Intel Core i7 4960X EE CPU / Asus X79-Deluxe Motherboard Review

1. Introduction2. Meet Ivy Bridge-E: The Core i7 4960X EE CPU3. X79 Meets 2013: Asus X79-Deluxe Motherboard4. Asus X79-Deluxe: BIOS5. Overclocking: Processor Frequencies6. Overclocking: Memory Frequency7. Testing Methodology8. System: PCMark 89. System: 3DMark 1110. System: 3DMark11. System: 3DMark Vantage12. System: Unigine Heaven Benchmark13. Processor: SiSoft Sandra14. Processor: Cinebench R11.5 64 bit15. Processor: Super Pi16. Processor: Hyper Pi17. Processor: WinRar18. Processor: MKV HD Performance19. Processor: Media Encoding - MediaEspresso20. Processor: Media Encoding - HandBrake21. Gaming: Battlefield 322. Gaming: Sleeping Dogs23. Gaming: Metro 203324. Motherboard: SATA Performance25. Motherboard: USB 3.0 Performance26. Comparison: Overclocked Performance Results27. Comparison: Clock-per-Clock Comparison28. Technical: Power Consumption29. Technical: Temperatures30. Closing Thoughts31. View All Pages

A little under two years after the release of Intel's high-end desktop – Sandy Bridge-E – LGA 2011 processors and the corresponding X79 chipset, it's time for the platform to go through its refresh phase. Featuring a die shrink from SB-E's 32nm to 22nm, Ivy Bridge-E is intended to take command of Intel's flagship line of high-end, consumer processors. Welcome Intel's newest hexacore processor sat at the summit of the company's desktop CPU line-up, the Core i7 4960X Extreme Edition.

With Intel's most recent launch – that of its ‘tock' phase in the form of Haswell – enthusiasts across the globe greeted the new line of chips with mixed receptions. To be fair, mixed receptions is putting it nicely. Haswell was faster than its predecessor – Ivy Bridge – and very good for mobile purposes, but those improvements came at the cost of overclocking headroom which tended to be far more limited than the 3xxx series models. Many enthusiasts, overclockers in particular, were not pleased.

With this in mind, Intel is hoping that its latest high-end desktop (HEDT) processors can restore faith in the hearts of enthusiasts. Representing a ‘tick' phase in Intel's development cycle, IVB-E is built on a 22nm fabrication process and fits into the current LGA 2011 socket. Asus has sent us its refreshed X79-Deluxe motherboard which features specific tweaks and optimisations for use with Ivy Bridge-E processors.

Is the 4960X a worthy heir to the consumer processor throne?

Processor Features:

• 3.60GHz base clock speed (up to 4.0GHz turbo frequency).
• 6 cores, 12 threads.
• 15MB Intel Smart Cache.
• Quad-channel DDR3 1866MHz native memory support.
• 130W TDP.
• Supports LGA 2011 socket, X79 Express chipset-based motherboards.
• 40 PCI-Express lanes.
• Unlocked core ratio multiplier.
• Intel Turbo Boost technology 2.0.
• Intel Hyper-Threading technology.

Motherboard Features:

• Dual Intelligent Processors 4 with 4-way optimisation.
• Support for up to 64GB of system memory with an 8-DIMM design.
• ASUS SSD Caching II.
• Wi-Fi GO! – Support for dual-band wireless 802.11 a/b/g/n/ac and Bluetooth v4.0.
• 3-Way SLI and Quad-GPU CrossFireX Support!

Intel shipped the CPU to us in a miniature box with protective foam. Retail packaging will look similar to that of previous Intel Core series processors.

Notably thinner than the box that LGA 115x chips are supplied in, Intel's LGA 2011 processors are not bundled with heatsinks. Intel's thought process must be that anybody sinking over $300 (in the 4820K's case) into an overclocking-geared processor is going to be using a cooling setup that offers considerably higher-performance than that of its diminutive heatsink and fan.

Parts forming the Ivy Bridge-E (IVB-E) line-up are ‘tick' processors, meaning that they feature a die shrink from previous products. Sandy Bridge-E was designed with a 32nm fabrication process, while Ivy Bridge-E's die shrink takes it down to 22nm.

Intel's Core i7 4960X Extreme Edition is a six core, twelve thread part. Lower power consumption, henceforth improved efficiency, is one of the IVB-E series of processors' biggest aims.

Given its ‘tick' status in Intel's development cycle, Ivy Bridge-E fits into the same LGA 2011 socket that its predecessor – Sandy Bridge-E – did. This means that currently-available X79 motherboards will support IVB-E chips, even if they do require a BIOS update and recent drivers from Intel.

Built on the 22nm fabrication process, Ivy Bridge-E's die shrink from its 32nm predecessor is considerable. According to Intel's information, an IVB-E processor's die area is 256.5 millimetres-squared, as opposed to SB-E's 434.7 millimetres-squared.

As the above processor die image shows, the Ivy Bridge-E 4960x (and presumably partnering hexacore models) is a full-blooded, six-core part, as opposed to a cut-down octacore model that Intel used for its SB-E Core i7 chips.

Some of Ivy Bridge-E's biggest updates come in the form of true PCI-Express 3.0 capabilities, a 63x maximum core ratio, and support for up to four channels of DDR3 at 1866MHz (one DIMM per channel). Real-time core overclocking, power limits, and turbo voltage control are new features for IVB-E.

Officially, the Core i7 4xxx series processors' integrated memory controller natively supports DRAM frequencies of up to 1866MHz, but if the desktop Sandy Bridge to Ivy Bridge changes are anything to go by, the improved IMC could show support for much higher memory speeds.

Natively operating at a 3.60GHz base clock speed, with turbo boost kicking the chip up to a maximum of 4.00GHz, the Core i7 4960X is one core ratio faster than Intel's previous consumer flagship – the 3970X. As is the standard for Intel X-series, Extreme Edition parts (and, more recently, the K-series), the Core i7 4960X features an unlocked core ratio multiplier.

The Core i7 4960X features 15MB of shared L3 cache and operates within a 130W TDP – the same as SB-E did. When idling (and operating under the necessary motherboard and OS settings), the processor will drop its core multiplier to 12x and operate at 1200MHz, hence reducing power consumption and CPU temperature.

A recent technique that has been widely adopted by motherboard manufacturers – ‘forced turbo' – pushes the Core i7 4960X processor's operating frequency to a constant 4.00GHz when the setting is applied. Our Asus X79-Deluxe motherboard automatically applied the ‘forced turbo' parameter as soon as we enabled XMP.

It's worth noting that CPU-Z version 1.66.1 (or later) is required to properly read the Core i7 4960X processor's information. Version 1.66 incorrectly reported ridiculously low core voltage levels.

Intel's latest Core i7 processor family now includes an additional three chips from the Ivy Bridge-E camp. Haswell's 4770K forms the current-generation family's only unlocked, non-LGA 2011 model.

Taking a quick look at the family comparison table, with particular attention paid to the pricing column, it's easy to see that the 4930K is going to be a favourite amongst enthusiast-level overclockers. Of particular interest is Intel's decision to give the quad-core IVB-E 4820K an unlocked multiplier, especially given that its 3820 predecessor was partially locked.

In the 4820K, Intel is offering a more alluring and affordable way of getting onto the LGA 2011 platform, but at the expense of its own mainstream processors, it would seem. Under-cutting the 4770K's retail price and offering 40 PCI-E lanes, as well as quad-channel memory support, we could see a fierce battle between the 4820K and its LGA 1150 brother, especially amongst dual-card gamers.

There's nothing new about the X79 Express chipset. We still see the same disappointing omission of native USB 3.0 support, and maximum of two SATA 6Gb/s ports.

