The following benchmarks are based on AutoCAD 2015, Cadalyst 2015, and three modules of SPECviewperf 2015. They were specifically chosen to represent the CPU performance in this context well, in spite of us using a Palit GeForce GTX 980 Super JetStream OC graphics card and not a workstation graphics card.
AutoCAD 2015 2D and 3D Performance
We’ve already described why and how we’re using AutoCAD in the integrated graphics section. Suffice it to say here that the CPU needs to help out quite a bit when it comes to 2D graphics acceleration, since this type of acceleration hasn’t existed via the GPU since Microsoft Windows Vista. Neither the driver model nor the unified shader architecture provides this functionality.
The positions in this benchmark are solely determined by the CPU, since the speed of the graphics is the same all around.
It’s interesting that the smaller CPU comes out a little bit ahead for 3D yet again. This trend reverses once SMT has been turned off, though.
With Autodesk’s Maya 2013, a current and popular application was added to the benchmark suite. Viewport 2.0 isn’t part of the results on purpose, since it’s based on DirectX. This means that this benchmark is based exclusively on OpenGL. The render modes used for this benchmark are shaded, ambient occlusion, multi-sample anti-aliasing, and transparency, and the model consists of 727,500 vertices.
Showcase 2013 is a DirectX-based benchmark. Autodesk might still be the only major company to have made the jump to DirectX, but many smaller developers have taken the plunge as well. The benchmark model used for this benchmark uses eight million vertices, as well as render modes such as shading, projected shadows, and self-shadowing. What it comes down to in the end is the clock frequency.
SolidWorks 2013 SP1
The different models for our SolidWorks 2013 workloads range in size from 2.1 to 21 million vertices. Individual tests use the software's many rendering modes, including a shaded mode, shaded with edges, ambient occlusion, shaders and environment maps. Compared to SPECapc for SolidWorks 2013, the CPU test is gone, the number of models is lower and a benchmark with parallax effects was added.
Last I heard Skylake was supposed to support DDR3 and DDR4. Was that just a rumor that wasn't the truth or will it actually support DDR3 as well?
hmm if intel whacked a few iris's on a gfx chip and did there thing they could possibly beat NVidia...and make even more money lol, hmm mutli 128mb ring buses and iris core's and hbm...delicious
i missed some things:
unlocked broadwells but no o.c. not even a little look into how these overclock and behave o.c.ed.
no comparison (gaming, power use, htpc etc.) with the desktop haswell i5 and i7 -R cpus' iris pro igpus. the amd comparisons were good though. i hope you guys test these against the haswell iris pro later.
in some of the charts, the core i7 5775 was written as i7 7557.
in the test setup page, system memory section, is it "transcend" instead of "transcent"?
one last thing: do these unlocked broadwell cpus really have 16x gen 3.0 lanes off the processor? i thought these were soc dies (with southbridge disabled) with 8x gen 3.0 lanes.