The nearest equivalent of the Core i7-4960X in the enterprise lineup is the Xeon E5-1660 V2. In terms of my testing at AnandTech, the i7-4960X represents the standard enthusiast processor that blitzes our benchmarks, and thus an opportunity to test something potentially faster is always welcome.

At the ultra-high-end of any CPU range, we can see a fight for cores against MHz to remain within the thermal design power limitations. Users can spend their money on more cores, which benefits parallel computation, or focus purely on MHz for single-threaded throughput. The downside of moving to higher MHz is usually efficiency, so the gains might not be as linear as expected.

This review tested two of the high end Intel E5-26xx processors – the 12-core 130W E5-2697 v2 and the 8-core 150W E5-2687W v2. The former is also the 12-core representative in the late 2013 Mac Pro, whereas the latter is the highest TDP processor that Intel makes in this segment. A few other CPUs share this honor, although they are part of the Ivy Bridge-EX E7-x8xx line. My goal was to find out where these two CPUs stand in what I consider ‘an enthusiast user’s scenario’, and as such we used the same benchmarks as in the AMD Kaveri launch article, involving gaming, compression, rendering, video conversion and 2D image to 3D modeling creation. Johan has dealt extensively on the enterprise server and high performance computing aspect of similar CPUs, and his deep dive into the functionality is worth a read if you have not already seen it.

Intel E5-2697 v2 - 12C/24T, 2.7 GHz (3.5 GHz Turbo), 130W

This processor is the most expensive E5-26xx CPU you can purchase, tipping the scales at $2614 (Intel price), but is expandable into dual socket systems. For the green we get 12-cores at a max loading of 3.0 GHz (base frequency + 3 turbo bins), which for most purposes should blitz through any multithreaded workload we can throw at it. The benchmarks tell the story, particularly when it comes to PovRay and the multi-threaded version of 3D Particle Movement – anything that can be subdivided up with no overhead benefits greatly from more cores over more MHz. But looking at other software that cannot take advantage of all the cores (Xilisoft seems to only use half cores on a single file at low resolution) then a processor with more MHz under the hood becomes the right choice.

Unfortunately anything over 6-core loading reduces it down to that lower 3.0 GHz mark, whereas single threaded speed is up at 3.5 GHz. Ultimately it is up to the motherboard to implement which turbo modes and P states are in use, and on the consumer line we often find motherboards using a form of ‘MultiCore Turbo’ (read our explanation here). If the E5-2697 v2 was put in this position, we would have 12 cores at 3.5 GHz, ready to blast through the workload.

At this level of single socket production, the price might seem outrageous to home users. However if we consider a workstation scenario (such as rendering at the office) which requires 256GB of DRAM and a beefy CPU, then the DRAM can easily be half the cost of the system – or even the software license can outstrip that. The E5-2697 v2 is the king of the 12-core Intel CPUs in the E5-26xx space. It makes me want to see the Haswell-E versions as soon as possible to see where we stand.

Intel E5-2687W v2 - 8C/16T, 3.4 GHz (4.0 GHz Turbo), 150W

At some point in the socketed processor space, we have to consider ‘what is the absolute thermal limit of a processor?’  Over the last couple of decades we have seen it rise from 20W to 40W, 95W, 115W, 130W, 150W and if we glance sideways to AMD, even 220W seems to be on the cards. The increase of power consumption is from more cores, more frequency and more voltage – as the high end is pushed, efficiency drops and we need more power to get a smaller increase in performance. However there are users who would pay for that extra 100 MHz all the time. This is why the E5-2687W exists – it is simply the 8 core version of the i7-4960X at the same clock speeds. But the power consumption for 33% more cores is actually only 15%, because Intel tightens up the frequency/voltage characteristics for these models.

While the E5-2687W v2 performs almost identical to the i7-4960X at single thread benchmarks, and then beats it in the variable-threaded scenario, it does come at a 2X cost. A user with an i7-4930K could argue that with a small overclock, their purchase could be up to 4X the value. But again, part of the added cost comes in the Xeon features – memory support, 2P system compatibility, virtualization and so forth.

I would actually go ahead and say that Intel has kind of shot themselves in the foot with this processor. The reason for this comment is based on another model in the product stack, the E5-2667 v2. If I line them up side by side, it should become obvious why:

Intel E5 SKU Comparison
  Xeon E5-2687W v2 Xeon E5-2667 v2
Release Date September 10th, 2013
Cores 8
Threads 16
Base Frequency 3400 3300
Frequency 1C 4000
Frequency 2C 3900
Frequency 3C 3800
Frequency 4C 3700
Frequency 5C 3600
Frequency 6C 3600
Frequency 7C 3600
Frequency 8C 3600
L3 Cache 25MB
Max TDP 150W 130W
Max Memory Size 256 GB 768 GB
Memory Channels 4
Memory Frequency DDR3-1866
PCIe Revision 3.0
PCIe Lanes 40
Multi-Processor 2P
VT-x Yes
VT-d Yes
TSX-NI Yes
Memory Bandwidth 59.7 GB/s
Price (Newegg) $2108 $2057

The E5-2667 v2 is the same speed at any core loading as the E5-2687W v2, the same cache, the same features, except it is slightly cheaper, uses less power and supports more memory. Sounds like an easy win for the E5-2667 v2.

