Page 5
SiSoft Sandra Benchmarks
SiSoftware Sandra (the System ANalyser, Diagnostic and Reporting Assistant) is an information & diagnostic utility. It should provide most of the information (including undocumented) you need to know about your hardware, software and other devices whether hardware or software. Sandra provides similar level of information to Norton SI, Quarterdeck WinProbe/Manifest, etc. The Win32 version is 32-bit and comes in both ANSI (legacy for Windows 98/Me systems) and native Unicode (Windows NT4/200X/.Net) formats. The Win64 version is 64-bit and comes in native Unicode format.
Do note that all the SANDRA benchmarks are synthetic and thus may not tally with real-life performance. The latter stands for whatever your environment is, i.e. which applications you run with what amount of data and so on. It is up to you to decide whether what Sandra measures is what you want to measure.
- Download SANDRA (special Guru3D.com edition)
Here you can find the scores of Sandra.
The test system used is an ABIT AA8XE. It uses a 3.6 GHz Pentium 4 Processor (aircooling). It of course has a quad pumped 4x200 MHz FSB and even better, a multiplier that can be set at 14 and 18. This allows us to manage 18x200 = 3600 MHz but also 14x256= ~3600 MHz or an ever higher FSB. Combined with some nice multiplier/divider option we can achieve a very high memory bandwidth and that's exactly the goal for today... to reach the highest memory bandwidth possible.
This DDR2 Memory is based on non-tweaked and SPD (4:7:4:4) timings. Memory timings as you probably know are good for this rule: the lower the better. Memory timings let you know how many cycles it takes for diverse operations internally to the memory with CAS being the most important one. One thing this memory did not take well where timing changes when running at high performance. But at default up-to 533 MHz for example this memory would even run a CAS3 setting! We'll leave timings at default as we'll focus on a high frequency today. The memory can take high voltages, allowing very high clock frequencies :)
As you can see from the results performance is extremely good and shows flexible memory. We increased the FSB until we could not go any further, unfortunately that absolute limit was roughly 275 MHz with the CPU multiplier at 14.
Overclocking at this level is nothing more then basic simple math versus finding the limit. In the first three runs (and this goes for all tests) we see the CPU clocked at 3600 MHZ with only the memory clock changing. The fourth result where it says 275 MHz is the most interesting one.
In combo with our Abit mainboard we can do some seriously tricky stuff. We hardware strapped the main board chipset faking a 1066 MHz CPU, then raised the FSB towards our maximum 275 MHz, lowered the CPU multiplier to 14, increased memory voltage towards 2.1 Volts and a some selected a very nasty divider (4:3). Suffice to say we got the memory running at 733 MHz, giving us bandwidth way over 7000 MB/sec. It even went above what can be expected yet test after test, the bandwidth remained steady.Let's go have a look at the PC Mark04 results.
PC Mark 2004
PCMark®04 is the latest version of the popular PCMark series. PCMark04 is an application-based benchmark and a premium tool for measuring overall PC performance. It uses portions of real applications instead of including very large applications or using specifically created code. This allows PCMark04 to be a smaller installation as well as to report very accurate results. As far as possible, PCMark04 uses public domain applications whose source code can be freely examined by any user.
Info and download - Download!
Another synthetic benchmark that displays memory performance. Testing real world performance for high performing memory is very difficult. At default system settings we expect this memory to perform above a 5000 score. And it does precisely that, but once we start overclocking... wowza.
