The Intel Broadwell Review Part 2: Overclocking, IPC and Generational Analysis
by Ian Cutress on August 3, 2015 8:00 AM ESTProfessional 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.
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.
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.
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.
121 Comments
View All Comments
Oxford Guy - Tuesday, August 4, 2015 - link
No gaming results.TheJian - Monday, August 3, 2015 - link
I was hoping any more coverage of broadwell would include ripping quality comparisons to haswell at least. Is it still fast but crappy, or have they fixed quality so I don't have to keep my gpu off? :( Throw some handbrake tests in please. Quicksync fixed yet?http://www.anandtech.com/show/7007/intels-haswell-...
Any changes since this? Or the review by anand that covered it (linked in there)? Or do we all just hope for a fix with skylake? I saw a recent software update for haswell, but not sure if that does anything about quality here.
Enterprise24 - Tuesday, August 4, 2015 - link
Grid Autosport with 290X show very strange result. I assume this game support AVX2 instruction set ? Since Sandy and Ivy have roughly the same performance. But jump to Haswell gain big improvement.StrangerGuy - Tuesday, August 4, 2015 - link
Yawn...At this point I'm more interested in much better utilisation of hardware through software like DX12 than sinking tens of billions into CPU die shrinks with next to zero real world benefit. The paradox here is of course how the former will make the latter even more irrelevant as it is.lukarak - Tuesday, August 4, 2015 - link
i7-920 waving...Still no reason to upgrade. It was released in 2008, bought it early 2009. and it has been quite sufficient for almost 6.5 years now.
HeJoSpartans - Tuesday, August 4, 2015 - link
Hello Ian,Unfortunately, your IPC increase charts on page 3 of the article reveal several mistakes. The benchmark charts show that the Broadwell chip is not the fastest on both the 3DPM:ST as well as the CBenchR15:MT benchmarks, while your IPC increase charts tell something completely different. Also, some of the other calculated IPC increase values stated in the charts are completely irreproducable on my side. Second, you made a systematic mistake in calculating the IPC increase for the "Lower is better" benchmarks. This is very crucial especially in the Dolphin benchmark, where the total IPC increase from SB to Broadwell would be 58.0% rather than 36.7%. To make this clear: Processor A, that takes half the time for performing a certain task compared to processor B, does not offer a 50% increase in IPC over B, but a total of 100%.
You should re-check your numbers. Hope this helps.
Ian Cutress - Tuesday, August 4, 2015 - link
Hi HeJoSpartans,Somehow the incorrect benchmark result graphs were placed in those spots and they were from the non 3GHz testing - they also had a different z-height. I have updated it - the IPC numbers for those benchmarks in the main graphs are still accurate. For those benchmarks at stock voltage, the balance between frequency and IPC as to which is more important plays out on a larger scale for sure.
Also, with the timed benchmarks. Arguably I actually labelled the axis in terms of Percentage Improvement rather than IPC improvement, despite the title of the benchmark. But you are correct - I mistakenly used the % improvement and the term IPC interchangeably. I have updated the results with a disclaimer.
Any other issues, let me know. I'm also contactable by email if urgent!
-Ian
Navvie - Tuesday, August 4, 2015 - link
No compelling reason to upgrade from my 4770k.Intel needs to get back to working on CPU development rather than GPU.
Speedfriend - Tuesday, August 4, 2015 - link
I have a question. I have seen a article on Intel that speculates that it intends to launch a new chip that has a larger eDRAM (1Gb) and then has 3d Xpoint (15GB) on. The large eDRAM will compensate for the lower write speed of the 3d Xpoint, however the overall chip will offer massive advantages in power consumption and having nonvolatile memory for sleep states. This would be extremely competitive for mobile computing and servers.Given this quote above "Cycling back to our WinRAR test, things look a little different. Ivy Bridge to Haswell gives only a 3.1% difference, but the eDRAM in Broadwell slaps on another 16.6% performance increase, dropping the benchmark from 76.65 seconds to 63.91 seconds. When eDRAM counts, it counts a lot." would that make sense.
The article says that this change in chip design is the reason we are seeing another tock for kaby lake.
Any view from CPU experts on here?
NeilPeartRush - Tuesday, August 4, 2015 - link
1999: Intel Celeron 300A @ 450MHz2003: AMD Athlon XP-M 2200+ @ 2.40GHz
2007: Intel Q6600 @ 3.60GHz
2011: Intel i5-2500K @ 4.8GHz
2015: Skylake?