
We run two scripted Photoshop benchmarks: one that is threaded and fully taxes our CPUs, and another that leverages OpenCL acceleration to offload work to our graphics cards.
Since previous tests have shown us that OpenCL-accelerated filters don't leverage multiple GeForce cards in SLI, the single-GPU GeForce GTX 780 Ti takes a clear win. The CPU-dependent results are much closer, however. Four physical cores appear to be a sweet spot, and Hyper-Threading has no impact whatsoever.

CPU and memory performance are the critical variables in the rest of these tests, which do not leverage the heterogeneous nature of OpenCL. The Acrobat benchmark is single-threaded, and like iTunes and LAME, is affected by memory bandwidth. Meanwhile, Premiere Pro sees benefit from the Core i7's Hyper-Threading technology.


- Taking The SBM Down A Different Road
- CPU, Motherboard, And Cooler
- Video Card, Power Supply, And Case
- Memory, Hard Drives, And Optical Storage
- System Assembly And Overclocking
- Test System And Benchmarks
- Results: Synthetics
- Results: Media Transcoding
- Results: Rendering And Productivity
- Results: Adobe Creative Suite
- Results: Compression Tools
- Results: Battlefield 4 And Arma 3
- Results: Grid 2 And Far Cry 3
- Power And Temperature
- A Core i7 And Flagship GPU Impress, Naturally
(1) You could include temperatures and acoustics performance in the overall assessment, given I think that is a big part of the case buying decision, and
(2) A way to factor in the intangibles (i.e. blu ray vs dvd, choice of SSD/HDD, etc), you could include a separate vote between this quarter's and last quarter's to see what the readers would choose for the best build given all the performance factors, aesthetics, and other components that do not contribute directly to performance. The reader's vote of this quarter vs. last quarter and/or an overall value winner for this quarter could be included in the final write-up.
I would also 2nd the vote for starting 4K testing. And also, why not 1440p? It seems those two resolutions are more relevant now in 2014 at the level of this competition than 1600x900 and 4800x900 resolutions.
Hmm.... What percentage of the performance measures in this article are for gaming?
I'm thinking a selection of CPUs as a fixed starting point, and GPU decisions based on remaining budget. Maybe an i7, i5, FX-8, and an APU.
Would be really interesting to see the performance differences across workloads by allocating budget between CPU and other components.
Already done for ITX. See here:
http://www.tomshardware.com/reviews/build-your-own-haswell-overclocking,3608.html
I'd second the uATX. In fact, I'd really like to see Crash attempt a uATX dual-gpu setup.
Frankly, it was the cheapest available card when the systems were ordered.
Nope.
The purpose is to have a resolution that the low-budget PC can operate at for the comparison article at the end of the week.