|Introduction and Specifications|
We've been talking about Intel's 4th Generation Core processor technology, codenamed Haswell, for over two years now. If you've been hanging around these parts and are like most self-respecting geeks, you've likely been feeling the build-up of Haswell for quite some time. Intel has disclosed lots of information on their latest processor technologies, from its 22nm 3D Tri-Gate transistors, to its beefed-up integrated graphics core. Sure, there's marketing hype and press releases but at some point you've got to have execution -- and today is all about execution for Intel.
Intel is officially launching Haswell today, their 4th Generation Core processor technology for the desktop market, with our follow-on coverage of dual core and mobile Haswell offerings coming up quickly on June 3rd. Haswell marks a number of firsts for Intel, including leading-edge technologies like on-chip voltage regulators, eDRAM, and their first graphics engine that fully supports the DX11.1 specification and OpenGL 4.0. With these new features alone, Haswell is primed to offer significant performance and feature enhancements, the likes of which should appeal first and foremost to mainstream consumers; but there's also plenty brought to the table for the performance enthusiast here as well. First, let's dig into the specifications of Intel's Haswell desktop variant and then we'll dive deeper into its underlying technologies going beyond just high level speeds and feeds.
As we noted earlier, Haswell is the codename of a brand new microarchitecture that will be foundation of an entire line of desktop, server, and mobile processors in 2013 and beyond. It is a “Tock” in Intel’s "Tick-Tock" CPU release cadence, which is to say it is a major micro architectural update and not a mild revision or shrink of an existing design.
Above we have a die shot of a Haswell Core processor, along with the features and specifications of the new Core i7-4770K processor we’ll be showcasing here. Some of the features will sound similar to previous Intel processor offerings, but make no mistake, these are completely new chips. They may leverage many technologies from previous Intel CPU microarchitectures, but with Haswell, Intel has redesigned virtually all of the execution engines, integrated a much more powerful graphics core, and built in support for new extensions that could drastically improve performance.
We've covered a number of the features inherent to Haswell and its companion 8-series chipset, so we won't be going in-depth again here. Intel's 8-Series chipsets will offer Intel Smart Response Technology (SRT) and Rapid Storage Technology (RST) which we covered in our Z68 launch piece. Our coverage of Intel's 22nm Tri-Gate transistor announcement gives some detail on the manufacturing process technology used with Ivy Bridge and the various Haswell labeled stories cover many of the details Intel has released in the months leading up to today's launch.
Intel’s 4th Generation Core processors featuring the Haswell microarchitecture look similar to the Sandy Bridge and Ivy Bridge processors that came before, but these new chips require a new socket, LGA 1150. Along with the new socket, Haswell-based 4th Generation Core processors require motherboards built around Intel’s 8-Series chipsets as well, but we should note that heatisnks compatible with socket LGA 1155/1156 will also work with LGA 1150 processors -- which is a good thing for the average consumer and OEMs alike.
The processor you see pictured here will be Intel’s flagship 4th Generation Core processor at launch, the Core i7-4770K. We’ll have more details on this particular processor on the pages ahead, but to quickly summarize, the Core i7-4770K is a quad-core part with a base clock of 3.5GHz and maximum Turbo clock of 3.9GHz. It features Intel HD 4600 series on-die graphics (formerly known as GT2), and since it is a K-SKU, it is fully unlocked. Along with the Core i7-4770K, Intel is launching a bevy of other processors for various form factors and price points. We’ll have more information on those other processors a little later on as well.
Underneath their integrated heat-spreader, quad-core 4th Generation Core processors sport a rectangular die of roughly 177mm2. The processors are manufactured using Intel’s advanced 22nm process node with tri-gate transistors and they feature more powerful on-die graphics than before, a new CPU microarchitecture, and number of other tweaks and enhancements as well. Haswell supports new power states and supports connected standby. It also features new AVX 2.0 instructions that can improve floating point performance with some workloads by 2X. Haswell also has a fully integrated voltage regulator as well, which is actually part of the CPU die itself. The fully integrated voltage regulator, or FIVR as it is known, frees up motherboard area and saves OEM some costs too, but FIVR also results in other benefits like more dynamic power arrangements and increased efficiency (FIVR is on all Haswell processors, not just mobile parts).
