Monday, March 27, 2006
Powerful, Small, Stealthy: Shuttle's SD11G5 XPC
Posted by Jason Dunn in "HARDWARE" @ 09:15 AM
Benchmarking the SD11G5
Because this is primarily designed to be a workstation, I didn't do any 3D benchmarking for gaming performance. I focused instead of productivity and digital media content creation benchmarks. To this end, I used two benchmarks: Business Winstone 2004, and Multimedia Content Creation 2004. Each benchmark installs a variety of applications and runs them though high-speed automatic macros, then measures the result. Benchmarks are like any other form of statistics: they only make sense in context. So in that light, I manipulated two variables on the SD11G5 to see how they impacted the benchmarks: I overclocked the CPU and used different hard drives. The results were interesting, and are presented here without detailed analysis (this review is long enough), just as raw data. The on-board Intel video was used in all these tests.
Business Winstone 2004
2.13 Ghz CPU, 7200 RPM hard drive. Score: 23
2.13 Ghz CPU, 10K RPM Raptor. Score: 23.7
2.13 Ghz CPU, 10K RPM Raptor X. Score: 24
2.47 Ghz OC'd CPU, 10K RPM Raptor X. Score: 26.1
Multimedia Winstone 2004
2.13 Ghz CPU, 7200 RPM hard drive. Score: 28.5
2.13 Ghz CPU, 10K RPM Raptor. Score: 29.4
2.13 Ghz CPU, 10K RPM Raptor X. Score: 29.8
2.46 Ghz OC'd CPU, 10K RPM Raptor X. Score: 33.7
The quick analysis? The fastest CPU still makes the biggest difference in these benchmarks. The Raptor X hard drive made less difference than I thought it would. It would have been interesting to see how a Core Duo CPU would have benchmarked under the same scenario.
I also used SiSoft Sandra 2005 to measure hard drive speed in the SD11G5 with three drives I had on hand. The results were interesting:
160GB Western Digital 7200 RPM: 42MB/s
74GB Western Digital Raptor 10,000 RPM: 55 MB/s
150GB Western Digital Raptor X 10,000 RPM: 72 MB/s
This shows that although the 74GB Raptor was 13MB/s faster (31%) than a normal 7200RPM drive, the Raptor X kicks it up another notch and is a massive 30MB/s faster – that's 71%. It's not like your programs will start 71% faster, but in large file copy functions, the Raptor X will get the job done quicker.
SiSoft Sandra 2005 also had a multimedia benchmark. I ran it and compared the numbers to other CPUs in the database, and it showed that the Pentium M CPU measures very favourably against other Intel CPUs on a per Mhz basis. It also shows how potent AMD CPUs are. The numbers below are "pure" benchmarks, meaning they are only valid in comparison to each other.
Pentium M CPU 1.86 Ghz: 17,694 Integer / 19,496 Floating Point
Pentium M CPU 2.13 Ghz: 20,431 Integer / 22,509 Floating Point
Pentium M CPU OC'd 2.5 Ghz: 23,950 Integer / 26,392 Floating Point
Pentium 4 CPU 3.2 Ghz: 24,459 Integer / 32,341 Floating Point
AMD 64 X2 4800+ CPU: 45,955 Integer / 49,481 Floating Point
A side benefit of using these benchmarks is that I was able to test how far I could push the overclocking on this CPU. I tend to do the type of overclocking that doesn't take real guts: I'll overclock the CPU bus speed, but I won't juice the CPU with extra voltage - it's just too risky with an expensive CPU that I paid for. The benchmarks were a good tool for system stability because they push the CPU and most sub-systems very hard. The system was stable under SiSoft Sandra at 2.5 Ghz, 2.47 Ghz with the Multimedia Windstone 2004, and ultimately in day to day use I have it overclocked to a stable 2.38 Ghz.