Despite the ageing chipset's shortcomings, we have no doubt that users will be pleased to see Intel lengthen its time spent on a single platform. X79 motherboards are widely available and early BIOS glitches have typically been ironed out. Using the same chipset also gives current X79 owners a direct upgrade path to IVB-E, albeit after a BIOS update.

Asus ships the X79-Deluxe motherboard in a black box which features gold tints and plenty of product information. A flap can be opened to reveal additional information regarding product features and to provide a viewing point for the motherboard (albeit through an anti-static bag).

A driver CD, user guide and Asus sticker form the supplied documentation material.

Asus provides a strong accessory bundle with the X79-Deluxe which includes a 3-way SLI bridge and a magnetic, dual-band, two-stream WiFi antenna that supports 802.11ac wireless technology, as well as Bluetooth 4.0.

The bundle consists of:

• 10x latching SATA cables.
• 1x 2-way flexible SLI bridge.
• 1x 3-way rigid SLI bridge.
• 1x dual-band, two-stream magnetic WiFi antenna.
• 2x Asus Q-connectors.
• 1x IO shield.

With Asus' refreshed X79-Deluxe motherboard, the company has given it a makeover, bringing its appearance in line with that of the company's 8 series boards. Feedback on our Facebook page and across the web has proven that the colour scheme is of a hate-it-or-love-it appeal. Personally, I like the smooth gold and dark black appearance, but I know that an equal numbers of readers will dislike it.

Measuring in at 12″ x 9.6″, the Asus X79-Deluxe motherboard conforms to the standard ATX form factor.

A metal strip beneath the CPU's VRM area helps to provide additional heat transferral away from the MOSFETs. The large LGA 2011 backplate is used to natively support many CPU coolers, including the waterblock and pump attachment for our Corsair H100i.

Adjacent to the 24-pin power connector is an internal USB 3.0 header and a speed-controlled 4-pin fan connection. The USB 3.0 header operates from ASMedia's ASM1042 chipset and provides a maximum of two ports, natively.

Given that native USB 3.0 support is omitted from the X79 chipset, Asus is forced to use the ASMedia controller which converts a PCI-E 2.0 lane to the pair of Universal Serial Bus 3.0 ports. Due to the fact that a PCI-E 2.0 lane is required, Asus has made the decision to supply one internal USB 3.0 header. The remaining PCI-E lanes are saved for rear USB 3.0 ports and other connections.

Found in close proximity to the right-side DIMM bank is the MemOK button. The toggle can be used to make the motherboard automatically apply memory settings that POST with stability. MemOK is one of Asus' innovative features that can sometimes be overlooked, but deserves all of the credit it gets, especially from users overclocking their memory.

A cluster of 4-pin fan headers (three, to be precise) is located on the board's upper-right corner. Two are given CPU duties and feature precise PWM control, while the other header is given chassis fan responsibilities but still provides accurate optimisation.

Positioned either side of the sizeable LGA 2011 CPU socket is a bank of four DIMM slots. Each single-latch slot is capable of holding a maximum sized memory stick of 8GB with operational speeds of up to 2800MHz (via overclocking). We will be putting the Asus X79-Deluxe motherboard's high-speed memory compatibility to the test with a 2800MHz set of ADATA's XPG V2 RAM, as well as a stick of G.Skill's 2933MHz TridentX.

The right side bank of DIMM slots features a two phase power delivery system which is identical to the number used for the left side (hidden beneath the left VRM heatsink).

Single-, dual-, triple-, and quad-channel memory configurations are supported by the Asus X79-Deluxe. Asus' motherboard can be used with up to 64GB of memory when all eight DIMM slots are populated.

Due to the fact that Intel's LGA 2011 processor socket is surrounded by eight DIMM slots which cause space-constraints, many motherboard manufacturers (Asus included) are forced to compress their VRM system into the remaining available area. Asus exploits the constricted space above the LGA 2011 socket to squeeze in a 10-phase power delivery system which feeds the 130W SB-E or IVB-E CPUs.

As well as high-performance electrical components, Asus also implements its Digi+ power control system to ensure that accurate voltage and current levels reach the CPU. The X79-Deluxe has received tweaks for optimised used with IVB-E processors, most notably to the EPU function.

A single 8-pin CPU power connector is found in its usual location – the motherboard's top-left area. For users with compact cases, the uppermost MOSFET heatsink may cause interference when using All-in-one liquid CPU coolers positioned in overhead 240mm mounting points.

The VRM heatsink clearance issue is true for most of today's high-performance motherboards, but with X79's congested CPU area, the block of metal typically spans to the PCB's uppermost edge, rather than sit a few centimetres beneath. Asus' inwards-sloping design for the central VRM heatsink allows interference with wide CPU coolers to be minimised.

Asus' X79-Deluxe features six PCI-E expansion slots, two of which are of x1 length, with the other four using a physical x16 size. The two x1 slots are rated for PCI-E 2.0 speed. Each of the four x16-length slots uses a quota of the IVB-E CPU's 40 PCI-Express 3.0 lanes.

The uppermost, second, and bottom PCI-E x16 slots support SLI, due to their link speed of x8 or above. The third PCI-E x16-length slot (sandwiched between two in the bottom cluster) is provided with an x4 connection at Gen 3 speed.

For dual-card operations, Asus wisely runs the top and bottom PCI-E full-length slots at x16 bandwidth, providing plenty of space for airflow and oversized coolers. When three cards are used, the top slot is given an x16 link speed, with the second and bottom connectors receiving eight lanes each.

Slot spacing on the X79-Deluxe is about as good as it gets for an LGA 2011 board. Additional memory banks force manufacturers to omit the slot in the uppermost position. Typically only the extreme motherboards geared towards hardened overclockers and benchmarkers feature support for 4-way graphics configurations (and seven expansion slots) by means of re-organising the motherboard PCB, or increasing its size to XL-ATX.

Most of the front panel headers are found in their usual locations, with the exception of HD audio which is positioned further to the right than normal. This actually aids cable management, as its typical position on the bottom-left corner of a motherboard is inconvenient when used with many cases.

A two-digit LED, power and reset buttons, and Asus' DirectKey button are all useful onboard features for overclocking circumstances. Four USB 2.0 headers and TPU and EPU switches are also convenient.

The red clear CMOS button is found further up the board than usual, sitting directly beneath the bottom PCI-E x16 slot. This is an inconvenient location as SLI or CrossFire users' bottom graphics card will instantly block it. Asus should have placed it along the motherboard's bottom edge, or at least out of the way of graphics cards.

Twelve SATA ports are provided on the Asus X79-Deluxe motherboard. On the right half are black-coloured ports operating from the X79 chipset. The furthest pair to the right operates at SATA 6Gb/s speed, while the two pairs to the left support 3Gb/s connections.

ASMedia's ASM1061 chipset controls the grey-coloured pair of SATA 6Gb/s ports situated furthest to the left of the set. The four grey 6Gb/s ports which feature a sticker on them are provided by Marvell's 88SE9230 controller which uses a PCI-E 2.0 x2 lane for the required bandwidth increase.

Asus opts for the Marvell ports as they provide SSD caching (also known as Intel Smart Response Technology on other Intel platforms) for up to four drives. The system, known as Asus SSD caching II, permits up to three SSDs to be used with a single HDD, for example.

As previously mentioned, the X79 chipset does not feature any native USB 3.0 ports. Asus implements six of the SuperSpeed connections on the X79-Deluxe motherboard's rear IO by means of two ASM1042 controllers, and a single ASM1074 hub. Four USB 2.0 ports are provided by the X79 chipset and are perfect for use with speed-irrelevant devices such as peripherals.

The pair of powered eSATA 6Gb/s ports is provided by an ASMedia ASM1061 controller and will therefore feature the same circa-400MB/s throughput performance cap that we are accustomed to seeing the chipset enforce. Audio is provided by Realtek's high-end ALC1150 chip – the same one that is found on many Z87 motherboards.