The Haswell microarchitecture leverages technology from previous generation Intel processors, but much of the CPU (and the platform) has been updated to increase performance and power efficiency, and to ensure good scaling. Intel has stated that Haswell scales so well, it will find its way into a wide range of form factors, ranging from Tablet PCs to big-iron servers. The key pipelines in the processor remain largely unchanged over Ivy Bridge, though Haswell features improved code fetch and better brand prediction and more Out-Of-Order resources. There are two additional dispatch ports in the processors, a larger L2 TLB (translation look-aside buffer), and the bandwidth to the processor’s cache has been doubled.
As we’ve mentioned, Haswell-based 4th Generation Core processors also feature Intel’s Advanced Vector Extensions 2.0, or AVX 2. AVX 2 essentially doubles up on the AVX instructions first introduced with Sandy Bridge. AVX 2.0 supports 256-bit integer vectors and adds hardware FMA units to the core. Haswell can perform double the number of single and double-precision FLOPs as Sandy Bridge and AXV 2 adds new integer and gather instructions as well. That double-bandwidth cache we mentioned earlier was introduced to keep the wider vector units in Haswell fed.
Another major addition coming with some Haswell-based 4th Generation Core processors in an on-package eDRAM cache. This 128MB eDRAM cache is one of the ways Intel was able to improve the performance of its top-end Iris Pro GT3e graphics engine, but it’s actually a fully coherent L4 cache that can be shared between the CPU and GPU. The eDRAM cache is co-located on the CPU package (it’s not part of the die) and offers 50+GB/s of read/write bandwidth. We’re told the eDRAM cache at idle uses only about 1/2w to 1w of power, and up to 3.5w to 4.5w when it’s fully utilized, but it can also be powered down when not in use. As of today though, none of Intel's socketed 4th Gen processors will feature this eDRAM--it's currently reserved for premium mobile parts.
|Haswell Microarchitecture - FIVR|
There's little question that current-generation microprocessors from Intel (and others) are complex designs that require highly-tuned, well-regulated power supplies of various voltages for multiple component blocks, along with both active and idle or standby power states. CPU cores, graphics cores, the DDR memory interface and PCI Express lanes, all need their own power planes, many of which are required to dynamically scale with varying workloads.
Within the Haswell architecture, Intel's 22nm 3D Tri-Gate 3D transistor technology certainly helps drive lower power consumption in general but Haswell also has a few other tricks up it sleeve to further reduce power consumption. Specifically, Haswell is the first X86 processor to incorporate on-die voltage regulators, which in turn allows the chip to reduce the number of VREG inputs from five to one simple input.
In this simplified block diagram, Intel is showing Haswell's new FIVR, or Full Integrated Voltage Regulator, technology. As it's shown here, Processor, Graphics, System Agent, IO and PLL (Phase Lock Loop / Clock) Voltage Regulators are now combined into a single input VR. Intel has kept its DDR memory VREG discrete, so it can migrate with industry standard changes as needed. Regardless, at a high level, Haswell's new FIVR power design allows for a dramatic reduction of complexity in motherboard design. All those Voltage Regulator Modules are now on-chip. However, the intrinsic benefits of an integrated Voltage Regulator design go far beyond just board real estate savings.
Haswell's integrated Voltage Regulator design also allows the chip to ramp voltages faster -- about 5X to 10X faster than the previous generation's external VR design. Haswell's FIVR array actually operates at 125MHz. As a result, Haswell is able to enter deeper sleep states and come out of those sleeps states more quickly. This may not result in as dramatic a power consumption on the desktop but Intel is claiming up to 50 percent longer battery life for Haswell notebooks, versus the previous generation of Ivy Bridge machines.
|Intel HD and Iris Graphics and Media Encoding|
|With the release of the Haswell architecture, Intel's new integrated graphics cores will be branded with the “Iris” moniker. Now, we’ll see Intel HD 5000 graphics along with the higher-end Intel Iris Graphics 5100 and Iris Pro Graphics 5200; previously, these were codenamed GT3 (15W), GT3 (28W), and GT3e, respectively. Iris graphics are primarily designed for mobile devices, and we’ll be seeing them in Intel-based ultrabooks, higher-end notebooks, and desktop all-in-one devices. Upcoming Haswell desktops will still be branded as "Intel HD Graphics." High-end desktop parts will carry the "HD 4600" label, with lower-end parts branded as "HD 4400" and "HD 4200."
Iris brings new features including integrated on-package eDRAM memory for the GT3e graphics core, DX11.1, OpenGL 4.0, and OpenCL 1.2 support, double the bandwidth with DisplayPort 1.2, support for a 3-screen collage display, faster Intel Quick Sync Video performance and fast JPEG decode and MPEG encode. Of note is the GT3e's eDRAM (embedded DRAM) cache complement that dramatically increases bandwidth and lowers latency with much faster access for the graphics and CPU cores. Again, this is a feature for mobile Iris Graphics variants only, however.
