Assembling Our Budget Box And Limited Overclocking
Building our $500 gaming rig was a simple, snag-free procedure. Most of the comments worth mentioning center on our $30 Rosewill microATX enclosure.
This rather small mid-tower case is noticeably heavier than the chassis we utilized last round. Double-boxed for added protection, its gross shipping weight was 50% greater. With more concessions for additional drives in a simislarly-sized enclosure, we were left with quite a bit less room for add-in cards. Our 9.5” GeForce GTX 560 Ti had to be angled a bit during installation. Once it was seated, though, it had about half of an inch of clearance between its shroud and the drive cage. Installed, the GeForce GTX 560 Ti does block the two center bays, though.
We again have to criticize Gigabyte's SATA port placement. Installing a dual-slot graphics card renders two of the H61 chipset's integrated ports inaccessible. The problem isn't huge, since we still have access to a pair of integrated connectors, along with two SATA 6Gb/s-capable ports, enabled by a Marvell controller.
Gigabyte bundled two cables with 90-degree SATA connectors in our retail package, rather than including one straight cable, as it did last quarter. This almost presented a problem. The cable either interfered with the graphics card or the bottom of the case, depending on the bay we used for mounting our hard drive.
Cable management was much easier this time around, though. In fact, we didn't use a single tie-strap in our system photos, just to show that it could be done. The enclosure itself also benefited from improved rear-panel rigidity, addressing one of our major concerns last time around. Unfortunately, slightly warped side panels were far more difficult to remove and install. Once they were locked down by thumbscrews, they wouldn't quite line up precisely with the front bezel.
Finally, although I was pleasantly surprised by the quietness of last quarter's 80 mm exhaust and 120 mm intake fans, it turns out that this enclosure's single 120 mm exhaust cooler isn't as quiet. It didn't hum or tick, but the sound of air turbulence made it the most obviously-audible fan in the system. As someone who appreciates a quiet PC, I would probably replace the cooler or experiment with a step down to 5, 7, or 9 V.
With no access to base clock settings, multiplier ratios, or memory data rates above 1066 MT/s, we’re limited once again to pushing lower latencies for any performance increase. Automatic memory timings dialed in 7-7-7-19 at 1.5 V. With a bump to 1.6 V, we found stability at 6-6-6-14 1T. Like last quarter, every attempt to manually adjust memory timings was answered by rapid beeps and a power cycle. The new timings were eventually applied after the restart.
MSI Afterburner served our purposes for overclocking ECS' GeForce GTX 560 Ti. Since I haven't been overly aggressive with our last few Radeon cards, staying within the bounds of AMD Overdrive, we didn't make any attempt to alter our GPU's 1.0 V core setting.
With core clock expectations between 900 and 950 MHz, I began stability testing at 880 MHz, bumping up the core in 10 MHz increments. Attempts to push beyond 900 MHz were punished with artifacts in DiRT 3’s game menus. Here's a helpful hint: if you aren't using DiRT 3, F1 2010, S.T.A.L.K.E.R.: Call of Pripyat, or Just Cause 2 for graphics stability testing, I highly recommend you add at least one of them. They have a tendency to demonstrate weaknesses in stability earlier than artifact scanners, synthetics, or other demanding titles like Metro 2033, Battlefield 3, or Crysis.
As I often do, I played it safe and quit overclocking the GDDR5 memory at 1125 MHz (4500 MT/s) before reaching the upper limits of stability. Our final testing frequencies were then dialed back to 891 MHz core and 1102.5 MHz (4410 MT/s) for the memory.