Q & A – File System/CD-ROM/DVD Benchmark
This document provides some frequently asked questions about Sandra. Please read the Help File as well!
Q: Why does it take so long to run the drive/file system benchmark?
A: This occurs when you have a large amount of memory and a rather slow disk. In order to make accurate measurements, the disk and software cache must be swamped by using a test file at least twice as big, as big as possible. Since Windows uses a dynamically allocated disk cache which can be as large 80% of your total memory, Sandra must use a huge file which takes a long time to process. There is no other way since the cache cannot be turned off.
Q: Why does the rating vary between sequential runs?
A: Since the disk cache is dynamic, the size will increase after many disk operations. Since a disk benchmark does only disk operations it is likely that the cache will increase between runs up to the maximum size possible. Generally the bigger the cache the larger the index, therefore the index should increase slightly between runs. When the size of the cache is constant the index should not vary appreciably.
Q: My (S)ATA/SCSI RAID array gets ~100MB/s burst rate and not higher!
A: If you are using a PCI RAID card or a controller connected to the PCI bus that may well be the maximum banwidth the PCI bus can give you. Use a PCI Express/PCI-X/64-bit adapter or use the South Bridge integrated RAID controller if it has a fast/non-PCI connection to the North Bridge.
Q: Changing the (S)ATA mode does not change the index! Is the benchmark bad?
A: The burst read/write speeds are not included in the index! Thus, unless the sequential & random read/write speeds are limited by the interface the index will stay the same. A slow drive is not faster in, say ATA66 mode than in ATA33.
Q: I’ve put the drive on a new controller and I get worse results! What’s going on?
A: The controller driver performance is paramount! A fast controller with a slow driver will be outperformed by the built-in controller using Microsoft/Intel drivers. The burst speeds do not influence the index value.
Q: Why are the SMP tests disabled?
A: We haven’t found a way to simulate a multi-threaded disk benchmark that would yield proper results yet. Once we do, we’ll release an update.