How we test
Comparison systems |
||||||
|---|---|---|---|---|---|---|
| Motherboard | Gigabyte Z68XP-UD3P | ASUS P8Z68-V PRO | ASUS P8P67 PRO | Gigabyte P67A-UD3 | ASUS P8P67 Deluxe | ASUS Crosshair IV Extreme |
| Motherboard BIOS | F4 | 8801 | 1401 | F5 | 1053 | 1102 |
| Processor | Intel Core i5 2500K (ES) | AMD Phenom II X6 1055T | ||||
| Chipset driver | Intel Inf 9.2.0.1019 and RST 10.1 (10.5 for Z68) | Catalyst 10.12 chipset driver | ||||
| Memory | Corsair 8GB (2 x 4GB) DDR3 | |||||
| Memory timings | 9-9-9-24-1T @ 1,600MHz | |||||
| Graphics | AMD Radeon HD 6950 2GB | |||||
| Graphics driver | Catalyst 10.12 | |||||
| Disk drive | Corsair V128 SSD | |||||
| Optical drive | Sony AD-7263S | |||||
| Chassis | Corsair Obsidian Series 700D | |||||
| Power supply | Corsair HX1000W | |||||
| Operating system | Windows 7 Ultimate 64-bit | |||||
CPU and memory benchmarks |
||||||
| AIDA64 v1.50.1200 | The successor to Everest - useful for measuring memory bandwidth and latency | |||||
| HEXUS.PiFast |
Our number-crunching benchmark stresses a single core by calculating Pi to 10m places | |||||
| wPrime 2.0.4 | Another number-crunching benchmark that stresses all available CPU cores/threads | |||||
| CINEBENCH 11.5 | Using Cinebench's multi-CPU render, this cross-platform benchmark stresses all cores | |||||
| 7zip 9.20 | We use the built-in benchmark in this open-source file-compression utility | |||||
| TrueCrypt 7.0a | An encryption/decryption benchmark that's partial to AES acceleration | |||||
GPU benchmarks |
||||||
| StarCraft II | DX9, 1,680x1,050 medium quality. | |||||
| Call of Duty: Black Ops | DX9, 1,680x1,050, 4xAA, high quality | |||||
| 3DMark Vantage b1.2 | Run at the default performance preset | |||||
General benchmarks |
||||||
| Storage performance | USB 2.0/3.0 read speed, SSD average read speed | |||||
| Power consumption | While idling and when running wPrime | |||||
Testing notes
Keeping parity with the other setups, a Radeon HD 6950 2GB card is the sole display for this round of benchmarking; the IGP's switched off in the BIOS. We'll discuss its performance separately.
Eagle-eyed readers may have noticed that the Turbo Boost frequencies shown in the BIOS shot on the previous page do not tally up with Intel's. The board would run a single multiplier across all cores, rather than the stepped pattern mandated by the CPU's spec. sheet. We changed the chip out to a full-retail Core i5 2500 with no success. Flashing the BIOS down from F4d to F4 didn't help.
However, contacting Gigabyte we learned that a single multiplier across all cores is actually a feature when running with DDR3-1,600 memory. The default multiplier is 37x, and the 2500K chip will, under load, run at this speed all the time.
As such, to keep parity with older boards - ASUS has since implemented the same-multiplier feature - the Gigabyte board has been tested with the multiplier maximised to 34x (all-core load). This means that light-load performance will suffer a touch - as the multiplier is limited. We could have chosen a multiplier of 37x for all cores, of course, but that would have been unfair on boards tested a few months' ago and included in this comparison.
