Intel i7-4960X Ivy Bridge-E Review

[SKYMTL]

Synthetic Gaming Benchmarks: 3DMark06 / 3DMark11

Synthetic Gaming Benchmarks

3DMark11

3DMark 11 is the latest in a long line of synthetic benchmarking programs from the Futuremark Corporation. This is their first foray into the DX11 rendering field and the result is a program that incorporates all of the latest techniques into a stunning display of imagery. Tessellation, depth of field, HDR, OpenCL physics and many others are on display here. While the GPU plays a primary role here, 3DMark 11 takes full advantage of multiple CPU threads so any bottlenecks caused by the CPU will also be evident. In the benchmarks below we have included the results (at default settings) for both the Entry and Performance presets.

3DMark06

While its DX9 tests may not seem to be completely relevant in a DX11 era, 3DMark06 still provides an excellent measuring device for both multi core CPU and GPU performance. For this benchmark we are using the standard preset.

RESULTS: Synthetic benchmarks don’t necessarily tell the whole story of in-game performance and these two versions of 3DMark certainly provide some interesting results. On one hand, 3DMark06 shows the Haswell architecture surging to the forefront, in some cases significantly outpacing Intel’s thousand dollar processor.

Meanwhile, 3DMark11’s Entry preset has three of Intel’s fastest processors literally neck and neck. The Performance preset is where some confusion lies since it is AMD’s processors which come out ahead, which seems to indicate that in some situations the Bulldozer architecture’s modular design has some significant…albeit limited….benefits.

 

[SKYMTL]

720P Gaming: Deus Ex: HR / Dirt 3 / Skyrim

720P Gaming Benchmarks

Processors play a huge roll in gameplay performance since they have to process copious amounts of information for the GPU. In the following tests, we use a simple 720P resolution and the lowest possible detail settings in an effort to remove the graphics processor from the equation and place additional pressure on the CPU. 720P was used since it is a resolution that is extensively used by gamers sporting lower end HDTVs and it doesn’t put as much stress upon the GPU as 1080P.

For every one of the following titles, a simple 1 minute gameplay walkthrough was used and the average frames per second was logged via FRAPS.

RESULTS: It seems that at lower resolutions, where processor speed plays a significant role in game performance, the i7-4770K continues to rule the roost. IVB-E simply can’t keep up, even in certain games which are supposedly multi-core aware.

 

[SKYMTL]

720P Gaming: Super Street Fighter IV / Torchlight

720P Gaming Benchmarks (pg.2)

Processors play a huge roll in gameplay performance since they have to process copious amounts of information for the GPU. In the following tests, we use a simple 720P resolution and the lowest possible detail settings in an effort to remove the graphics processor from the equation and place additional pressure on the CPU. 720P was used since it is a resolution that is extensively used by gamers sporting lower end HDTVs and it doesn’t put as much stress upon the GPU as 1080P.

For every one of the following titles, a simple 1 minute gameplay walkthrough was used and the average frames per second was logged via FRAPS.

RESULTS: In SSF IV and Torchlight, the i7-4960X’s fortunes change somewhat with ties against the i7-4770K. However, once again, there is very little difference between Ivy Bridge-E and the outgoing Sandy Bridge-E processors.

 

[SKYMTL]

1080P Gaming: Deus Ex: HR / Dirt 3 / Skyrim

1080P Gaming Benchmarks

While lower resolution gaming highlights processor bottlenecks, most people use slightly higher resolution monitors and want to play with increased detail settings. In these situations, the CPU tends to take a back seat to the graphics processor but even at 1080P (ie: 1920x1080) a slower CPU can still have a drastic impact upon in-game performance. In order to illustrate this, we have carried over the games from our previous tests, pumped detail levels to their max and used the increasingly popular 1080P resolution standard.

RESULTS: With the CPU taking a bit more of a back seat in the higher resolution tests, the situation becomes a bit less defined but once again, we’d say the Haswell architecture has a distinct advantage. Ivy Bridge-E does remain near the top of our charts but it never becomes a clear winner, often featuring less than a 3% improvement over its predecessors.

 

[SKYMTL]

1080P Gaming: Super Street Fighter IV / Torchlight

1080P Gaming Benchmarks (pg.2)

While lower resolution gaming highlights processor bottlenecks, most people use slightly higher resolution monitors and want to play with increased detail settings. In these situations, the CPU tends to take a back seat to the graphics processor but even at 1080P (ie: 1920x1080) a slower CPU can still have a drastic impact upon in-game performance. In order to illustrate this, we have carried over the games from our previous tests, pumped detail levels to their max and used the increasingly popular 1080P resolution standard.

