work with Thunderbolt, as well.
Desktop Broadwell caches in
I mentioned the desktop variant of Broadwell earlier, and here is the fastest model: the Core i7-5775C. This is a four-core, eight-thread desktop processor with 6MB of L3 cache, and like all Broadwell chips, it’s built on Intel’s latest 14-nm process. With only a 65W TDP, the 5775C isn’t a speed demon; its base and peak clocks are 3.3 and 3.7GHz.
This poor devil has led a neglected life. Intel unveiled it at Computex in June, but we didn’t get a review sample for weeks. Then our review unit sat for a while in Damage Labs, untouched, while I labored away on Radeon Fury reviews. Meanwhile, these chips still aren’t broadly available in North America, and now the new hotness of Skylake has arrived.
Regardless, the 5775C is intriguing for several reasons. First, it drops into existing Z97 motherboards and could be an upgrade option for current Haswell owners. Second, it has built-in Iris Pro graphics, which are pretty potent as these things go. As a 65W chip with an integrated GPU, the 5775C could fulfill some unique missions. Last and definitely not least, the 5775C has 128MB of eDRAM situated in a separate chip on the CPU package. This eDRAM helps to accelerate the Iris Pro graphics core, but it’s also just a big frickin’ L4 cache for the entire CPU. Any application whose working data set will fit into a 128MB cache could stand to benefit from its presence. We saw hints of greatness from a similar chip with an eDRAM cache back when we reviewed the first Haswells, but we weren’t able to use it with a discrete GPU. The 5775C has no such limitations, and you may be surprised by its performance.
One qualifier, though: you will pay for the privilege of owning a 5775C. The current list price is $366, 27 bucks more than a Skylake 6700K. Assuming you can find one in stock.
Since we have a Haswell, Broadwell, and Skylake on hand, and since Windows 10 is out, I figured we should test a whole range of Intel processors against one another—so that’s just we did, spanning five generations back to Sandy Bridge.
Our testing methods
As usual, we ran each test at least three times, and we’ve reported the median result. Our test systems were configured like so:
|Processor||AMD FX-8370||Core i7-2600K||Core
Fatal1ty 990FX Killer
|Asus P8Z77-V Pro||Asus P8Z77-V Pro|
|Memory size||16 GB (2 DIMMs)||16 GB (2 DIMMs)||16 GB (2 DIMMs)|
|1333 MT/s||1600 MT/s|
|Memory timings||9-10-9-27 1T||8-8-8-20 1T||9-9-9-24
|Memory size||16 GB (2 DIMMs)||16 GB (2 DIMMs)||16 GB (4 DIMMs)|
|Memory speed||1600 MT/s||2133 MT/s||2133 MT/s|
They all shared the following common elements:
|Hard drive||Kingston HyperX SH103S3 240GB SSD|
GTX 980 4GB with GeForce 353.62 drivers
|Power supply||Corsair AX650|
Thanks to Corsair, Kingston, MSI, Asus, Gigabyte, Asrock, Cooler Master, Intel, and AMD for helping to outfit our test rigs with some of the finest hardware available.
Some further notes on our testing methods:
- The test systems’ Windows desktops were set at 1920×1080 in 32-bit color. Vertical refresh sync (vsync) was disabled in the graphics driver control panel.
- We used a Yokogawa WT210 digital power meter to capture power use over a span of time. The meter reads power use at the wall socket, so it incorporates power use from the entire system—the CPU, motherboard, memory, graphics solution, hard drives, and anything else plugged into the power supply unit. (The monitor was plugged into a separate outlet.) We measured how each of our test systems used power across a set time period, during which time we encoded a video with x264.
- After consulting with our readers, we’ve decided to enable Windows’ “Balanced” power profile for the bulk of our desktop processor tests, which means power-saving features like SpeedStep and Cool’n’Quiet are operating. (In the past, we only enabled these features for power consumption testing.) Our spot checks demonstrated to us that, typically, there’s no performance penalty for enabling these features on today’s CPUs. If there is a real-world penalty to enabling these features, well, we think that’s worthy of inclusion in our measurements, since the vast majority of desktop processors these days will spend their lives with these features enabled.
The tests and methods we employ are usually publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Memory subsystem performance
Since we have a new chip architecture and a new memory type on the bench, let’s take a look at some directed memory tests before moving on to real-world applications.
