First up: the synthetic benchmarks. Most of the following testing utilities use various memory intensive operations to uncover transfer speeds of the different memory kits. These numbers can vary depending on the CPU used, since most processors now contain their memory controllers on the CPU die.
I tested the HyperX Fury RAM at their stock timings (1866MHz 10-11-10-30 1.5V), which were detected almost perfectly by my motherboard – although I’ll chalk that up to the older Z68 chipset (incidentally, the Gigabyte GA-F2A88XN-Wifi board used by the Kaveri A10-7850K detected the stock timings perfectly). Originally detected at the correct speed but slightly relaxed timings (1866 MHz 11-11-11-30 1.582V), typing in the correct parameters in the BIOS caused zero issues in stability. A quick multiplier change and relaxed tRAS timing resulted in an easy overclock to 2133 MHz (11-11-11-35 1.5V). Since I didn’t need to touch the voltage or relax timings very much, those users with a newer chipset may be able to reach even higher clockspeeds than I did.
Motherboard: ASRock Z68 Extreme3 Gen3, BIOS P1.20
System Memory: (See charts for tested memory, all modules used a command rate of 1)
Processor: Intel Core-i5 2500K @ 4.6 GHz
Audio: Asus Xonar DG
Video: XFX Radeon R9 290
Disk Drive 1: WD Black 1TB 7200 RPM
Disk Drive 2: OCZ Vertex 2 50GB (SRT Cache)
Enclosure: NZXT Phantom 820
PSU: Cooler Master V700
Monitor: HannsG 27″ 1920×1200 LCD
Operating System: Windows 7 Ultimate 64-bit
AIDA64 is one of Benchmark Review’s most popular utilities. For the purposes of reviewing DDR3 SDRAM, AIDA64 has a built-in Cache and Memory Benchmark that – well, I’ll just let them tell you: Memory bandwidth benchmarks (Memory Read, Memory Write, Memory Copy) measure the maximum achievable memory data transfer bandwidth. The code behind these benchmark methods are written in Assembly and they are extremely optimized for every popular AMD, Intel and VIA processor core variants by utilizing the appropriate x86/x64, x87, MMX, MMX+, 3DNow!, SSE, SSE2, SSE4.1, AVX, and AVX2 instruction set extension.
The Memory Latency benchmark measures the typical delay when the CPU reads data from system memory. Memory latency time means the penalty measured from the issuing of the read command until the data arrives to the integer registers of the CPU.
This is a synthetic benchmark, so these numbers are firmly in the “theoretical performance” arena. The memory controllers on newer Ivy Bridge and Haswell chips would probably see a slight improvement over these numbers, but at least they are useful for comparing modules to each other. The main take-away here is the relationship between latency and speed – the Fury modules perform as expected, but it seems lower latency modules can achieve similar performance.
SiSoft Sandra Lite
SiSoft’s Sandra benchmarking tool is a System Analyzer, Diagnostic and Reporting Assistant. While it contains an entire suite of tests, for memory we’re only concerned with the Memory Bandwidth benchmark.
Sandra shows a similar story – essentially running the same test as AIDA64, the memory bandwidth numbers follow the same trend.
Performance Test 8
Fast, easy to use, PC speed testing and benchmarking. PassMark PerformanceTest ™ allows you to objectively benchmark a PC using a variety of different speed tests and compare the results to other computers. PassMark’s PerformanceTest benchmark allows you to easily compare your results to a huge database online. For our purposes, we’re interested in the Read Uncached and Write memory tests.
Again, you’d be hard-pressed to notice the difference with these numbers. If you were to go by synthetics alone, price would probably end up as the most important factor. Other than very memory-intensive scientific or computational applications (which probably wouldn’t use a Z68 chipset…) you just can’t “feel” a difference between most sets and speeds of RAM.
Unless you’re gaming on an integrated GPU, that is…