Scientific and engineering analysis

No "blue flags", only four red ones, three of which refer to Hyper-Threading. Turbo Boost desperately pulls the total score up, but it sometimes fails to cope with heavy performance drops caused by HT: pay attention to results in Mathematica. Engineering packages perform very strangely: +8% from TB and -6% from HT in SolidWorks transform into the same +8% from both technologies enabled. It looks as if Hyper-Threading does not affect performance in this application without Turbo Boost. But the bad influence disappears as soon as TB is enabled. Pro/ENGINEER also behaves in a strange way: performance gain from TB equals 8%, from HT -- 1%, and the total performance gain... is just 4%.

Raster graphics

Microsoft.NET platform (represented by Paint.NET) proves its positive attitude to various "boost" technologies again. However, the overall situation in the bitmap graphics group is not very peachy: ACDSee acts very strangely, PaitShop Pro, PhotoImpact and Photoshop respond negatively to HT. Only Photoshop manages to use TB properly, even though it also has good multithreading optimizations. These results also lack logic and common sense: two minuses in ACDSee add up to a plus. Photoshop also shows funny arithmetics: 13 minus 1 equals 18 there.

Data compression

It's a logical situation: no archiver can use more than two threads to pack files into its own format, so Turbo Boost is much more useful here than Hyper-Threading. But the latter gives a surprise again: WinRAR inexplicably benefits from enabled HT -- but we know for sure that this archiver cannot use even four cores, to say nothing of eight.

You might have already noticed that our study of how HT and TB interact with real software resembles riding a jibbing horse: nobody knows what's inside its head and where it will bolt the next moment. So the only thing you can sometimes do is to relax and enjoy the view, trying to remain in the saddle.

Compiling

The new compile test handles four physical cores well, and it did well with eight virtual ones: Hyper-Threading yields 23% of performance gain.

Audio encoding

Audio codecs running in parallel (as many as cores) should have been the ideal environment to reveal all positive traits of Hyper-Threading: as we have already determined from indirect signs, their computing core is most likely stored in cache, and their algorithms should agree with the HT concept well. Test results confirm our assumptions: performance gain from HT amounts to 28%. Turbo Boost yields smaller dividends, but they are still good. But the most successful results are demonstrated, when these technologies work together: 45% performance gain! That's the highest result in our tests here, by the way.

Write a comment below. No registration needed!