The fancy plot above mainly looks at cache bandwidth. This test is multithreaded, so the numbers you see show the combined bandwidth from all of the L1 and L2 caches on each CPU.
The most notable result above involves the comparison between the, uh, sky-blue Skylake 6700K line and the yellow Haswell 4790K one. At the 256KB to 1MB block sizes, where we’re accessing the four 256KB L2 caches in these chips, Skylake achieves substantially higher transfer rates at the same basic clock frequency as Haswell.
Oh, and get used to the Core i7-5960X taking the top spot in almost every benchmark. That CPU has eight cores, 16 threads, quad channels of DDR4, and a 20MB L3 cache. It also costs a grand. The 5960X is not in the same class as the rest of these chips. It’s just here for reference.
Now let’s zoom in on a portion of the graph above.
My main motivation for including this strange plot is to get you to consider the 64MB test block size. There, the purple line for the 5775C indicates higher bandwidth than any other CPU tested; the 5775C’s bandwidth at this block size is roughly double the 6700K’s. This data point is one spot where we can see the impact of the 5775C’s 128MB L4 cache. Ooh, ahh.
Interesting. You can see the impact of the 6700K’s higher-bandwidth DDR4 memory easily in Stream. That wasn’t the case with Haswell-E compared to Ivy-E. My suspicion is that Skylake may be more aggressive about speculatively pre-fetching data into its caches than prior architectures. That would explain its ability to take immediate advantage of DDR4’s additional bandwidth.
Looks to me like the Broadwell 5775C’s large L4 cache is a boon in AIDA’s memory copy test. The 5775C doesn’t look like anything special in isolated read or write tests, but when asked to do both, having that big cache on hand appears to help.
Next up, let’s look at access latencies.
SiSoft has a nice write-up of this latency testing tool, for those who are interested. We used the “in-page random” access pattern to reduce the impact of pre-fetchers on our measurements. This test isn’t multithreaded, so it’s a little easier to track which cache is being measured. If the block size is 32KB, you’re in the L1 cache. If it’s 64KB, you’re into the L2, and so on.
Despite the higher transfer rates of Skylake’s L2 cache, its access latencies have barely risen. L2 accesses on the Haswell 4790K take 11 cycles using this tool, and they take 12 cycles on the Skylake 6700K.
The move to DDR4 at 2133 MT/s carries only a slight penalty for the 6700K—it’s four nanoseconds slower than the 4790K with DDR3. That’s not bad at all, and I suspect that penalty could evaporate pretty quickly as DDR4 clock speeds ramp up.
Some quick synthetic math tests
The folks at FinalWire have built some interesting micro-benchmarks into their AIDA64 system analysis software. They’ve tweaked several of these tests to make use of new instructions on the latest processors. Of the results shown below, PhotoWorxx uses AVX2 (and falls back to AVX on Ivy Bridge, et al.), CPU Hash uses AVX (and XOP on Bulldozer/Piledriver), and FPU Julia and Mandel use AVX2 with FMA.
These quick tests give us a nice starting sense of how the Skylake-based 6700K may compare to its 4790K predecessor. In Photoworxx, the 6700K manages a pretty dramatic gain over the 4790K. The 5775C even gets in on the action, with the Broadwell chip taking the third spot ahead. However, the 6700K’s advantage over the 4790K grows slimmer as we move to different workloads. Skylake’s improvements in per-clock performance can be substantial in the right circumstances, but not every application will benefit equally. Some may not benefit much at all.
Power consumption and efficiency
The workload for this test is encoding a video with x264, based on a command ripped straight from the x264 benchmark you’ll see later.
Our Core i7-6700K-based test system draws a little more power under load than the 4790K-based equivalent, which is what we’d expect given these processors’ respective TDP ratings of 91W and 88W. Our 6700K system is relatively frugal at idle, too, although it’s nothing special there. The Core i7-5775C is in a class of its own on that front.
We can quantify efficiency by looking at the amount of power used, in kilojoules, during the entirety of our test period, when the chips are busy and at idle.
Perhaps our best measure of CPU power efficiency is task energy: the amount of energy used while encoding our video. This measure rewards CPUs for finishing the job sooner, but it doesn’t account for power draw at idle.
Even though it draws a little more power at peak, the 6700K-based system requires less energy to encode this video than the 4790K-based one. The difference between the two isn’t dramatic, but Intel’s architects appear to have succeeded in improving Skylake’s power efficiency over Haswell’s.