Wired network connectivity comes in the form of two gigabit Ethernet ports, one provided by Intel's 82579V controller, and the other from Realtek's 8111GR. Asus' updated Wi-Fi Go! features Bluetooth 4.0 support, as well as dual-band, two-stream 802.11ac wireless connectivity. Wireless speeds of up to 867Mb/s can be obtained from the dual-stream antenna.

A quick test of the wireless connectivity gave us a perfect 300Mb/s connection speed, as outlined by Windows 7, on the 5GHz band (the maximum speed that our 802.11n Tenda N60 router is capable of). We didn't suffer any drop-outs or signal problems, albeit while testing from a range of no more than four metres.

One notable omission from the rear IO ports is a clear CMOS button. This will inevitably irritate overclockers who are forced to revert back to stock settings due to a failed attempt. And given the X79-Deluxe motherboard's target audience – a group which is likely to use cases, not test benches – the decision to omit the rear IO clear CMOS button seems very odd.

Motherboard rear ports:

• 4x USB 2.0.
• 6x USB 3.0.
• Realtek LAN (RJ-45) port.
• Intel LAN (RJ-45) port.
• USB BIOS Flashback button.
• 2x powered eSATA 6Gb/s ports.
• 2x antenna connections for dual-band 802.11ac wireless and Bluetooth 4.0.
• Optical S/PDIF output.
• Audio jacks.

Asus' distribution of fan headers around the X79-Deluxe is very good. Each of the six fan headers is of the 4-pin variety and features control for speed-adjustment, based on CPU temperatures, via the UEFI BIOS or Asus' software.

Firstly, we are pleased to report that our Leetgion Hellion mouse worked to its usual standard in the Asus X79-Deluxe motherboard's UEFI BIOS. This is a positive point as we have experienced problems when using certain mice in the UEFI BIOS of other motherboards in the past.

Diving into the Asus X79-Deluxe motherboard's UEFI BIOS reveals an interface that is very similar to the company's previous iterations. The opening page displays system information, such as frequencies and memory size, as well as fan speed information and boot device priorities.

The favourites page can be used to obtain quick access to frequently-edited settings.

Additional system information is highlighted on the ‘Main' page.

Asus' AI Tweaker provides plenty of flexibility in regards to applicable voltage and frequency settings.

The built-in overclock tuner can be used to apply a ‘BCLK First' overclock which, unsurprisingly, adjusts the base clock frequency. A simple multiplier adjustment is provided by the ‘Ratio First' overclock. We will outline the automated overclocking procedures and settings later in the review.

Additional power control sub-sections allow more precise tweaks to be applied to the processor, memory, and motherboard.

DRAM frequencies up to 3200MHz can be applied, although we would be shocked if any IVB-E CPU can make use of the necessary memory multiplier. A long list of memory timings can be adjusted via its own section.

The ‘Advanced' section can be used to gain detailed information regarding the installed processor and to adjust some of its features. PCI-E link speeds and controller operations can also be adjusted via the ‘Advanced' sub-sections.

As we have come to expect from Asus' motherboards, the UEFI BIOS features a comprehensive fan control and monitoring system. A number of pre-defined operating modes can be applied, or manual adjustment for the target CPU temperature and related fan speed percentages can also be set.

Boot settings can be adjusted.

Asus' standard set of tool is featured on the refreshed X79-Deluxe. The EZ Flash 2 utility allows quick and simple BIOS updates to be made. Up to eight overclocking profiles can be saved and named, as well as exported to (or imported from) a USB flash drive.

One specific tool which was of particular importance to us was the DRAM SPD information. Our original set of memory wasn't functioning correctly so Asus' tool was used to identify exactly which stick was at fault. Thankfully we were able to switch to a different kit of memory to conduct the review's performance testing.

A note can be applied to the BIOS profile – a particularly useful tool for overclockers testing out many different voltage and frequency settings. The ‘last modified' log is of equal convenience for the aforementioned reasons.

In similar style to the ‘last modified' text box, a log of the settings changed on a specific visit to the BIOS is shown after hitting save & exit. The log is useful for identifying wrongly-applied settings such as obscene CPU speeds when switching between the 1.00x and 1.25x CPU BCLK strap, for example.

As I have said in the past in regards to Asus' ROG motherboards' UEFI BIOS implementations, there is no other way to describe the company's interface than excellent. The lay-out is practically faultless and the overall appearance is modern and easy to grasp.

Functionality-wise, the Asus X79-Deluxe motherboard's UEFI BIOS settings are highly appropriate for its target audience and should suffice for novices all the way through to experienced enthusiasts.

As is the case with SB-E and Haswell chips, IVB-E features CPU clock straps which allow its native base clock to be multiplied by 1.00 and 1.25, or 1.67 if the chip supports it. Put simply, it allows the processors to operate at high base frequencies, such as 125MHz, without having negative implications on link speeds to the rest of the system's components.

Memory is also fed by the multiplied base frequency, giving it the potential to overcome a CPU's memory multiplier limits (as we will see later in the review).

Automatic CPU Overclocking – BCLK First:

Using the ‘BCLK First' automated overclocking mode, Asus' X79-Deluxe boots the i7 4960X CPU with a frequency of 4125MHz. The 4.125GHz CPU speed is achieved by way of 125MHz base clock (using the 1.25x CPU strap) and a 33x core ratio. Our 2133MHz Patriot memory was set to operate at 2000MHz with its default timings of 11-11-11-27-3T.

For this setting, Asus' motherboard used the ‘Auto' voltage parameters. With our system, this resulted in a CPU VCore of 1.280V for idle and load scenarios.

Automatic CPU Overclocking – Ratio First:

Using the ‘Ratio First' automated overclocking mode, the X79-Deluxe motherboard opted for a dynamic overclock which used varied turbo ratios and the standard 100MHz base clock. The maximum core ratio was 43x, with a 41x setting being applied when all cores were loaded. Memory was set to 1866MHz and the default timings, which is slightly disappointing, especially given that the CPU and motherboard both support the 2133MHz divider.

Under idle conditions, the CPU frequency would drop to 1200MHz (12 x 100MHz) and a voltage of around 1.168V. When a multi-core load was applied, the CPU frequency jumped to 4100MHz (41 x 100MHz) and voltage increased to around 1.344V.

With the applied load voltage, our Corsair H100i kept the CPU temperature in the low-60s. Lesser performing CPU coolers may have a tougher time taming the heat from a 1.344V VCore. A good-value, sub-£30 model, such as the Cooler Master Hyper 212 EVO, should be capable of keeping temperatures at a reasonable level.

Manual CPU Overclocking (1.00x CPU Strap):

In order to push for the maximum frequency of our 4960X engineering sample chip, we applied a number of power tweaks and voltage increases.

The tweaks we applied included; Extreme LLC, 140% CPU current capability, and Extreme CPU power duty and phase control. We also enabled PLL overvoltage and PLL Termination, and disabled CPU spread spectrum. A BIOS-set VCore of 1.425V was used – a value which peaked at 1.456V under heavy load.

The Asus X79-Deluxe motherboard managed to push our Intel Core i7 4960X EE sample to 4.60GHz by simply increasing the multiplier to 46x.

Despite being able to boot into Windows, we could not obtain stability at 4.70GHz, no matter how many power settings we tweaked. We tried a CPU VCore of 1.475V, PLL voltage of 2.00V, system agent voltage of 1.300V, and VTTCPU voltage of 1.15V, but we still could not garner stability at 4.70GHz.

Our best overclocking result (using the 100MHz base clock) was 4.60GHz (46 x 100MHz) at a BIOS-set voltage of 1.400V. This setting was perfectly stable and temperatures were acceptable for 24/7 usage with our Corsair H100i CPU cooler.