Intel is claiming overall performance improvements of up to 2X for Iris Graphics, over previous 3rd Gen Ivy Bridge integrated graphics. The company also boasts that Iris will bring triple the 3D performance for desktop R-series chips. The R-series chips are the BGA-only (non-socketable) variants -- conventional desktop parts will be confined to GT2-class graphics.
Source: Intel - Iris Graphics Desktop Performance
Source: Intel - Iris Graphics Ultrabook Performance
Intel will also offer Iris switchable, hybrid graphics implementations with NVIDIA and AMD GPUs, in systems with a discrete graphics card. Iris will also supports 4k UltraHD display output.
We'll be stepping through benchmark numbers of the desktop Intel HD 4600 series devices in the pages ahead but you'll have to wait for our Haswell mobile coverage to get a glimpse of Iris graphics in action, along with our direct confirmation of Intel's performance claims you see above. In the meantime, in case you missed it, you can feast your eyes our quick side-by-side demo of Haswell GT3 Iris graphics in action versus NVIDIA discrete graphics, that we shot earlier this year at CES 2013.
Side-by-side Haswell graphics vs. NVIDIA demo
So, as you can see, Haswell definitely has the chops to at least hang with discrete graphics in Dirt 3 but we'll have more detailed numbers for you in the days ahead as we get Iris Graphics-powered machines in house for testing.
|4th Gen Intel Core Processor Lineup|
As you probably expect, Intel’s got a host of Haswell-based 4th Generation Core processors coming down the pipe. We’ll have more new to share with you soon regarding mobile and dual-core parts, but here’s a breakdown of the desktop quad-core parts that are on the immediate horizon.
This first batch of Core i5 processors all offer the same max Turbo frequency of 3.6GHz and feature the same on-die Intel HD 4600 series graphics as the Core i7-4770K we’ll be showing you here. Note, however, that the maximum graphics frequency of these Core i5 parts of 1150MHz is a full 100MHz below the 4770K. HyperThreading is not supported on the true quad-core chips, but it is on the dual-core Core i5-4570T (far right). TDPs range from 35w to 84w.
This next batch of Core i5 processors are all quad-core parts, with the same cache and graphics configurations. The only differences between these chips are their base and max Turbo frequencies and their test TDPs. The K-SKU is also fully unlocked for easier and more flexible overclocking. These processors also feature Intel HD 4600 series graphics, but with a maximum GPU frequency of 1200MHz.
And here we have Intel’s 4th Generation Core i7 processor line-up at launch. All of these processors are quad-core parts with support for HyperThreading. The Core i7-4770K is the enthusiast targeted product with unlocked multipliers, and for some inexplicable reason, vPro features disabled. Also note that the Core i7-4770R is the only chip here with Iris-branded graphics. The Core i7-4770R features Intel Iris Pro 5200 series graphics with a max GPU frequency of 1300MHz. The rest of the i7 desktop processors have Intel HD 4600 series graphics with support for max frequencies in the 1200MHz -1250MHz.
Intel will also be offering a single-chip 4th Generation Core processor with the CPU and PCH integrated into single BGA package. The 4th Gen Core Y-Processor line as it’ll be known will include 15W and 28W TDPs, with S0ix support, and supports LPDDR3 and DDR3L memory. These chips are designed for Ultra-Thin devices smaller form factors.
|Z87 Chipset and Intel DZ87KLT-75K|
We tested the Core i7-4770K on a quartet of Intel Z87 Express based motherboards from Intel, ASUS, MSI, and Gigabyte. The Z87 Express chipset was designed to be the companion to Intel’s high-end Haswell-based desktop processors, but there are a slew of other 8-series chipsets planned as well. The main differences between the various 8-series chipsets is in their support for things like overclocking, Intel Smart Response Technology, and management and vPro-related features. Some of the chipsets also offer support for flexible PCI Express lane configurations and vary in their number of available SATA 6.0Gbs and USB 3.0 ports.
Since we’re focusing on the Core i7-4770K here, we’ll discuss the Z87 Express a bit further. The high-level block diagram above gives a good visual representation of the Z87 Express chipset’s main features. Like the previous-gen Z77 (and a few generations to come before it), the new Z87 Express chipset is essentially an I/O hub, as all of the traditional Northbridge functionality previously found in a Northbridge chip has been integrated into the processor itself.
