Professional Performance: Windows

Agisoft Photoscan – 2D to 3D Image Manipulation: link

Agisoft Photoscan creates 3D models from 2D images, a process which is very computationally expensive. The algorithm is split into four distinct phases, and different phases of the model reconstruction require either fast memory, fast IPC, more cores, or even OpenCL compute devices to hand. Agisoft supplied us with a special version of the software to script the process, where we take 50 images of a stately home and convert it into a medium quality model. This benchmark typically takes around 15-20 minutes on a high end PC on the CPU alone, with GPUs reducing the time.

Agisoft PhotoScan Benchmark - Total Time

The presence of Crystal Well had a small effect on Photoscan, occurring mostly in the second phase of the calculation which is the one that also has an option to enable the GPU, indicating that memory bandwidth is an potential limitation in that segment.

Cinebench R15

Cinebench is a benchmark based around Cinema 4D, and is fairly well known among enthusiasts for stressing the CPU for a provided workload. Results are given as a score, where higher is better.

Cinebench R15 - Single Threaded

Cinebench R15 - Multi-Threaded

Cinebench is a historically CPU-limited benchmark, and the results show this again here. The fact that the 3.6GHz Broadwell-based i5-5675C performs so closely to the 3.9GHz Haswell-based i5-4690 is a promising sign here, as it means that despite being a mere "tick" in Intel's development efforts, there are tangible IPC increases on the desktop from Broadwell.

HandBrake v0.9.9: link

For HandBrake, we take two videos (a 2h20 640x266 DVD rip and a 10min double UHD 3840x4320 animation short) and convert them to x264 format in an MP4 container.  Results are given in terms of the frames per second processed, and HandBrake uses as many threads as possible.

HandBrake v0.9.9 LQ Film

HandBrake v0.9.9 2x4K

While no obvious improvement was seen in the low quality conversion, the double UHD conversion put the i7 above what was otherwise expected.

Hybrid x265

Hybrid is a new benchmark, where we take a 4K 1500 frame video and convert it into an x265 format without audio. Results are given in frames per second.

Hybrid x265, 4K Video

Unlike the Handbrake H.264 tests, the Hybrid x265 tests show a clear uptick in performance on the Broadwell processors. It is not fast enough to catch the i7-4790K and its 4.4GHz turbo clockspeed, but we see the i5-5675C shoot well past the i5-4690 despite the clockspeed deficit. Whether this is due to Broadwell architecture enhancements, Crystal Well acting as an L4 cache, or a combination of the two is difficult to determine, but the end result is substantial.

Office and Web Performance Professional Performance: Linux
Comments Locked

196 Comments

View All Comments

  • extide - Tuesday, June 2, 2015 - link

    Have you ever overclocked a CPU before? Those limits are easily raised, and a properly O/C'd build will not throttle...
  • DCide - Tuesday, June 2, 2015 - link

    It's hardly elite - just live video. But if the CPU saturates or crashes, everybody will know.

    I'm not skilled at overclocking. But for this application, stability is paramount. I wouldn't want to push it close to the limit, only to have it crash 5 hours into the stream. When I tested 4-core CPUs I had to run at low resolutions, or else the video would stutter. And I even at that I couldn't use all the features of the application.

    With the same low-profile case and the same heatsink/fan the 4790K would throttle in less than 30 minutes while the 5820K and 5960X never would. They were all running at 4GHz. I'm sure someone more skilled could make all 3 CPUs run faster. But the 8-core is still going to be about twice as fast as the 4-core.

    On the topic of the new CPUs, I won't be surprised if they O/C well enough to make them dominate the previous generations. But they won't be a match for the 2011-3 CPUs.
  • vision33r - Tuesday, June 2, 2015 - link

    If you're into server workloads, 8 cores would nice. For gaming, more cores only adds more heat and hardly any benefit to gaming performance. For gaming, quad-core is good enough with no integrated GPU. All this integrated GPU stuff is just a waste of resources. Most of us into performance and gaming don't want the extra heat and power consumption.
  • FlushedBubblyJock - Sunday, June 14, 2015 - link

    If the kiddies can play WoW on it, amd and intel are happy. They will never stop.
  • Galatian - Tuesday, June 2, 2015 - link

    Would like to see some Quicksync Benchmarks! Anand noted fast Quicksync performance on the Iris Pro 5200, so I sure would like to see a) the differences between GT3e (Haswell) and GT3e (Broadwell) and b) the differences between GT2 (Haswell) and GT3e (Broadwell).
  • Refuge - Tuesday, June 2, 2015 - link

    I think that is coming in part 2 judging by the first two pages of the article.
  • Kevin G - Tuesday, June 2, 2015 - link

    I'll second the request for QuickSync testing. I do a bit of video editing with Vegas and do some bulk conversions with Handbrake so I have a keen interest on this topic.
  • dj_aris - Tuesday, June 2, 2015 - link

    I'd like to see the 65w 4770R in the charts to measure the delta over a a generation (okay half a generation). Meanwhile, considering that i5-5675C is $276 but games much worse than a Haswell Pentium + a decent $200 discrete card I really don't see the point in traditional desktops. The only scenario I see fit would be in a tiny case using a super low profile thin mini-ITX board, so as to create a system similar to, say, an Alienware Alpha. But wait, there aren't any thin mini-ITX H or Z97 boards. Whereas back with Haswell we had thin mini boards but only non-Iris iGPUs. Oh Intel, there's always a part missing!
  • DCide - Tuesday, June 2, 2015 - link

    I think you mean you don't see the point for a gaming desktop. A Pentium + $200 dGPU would be a poor choice for most traditional desktops.
  • Mech0z - Tuesday, June 2, 2015 - link

    With DX12 supporting multiGPU much much better, could it be theorized that these will be very good for DX12 games together with a dGPU?

Log in

Don't have an account? Sign up now