Unfortunately I could not find the E5-2667 v2 for sale as easily as the E5-2687W v2. The sole UK retailer I found with an E5-2667 v2 was not one I was familiar with; however Newegg will sell you the E5-2687W v2 for $2200. This feeds back into another issue with Intel’s SKU policy – only certain SKUs will be sold direct to the public, while others might go only to OEMs and system integrators, like SuperMicro, Dell, HP and so on. We find this issue on the LGA1150 Xeons as well, where the low power SKUs like the Xeon E3-1230L v3 are not on general release. An ideal solution for this would be for Intel to sell direct to the consumer, rather than regional sales offices deciding which models each region needs (and thus limiting our selection).

Gaming Benchmarks: Sleeping Dogs, Company of Heroes 2 and Battlefield 4
Comments Locked

71 Comments

View All Comments

  • XZerg - Monday, March 17, 2014 - link

    this bench also shows that the haswell had almost no CPU related performance benefits over IVB (if not slowed down performance) looking at 3770k vs 4770k and that haswell ups the gpu performance only.

    i really question intel's skuing of haswell...
  • Nintendo Maniac 64 - Monday, March 17, 2014 - link

    Emulation?
  • BMNify - Monday, March 17, 2014 - link

    its a shame they didn't do a UHD x264 encode here as that would have shown a haswell AVX2 improvement (something like 90% over AVX), and why people will have to wait for the xeons to catch up to at least AVX2 if not AVX3.1
  • psyq321 - Wednesday, March 19, 2014 - link

    There is no "90% speedup over AVX" between HSW and IVB architectures.

    AVX (v1) is floating point only and thus was useless for x264. For floating point workloads you would be very lucky to get 10% improvement by jumping to AVX2. The only difference between AVX and AVX2 for floating point is the FMA instruction and gather, but gather is done in microcode for Haswell, so it is not actually much faster than manually gathering data.

    Now, x264 AVX2 is a big improvement because it is an integer workload, and with AVX (v1) you could not do that. So x264 is jumping from SSE4.x to AVX2, which is a huge jump and it allows much more efficient processing.

    For integer workloads that can be optimized so that you load and process eight 32-bit values at once, AVX2 Xeon EPs/EXs will be a big thing. Unfortunately, this is not so easy to do for a general-purpose algorithms. x264 team did the great job, but I doubt you will be using 14 core single Haswell EP (or 28 core dual CPU) for H.264 transcoding. This job can be done probably much more efficient with dedicated accelerators.

    As for the scientific applications, they already benefit from AVX v1 for floating point workloads. AVX2 in Haswell is just a stop-gap as the gather is microcoded, but getting code ready for hardware gather in the future uArch is definitely a good way to go.

    Finally, when Skylake arrives with AVX 3.1, this will be the next big jump after AVX (v1) for scientific / floating point use cases.
  • Kevin G - Monday, March 17, 2014 - link

    Shouldn't both the Xeon E5-2687W v2 support 384 GB of memory? 4 channels * 3 slots per channel * 32 GB DIMM per slot? (Presumably it could be twice that using eight rank 64 GB DIMMs but I'm not sure if Intel has validated them on the 6 and 10 core dies.) Registered memory has to be used for the E6-2687w v2 to get to 256 GB, just is the chip not capable of running a third slots per channel? Seems like a weird handicap. I can only imagine this being more of a design guideline rule than anything explicit. The 150W CPU's are workstation focused which tend to only have 8 slots maximum.

    Also a bit weird is the inclusion of the E5-2400 series on the first page's table. While they use the same die, they use a different socket (LGA 1356) with triple memory support and only 24 PCI-e lanes. With the smaller physical area and generally lower TDP's, they're aimed squarely the blade server market. Socket LGA 2011 is far more popular in the workstation and 1U and up servers.
  • jchernia - Monday, March 17, 2014 - link

    A 12 core chip is a server chip - the workstation/PC benchmarks are interesting, but the really interesting benchmarks would be on the server side.
  • Ian Cutress - Monday, March 17, 2014 - link

    Johan covered the server side in his article - I link to it many times in the review:
    http://www.anandtech.com/show/7285/intel-xeon-e5-2...
  • BMNify - Monday, March 17, 2014 - link

    a mass of other's might argue a 12 core/24 thread chip or better is a potential "real-time" UHD x264 encoding machine , its just out of most encoders budgets, so NO SALE....
  • Nintendo Maniac 64 - Monday, March 17, 2014 - link

    Uh, where's the test set up for the 7850K?
  • Nintendo Maniac 64 - Monday, March 17, 2014 - link

    Also I believe I found a typo:

    "Haswell provided a significant post to emulator performance"

    Shouldn't this say 'boost' rather than 'post'?

Log in

Don't have an account? Sign up now