The processors are linked to the chipset via Intel’s FDI (Flexible Display Interface) and 20Gb/s DMI 2.0 interface. The chipset itself is outfitted with 8 more PCIe 2.0 lanes, along with various other I/O, like six SATA ports and an integrated Gigabit MAC. You’ll also find native USB 3.0 support, with up to six USB 3.0 and fourteen USB 2.0 ports built in. The chipset has support for Intel’s Rapid Storage Technology (RST) as well, along with RAID, Smart Response Technology, or SRT, and other features like FastBoot and HD Audio.
Intel’s top-of-the-line Z87 Express-based motherboard is the DZ87KLT-75K, pictured here. The DZ87KLT-75K is a member of Intel’s Extreme Series of motherboards, and as such, features the black and dark blue color scheme of its predecessors and a full complement of overclocking capabilities.
The DZ87KLT-75K also supports Extreme Memory Profiles for easy high-speed memory configuration and even includes an extensive set of overclocking tools via an excellent UEFI Intel is calling their Visual BIOS. We should point out that those overclocking tools are not only available via the UEFI, but also through a Windows-based application called the Intel Extreme Tuning Utility, or Intel XTU. XTU gives users the ability to tweak numerous performance-related options and monitors system temperatures and fan speeds. “One Touch” overclocking options are also available via a simple slider for those that want a quick and easy speed boost, without doing much tweaking.
|Z87 Motherboards: MSI|
Next up we have a couple of boards from MSI, the Z87-GD65 Gaming and the Z87 M-POWER MAX. While we picture two MSI motherboards here, we should note that we used only the Z87-GD65 for all of the MSI benchmarks you’ll see a little later in this article. Since we had the M-POWER MAX on hand though, we thought many of you would like to see the board as well.
Like the other Z87 Express-based motherboards we’ll be featuring here, all of the Z87 chipset’s features are available on the MSI Z87-GD65, but MSI works a bit of their own magic as well. The MSI Z87-GD65 features a mouse-friendly that’s quick and easy to navigate and offers a full complement of overclocking and performance-related options as well. The board is also in MSI’s “Military Class 7” family and features super ferrite chokes, highly conductive polymer capacitors and / or solid capacitors throughout, which should offer increased stability and longevity, as well as lower total power consumption. The Z87-GD65 also has a digital PWM for more efficient power delivery.
The MSI Z87-GD65 also supports “1 second overclocking” thanks to its OC Genie 4 feature and button, but the UEFI sports all of the overclocker-friendly features we’ve come to expect from MSI for manual tweaking as well. MSI also includes their own Control Center software which gives users the ability the monitor and control system parameters from within Windows, with no need to reboot when making changes, and the board's got voltage check points too.
The MSI Z87 M-POWER MAX incorporates virtually all of the features on the Z87-GD65 above, and then some. The first thing you’ll notice about the board is that is sports a completely different color scheme. This puppy is black and yellow and is outfitted with more elaborate PWM cooling. Before we talk about some of the more concrete features of the Z87 M-POWER MAX though, we want to point out something unique MSI is going with these boards. MSI Z87 M-POWER MAX motherboards all go through a 24-hour Prime95 burn-in test, using an overclocked, liquid-cooled CPU with no other cooling. MSI is taking this extra step to ensure the boards perform well and are stable while overclocked.
As we’ve mentioned, the Z87 M-POWER MAX has a similar feature set to the Z87-GD65 that includes Audio Boost, Sound Blaster Cinema Audio, Killer E2200 networking, OC Genie 4 with easy OC Buttons, Military Class 4 components, an all digital PWM, etc. But the Z87 M-POWER MAX incorporates a built-in WiFi / BlueTooth module that supports Intel’s WiDi Wireless Display technology as well.
|Z87 Motherboards: ASUS and Gigabyte|
Next up, we have the Asus Z87-Deluxe. The Asus Z87-Deluxe is one of the more feature rich motherboards we’ve tested, both in terms of both its hardware and software / UEFI. Asus clearly puts a ton of effort into UEFIs on their boards and it shows with the Z87-Deluxe. This board also features Asus’ new black and gold color scheme, which has received somewhat of a mixed reaction from enthusiasts.