4.60GHz was also largely stable at 1.375V (in the BIOS), but Handbrake conversion tests repeatedly caused BSODs, hence why we settled for a 1.400V VCore.

The Asus X79-Deluxe motherboard sent 1.408V to the 4960X when idling. As soon as a load was applied, the processor's core voltage increased to a maximum of 1.424V.

We did manage 4646MHz using a 46 x 101MHz setting, but the memory multiplier had to be dropped (to stay below our sticks' 2133MHz rating), so performance gains from the extra 46MHz CPU frequency would be negated by the decreased RAM speed.

Our 4.6GHz validation can be viewed here. Unfortunately, the CPU-Z validator destroyed our link when uploading a later validation soon after.

Manual CPU Overclocking (1.25x CPU Strap):

Hoping to achieve a greater CPU frequency by using a high base clock and lower multiplier, we switched to the 1.25x CPU strap. All voltage and power settings remained the same as those using for the 1.00x strap (including the increased values when searching for the maximum stable frequency).

The memory speed was dropped to 2000MHz to eliminate RAM instabilities from the overclocking equation (and because 2133MHz is not possible with a 125MHz base clock).

Our highest stable clock speed using the 1.25x CPU strap was 4625MHz (37 x 125MHz). We tried to push past the 4650MHz barrier by using a 36 x 130MHz setting, but could not do so with our sample processor.

When a heavy load (such as Prime95) was applied to the processor, our settings for the Asus X79-Deluxe motherboard provided a maximum CPU VCore of 1.456V (1.425V BIOS-set). Temperatures were safe for 24/7 usage with our Corsair H100i.

A lower CPU VCore may have provided ample stability, but with decreased temperatures. We didn't search for the lowest stable VCore because the 4.60Ghz overclock using the 1.00x CPU strap was already more desirable due to the fact that our memory stayed at its rated speed of 2133MHz.

Our 4.625GHz (37 x 125MHz) validation can be viewed here. As can be seen by the 4.60GHz and 4.625GHz validations' identical codes and links, the CPU-Z validator would only allow us to link to the one that was updated last (the 4625MHz validation).

Asus X79-Deluxe Overclocking (SB-E):

Referring to the Asus X79-Deluxe motherboard's overclocking performance with SB-E, we also had no problems taking a retail 3930K processor to 5.0GHz with sensible voltage levels.

The system wasn't 100% stable at 5.0GHz (the odd benchmark would fail), but a little extra voltage and enhanced cooling would have solved BSODs.

Memory Overclocking:

A system’s maximum memory frequency is heavily related to the performance of an individual CPU’s memory controller. The motherboard’s performance can also help to obtain higher speeds, especially when XMP settings are taken into account.

We switched to a 2800MHz set of ADATA XPG V2 memory, as well as a single stick of G.Skill's 2933MHz Trident X F3-2933C12D-8GTXDG (for speed validation purposes), to test the performance of the Core i7 4960X CPU's integrated memory controller (IMC) and the Asus X79-Deluxe motherboard’s high-speed memory support.

We didn't expect the 4960X to be able to reach either set's maximum data rates (Haswell or extreme overclocking is required for that), but the high DRAM frequencies allowed us to move the speed bottleneck away from the memory and onto the CPU.

A 2400MHz DRAM frequency proved too easy for the IVB-E CPU's strong (and seemingly improved, in comparison to SB-E) IMC. Using the 1.00x CPU strap with a 100MHz base clock, 2400MHz was as far as we could go. Despite the motherboard offering higher dividers, none of them would reach POST, indicating that the 4960X's maximum (realistically-achievable) memory multiplier is 18x (100:133.3 * 18 = 2400MHz).

To reach 2400MHz, we did nothing more than manually configure the memory timings and voltage, before selecting the 2400MHz divider. No power tweaks or additional voltage parameters were changed.

After switching to the 1.25x CPU strap and using a 125MHz base clock, the 4960X CPU's IMC was able to push our memory sticks up to 2666MHz. Again, we didn't change any power settings or voltage parameters (other than a manually-set 1.65V DRAM voltage) to reach 2666MHz.

3000MHz (a speed which both kits are capable of with our loosened timings) proved a step too far for our IVB-E chip's IMC. Even tweaking the power settings and memory controller voltages had no affect on the system's inability to POST at a 3000MHz DRAM frequency.

Given that SB-E processors started to provide problems (and usually required voltage increases) when using memory speeds around the 2400MHz mark, IVB-E's ability to boot and hold stability with 2666MHz memory at its stock voltages is impressive. It's not Haswell-calibre memory support, but being able to easily use a 2666MHz DRAM frequency is a good improvement from what SB-E offered.

Our validation with the ADATA XPG V2 2800MHz memory running at 2666MHz with the Core i7 4960X EE processor can be viewed here.

To test the Intel Core i7 4960X Extreme Edition processor, we paired it with Asus‘ refreshed X79-Deluxe motherboard. We will be comparing the 4960X processor's performance to that of SB-E's 3930K (due to its immense popularity), the Haswell 4770K, and AMD's previous flagship – the FX-8350 Black Edition. All processors are partnered with identical hardware and software (except for the motherboard and its drivers), so the results are directly comparable.

By default, the Asus X79-Deluxe motherboard (as well as many other X79 parts) forces the Intel Core i7 4960X EE to constantly operate at its maximum turbo speed of 4.0GHz when XMP is enabled. This will be displayed as the ‘stock’ setting. We could have turned the option off, but as the same process is applied to the other Intel processors that we are using as comparisons, we left it enabled.

On one hand, this shows the Intel processors operating slightly faster than they would if ‘forced turbo' was disabled. On the other hand, this setting is automatically applied to most motherboards (especially when XMP is used), therefore testing with it activated is arguably more representative of real world performance and power consumption figures.

We will also be running some of the most CPU-intensive tests with each processor overclocked to its maximum, widely-achievable level. Research tells us that 4.5GHz is a safe bet for many Core i7 4770K chips, 4.8GHz is achievable with many 3930K processors, and many of AMD's FX-8350 CPUs are capable of 4.6GHz. We didn't want to go any higher on the comparison processor frequencies as that may see us entering territory that many end users cannot reach.

The Intel Core i7 4960X Extreme Edition processor will be operating at 4.60GHz for overclocked performance testing.

Another test that we will use is a clock-per-clock comparison. This comes in the form of two single-threaded benchmarks that are run by each processor operating at the same frequency of 4.40GHz. Using single-threaded benchmarks allows us to take core count out of the equation and give an insight into any architectural performance differentials.

Test System – Common Components:

• Memory: 16GB (4x 4GB) Patriot Viper 3 2133MHz CL11 (quad channel for X79, dual channel for Z87 and 990FX).
• Graphics Card: nVidia GTX 760 2GB.
• System Drive: 120GB Kingston HyperX 3K SSD.
• CPU Cooler: Corsair H100i.
• Case: NZXT Phantom 630.
• Power Supply: Seasonic Platinum 1000W.
• Operating System: Windows 7 Professional 64-bit (with SP1).

Comparison Processor and Motherboard Configurations (stock speeds):

• Intel Core i7 3930K (3.20GHz / 3.80GHz forced Turbo) & Asus X79-Deluxe.
• Intel Core i7 4770K (3.50GHz / 3.90GHz forced Turbo) & ASRock Z87 OC Formula.
• AMD FX-8350 Black Edition (4.00GHz) & ASRock 990FX Extreme9.

Comparison Processor Configurations (overclocked speeds):

• Intel Core i7 4960X (4.60GHz @ 1.400V, Extreme LLC).
• Intel Core i7 3930K (4.80GHz @ 1.400V, Ultra High LLC).
• Intel Core i7 4770K (4.50GHz @ 1.350V, Level 1 LLC).
• AMD FX-8350 Black Edition (4.60GHz @ 1.575V, 100% LLC, 2000MHz HT Link speed [set by motherboard]).

