Overall System Performance Benchmarks
With PC Mark 8 being in a transitional period, we still provide scores for PC Mark 7. The default test suite in PC Mark 7 is heavily biased to video encoding and reliant on the CODECs that are built into Windows 7. Because of this, when run on Windows 8 we see performance biased towards GeForce thanks to its more robust and polished 'NVENC' Video Encoding Processor.
PCMark 8 Incorporates OpenCL for the first time however in our testing we feel the feature was no optimally integrated in the first release of this benchmark.
PC Mark 8 incorporates both simulated test cases such as content creation and web, relying on the benchmarks own engine to run these. Additionally, PC Mark 8 can take advantage of Microsoft Office and Adobe Creative Suite and run benchmarks using these installed applications. The latter scores are independent of the main benchmark. The actual tests/tasks being run as well as he version of the Adobe app being used are very critical when considering
OpenCL performance as Adobe is constantly improving their OpenCL Support. This will change as both Adobe Creative Suite and PCMark 8 evolve. This review was our first use of PC Mark 8. Adobe has promised significant upgrades to Creative Suite plus their compute implementation also scales with number of GPUs.
It is important to note that CUDA and OpenCL are not 'uber accelerators', they can either offload real-time tasks to the GPU or it can offload highly parallel tasks, they cannot accelerate any random task.
The best demonstration of OpenCL is non-linear video editing where effects such as filters, graphics and colour adjustments can be applied in parallel in real time. The speed of the CPU in our test bed combined with the workload means both cards give similar performance in the benchmark, nulling the GPU's advantage.
We provide some other test suite overall scores such as Passmark 8 and Sandra 14 both of which incorporate graphics and compute testing and both of these tests are very similar thanks to the overall power of the GPUs.