In addition to exploiting all of the features inherent to the Z87 Express chipset, the Asus Z87-Deluxe offers Bluetooth and WiFi connectivity, additional SATA 6Gbps ports, and what Asus calls its “Dual Intelligent Processors 4”. The Dual Intelligent Processors consist of Asus’ EPU unit, which we’ve covered in the past, and the TPU processing unit. The processors work together with the Asus Z87-Deluxe’s programmable 16+2 power phase digital VRM (DIGI+ VRM) and give users the ability to monitor and adjust power delivery. According to Asus, the combination of the programmable digital VRM and Dual Intelligent Processors 4 results in superior power efficiency and improved longevity over less sophisticated motherboards.
The Asus Z87-Deluxe is also outfitted with a PLX PCI Express 3.0 switch, which give the board to the ability to have all of its PCIe expansion slots enabled with full bandwidth at all times, regardless of the CPU installed. In addition to the aforementioned items, the Asus Z87-Deluxe has quite a few other interesting features as well. For example, you can flash this board’s BIOS even if there is no CPU or memory installed via its “USB BIOS Flashback” feature. It also supports the USB Attached SCSI Protocol, which can be enabled via Asus’ AI Suite software, to boost USB 3.0 throughout significantly.
And here we have the Gigabyte Z87X-UD3H. Like the other motherboards featured here, the Z87X-UD3H is an enthusiast class, Z87 Express-based motherboard for socket 1150 Haswell-based 4th Gen Core processors, though the Z87X-UD3H isn’t quite as loaded up with the numerous bells and whistles of the others.
In terms of more tangible features, the Gigabyte Z87X-UD3H offers a total of 16 USB ports (10 USB 3.0 and six USB 2.0), eight SATA 6Gbps connectors, CrossFireX and SLI support, DualBIOS technology, and On/Off Charge support. The Z87X-UD3H also offers every display output option available (VGA, DVI, HDMI, and DP).
|Haswell Memory and Coolers|
Anytime Intel launches a new family of processors, component makers usually launch a fresh line of accessories to coincident with the processors’ release. In the leadup to this review, we got our hands on some new memory from G.SKILL qualified for use with Haswell and some nice coolers from Noctua as well.
Like previous gen Sandy and Ivy Bridge processor, Haswell also has support for the latest XMP (Extreme Memory Profiles) code, for easy memory optimization and overclocking. You can expect an influx of fresh memory kits to arrive alongside Haswell, like the G.SKILL kit we have here. The G.SKILL Trident X memory kit pictured above consists of two, 4GB, DDR3-2666 sticks of memory (total 8GB), model number F3-2666C11D-8GTXD. The sticks run at 1.65v with CL11-13-13-35 timings, and are XMP compatible. A dual-fan active cooler is also included with the kit, which snaps right over the DIMM retention clips on most motherboards.
The Noctua NH-U12S and NH-U14S coolers you see here were released previously, but will work with Haswell-based 4th Generation Intel Core processors. These coolers feature a similar design with a dense array of aluminum fins, linked to a copper heatplate via multiple heatpipes.
The heatsinks are nickel plated and include fan with self-stabilizing, oil-pressure bearings, which provide long life and quiet operation. These coolers can also be arranged with dual-fans in a push-pull configuration if you so choose. The major difference between these two coolers is simply their size. The U12S works with 120mm fans, while the U14S is larger and works with 140mm fans.
|Vital Signs and Overclocking|
Below are a couple of screenshots from the latest version of CPU-Z that detail some of the new Core i7-4770K processor’s inner-workings.
As you can see, the Core i7-4770K utilizes the new socket 1150, it's built using Intel’s 22nm process node, and has a TDP (Thermal Design Power) rating of just 84w. The chip we tested, and that will hit retail shelves, is stepping 3 and it features a stock 100MHz BCLK, which is incorrectly labeled bus speed in the first image above.
Like the Ivy Bridge-based third generation Intel Core processors, new Haswell-based processors offer limited flexibility when overclocking via BCLK manipulation. If you want to tweak CPU and memory frequencies via the BCLK, it can only be increased by a few MHz at a time. Intel claims about 4-%-7% in either direction, but that’s just a rough guideline--we've seen some go higher and some go lower. With Haswell, however, additional BCLK multipliers or straps are also available. So, while you only have a few MHz in each direction to change the BCLK, the additional straps--which will be available on K-SKUs only--give users the ability to set the BCLK to 100MHz, 125MHz, 167MHz, or 250MHz.