RESULTS: The trend of middling performance increases continues here with the 4960X featuring limited benefits over other current generation Intel processors. Even against Sandy Bridge-E, the benefits are slim to none.

 

[SKYMTL]

1080P Gaming; Current Games

1080P Gaming; Current Games

While our standard in-game testing consists of slightly older but still popular titles, we decided to throw in a number of additional, newer games. Each of these puts a significant amount of load on the CPU but many are also multi-core optimized.

All of the standard testing guidelines apply here as well though there is one major difference: instead of using FRAPS, we are utilizing the FCAT tool. This captures frames in real time before analyzing them, allowing for a much more accurate depiction of what will eventually be displayed onscreen. Once again, a GTX 670 has been used.

The first few newer games show a trend towards the same results we saw in the previous tests. The i7-4770K is often the fastest processor around since most titles can’t take advantage of more than four cores.

On the other hand, Crysis 3 uses additional CPU cores for various physics tasks and Company of Heroes 2 is fully multi core aware, making use of up to 12 threads for post processing calculation and AI. As a result, these two games allow the i7-4960X to surge ahead, though it doesn’t beat the 12-thread i7-3960X by all that much. Even AMD’s FX-9590 turns in an impressive showing every now and then.

 

[SKYMTL]

1080P Gaming; Current Games (pg.2)

1080P Gaming; Current Games (pg.2)

While our standard in-game testing consists of slightly older but still popular titles, we decided to throw in a number of additional, newer games. Each of these puts a significant amount of load on the CPU but many are also multi-core optimized.

All of the standard testing guidelines apply here as well though there is one major difference: instead of using FRAPS, we are utilizing the FCAT tool. This captures frames in real time before analyzing them, allowing for a much more accurate depiction of what will eventually be displayed onscreen. Once again, a GTX 670 has been used.

In our last few gaming tests, the i7-4770K remains ahead of the pack more often than not. It does indeed seem like the Haswell architecture’s IPC improvements have led directly to better in-game framerates. Once again though, Tomb Raider does buck that trend and allows Ivy Bridge-E to eke out a single win.

 

[SKYMTL]

Clock for Clock SB-E vs. IVB-E

Clock for Clock SB-E vs. IVB-E

Intel claims Ivy Bridge-E includes several minor architectural improvements which, when combined, allow it to perform slightly better than its predecessor even without a clock speed advantage. In order to test this, we took an i7-3960X and the i7-4960X, clocked them both to 4GHz (40 x 100MHz) and disabled Turbo, C-States and other power saving features to ensure the both ran at a constant speed. All other settings remained identical across both platforms.

Evening things out shows a literal statistical dead heat between the two architectures but there is a clear trend that shows Ivy Bridge-E to be slightly faster. This could be due to any number of reasons but we’ll side with Intel on this one and validate their architectural claims for the time being.

It is great to see that IVB-E buyers will indeed be getting more for their money, most of the time.

 

[SKYMTL]

Power Consumption / Temperature Testing

System Power Consumption

Our power consumption numbers are broken down into two categories: one which simply stresses all of the CPU cores with WPrime and another which puts a high amount of load on both the CPU cores and the IGP. The latter will only be included if a given processor includes a dedicated internal graphics sub-processor.

For the CPU power consumption test, we use the standard testing system (with an NVIDIA GTX 670 installed) and wait until the system and discrete GPU are at idle speeds in order to log the idle power consumption. After this, WPrime 1024M is looped for 15 minutes while the power consumption is logged with a calibrated power meter to determine the peak watts.

Please note that after extensive testing, we have found that simply plugging in a power meter to a wall outlet or UPS will NOT give you accurate power consumption numbers due to slight changes in the input voltage. Thus we use a Tripp-Lite 1800W line conditioner between the 120V outlet and the power meter.

While the i7-4960X has the same 130W TDP as the outgoing i7-3960X, the 22nm manufacturing process significantly reduces power consumption regardless of the processor’s load.

In our testing, we noticed an approximate 30W drop at full load and nearly identical efficiency ratios at idle. This allows IVB-E to boast an excellent performance per watt ratio in many tasks, though Hawell processors do have it beat in some instances. It does seem like the 22nm technology is working exactly how Intel intended it to.

