The Core i7-5960X is Intel’s gleaming desktop flagship, with no compromises. It has the full-fat Haswell-E silicon, so it has a fearsome specification, including eight cores, 20MB of shared L3 cache and Hyper-Threading. However, the eight cores have resulted in a slight reduction in clock speed. The 5960X runs at 3GHz, which is less than the 3.6GHz from last year’s 4960X, and it’s also less than the frequencies of both the 5930K and even the 5820K.
As such, benchmark results aren’t always clear-cut. While this new chip is an undeniable beast in most of our tests, its lower clock speed and higher core count mean it isn’t always the best option in applications that aren’t heavily multi-threaded.
For example, in our image editing benchmark, the 5960X scored 49,038 points – but the two other Haswell-E parts, with their lower core counts and higher clock speeds, returned better results. At the top of the heap is the quad-core 4790K, which is cheaper than any Haswell-E chip, but has a higher clock speed. The 5960X flexed its multi-core muscle in our multi-threaded video encoding test, though, where it scored 405,115 – a long way beyond the 5930K’s 324,929, and miles ahead of every other CPU we’ve tested.
The 8-core chip further demonstrated its prowess in the heavy multi-tasking benchmark, where its score of 170,329 was almost 7,000 points ahead of the nearest challenger. It all adds up to a total system score of 158,169 – a fair distance ahead of the 137,066 from the i7-5930K.
Not surprisingly, the 5960X stormed through our heavily multi-threaded 3D content creation tests too. Its 1,392 score in Cinebench R15 was more than 300 points better than its nearest rival, and it led by almost eight seconds in Lightwave. The i7-5960X was also the only chip to hit a double-figure cycle frequency in Euler3D, and it took a healthy lead in Terragen as well.
The presence of eight powerful cores makes overclocking hard though. We raised the base clock to a modest 4.3GHz using a 1.4V vcore and the chip immediately throttled its speed. We only got the chip stable by reducing the vcore to 1.3375V, and couldn’t get the CPU to run at 4.4GHz with any amount of voltage. Heat was a constant theme when overclocking – the i7-5960X would certainly benefit from water cooling.
It isn’t the biggest boost, but the i7-5960X still demonstrated sizeable benchmark improvements. Its score jumped up by 70,000 in the video encoding test, and its overall system score jumped from 158,169 to 185,541. The gap between the i7-5960X and the i7-5930K widened when overclocked too. We saw similar improvements in the 3D content creation tests. The overclock added almost 300 points in the Cinebench test, and trimmed nearly four seconds from its Lightwave time.
Not surprisingly, though, the 5960X’s extra cores didn’t make much difference in our gaming tests. When running at stock speed, the chip’s minimum of 32fps in Shogun 2: Total War was only 1fps quicker than the other two Haswell-E parts. When overclocked, the i7-5960X’s minimum rose to 39fps – but that was only 2fps better than the i7-5820K.
On the plus side, the lower clock speed has a positive impact on idle power consumption. The i7-5960X’s idle figure of 64W was lower than both the other Haswell-E chips, and the peak consumption of 119W at load was only a few watts ahead of the other two Haswell-E chips. However, the 8-core chip was exceedingly power-hungry when overclocked, where its peak total system power draw was a mighty 493W.
There’s no denying the i7-5960X’s power in heavily multi-threaded workloads, but it isn’t dominant in every test – in applications where clock speed is preferable to cores, cheaper chips proved better bets. The stratospheric price also means that this chip is only suitable for those who perform an awful lot of 3D content creation and have deep wallets. If that’s you, though, this chip offers an incredible amount of power.
SPECIFICATIONS:
Frequency 3GHz
Turbo Boost frequency 3.5GHz
Core architecture Haswell-E
Manufacturing process 22nm
Number of cores Eight physical, 16 logical
Memory controller Quad-channel DDR4, up to 2,133MHz
Cache L1: 32KB (each core), L2: 256KB (each core), L3: 20MB (shared)
Packaging LGA2011-v3
Thermal design power 140W
Features SSE, SSE2, SSE3, , SSE4, SSE4.1, SSE 4.2, Turbo Boost 2, Hyper-Threading, Smart Cache, EIST, AES-NI LABS TEST / LGA2011-V3 CPUs
VERDICT: A monster chip, but its high price means it’s only for those who will use all of its power.