Also like previous-gen Intel Core processors, different product families based on Haswell have differing overclocking options enabled. K-series SKUs, like the Core i7-4770K we’re featuring here, are unlocked for easy overclocking via multiplier or BCLK strap manipulation. But non-K SKUs will have “limited” unlocking which will allow for multipliers to be increased by up to 4 speed bins above the processor’s peak, official Turbo frequency. And processors that don’t offer Turbo, can only be overclocked by altering the BCLK. Now, while there’s not nearly as much BCLK headroom as previous architectures, that doesn’t mean it can’t be altered at all. As we’ve mentioned, there’s a bit of wiggle room there. As far as the other overclocking related features go, 4th Gen Core processors also support the latest XMP memory reference code.
We were actually able to boot into Windows and run benchmarks at clocks above 4.6GHz, but our CPU was breaking the 90’C mark, throttling to keep temperatures in check, and almost immediately throwing a blue screen. We should also mention that at the speeds we hit, with the stock cooler, temperatures were not an issue. While overclocked using a stock cooler, the 4770K idled in the high 40'C range and peaked in the low 80'C range. With such a small die though, thermal density will likely be an issue for overclockers. If you plan to overclock a Haswell chip, despite their relatively low-power consumption characteristics, you'll need a highly efficient cooler to pull heat away from the chip as fast as possible. We should also note that 4th Gen Core processors are using the same TIM between its integrated heat spreader and the surface of the chip as Ivy Bridge, which isn't terribly high performing.
While we had the chip overclocked, we ran the multi-threaded Cinebench R11.5 benchmark to illustrate the kind of performance gains possible and were able to increase performance from the Core i7-4770K’s stock score of 7.91 to an impressive 8.77. The higher clocks we were able to achieve on the overclocked Core i7-4770K almost allowed it to overcome the two-core advantage of the Core i7-990X, but not quite.
We also monitored out test system's power consumption while we had the Core i7-4770K overclocked and saw a relatively small increase in power consumption. The additional 500MHz over stock we were able to achieve resulted in an increase in power consumption of about 22W.
|Test Setup and SiSoft SANDRA|
Test System Configuration Notes: When configuring our test systems for this article, we first entered their respective system BIOSes and set each board to its "Optimized" or "High performance Defaults". We then saved the settings, re-entered the BIOS and ensured the memory speed to DDR3-1600, or DDR3-1333 on the Intel systems and DDR3-1866 on the AMD systems (their maximum, officially supported speeds).
The solid state drives in the test systems were then formatted, and Windows 7 Ultimate x64 was installed. When the Windows installation was complete, we fully updated the OS, and installed the drivers necessary for our components. Auto-Updating and Windows Defender were then disabled and we installed all of our benchmarking software, performed a disk clean-up, cleared any prefetch and temp data, and ran the tests.
We began our testing with SiSoftware's SANDRA 2013, the System ANalyzer, Diagnostic and Reporting Assistant. We ran four of the built-in subsystem tests that partially comprise the SANDRA 2013 suite with Intel's new Core i7-4770K processor (CPU Arithmetic, Multimedia, Memory Bandwidth, and Cache and Memory). All of the scores reported below were taken with the processor running at its default clock speed of 3.5GHz (3.9GHz Turbo) with 8GB of DDR3-1600 RAM running in dual-channel mode on the Intel DZ87KLT-75K motherboard.
|Futuremark PCMark 7 Tests|
Futuremark's PCMark 7 is the latest version of the PCMark whole-system benchmarking suite. It has updated application performance measurements targeted for a Windows 7 environment and uses newer metrics to gauge relative performance. Below is what Futuremark says is incorporated into the base PCMark suite and the Entertainment, Creativity, and Productivity suites--the four modules we have benchmark scores for you here.
The PCMark test is a collection of workloads that measure system performance during typical desktop usage. This is the most important test since it returns the official PCMark score for the system
From this point forward in this review, we'll not only be comparing the Core i7-4770K's performance to a handful of other high-end processors, but will also be comparing the performance of a quartet of motherboards. As we mentioned in the test setup section on the previous page, the Intel board used for testing was the DZ87KLT-75K, the Asus board is the Z87-Deluxe, the MSI board is the Z87-GD65, and the Gigabyte board is the Z87X-UD3H.
|LAME MT and SunSpider|
In our custom LAME MT MP3 encoding test, we convert a large WAV file to the MP3 format, which is a popular scenario that many end users work with on a day-to-day basis to provide portability and storage of their digital audio content. LAME is an open-source MP3 audio encoder that is used widely in a multitude of third party applications.
In this test, we created our own 223MB WAV file (a hallucinogenically-induced Grateful Dead jam) and converted it to the MP3 format using the multi-thread capable LAME MT application, in both single and multi-thread modes. Processing times are recorded below, listed in seconds. Shorter times equate to better performance.