Project Cars is beautiful. I could race around Road America in a Formula C car for hours and be thoroughly entertained. In fact, that’s pretty much what I did in order to test.
Click the buttons above to cycle through the plots. We capture rendering times for every frame of animation so we can better understand the experience offered by each solution.
Whoa. That’s different.
So, first things first: the Skylake 6700K takes a clear lead over the Haswell 4790K with a markedly higher FPS average. The differences are a little smaller if you switch to a more advanced metric like our 99th-percentile frame time, but either way, Skylake beats Haswell cleanly.
Things get weird, though, with the Core i7-5775C in the picture. The Broadwell-based CPU with the 128MB L4 cache turns in the top performance in Project Cars, outdoing even Skylake. Looks like that big cache can help with gaming performance, even with a discrete GPU.
We can understand in-game animation fluidity even better by looking at the entire “tail” of the frame time distribution for each card, which illustrates what happens with the most difficult frames.
These “time spent beyond X” graphs are meant to show “badness,” those instances where animation may be less than fluid—or at least less than perfect. 16.7 ms correlates to 60 FPS, that golden mark that we’d like to achieve (or surpass) for each and every frame. And 8.3 ms corresponds to 120 FPS, an even more demanding standard that may become more popular with the growing popularity of PC displays with high refresh rates.
Someone told me Project Cars was seriously CPU-bound. Maybe that was true in older versions of Windows. Maybe Win10 has introduced some magic that changes the math. I dunno. What I do know is that even the slowest CPU here, the FX-8370, spends less than half a millisecond beyond our 16.7-ms threshold. In other words, that CPU pumps out frames at a constant 60Hz throughout almost the entire test run. The faster Intel processors aren’t too far from delivering a constant 120Hz.
The Witcher 3
Uh, sorry. This game has a lot of settings.
After testing this game quite a bit for recent GPU reviews, I expected it to be somewhat CPU-bound. I suppose it is, in the sense that we can show that faster CPUs seem to perform better in this game than slower CPUs. The 6700K looks strong here, and the 5775C’s magical gaming prowess continues. Still, all of the processors are doing a great job of producing smooth animation in conjunction with our GeForce GTX 980 graphics card. The FX-8370 produces 99% of all frames in 11.2 ms or less, roughly the equivalent of 90Hz. Every other CPU here is even faster.
Civilization: Beyond Earth
Since this game’s built-in benchmark simply spits out frame times, we were able to give it a full workup without having to resort to manual testing. That’s nice, since manual benchmarking of an RTS with zoom is kind of a nightmare.
The 6700K proves to be just a little slower than the 4790K in each of our metrics here—by an eyelash. Meanwhile, that crazy Broadwell 5775C embarrasses them both with the help of its beefy L4 cache.
Far Cry 4
You can see some minor spikes in the frame time plots above. I think, in this case, the 99th-percentile frame time does the best job of sorting out some closely matched contenders. The 4790K, 6700K, and 5775C are essentially tied at the top of the pack. They’re all exceptionally fast, nearly producing a constant 85Hz stream of frames. Then again, even with our advanced metrics helping tease out any differences, this game appears to be largely CPU-bound among the top Intel processors.
Dat 5775C, tho.
Shadow of Mordor
I had hoped this quick, FPS-only built-in benchmark from Shadow of Mordor would shed some new light on the contest between the CPUs. Instead, this result seems to be driving home the point that even many of today’s most demanding PC games simply are not meaningfully CPU-bound—provided you have a GeForce graphics card and a fast Intel quad-core CPU from the past four or five years. At this rate, we may have to start hunting explicitly for CPU-bound games and scenarios in order to stress test CPUs in the future.
Also, I may have made a mistake in switching to GeForce graphics cards on our CPU test rigs. My sense is that Radeons tend to be quite a bit more CPU-bound, especially when measured with advanced metrics.
Compiling code in GCC
Our resident developer, Bruno Ferreira, helped put together this code compiling test. Qtbench tests the time required to compile the QT SDK using the GCC compiler. The number of jobs dispatched by the Qtbench script is configurable, and we set the number of threads to match the hardware thread count for each CPU.
As we switch away from gaming, the 6700K turns in a dominating performance versus the other quad-core processors in the bunch. Is this result the start of a trend among our other productivity tests?
