Asus A7A266-E in the test: The 3rd generation of the ALi MAGiK 1
- 1 Foreword
- The chipset: ALi MAGiK 1 - 'C1 Stepping'
- 2 Scope of delivery and documentation
- 3 Features
- 4 Test system
- 9 Conclusion A7A266-E
Since the benchmark suitability of the SiSoft Sandra tool has often been questioned in the last few weeks, we want to anticipate any doubts about our measurement get rid of the world. It is true that new versions of the software and the measured values achieved in them cannot always be linked one hundred percent to their predecessors. For precisely this reason, we continue to benchmark with the version with which we once carried out the first mainboard test.
In view of the increase in performance of the KT266A of up to 40% after this test result the disillusionment spread. Only 8% in the ALU and a meager 3.5% in the FPU rating, should this be a legitimation for the chip remake? Neither the chip competition of the '1st generation' could be left behind, nor could the VIA chip be brought into trouble at all. And since the results achieved in Sandra in the end largely also apply to all othersBenchmarks were valid, we prepared ourselves for a rather leisurely performance.
The second discipline was again 'chipset tuning'. Even if the values achieved here are not included in the outcome of the following benchmarks, we at least hoped to get the ALi MAGiK 1 going with a few changed BIOS settings.
Well something, was our first thought. With the sharpest timing, the A7A266-E managed to gain quite a bit in both ratings. However, the FPU value was only slightly above the standard values of the KT266. AMD 760 and SiS735 left this noticeably unimpressed - not to mention the KT266A. The ALU value, which we already noticed with the A7A266 with its quite good performance, could even almost reach the standard value of the 2nd VIA generation through the tuning.
In order to test the results obtained in Sandra for their suitability for everyday use, we have included the widespread packer WinACE as a benchmark in our test for several reviews. The test candidates have to deal with a wild mix of bitmaps, documents, HTML files and already compressed MP3 files. A total of 359MB needs to be compressed in the shortest possible time.
Measurement - Reboot - Measurement. But despite all the doubts, the result seemed to be solid. The 35% deficit on the KT266A seemed, at least as far as the EPoX board was concerned, like blown away. And the KT266, which is still superior in Sandra, not only had to lose its lead; the ALi MAGiK 1 won with a 34Seconds better packing performance even in front of this chip. The lead of the A7V266-E that was still achieved in Sandra shrunk to almost 7%.
If you look at them with WinACE once again in relation to the Sandra performance, so one comes to the conclusion that ALi did not concentrate on optimizing the streaming performance. Much more routines have apparently been optimized that have a greater impact in practice than in the pure streaming test Sandra. First and foremost, the so-called 'latency', i.e. the delay with which inquiries and the sending of data are processed, is clearly stated. Acer Labs Industries seems to have made up some ground here. It's actually a shame that ALi has remained silent on this subject so far. Or should the result in WinACE just have been a 'slip'?
On the next page: Quake3Arena