Settings:

• Asus X79-Deluxe BIOS 0253.
• GeForce 320.49 VGA drivers.
• Intel 9.3.2.1010 chipset drivers.

Software Suite:

• PCMark 8 (System)
• 3DMark (System)
• 3DMark 11 (System)
• 3DMark Vantage (System)
• Unigine Heaven Benchmark 4.0 (System)
• SiSoft Sandra 2013 SP4 (CPU)
• Cinebench 11.5 64 bit (CPU)
• Hyper Pi (CPU) (Except AMD FX-8350 due to x87 issues)
• Super Pi (CPU) (Except AMD FX-8350 due to x87 issues)
• WinRAR 4.20 (CPU)
• VLC Media Player 2.0.8 (CPU)
• CyberLink Media Espresso 6.7 (CPU)
• HandBrake 0.9.9 (CPU)
• ATTO (Motherboard)
• Battlefield 3 (Gaming)
• Metro 2033(Gaming)
• Sleeping Dogs(Gaming)

PCMark 8 is the latest version in the popular series of PC benchmarking tools. Improving on previous releases, PCMark 8 includes battery life measurement tools and new tests using popular applications from Adobe and Microsoft. Whether you are looking for long battery life, or maximum power, PCMark 8 helps you find the devices that offer the perfect combination of efficiency and performance for your needs.

Starting with Futuremark's all-round test, PCMark 8, the 4960X manages to take second place, bested only by the Haswell 4770K. The 4770K's advanced architecture gives it a 1.9% performance lead over the IVB-E part.

The IVB-E 4960X does manage to beat its other 6 core, 12 thread competitor (SB-E's 3930K) by just over 5%.

3DMark 11 is designed for testing DirectX 11 hardware running on Windows 7 and Windows Vista the benchmark includes six all new benchmark tests that make extensive use of all the new features in DirectX 11 including tessellation, compute shaders and multi-threading.

After running the tests 3DMark gives your system a score with larger numbers indicating better performance. Trusted by gamers worldwide to give accurate and unbiased results, 3DMark 11 is the best way to test DirectX 11 under game-like loads.

If you want to learn more about this benchmark, or to buy it yourself, head over to this page.

Intel's 6 core, 12 thread Ivy Bridge-E part offers the highest physics score in 3DMark 11. The IVB-E chip is 12.3% faster than the 3930K, according to 3DMark 11's physics test.

Haswell's 4770K is beaten by 30.5% in the physics test, although it does provide a slightly better graphics score, which could also be motherboard related.

3DMark is Futuremark's latest benchmark. It can be used to benchmark and compare everything from mobile devices, such as smart phones, tablets and laptops, to high-end gaming systems. The benchmark is available for Windows, Windows RT Android and iOS.

With 3 separate tests, each of which is intended to be used alongside a specific classification of hardware, 3DMark is a very versatile benchmark. Ice Storm is intended to be used with mobile devices, Cloud Gate is good for use with laptops and home PCs, and Fire Strike can be used to push the performance of gaming PCs.

We used the ‘Fire Strike' benchmark which is designed to be used on gaming PCs. We opted for the Normal setting, NOT the Extreme mode.

3DMark also shows the 4960X to have the strongest physics performance, but this time its lead over the popular 3930K is lowered to 9.4%.

The 4960X is faster than Intel's Haswell 4770K by a strong 34.7%, according to 3DMark's physics test results.

Futuremark released 3DMark Vantage, on April 28, 2008. It is a benchmark based upon DirectX 10. This is the first edition where the feature-restricted, free of charge version could not be used any number of times. 1280×1024 resolution was used with performance settings.

IVB-E's flagship part scores four-and-a-half thousand more points than the 3930K in 3DMark Vantage's CPU-only tests. This gives the 4960X an 11.7% performance lead over the six-core 3930K.

The 4960X registers a CPU-testing score that is 47.5% greater than that of the 4770K's in 3DMark Vantage. Moving from an FX-8350 to the i7 4960X would provide a performance increase of almost 94%.

Unigine provides an interesting way to test hardware. It can be easily adapted to various projects due to its elaborated software design and flexible toolset. A lot of their customers claim that they have never seen such extremely-effective code, which is so easy to understand.

Heaven Benchmark is a DirectX 11 GPU benchmark based on advanced Unigine engine from Unigine Corp. It reveals the enchanting magic of floating islands with a tiny village hidden in the cloudy skies. Interactive mode provides emerging experience of exploring the intricate world of steampunk.

Efficient and well-architected framework makes Unigine highly scalable:

• Multiple API (DirectX 9 / DirectX 10 / DirectX 11 / OpenGL) render
• Cross-platform: MS Windows (XP, Vista, Windows 7) / Linux
• Full support of 32bit and 64bit systems
• Multicore CPU support
• Little / big endian support (ready for game consoles)
• Powerful C++ API
• Comprehensive performance profiling system
• Flexible XML-based data structures

The graphics-heavy Unigine Heaven benchmark seems to prefer Haswell's architectural enhancements over the IVB-E chip's extra cores and threads, although the performance difference between each of the processors is barely 1% (and presumably within the Heaven benchmark's error margin).

What this test better shows is that none of the processors being compared today are likely to bottleneck a graphics configuration comparable to the GTX 760 in Unigine's Heaven benchmark.

SiSoftware Sandra (the System ANalyser, Diagnostic and Reporting Assistant) is an information & diagnostic utility. It should provide most of the information (including undocumented) you need to know about your hardware, software and other devices whether hardware or software.

Sandra is a (girl’s) name of Greek origin that means “defender”, “helper of mankind”. We think that’s quite fitting.

It works along the lines of other Windows utilities, however it tries to go beyond them and show you more of what’s really going on. Giving the user the ability to draw comparisons at both a high and low-level. You can get information about the CPU, chipset, video adapter, ports, printers, sound card, memory, network, Windows internals, AGP, PCI, PCI-X, PCIe (PCI Express), database, USB, USB2, 1394/Firewire, etc.

Native ports for all major operating systems are available:

• Windows XP, 2003/R2, Vista, 7, 2008/R2 (x86)
• Windows XP, 2003/R2, Vista, 7, 2008/R2 (x64)
• Windows 2003/R2, 2008/R2* (IA64)
• Windows Mobile 5.x (ARM CE 5.01)
• Windows Mobile 6.x (ARM CE 5.02)

All major technologies are supported and taken advantage of:

• SMP – Multi-Processor
• MC – Multi-Core
• SMT/HT – Hyper-Threading
• MMX, SSE, SSE2, SSE3, SSSE3, SSE 4.1, SSE 4.2, AVX, FMA – Multi-Media instructions
• GPGPU, DirectX, OpenGL – Graphics
• NUMA – Non-Uniform Memory Access
• AMD64/EM64T/x64 – 64-bit extensions to x86
• IA64 – Intel* Itanium 64-bit

Sandra's set of tests show the IVB-E part to have an across-the-board performance improvement of around 5 to 11%, in comparison to the popular SB-E 3930K. Sandra's processor arithmetic tests show the biggest improvement – an average percentage increase of almost 10% when going from the 3930K to 4960X.

Memory bandwidth increases by 5% when going from the SB-E's former second-in-command part to IVB-E's flagship chip. The IVB-E processors natively support memory speeds of up to 1866MHz, giving a bandwidth of 14.9GB/s for each of the four supported channels. Performance motherboards will typically allow for memory speeds greater than 2GHz to be used.

Neither of Intel's HEDT processors can match the Sandra multimedia integer performance of the Core i7 4770K, thanks to its AVX2 support. The same can be said for hashing bandwidth numbers in Sandra's processor cryptography test.