Our custom LAME MT benchmark uses only one or two threads. As a result, the Core i7-4770K's ability to Turbo to higher frequencies than the other processors, in addition to the slight IPC enhancements of the architecture, allow it to outpace all of our reference systems. All of the motherboards performed similarly, however.
All of the systems were tested using the latest version of Internet Explorer 9, with default browser settings, on a clean install of Windows 7 Ultimate x64.
All of the Core i7-4770K powered systems performed similarly in the SunSpider benchmark, tightly grouped at the head of the pack.
|Cinebench R11.5 and POV-Ray|
Cinebench R11.5 is a 3D rendering performance test based on Cinema 4D from Maxon. Cinema 4D is a 3D rendering and animation suite used by animation houses and producers like Sony Animation and many others. It's very demanding of processor resources and is an excellent gauge of pure computational throughput.
This is a multi-threaded, multi-processor aware benchmark that renders a photorealistic 3D scene (from the viral "No Keyframes" animation by AixSponza). This scene makes use of various algorithms to stress all available processor cores. The rate at which each test system was able to render the entire scene is represented in the graph below.
The new Core i7-4770K put up the best scores of any quad-core processor in the multi-threaded benchmark, but couldn't quite keep pace with the hex-core Intel CPUs. In the single-threaded test, however, the Core i7-4770K was the fastest of the bunch, due to its architectural enhancements and high Turbo frequency. All of the motherboards we tested, once again, performed similarly, though the third-party boards all finished ahead of the Intel-built board.
POV-Ray, or the Persistence of Vision Ray-Tracer, is an open source tool for creating realistically lit 3D graphics artwork. We tested with POV-Ray's standard 'one-CPU' and 'all-CPU' benchmarking tools on all of our test machines, and recorded the scores reported for each. Results are measured in pixels-per-second throughput; higher scores equate to better performance.
The POV-Ray benchmark showed 4th Gen Core i7-4770K with some nice performance gains over 3rd Gen Core i7-3770K. In both the single and multi-threaded tests, the Core i7-4770K was clearly the fastest quad-core processor--only the hex-core Core i7-3960X was faster.
|Low-Res Gaming: Crysis and ETQW|
For our next set of tests, we moved on to some in-game benchmarking with Crysis (DirectX) and Enemy Territory: Quake Wars (OpenGL). When testing processors with Crysis or ET:QW, we drop the resolution to 1024x768, and reduce all of the in-game graphical options to their minimum values to isolate CPU and memory performance as much as possible. However, the in-game effects, which control the level of detail for the games' physics engines and particle systems, are left at their maximum values, since these actually place some load on the CPU rather than GPU.
The Core i7-4770K rocked our low-res game tests. In the Crysis CPU benchmark, our various Core i7-4770K based systems sandwiched the Core i7-3960X, with the Asus and MSI build motherboards coming out on top. In the Enemy Territory: Quake Wars benchmark, the Core i7-4770K led all the way. The higher-bandwidth first level caches, increased IPC, and high Turbo core frequency, seem to give the 4770K a nice edge in performance here.
|HD Graphics vs. Discrete GPU Performance|
For this next set of tests, we pit the integrated Intel HD 4600 series processor graphics incorporated into the Haswell-based Intel Core i7-4770K against a couple of the least expensive, discrete GPUs from NVIDIA and AMD, the GeForce GT 430, Radeon HD 6450 and Radeon HD 5550, respectively. A couple of Sandy Bridge-based processors with integrated Intel HD 3000 series graphics and recently released AMD APUs with on-die Radeon HD 7000 series GPUs are thrown into the mix as well.
We also ran a few tests with 3DMark Vantage and with Cinebench's OpenGL benchmark. According to 3DMark, the Intel HD 4600 is evenly matched with the Radeon HD 7660D in the A10-5800K, though the AMD part still had a slight edge. Though it's performance advantage looks very big versus the HD 4000 engine here, keep in mind that CPU performance is also factored into the 3DMark Vantage results. In the Cinebench OpenGL test, the AMD APUs finished well ahead once again. Versus Intel's previous-gen integrated graphics though, the Core i7-4770K and its Intel HD 4600 series graphics engine perform significantly better.
|Quick Sync vs. CPU vs. Discrete GPU|
As we've mentioned, the new Haswell-based Core i7-4770K processor features an updated Quick Sync media encoding engine, that's available when the integrated Intel HD graphics core is enabled. We tested the Intel Quick Sync encoder using Cyberlink's MediaEspresso, which can take advantage of the technology, as well as leverage NVIDIA's CUDA / NVENC technology, or simply run on the CPU alone.
