PCMark05 suspected special optimizati n for Intel processors do
Arstechnica. Com recently published an article VIA Nano and Intel Atom processor evaluation articles, PCMark05 a particularly interesting because from the results of a special Intel processor seems to be preferential treatment. Nothing seems wrong: the total score Atom slight lead, child Nano processor significantly dominant, and sub-graphics memory is Atom strengths. So really it? The industry has been a saying that PCMark05 test scores will vary according to CPUID, and now we have a chance to test it. Explain, CPUID is based on different processors and different specific string stored with the manufacturer, model, specifications, features and other information, equivalent to the processor ID. CPUID flags from different manufacturers vary, such as Intel, GenuineIntel, AMD is AuthenticAMD. Under normal circumstances CPUID is locked, can not be changed, but VIA is an exception, that is, you can get recognition software to VIA processor as Intel or AMD . Yi-Sung nest delicate flesh pure disaster Egypt Tuo CMark05 the White Paper a few words: “transparency and neutrality is the cornerstone of our design, we strive to generate a benchmark record each step. … We always ensure the highest standards of neutrality nature of any group are impartial. “score system and processor, graphics, hard disk sub-tests everything looks normal, the score difference within the error. But memory is not the same child, Nano into AMD processors after the score increased by 9.1%, while the Intel processor into the rise as high as 47.5%, even more than the Atom. This is clearly not the error caused by the repeated re-testers in the test, reinstall Windows XP system the result remains the same. If this calculation, the total score that VIA Nano will be more than Intel Atom, the situation is completely different. This at least shows Futuremark PCMark05 time in the production of serious negligence, because PCMark05 look different depending on the CPUID processor optimized open or shielded, not really to check to make sure the processor itself, make a decision. It can be seen, PCMark05 memory subkey does not seem to know VIA Nano processor supports SSE2 and SSE3 instruction set, therefore the implementation of the code is not optimized path. But the problem is not that simple. First, PCMark05 was released in 2005, and was born the same year, the VIA C7 processor has support for SSE2 / 3. PCMark, 3DMark benchmark tool that not only to ensure awareness of existing products, but also must adapt to the situation after the release of 2012, but PCMark05 failed to do so, even after upgrading the 1.2.0 test version as well. Second, if the VIA Nano processor from AMD and Intel face into the same score after that there is less to say, the work can only be Futuremark mistakes, but Intel CPUID bring even more performance for nearly 40 percent, which to be justified. The total does not know K8 Futuremark has long supported SSE2 / 3, right? According to the information we know, Futuremark PCMark05 seems three years to prepare the code paths, one for Intel, the second for AMD, the three other treatment for device. Optimized code path is not wrong, but to optimize varying degrees, bad explanation.