«

»

Intel Core i7-5960X Extreme CPU Performance Review

PAGE INDEX

<< PREVIOUS            NEXT >>

AIDA64 Engineer Tests

AIDA64 is FinalWire’s full 64-bit benchmark and test suite utilizing MMX, 3DNow! and SSE instruction set extensions, and will scale up to 32 processor cores. An enhanced 64-bit System Stability Test module is also available to stress the whole system to its limits. For legacy processors all benchmarks and the System Stability Test are available in 32-bit versions as well.

All of the benchmarks used in this test- Queen, Photoworxx, ZLib, and hash- rely on basic x86 instructions, and consume very little system memory while also being aware of Hyper-Threading, multi-processors, and multi-core processors. Of all the tests in this review, AIDA64 is the one that best isolates the processor’s performance from the rest of the system. While this is useful in that it more directly compares processor performance, readers should remember that virtually no “real world” programs will mirror these results.

aida64

The Queen and Photoworxx tests are synthetic benchmarks that iterate the function many times and over-exaggerate what the real-world performance would be like. The Queen benchmark focuses on the branch prediction capabilities and misprediction penalties of the CPU. It does this by finding possible solutions to the classic queen problem on a chessboard. At the same clock speed theoretically the processor with the shorter pipeline and smaller misprediction penalties will attain higher benchmark scores.

Like the Queen benchmark, the Photoworxx tests for penalties against pipeline architecture. The synthetic Photoworxx benchmark stresses the integer arithmetic and multiplication execution units of the CPU and also the memory subsystem. Due to the fact that this test performs high memory read/write traffic, it cannot effectively scale in situations where more than two processing threads are used, so quad-core processors with Hyper-Threading have no real advantage. The AIDIA64 Photoworxx benchmark performs the following tasks on a very large RGB image:

  • Fill
  • Flip
  • Rotate90R (rotate 90 degrees CW)
  • Rotate90L (rotate 90 degrees CCW)
  • Random (fill the image with random colored pixels)
  • RGB2BW (color to black & white conversion)
  • Difference
  • Crop

The 5960X acquits itself well here, turning in scores 10% higher than the 3960X at stock clocks and a startling 47% higher when overclocked. As we’ve seen before, Photoworxx is relatively insensitive to clock speed but does make effective use of more cores.

aida64_zlib_hash

The ZLIB scores are compressed by the scale of the Hash scores, but the numbers give the 5960X a 17% better score than the 3960X. The real surprise here, though, are the Hash scores, where the 5960X returns performance much further above the 3960X than just two extra cores would lead you to expect. Its score of 6763 is almost 60% higher than that of the 3960X.


SKIP TO PAGE:

<< PREVIOUS            NEXT >>

11 comments

Skip to comment form

  1. Chris

    On the last page, you wrote 48 vs 24 PCI-E lanes. I believe that should be 40 or 28 PCI-E lanes versus 16.

    In terms of value, it’s hard to justify this processor for most people. Only people with multi-threaded work can really benefit from this.

    I think though that the 5820k might be a decent value though. For perhaps $80 in a CPU compared to a 4790K and around $50-$100, you do get another 2 cores, which might be useful, although there will be a premium you have to pay for DDR4.

    1. David Ramsey

      What I said was “48 (total) PCI-E lanes as compared to the 24 lanes of an LGA1150 system”. Since I was talking about systems rather than CPUs, I included the PCI-E lanes provided by the chipsets as well.

  2. Ethan

    1. Page 2, you say Z79 instead of X79
    2. Page 8 says the RAM on the 3960X is running at 1066 while on page 2, you say that you are running 1600. Which is it?
    3. Page 8, you typed 3096X instead of 3960X.
    4. Why no clock for clock comparison? I mean in both the CPU and RAM speed, especially since you are giving tests scores with the 5960X being overclocked?

    1. David Ramsey

      Thanks for the corrections; I’ve updated the article.

      Clock for clock comparisons are are interesting if you’re into CPU architecture, or like to make people think you are. But there are so many other factors– amount of cache, clock speed which varies based on number of active cores, and so forth– that I think real-world performance tests are more useful.

      Overclocking results are never guaranteed. Of course I always include overclock results for the CPU I’m testing, and I’ve gone back and forth on whether or not to include overclocking results from previously-tested comparison CPUs. Currently I don’t include them since it introduces another uncertainty into the comparison.

  3. Tradesman

    Why 1600 DRAM on 1150 and X79 – 2133 would have been more appropriate?

    1. David Ramsey

      Two reasons:

      — Neither the 4770K nor 3960X officially support DDR3-2133. It is a supported speed for Haswell-E.

      — In any case I didn’t have any DDR3-2133 available.

  4. Tradesman

    If you can put your hands on some 2133 (decent sticks) you’d be surprised at the change you’ll see, and both the 4770K and the 3960X both easily run 2133…

    ” 46.5 gigabytes per second is about 20% higher than we see from the late-2011 Core i7-3960X. As usual, overclocking the CPU has no effect on memory bandwidth.”

    OCing won’t have any real effect, but it’s an apples and oranges difference in the 3960X results when also running at 2133 (and that’s where the bandwidth differences in your charts come from – 2133 vs 1600)…

    The slow CLs in DDR4 have caused many to reconsider moving to X99

    1. David Ramsey

      Benchmark Reviews has been testing memory for many years, and we’ve never seen much real-world difference with expensive, high speed enthusiast memory vs. standard memory. Synthetic benchmarks, of course, will be different.

      There’s another reason I wouldn’t do this, though: when we test components, we try to isolate the performance of the component as most users would see it, not as most users with lots of money who will equip their systems with high end memory, SSDs, and so forth. By sticking to the supported memory speeds for each platform, I’m providing a more accurate look at relative CPU performance, rather than “Haswell-E with stock memory vs. Sandy Bridge E with unsupported high speed memory”.

      Still, testing each CPU with high speed memory on enthusiast motherboards would make an interesting article in its own right. Maybe someday…

    2. Caring1

      I run 2133MHz Ram in an Ivy Bridge system with an i5K series, it has no problems running in that, but as David says, there is no appreciable difference to 1600MHz, but it does give me that warm and fuzzy feeling knowing I have fast Ram. 😉

  5. spikey27

    It’s looking like the end of the line has been reached, with regard to heat-related issues, the overclocking tied to the heat, cost, and actually performance.

    So, Intel’s predictions about the demise – or maybe more correctly, the end of the development road – is rapidly coming into view.

    Except for the guys who actually need a zillion of everything (cores, PCI-E lanes) and just about everything else that keeps climbing with each new chip issue, and maybe the highest calibre gamers, it looks like everything has been invented, and the gravy offered by Haswell-E, etc. may not be such a necessity after all.

    Just my 2cents worth.

    1. Olin Coles

      Your comment reminds me of an editorial I wrote for this website nearly four years ago:
      Intel Sandy Bridge CPUs Chill Aftermarket Cooling. I argued much the same, and lost some sponsors in the process. Read more: http://archive.benchmarkreviews.com/index.php?option=com_content&task=view&id=13488&Itemid=8

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

*