TrueCrypt disk encryption
TrueCrypt supports acceleration via Intel’s AES-NI instructions, so the encoding of the AES algorithm, in particular, should be very fast on the CPUs that support those instructions. We’ve also included results for another algorithm, Twofish, that isn’t accelerated via dedicated instructions. (Yes, the TrueCrypt project has fallen on hard times, but these results will come in handy later, as you’ll soon see.)
7-Zip file compression and decompression
The Skylake 6700K is the fastest Intel quad-core CPU pretty consistently, with the lone exception of the 7-Zip decompression test. Generally, though, it’s only a smidgen quicker than the Haswell-based 4790K. The 5775C performs respectably in these productivity tests, but it doesn’t continue the surprising excellence we saw in our gaming tests.
x264 HD video encoding
Our x264 test involves one of the latest builds of the encoder with AVX2 and FMA support. To test, we encoded a one-minute, 1080p .m2ts video using the following options:
–profile high –preset medium –crf 18 –video-filter resize:1280,720 –force-cfr
The source video was obtained from a repository of stock videos on this website. We used the Samsung Earth from Above clip.
Handbrake HD video encoding
Our Handbrake test transcodes a two-and-a-half-minute 1080p H.264 source video into a smaller format defined by the program’s “iPhone & iPod Touch” preset.
One of the most notable outcomes in our video encoding tests is simply that the eight-core 5960X performs so well. Last time we checked, x264 didn’t benefit much from having 8 cores and sixteen hardware threads on tap. This latest build certainly does.
The 6700K outperforms the other quad-core processors here, too.
The Panorama Factory photo stitching
The Panorama Factory handles an increasingly popular image processing task: joining together multiple images to create a wide-aspect panorama. This task can require lots of memory and can be computationally intensive, so The Panorama Factory comes in a 64-bit version that’s widely multithreaded. I asked it to join four pictures, each eight megapixels, into a glorious panorama of the interior of Damage Labs.
Sometimes, Skylake is clearly faster than Haswell. Other times, well, this happens.
Because LuxMark uses OpenCL, we can use it to test both GPU and CPU performance—and even to compare performance across different processor types. OpenCL code is by nature parallelized and relies on a real-time compiler, so it should adapt well to new instructions. For instance, Intel and AMD offer integrated client drivers for OpenCL on x86 processors, and they both support AVX. The AMD APP driver even supports Bulldozer’s and Piledriver’s distinctive instructions, FMA4 and XOP. We’ve used the AMD APP ICD on the FX-8370 and Intel’s latest OpenCL ICD on the rest of the processors.
We tested with LuxMark 3.0 using the “Hotel lobby” scene.
We’ll start with CPU-only results.
The 6700K manages a nice gain over the 4790K here. Now, if we switch to only using the GPU to render and just letting the CPU feed it, here’s what happens.
Chaos, mostly, but with higher scores thanks to the GPU’s number-crunching prowess.
We can try combining CPU and GPU computing power by asking both processor types to work on the same problem at once.
Sharing the load between the two processor types leads to the highest overall scores. It also erases the 6700K’s advantage over the 4790K from the CPU-only test, oddly enough.
The Cinebench benchmark is based on Maxon’s Cinema 4D rendering engine. This test runs with just a single thread and then with as many threads as CPU cores (or threads, in CPUs with multiple hardware threads per core) are available.
The Skylake 6700K achieves the highest per-thread performance of any of the CPUs tested. It’s roughly 80% faster than in the single-threaded test than AMD’s FX-8370, which is either awesome or depressing. Maybe a little of both. With all eight threads active, the 6700K only slightly outperforms the 4790K overall.
No surprises in POV-Ray. The 6700K continues to be just a bit faster than its predecessor.
STARS Euler3d computational fluid dynamics
Euler3D tackles the difficult problem of simulating fluid dynamics. Like MyriMatch, it tends to be very memory-bandwidth intensive. You can read more about it right here.
Probably thanks to its higher-bandwidth L2 cache and DRAM, the 6700K easily outruns the 4790K in this fluid dynamics simulation. The real star of this show is the Broadwell 5775C, though, whose L4 cache finally helps out rather dramatically in something other than a game. The Haswell equivalent also performed well in this test.
Many of you have asked for broader comparisons with older CPUs, so you can understand what sort of improvements to expect when upgrading from an older system. We can’t always re-test every CPU from one iteration of our test suite to the next, but there are some commonalities that carry over from generation to generation. We might as well try some inter-generational mash-ups.
