Intel Z68 Test Setup
This article is designed to test the two new features of the Z68 chipset – the Lucid Virtu hybrid graphics software and Intel’s Smart Response SSD caching technology. As we didn’t want one technology to interfere with the other, we tested each in isolation. This meant installing Windows on just a hard disk for Virtu testing, then testing the performance of that solitary hard disk, and then scrubbing the install to create the Smart Response set up.
We tried to keep the test hardware as consistent as possible between the two sets of tests in order to keep the number of variables to a minimum.
- 3.3GHz Intel Core i5-2500K (Turbo Boost disabled)
- Gigabyte Z68X-UD3H-B3 motherboard
- 4GB of Corsair XMS3 1,600MHz DDR3 memory (at default SPD settings of 1,333MHz, CL9)
- 1TB Western Digital Caviar Black hard disk
- 20GB Intel Solid-State Drive 331 (for Smart Response testing only)
- Asus GeForce GTX 590 3GB graphics card (GeForce 270.61 WHQL driver)
- Nvidia GeForce GTX 570 1.3GB graphics card (Virtu testing, GeForce 270.61 WHQL driver)
- Corsair Hydra H70 CPU cooler
Lucid Virtu Test Methodology
There are a few key questions to answer regarding the hybrid graphics of Lucid Virtu, mainly based around how much power it can save by sending basic tasks and video encoding to the Sandy Bridge Processor Graphics, and whether there’s a performance drop when using that discrete graphics card with Virtu enabled.
We therefore tested the system with no graphics card installed in a range of tests to gauge the performance and power draw of the PG. We then uninstalled the PG driver and ran the same tests with just the discrete graphics card. Finally, we re-installed the PG driver, installed Lucid Virtu and re-ran the tests for a third time.
We used 3DMark11 and Arma II: Operation Arrowhead to gauge how much performance was lost (if any) by using the Virtu software, as both are sensitive to system-wide performance levels. The power draw of the system was measured at the wall, and we measured the power when idle, playing video and video encoding (to see whether it was the PG or the GTX 590 3GB that was being used).
We used BadaBoom 2.0 for the video encoding test, as this can address both the GTX 590 3GB and the PG. We used the YouTube preset, but forced the Quality slider to maximum, giving us an output of 8Mb/sec. We also entered the Advanced menu and set Hardware to Quality (although the GTX 590 3GB doesn’t support this, so we couldn’t force this for that test). For the video playback test we used our most taxing video clip, a 33MB/sec 1080p .MKV file affectionately called Killa Sample, which we played via VLC 1.1.9.
We’ve presented all three tests – representing pure PG performance, pure discrete graphics performance and Virtu performance - where appropriate in our graphs. However, as the PG doesn’t support DirectX 11 or Arma II, the gaming performance tests are purely a comparison of discrete graphics card performance versus Lucid Virtu performance.
To test Smart Response, we first gauged the performance level of the Western Digital Caviar Black on its own by using HD Tach 184.108.40.206. This application had to be run in Windows XP (SP3) compatibility mode. We also manually timed how long it took to launch Arma II: Operation Arrowhead, stopping the timer as soon as the main menu appeared. We repeated each test three times, ignoring anomalies and averaging the remaining data to give an accurate time.
We then added the Intel Solid-State Drive 331 SSD and re-installed Windows to ensure all the RAID options were set correctly. Once we’d re-installed all the drivers (in the same order as before) we re-ran the tests. As Smart Response is an intelligent caching technology, we recorded the times of each of the three runs and presented these on the graphs. Run 3 is representative of a mature cache, while Run 1 shows the level of performance as the cache is being constructed for the first time.