|
SiSoftware Sandra Lite 2011.10.17.80
ó·±
Á¶È¸ :
832 ,
2011/09/29 21:44 |
|
|
÷ºÎ ÆÄÀÏÀÌ ¾ø½À´Ï´Ù
º¸µå³ª¶ó ÀÚ·á½ÇÀº Àü¿ë °í¼Óȸ¼±À¸·Î ÃÖ´ë 100MbpsÀÇ ÃÊ°í¼Ó ´Ù¿î·Îµå¸¦ Á¦°øÇÕ´Ï´Ù
IE9 »ç¿ëÀÚ´Â ´Ù¿î·Îµå°¡ ÀÀ´äÇÏÁö ¾ÊÀ»°æ¿ì µµ±¸-¿É¼Ç-»èÁ¦ Ŭ¸¯ÇÏ¿© ij½Ã »èÁ¦ÈÄ Å¬¸¯Çϼ¼¿ä
LINK.#01
2011³â 9¿ù 19ÀÏÀÚ SiSoftware Sandra Lite 2011.10.17.80 SP5 ÀÔ´Ï´Ù.
- ½Ã½ºÅÛ Á¤º¸ ¹× Áø´Ü À¯Æ¿¸®Æ¼
º¯°æ»çÇ×:
1. Clarified benchmark scores using base 2 or base 10; for example 1GB/s = 1024MB/s = 1024*1024kB/s rather than 1GB/s = 1000MB/s, etc. While other OSes use the GiB/MiB units to denote base 2, these are not widespread in Windows.
2. FMA3 (Intel) and FMA4 (AMD) support where faster than AVX; do note that as with GPUs, FMA is not always faster than a MUL-ADD thus we may need instrumentation upon testing on production hardware to optimise the code-paths.
3. Further support in hardware monitoring with more chipset/CPU/video/battery sensors detected and displayed.
4.Improved I/O in Physical Disks & File Systems Benchmarks: higher queue depth based on system thread availability (2-64). This helps SSD and RAID systems we tested score ~5-8% better. HDD performance does not change significantly.
- Note: With Options you can disable overlapped/async I/O, i.e. use synchronous I/O (no queue). As a side-effect the min. performance is very low as blocks have to wait the whole queue to execute.
5. GUI: Paint all graphs back-to-front, i.e. current results (red) are painted last, i.e. over the reference results. This ensures you always see the current result. Previously the reference results could obscure it.
6. Update: AMD STREAM 2.5 for GPGPU.
Áö¿ø OS:
Windows XP
Windows Server 2003/R2
Windows Home Server
Windows Vista
Windows Server 2008/R2
Windows 7
|
|
|
|