CINEBENCH R11.5 64 Bit is a real-world cross platform test suite that evaluates your computer’s performance capabilities. CINEBENCH is based on MAXON’s award-winning animation software CINEMA 4D, which is used extensively by studios and production houses worldwide for 3D content creation. MAXON software has been used in blockbuster movies such as Spider-Man, Star Wars, The Chronicles of Narnia and many more.

CINEBENCH is the perfect tool to compare CPU and graphics performance across various systems and platforms (Windows and Mac OS X). And best of all – it’s completely free.

Cinebench is a very well-threaded program, as proven by the six-core, hyper-threading parts' dominance.

The 4960X's 200MHz speed boost and 25% extra L3 cache (and perhaps enhanced micro-architecture) give it a greater-than-10% performance benefit over the 3930K.

The 4770K's result outlines the performance improvements to be obtained from the Haswell architecture; with a greater number of threads taken out of the equation, the Haswell 4770K beats all other chips in Cinebench's single-core test.

Super Pi is used by a huge audience, particularly to check stability when overclocking processors. If a system is able to calculate PI to the 2 millionth place after the decimal without mistake, it is considered to be stable in regards to RAM and CPU. Super Pi is a single-threaded benchmark.

We used Super Pi's '32M' benchmark setting.

Super Pi outlines the increased performance of Intel's Haswell architecture. Despite its additional L3 cache and 100MHz speed boost, the 4960X's Super Pi performance trails that of the 4770K's by almost 2.5%. The IVB-E part does manage to outperform the slower SB-E chip by over 7%, however.

The FX-8350 is omitted from the Super Pi test as AMD's parts are known to have an unfair disadvantage performing with Super Pi's x87 code.

Hyper Pi uses a workload very similar to Super Pi, but is adapted to make use of more than one processor thread.

We used Hyper Pi's ‘8M' benchmark setting and ran it on the maximum number of threads for each processor, with the ‘Realtime' priority setting.

Despite placing a single-threaded, pi-calculating benchmark on each processor thread, the quad-core, hyper-threading 4770K still manages to outperform IVB-E's 6 core, 12 thread part by around 10%. A similar performance advantage (to that of Super Pi) of around 7% is observed for the 4960X over the 3930K.

As was the case with Super Pi, the FX-8350 is omitted from the Hyper Pi test as AMD's parts are known to have an unfair disadvantage performing with Hyper Pi's x87 code.

WinRar is one of the most popular archive manager programs available. It can backup your data and reduce the size of email attachments, decompress RAR, ZIP and other files downloaded from Internet and create new archives in RAR and ZIP file format. You can try WinRAR before buy, its trial version is available in downloads.

WinRar's built in benchmark and hardware test can help us outline the performance differentials between each processor.

Thanks to the WinRAR benchmark's ability to make use of hyper-threading, the twelve-thread 4960X outperforms all of the competing processors that we tested. The IVB-E flagship part's increased clock speed, 25% extra L3 cache, and newer micro-architecture allow it to outperform the SB-E 3930K by almost 15%.

Haswell's 4770K can't compete with the IVB-E chip that has an additional two physical cores and four threads. The 4770K offers around 65% of the performance of Intel's IVB-E Extreme Edition processor.

The Matroska Media container is a very popular, open standard multimedia container which is usually found as .MKV files. It is a very popular format in enthusiast circles and can be played directly in VLC or Windows Media Player with suitable codecs installed.

We played our 1080P MKV rip of The Dark Knight using the latest version of VLC Media Player.

Playback of Full-HD MKV is a simple task for all of the processors on test today. With a rounded-down result of 3% CPU utilisation, the 4960X has the easiest time decoding the high-definition movie.

CyberLink MediaEspresso 6 is the successor to CyberLink MediaShow Espresso 5.5. With its further optimized CPU/GPU-acceleration, MediaEspresso is an even faster way to convert not only your video but also your music and image files between a wide range of popular formats.

Now you can easily playback and display your favourite movies, songs and photos not just on your mobile phone, iPad, PSP, Xbox, or Youtube and Facebook channels but also on the newly launched iPhone 4. Compile, convert and enjoy images and songs on any of your computing devices and enhance your videos with CyberLink’s built-in TrueTheater Technology.

New and Improved Features

• Ultra Fast Media Conversion – With support from the Intel Core i-Series processor family, ATI Stream & NVIDIA CUDA, MediaEspresso’s Batch-Conversion function enables multiple files to be transcoded simultaneously.
• Smart Detect Technology – MediaEspresso 6 automatically detects the type of portable device connected to the PC and selects the best multimedia profile to begin the conversion without the need for user’s intervention.
• Direct Sync to Portable Devices – Video, audio and image files can be transferred in a few easy steps to mobile phones including those from Acer, BlackBerry, HTC, Samsung, LG, Nokia, Motorola, Sony Ericsson, and Palm, as well as Sony Walkman and PSP devices.
• Enhanced Video Quality – CyberLink TrueTheater Denoise and Lighting enables the enhancement of video quality through optical noise filters and automatic brightness adjustment.
• Video, Music and Image File Conversion – Convert not only videos to popular formats such as AVI, MPEG, MKV, H.264/AVC, and FLV at the click of a button, but also images such as JPEG and PNG and music files like WMA, MP3 and M4A.
• Online Sharing – Conversion to video formats used by popular social networking websites and a direct upload feature means posting videos to Facebook and YouTube has never been easier.

For our testing today we are converting a 4.4GB 720p MKV file (1h:58mins) to Apple Mp4 format for playback on a portable device. This is a common procedure for many people and will give a good indication of system power.

Hardware acceleration is disabled to provide an accurate interpretation of the CPU performance.

Media Espresso isn't very well optimised for hyper-threaded parts, nor does its default setting demand a high CPU utilisation percentage. The 4960X outperforms the other hexacore processor by 11%, but Haswell's 4770K manages second place with a result that is 6% behind that of the IVB-E chip's.

Despite its eight physical cores and high clock speed, the AMD FX-8350 cannot compete with Intel's parts.

HandBrake is a fantastic free program that can be used to convert video files to many common formats for portable devices. HandBrake is an open-source, GPL-licensed, multi-platform, multi-threaded video transcoder, available for MacOS X, Linux and Windows.

We used the latest V 0.9.9 version.

For our testing today we are converting a 4.4GB 720p MKV file (1h:58mins) to MP4 format, using HandBrake's ‘Normal' profile, for playback on High-Resolution devices. This is a common procedure for many people and will give a good indication of system power.

In contrast to Media Espresso, Handbrake is very well optimised for multi-threaded processors and demands a high CPU utilisation percentage.

Both of the twelve-thread parts take assertive victories in our Handbrake video conversion test. The faster 4960X outperforms the 3930K by around 10%, while it beats the 4770K's performance by over 20.5%.

According to EA, Battlefield 3 garnered 3 million pre-orders by the day of its release. It is unknown at present whether these figures are worldwide or just for the US. The pre-order total makes it “the biggest first-person shooter launch in EA history”, according to the publisher. The engine is beautiful on the PC and very demanding of the partnering hardware.

We used the game's demanding ‘Ultra' setting and a 1920 x 1080 resolution to push today's gaming hardware.

Battlefield 3 gives a clear indication that, once you pass a certain point, cores, threads and cache mean very little in modern games, at a 1920 x 1080 resolution, at least. The micro-architectural improvements given to Intel's latest mainstream part allow it to gain an additional frame over the higher-core-count models that we pitted it against.

Despite being beaten by the 4770K, the single frame per second difference is unlikely to be of much concern when using the 4960X with test configurations similar to ours. The graphics card provides the biggest bottleneck in our system.

Sleeping Dogs started development as an original title, but was announced in 2009 as True Crime: Hong Kong, the third instalment and a reboot of the True Crime series.