Temperature Testing

In order to test temperatures, we set all processors on a loop of Prime95’s Blend Test for 25 minutes and logged the peak temperature core over that period of time using Intel’s XTU 4.2 utility. Ambient temperatures were kept at a constant 24°C. For cooling, a Noctua NH-U14S with two 140mm fans was used.

One of the major concerns with Intel’s 22nm 3D technology is its heat output due to stacked transistors focusing their energy into one condensed area. Many are experiencing this with Haswell, which tends to pump a ridiculous amount of heat into a small area, causing fits for heatsinks and water cooling systems alike.

Due to some changes in its IHS design, Ivy Bridge-E effectively spreads the heat from its 1.86 billion transistors over a larger area, allowing it to be dissipated faster. As a result, our i7-4960X sample posted some relatively low temperature results at idle and full load. It was even able to beat out the i7-3960X in this respect, which certainly wasn’t something we expected.

 

[SKYMTL]

Overclocking Results & Intel’s XTU

Overclocking Results & Intel’s XTU

Overclocking the i7-4960X is a relatively straightforward affair which is made easier with Intel’s Xtreme Tuning Utility, which has been updated to support Ivy Bridge Extreme processors. With that being said, there are some major architectural limitations which may cause overclocks to top out at the exact same levels as many Sandy Bridge Extreme processors.

According to our conversations with motherboard manufacturers, whose overclocking teams have been testing dozens of IVB-E samples, the vast majority of these CPUs will hit a clock speed wall between 4.5GHz and 4.7GHz when fed 1.4V. Meanwhile, only 2% or so will go to 4.8GHz and beyond so if you’re trying to hit extreme clock speeds, be prepared to bin heavily or hit the processor with extreme voltages. In our overclock testing, we used 1.4V (with LLC on Extreme) as our absolute maximum voltage since anything above that would likely lead to degradation if used on a 24/7 basis.

Anyways, back to Intel’s Xtreme Tuning Utility. Truth be told, we’ve never used its previous iterations but Intel has really put some effort into this piece of software, and it shows. Much like ASUS’ excellent AI Suite XTU gives you all of the overclocking tools within a simple interface that has been broken down into a number of tabs for CPU, memory and voltages. Alternately, the main page can be used for a more straightforward, simplified approach to overclocking.

Obviously the most important aspect XTU brings to the table is its ability to offer on-the-fly clock speed adjustments without hopping back into the BIOS. It even includes a beta version of AppTune which allows you to designate application-specific clock speed profiles.

After using it for some time, we still prefer overclocking in the BIOS but the XTU does provide a solid basis for someone who feel more comfortable with a Windows-based interface. Just be aware that you’ll need to understand that some of its functions need to be enabled by icons that are scattered throughout the main screen a subsequent sub-sections.

Ivy Bridge Extreme processors bring some new overclocking features to the table but for the most part, accessing the majority of their clock speed overhead is quite easy. All a beginner needs to do is increase core voltages and use the unlocked multiplier to boost clock speeds. Fine tuning can be done via the Base Clock or via ratio multipliers provided additional yet minor memory tuning is done.

Those “new” features we alluded to amount to unlocked Turbo ratios that have been expanded to encompass modifiers of up to 63x (in 100Mhz increments) and the ability to individually clock each core on a real-time basis. Neither will likely affect most enthusiasts since actually needing multipliers above 57x will go hand in hand with extreme cooling but overclockers will find uses for them.

Click on image to enlarge​

So, with all of this in mind, how was our overclocking experience with the i7-4960X? Surprisingly straightforward but quite frustrating as well. Even though our final overclock of speed 4.55GHz was 24/7 stable, we couldn’t seem to break through the 4.6GHz barrier while maintaining adequate stability at 1.4V. We could easily boot into windows and run a few applications through their paces at speeds up to 4.61GHz but our overclocking norms state that a speed has to be completely solid before we claim it as a viable overclock. So, with that in mind, 4.55GHz was where we remained.

It wasn’t heat that held us back since the CPU’s cores never went above the 80°C mark with the Corsair H100i working its magic, even with voltage turned up to 1.4V. A bit of extra voltage (1.45V) didn’t allow for meaningfully higher speeds either, though it did significantly elevate temperatures.

All in all, Ivy Bridge-E should be able to achieve the same frequencies as its predecessor with one exception: memory speeds. While Sandy Bridge-E had issues hitting anything above DDR3-2133 with 32GB installed, in our experience, the i7-4960X can easily hit 2400MHz provided your memory kit doesn’t hold thing back. DDR3-2666 should also be within reach, though looser timings will likely be needed.