Now, these comparisons won’t be as exact and pristine as our other scores. Our new test systems run Windows 10 instead of Windows 8.1 and 7, for instance, and have higher-density RAM and larger SSDs. We’re using some slightly different versions of POV-Ray and 7-Zip, too. Still, scores in the benchmarks we selected shouldn’t vary too much based on those factors, so… let’s do this.
Our first set of mash-up results comes from our last two generations of CPU test suites, as embodied in our FX-8350 review from the fall of 2012 and our original desktop Haswell review and our Haswell-EP review from last year. This set will take us back at least four generations for both Intel and AMD, spanning a price range from under $100 to $1K.
Aw, we can do better than that. Turn the page.
Legacy comparisons, continued
That was a nice start, but we can go broader. This next set of results includes fewer common benchmarks, but it takes us as far back as the Core 2 Duo and, yes, a chip derived from the Pentium 4: the Pentium Extreme Edition 840. Also present: dual-core versions of low-power CPUs from both Intel and AMD, the Atom D525 and the E-350 APU. We retired this original test suite after the 3960X review in the fall of 2011. We’ve now mashed it up with results from our first desktop Haswell review, our Haswell-EP review, and from today.
Never forget: in April of 2001, the Pentium III 800 rendered this same “chess2” POV-Ray scene in just under 24 minutes.
Let’s summarize our results with a couple of our famous power-performance scatter plots. The first one is based on a geometric mean of all of our non-gaming application tests, and the second one focuses on our frame-time-based game performance results. As ever, the best values will tend toward the top left corner of the plot. Since they’re discontinued products, I’ve used the original introductory prices for the 2600K and 3770K in order to provide context.
Intel’s annual parade of incremental progress marches on. For non-gaming applications, the Skylake 6700K offers a slight boost in performance overall compared to last year’s model, the Haswell 4790K. What the summary plot can’t tell you is that this advantage is sometimes fairly substantial, while other times the 6700K is no faster than its predecessor. The plot also neglects to mention that the gains from Haswell to Skylake come in my favorite flavor: per-thread performance, which is the most important and most difficult sort of CPU performance to achieve. I’m always happy to see improvements on this front, even if they’re relatively modest.
The gaming plot tells a similar story, but here, the 6700K is in the running for the fastest gaming CPU on the planet—and it would’ve won, too, if it weren’t for the pesky Broadwell 5775C and its magic L4 cache. The 6700K improves on the 4790K by a tad, but the 5775C upstages it with a freakish string of gaming performance wins, even though its prevailing clock speed is ~500MHz lower.
What should we make of these results? If you’ve been nursing along a system based on a Sandy Bridge processor like the Core i5-2500K or Core i7-2600K, then perhaps Skylake has enough to offer to prompt an upgrade. Cumulatively, Intel has made quite a bit of progress in the past several years—and the Skylake platform with the Z170 chipset is a considerable upgrade in terms of I/O bandwidth, too. Those motherboards bristle with storage options and high-speed USB ports and such. So there’s that. What’s jarring about our gaming results is that the Sandy Bridge-based 2600K remains a very competent processor for running the current PC games we tested. You’ll probably want to avoid thinking about that fact when it comes time to pull out the credit card.
One thing I haven’t addressed yet is the Core i7-6700K’s overclocking potential. Sorry, but I simply ran out of time after testing seven different processors and making this review happen in the short span of time since the Windows 10 launch. I’ve heard that some folks are reaching as speeds as high as 5GHz with Skylake parts, while others are reporting clock speeds around 4.7-4.8GHz, similar to what we’ve seen from most Haswells. Mark’s 6700K topped out at 4.6GHz aboard the Asus Z170-A. At the end of the day, if only a few hundred megahertz are at stake between Haswell and Skylake, well, that ain’t much in the grand scheme. Then again, we need to explore the bandwidth potential of higher DDR4 memory frequencies, too. We’ll have to overclock our Skylake chip soon and write it up in order to provide another data point.
I figure we should overclock the 5775C, as well, to see what it can do. Heck, if you’re a gamer sporting a Haswell-compatible motherboard and looking for an upgrade, this little desktop Broadwell may be a better choice than the 6700K. So long as your motherboard is Broadwell-compatible via a BIOS upgrade, the 5775C could deliver gaming performance that’s superior to Skylake, provided your games of choice benefit as much from that L4 cache as the ones we tested did.
Enjoy our work? Pay what you want to subscribe and support us.