As a result of the game’s high development budget and delays, it was cancelled by Activision Blizzard in 2011. Six months later, it was announced that Square Enix had picked up the publishing rights to the game, but the game was renamed Sleeping Dogs in 2012 since Square Enix did not purchase the True Crime name rights.

Our chart for Sleeping Dogs makes it seem that the 4960X may have a problem with its gaming performance. In reality, the difference between the two hexacore models' performance is one-fifth of a frame per second, but the 4960X's result got rounded down, while all of the other processors passed the point of rounding up.

Metro 2033 is a first-person shooter video game with survival horror elements, based on the novel Metro 2033 by Russian author Dmitry Glukhovsky. The game is played from the perspective of Artyom, the player-character. The story takes place in post-apocalyptic Moscow, mostly inside the metro system, but occasionally missions bring the player above-ground.

We used the game's built-in benchmark set to ‘Very High' quality to offer an intense challenge for the gaming hardware while also making playable frame rates a possibility.

Metro 2033 remains a demanding game, even for modern systems. As was the case with Battlefield 3, Metro 2033 prefers Haswell's architectural improvements over SB-E's and IVB-E's higher core counts and additional cache.

The ATTO Disk Benchmark performance measurement tool is compatible with Microsoft Windows. Measure your storage systems performance with various transfer sizes and test lengths for reads and writes. Several options are available to customize your performance measurement including queue depth, overlapped I/O and even a comparison mode with the option to run continuously.

Use ATTO Disk Benchmark to test any manufacturers RAID controllers, storage controllers, host adapters, hard drives and SSD drives and notice that ATTO products will consistently provide the highest level of performance to your storage.

Our testing uses a Samsung 840 Series SSD.

Native SATA 6Gb/s ports on the Asus X79-Deluxe motherboards provide restriction-free SSD performance. The only disappointment is that the X79 Express chipset only provides two SATA 6Gb/s connections, with the other four operating at 3Gb/s speeds.

Seemingly using a PCI-E 2.0 x2 connection, the SATA 6Gb/s ports operating from Marvell's 88SE9230 controller also manage to provide enough bandwidth to ensure that SSD performance is not noticeably hampered. We only had one high-speed SSD available for testing, so what happens when three or all four ports are used remains unclear. We would assume that drive throughput would drop due to the saturation point of a PCI-E 2.0 x2 connection being approached or surpassed.

ASMedia's SATA 6Gb/s ports provide the same circa-400MB/s throughput limitations that we are accustomed to seeing the ASM1061 chipset enforce on Z87 motherboards. When used on an X79 motherboard, though, they still offer better throughput than the chipset's native SATA 3Gb/s ports, so are arguably more useful than the ones used on Z87 parts.

With its emergence as the new standard for high-speed portable devices, USB 3.0 performance on a modern motherboard needs to be good to ensure that data transferral bottlenecks aren't created.

We tested USB 3.0 performance using an ADATA SP300 SSD connected to an Icy Box IB-223StU3 USB 3.0 enclosure (ASMedia ASM1051 controller). ATTO was the benchmark used.

Despite being provided with UASP-activating software support, the Asus X79-Deluxe motherboard's USB 3.0 ports operating from ASMedia's ASM1042 chipset aren't as fast as those found on the company's Z87 Maximus VI Formula. Fatigue on our ADATA SSD since testing the Maximus VI Formula could have a slight skewing effect on the data, however.

On the other hand, a motherboard that doesn't offer UASP support will limit its USB 3.0 ports' throughput speeds to around 200MB/s – a few Megabytes per second less than the throughput offered by ports on Asus' X79-Deluxe.

The following tests consist of CPU-intensive experiments conducted with each processor overclocked to its maximum, widely-achievable level.

Research tells us that 4.5GHz is a safe bet for many Core i7 4770K chips, 4.8GHz is achievable with many 3930K processors, and many of AMD's FX-8350 CPUs are capable of 4.6GHz. We didn't want to go any higher on the comparison processor frequencies as that may see us entering territory that many end users cannot reach.

With a 600MHz overclock (in comparison to the forced-turbo stock speed) applied, the 4960X gains more than 6000 points from 3DMark Vantage's CPU tests. The higher-clocked 3930K is not far behind the 4.6GHz 4960X in terms of performance, being narrowly defeated by just over 1000 points (around 2.6%).

With the maximum (safe and 24/7 usable) overclocks applied, the 4960X narrowly beats the 200MHz higher-clocked 3930K in Cinebench's tests. IVB-E's tweaked architecture and the Extreme Edition part's 25% extra L3 cache allow it to outperform the SB-E 3930K, despite its 200MHz clock-speed deficit.

Neither of Intel's hexacore chips can match the 4770K's single-threaded performance when overclocked. Haswell's architectural improvements are to thank for the overclocked 4770K's single-core victory.

Haswell's improved architecture again shows its strength by beating both of the faster parts, which also posses more L3 cache, in Hyper Pi. The overclocked 4960X trims six seconds off the 4.8GHz SB-E chip's time, registering a 3.6% performance improvement.

Super Pi's single-threaded workload turns the table for the hexacore chips' competition. This time, the 200MHz-faster 3930K is able to outperform the 4.6GHz 4960X by four seconds (almost 1%). Again though, the overclocked Haswell chip has the potential to breeze past the performance of each hexacore part.

Overclocking the 4960X provided large performance numbers from WinRAR's benchmark tool. The 4.6GHz IVB-E processor outperforms the 4.8GHz 3930K by over 12%. Additional cache on the Extreme Edition chip, in comparison to the 3930K, has also helped to boost the 4960X's performance against the SB-E part.

WinRAR didn't like the sub-stock HT Link speed of 2000MHz that was automatically applied to the FX-8350 system by its ASRock 990FX Extreme9 motherboard. With less than half the performance of Intel's overclocked hexacore chips, it made little difference to the interpretation of our results.

The well-optimised Handbrake media conversion program makes good use of the 4960X's additional L3 cache, allowing it to maintain its performance lead over the higher-clocked 3930K. The IVB-E flagship processor outperforms our comparison SB-E part by just over 2%.

The clock-per-clock comparison comes in the form of two single-threaded benchmarks that are run by each processor operating at the same frequency of 4.40GHz. Using single-threaded benchmarks allows us to take core count out of the equation and give an insight into any architectural performance differentials.

Despite its smaller amount of cache, the 4770K is able to outperform Intel's Ivy Bridge-E flagship processor by almost 4% in the single core Cinebench test. It is clear that Haswell has the greater architectural performance and not even substantially larger amounts of cache can overthrow that.

On a clock-per-clock basis, the IVB-E 4960X is just over 5% faster than the SB-E 3930K. An extra 3MB of L3 cache on the IVB-E chip is partly to thank for its improved performance, but architectural updates also assist.

The same picture is painted with Super Pi's single-threaded workload. Haswell's 4770K is around 4.8% faster than the IVB-E 4960X. The 4960X is, in turn, around 2.5% faster than the SB-E 3930K.

We measured the power consumption with the system resting at the Windows 7 desktop, representing idle values.

The power consumption of our entire test system is measured at the wall while loading only the CPU using Prime95's Small FFTs setting. The rest of the system's components were operating in their idle states, hence the increased power consumption values (in comparison to the idle figures) are largely related to the load on the CPU and motherboard power delivery components.

With the Asus X79-Deluxe motherboard applying a CPU VCore of 1.248V under load, the 4960X system draws a maximum of 212W. That's pretty impressive for a system featuring a high-performance hexacore processor that operates under forced-turbo conditions.

The 3930K system draws an extra 40W under load, which is around 19% more than the identical 4960X-based machine consumes. With a constant-load system usage of 25 hours per week, the 4960X system will use about £7.80 less electricity than the 3930K-based machine over the course of a year (based on a £0.15 per kWh electricity price).

