Dan3: I was wrong: You were wrong and I was right.
heheh
333mhz * 8 bytes = 2,664,000,000 Bytes per second.
SiSoft is reporting 2,714,763,264 Bytes per second - which exceeds the theoretical bandwidth of the RAM.
If they meant million Bytes, they should say so - and they don't. A megabyte is defined as 2^20 or 1,048,576 Bytes.
No. A megabyte is defined as either 10^6 or 2^20 bytes. Both are valid. For examples of decimal megabytes, check your hard drive ;-). Maybe it's time to adopt the IEEE proposal of using Ki, Mi and Gi for the binary prefixes instead messing things up by using SI prefixes incorrectly...
All that aside, I still don't see any reason why SiSoft have necessarily tampered with the benchmark. It's likely, IMHO, that cache accesses (which have far higher bandwidth) are contributing *slightly* (say 10%) to the results, pushing them over the theoretical limits.
-fyo |