In this test, we took a 277MB AVCHD MTS file recorded using a Canon HD camcorder and converted it to an H.264 encoded MP4 compatible file designed for use with an iPhone / iPad (or other portable media playback device).
Intel's Quick Sync engine offered excellent performance on Sandy Bridge and Ivy Bridge, but with Haswell it is even better. With Quck Sync enabled on the Core i7-4770K, it was able to encode our video in only 5 seconds; without Quick Sync, that score jumps to 27 seconds. Please note, we've included scores for both the "Fast" and "Better Quality" media conversion options in MediaShow Espresso. We've also included scores using CUDA running on a GeForce GTX 680--which comes within striking distance of the original Quick Sync engine of Sandy Bridge.
We have yet to do an in depth quality comparison of various video encoders, but according to Intel the sum total of the enhancements to Haswell’s Quick Sync engine are not only increased performance and quality, but greater flexibility for developers as well.
We also played back numerous video types on the new Core i7-4770K while using their integrated Intel HD 4600 series graphics core to test its multimedia prowess, including DVDs and a myriad of SD and HD clips of varying file types. All of the local content (or content streamed from a NAS) played back perfectly with very low CPU utilization. High resolution Flash videos streamed from the web also played back perfectly, as did low-resolution content that was scaled up to fill our screen. In the screenshot above, a 1080p video of the Thor 2 trailer was being streamed from YouTube and scaled to full-screen. CPU utilization remained close to zero the entire time.
|Total System Power Consumption|
Throughout all of our benchmarking and testing, we also monitored how much power our test systems consumed using a power meter. Our goal was to give you all an idea as to how much power each configuration used while idling and while under a heavy workload. Please keep in mind that we were testing total system power consumption at the outlet here, not just the power being drawn by the processors alone.
Power consumption with the Haswell-based Core i7-4770K is excellent. When using the processor's integrated graphics, our test system consumed only 37 watts at idle and 104 watts under load. That's the lowest idle power we've seen to date, and only 2W more than the 3770K, which doesn't perform as well as the new 4770K. With a discrete GPU installed, however, the Core i7-4770K-based system consumed a bit more power than its Ivy Bridge-based counterpart, the Core i7-3770K, which was somewhat surprising. We believe the deltas have more to do with the different motherboards that has to be used for testing though, rather than the processors themselves.
|Performance Summary and Conclusion|
Performance Summary: Summarizing the Core i7-4770K’s performance is somewhat difficult, simply because there is so much data to cover related to CPU and GPU performance and power. Strictly talking in terms of CPU performance, the Haswell-based Core i7-4770K performs on par with, or somewhat better, than the Ivy Bridge-based Core i7-3770K, but overall the differences are not dramatic. In some situations, the 4770K’s performance is flat versus the 3770K, but in others, like POV-Ray or our low-res game tests, for example, the Core i7-4770K finished well ahead.
Intel's new 4th Generation Haswell microarchitecture holds a lot of promise. From a graphics standpoint alone (and we've only gotten a glimpse of what Intel has to offer, as we'll be following-up with Iris Graphics coverage), Haswell is an obvious leap forward in performance, capability and compatibility. Factor in Haswell's Quick Sync video transcode gains and core CPU performance-per-watt advantage and Haswell continues to show its value proposition. That said, Intel's 4th Generation quad-core desktop variant, the Core i7-4770K that we've shown you here today, doesn't offer huge gains over previous generation Ivy Bridge processor performance, at least with current software. To us, Haswell currently feels like its biggest play will be in the mobile space, at least until higher-end Haswell desktop CPUs come to market. Haswell mobile CPUs with Iris and Iris Pro Graphics engines, will undoubtedly shine brighter, especially when you look at performance-per-watt and battery life metrics.
All told, Haswell is the proverbial foundation for Intel's next round of products and product life cycle. Intel's 4th Generation Core processor technology has been enabled with many new functional blocks and features that will provide much-needed platform hooks for future generation enhancements and performance gains. We've seen this before; when it comes to Intel's "tock" cadence, the manufacturing machine is just getting warmed up. Stay tuned in the days and weeks ahead as we get more Haswell-based hardware in here for testing and review. Hot Hardware's Haswell Mobile coverage will be coming along very shortly and we'll be looking at Haswell overclocking and other performance characteristics as well--so stick around!