When both hexacore chips are overclocked (using the same voltage, but different LLC settings), the 4960X demands a substantially lower amount of energy – 145W (33%) less, to be precise. With the same 25 hours per week constant-load system usage, that amounts to an electricity saving of just over £28 per year for the IVB-E system, in comparison to the SB-E machine.

Yes, the 3930K-based system is clocked higher, but that doesn't translate into better performance. While lowered power consumption is nice, people interested in buying an Intel Extreme Edition chip (or a 3930K-level part, for that matter) are unlikely to show much concern over £28 extra electricity per year (in the above usage scenarios).

What will be of more importance to enthusiasts is the lower CPU temperature that decreased power consumption tends to imply. Let's check the CPU temperatures.

We measured the Core i7 4960X CPU's temperatures using HWMonitor version 1.23. The readings from CPUID's monitoring software varied to those from the Asus motherboard's own utility by up to 10°C. This is something to bear in mind.

We measured the CPU's temperature after five minutes of idling and five minutes of Prime95 load (small FFTs setting). The Corsair H100i CPU cooler's fans were operating from Asus' ‘standard' PWM profile. The maximum fan speed was around 2300 RPM under load.

Our graphs show actual temperatures of the CPU, not delta temperatures. Ambient temperature was maintained at 22°C.

IVB-E seems to operate at pleasing temperatures, even when overclocked using relatively high voltages. By comparison, we put the same 1.40V through our retail 3930K chip (which, due to different LLC settings, was actually 1.408V when loaded, as opposed to the 4960X's 1.424V) and registered a load temperature of 84°C when overclocked to 4.80GHz, and the H100i fans' speeds were higher.

At stock, SB-E received 48mV less voltage than our 4960X sample, but also registered a load temperature that was 1°C higher. CPU cooler fan speeds for either chips were almost identical. Use our temperature readings as a reference only, though, because they were recorded using the Asus motherboard's default PWM profile.

Based on our sample, Ivy Bridge-E seems more forgiving in regards to temperatures than SB-E is. This is good news for users looking to push their CPU voltage higher to achieve greater overclocks, as it doesn't mean that high-end watercooling setups are a necessity. Coupled with the IVB-E processor's lowered power consumption, it also means that watercooling users are given fewer headaches when adding their GPU(s) or motherboard into the loop.

Intel Core i7 4960X Extreme Edition Processor

While Intel's Ivy Bridge-E processor does provide a noticeable performance improvement over its Sandy Bridge-E predecessors (based on the 3930K's performance, at least), it isn't that huge leap that we were all (optimistically) hoping for. Decreased power consumption and reduced temperatures are nice, but they are not the performance improvements that enthusiasts had been wishing for.

Intel has achieved its goal of shrinking Ivy Bridge-E onto a 22nm fabrication process, and still offering a performance bump over Sandy Bridge-E. Unfortunately, based on our clock-per-clock results, the performance increase isn't anywhere near the greater-than-15% level that would have been welcomed by many enthusiasts.

CPU overclocking also seems more limited in comparison to Sandy Bridge-E, but we can't confirm that point based on the frequency achieved by our sample alone. Given that 3930K chips generally seem to hit 4.7/4.8GHz without problems and our 4960X couldn't reach much more than 4.6GHz, perhaps the huge headroom variations that are plaguing Haswell processors will be found on the IVB-E chips. Perhaps they won't, and we just got not-so-great clocker.

On the other hand, IVB-E's improved IMC seems to offer greater memory overclocking headroom than SB-E does. We managed to hit a 2666MHz memory frequency without any settings changes, which is very good given that SB-E needed tweaking once the 2400MHz-mark was passed.

Don't get me wrong, there are some positives to expect from the Ivy Bridge-E chips, based on our 4960X sample. Temperatures seem to be far more forgiving than those of SB-E. Power consumption has also been decreased – and by a significant margin when a processor is overclocked.

Jumping back to the positive points of performance, memory bandwidth shows increases and the processors now support DDR3 1866MHz, natively. Our benchmark results also provided indications that IVB-E's clock-per-clock performance is faster than SB-E's. True PCI-E 3.0 support is another positive for the IVB-E chips.

With a retail price of $990 USD, which should translate into around £700-800 in the UK, the Intel Core i7 4960X Extreme Edition processor is a major investment. But then so is every other Extreme Edition part. Given that the 4960X should enter the market at the SB-E 3970X's current price point, purchasing the IVB-E chip is a ‘no-brainer' to customers formerly wanting the SB-E flagship.

If, however, you currently have a Sandy Bridge-E machine, there is little point in upgrading to Ivy Bridge-E for anything other than decreased power consumption and improved high-speed memory frequency support.

The bottom line is: if you want the fastest consumer processor that money can buy, Intel's Core i7 4960X Extreme Edition processor is the chip to purchase. If you're looking for strong performance in multi-threaded applications, the 4960X is a good choice, but its $435-cheaper (and largely identical) 4930K sibling will undoubtedly steal the show.

Pros:

• Impressive power consumption figures.
• Temperatures seem to be lower than those on SB-E processors.
• Currently the fastest consumer processor on the market.
• Fits in current LGA 2011 motherboards.
• Improved memory controller.
• Offers proper support for PCI-E 3.0.

Cons:

• Very expensive.
• Not much of a performance improvement over SB-E.
• The significantly-cheaper 4930K will undoubtedly steal the show.

KitGuru says: The fastest consumer processor that money can buy, Intel's Core i7 4960X Extreme Edition is a performance king with a price tag to match. Unfortunately the performance increases over Sandy Bridge-E chips aren't worth upgrading over, and the cheaper 4930K will steal the limelight.

Asus X79-Deluxe Motherboard

Asus' refreshed X79-Deluxe motherboard is an excellent product that has a great UEFI BIOS implementation, solid overclocking performance, and plenty of features that bring the X79 platform into the year 2013.

Specifically refreshed for IVB-E performance tweaks and optimisations, the Asus X79-Deluxe had no problem taking our Core i7 4960X chip to a little over 4.6GHz. We managed an easy 5GHz using a 3930K, too. Memory overclocking performance was also strong; we were able to hit 2666MHz with ease.

Retaining the support for 3-way SLI and CrossFire, the X79-Deluxe can be used to provide the foundation for a powerful system. With its numerous onboard features, many of which have entered the limelight subsequent to the X79 Express chipset's introduction, Asus' X79-Deluxe gives LGA 2011 users many of the ‘goodies' that are found on premium Z87 motherboards.

Those modern features included dual-band, two stream 802.11ac WiFi, Realtek's ALC1150 audio chip, and support for SSD caching with up to three drives. One of the more opinion-splitting updates is the board's implementation of Asus' newly-adopted black and gold colour scheme.

Asus' UEFI BIOS has yet again proven its dominance on the X79-Deluxe. The interface is attractive, features an excellent layout, and is very easy to use. Plenty of tweaking options are also available for overclockers.

Our biggest annoyance regarding the X79-Deluxe is Asus' positioning of the clear CMOS button. Access to the button will be hampered by the second graphics card for dual-VGA users. And given that the X79-Deluxe is likely to be housed inside a chassis, a clear CMOS button should have been placed on the rear IO panel.

Asus told us that the refreshed X79-Deluxe is expected to retail for the same price as its successor – the P9X79 Deluxe. That should make it available for around the £280-mark from retailers such as Overclockers UK.

Pros:

• Solid overclocking performance, especially on the memory side.
• Support for 3-way SLI/CrossFire and good two-card spacing.
• Good added features for the refreshed version  (802.11ac WiFi and a high-end audio chip).
• Supports Asus SSD Caching II with up to four drives.
• Excellent UEFI BIOS implementation.
• Tweaked for IVB-E support.

Cons:

• No clear CMOS button on the back panel and the internal one is awkward to reach.

KitGuru says: An excellent high-performance motherboard that brings the X79 platform into the year 2013.

Become a Patron!