Tests
For your convenience the results are represented in percent, 100% being the result of AMD Athlon II X4 620. Absolute results are provided in this Excel spreadsheet.
3D visualization
These tests generate a lot of threads, so there's no particular difference between Sandy Bridge DC and QC. Both platforms are equally the most efficient here. As for other processors, neither clock rates, nor cache, nor (obviously) extra cores help them catch up. Core 2 Quad Q8200 is the last, because while it has four cores, the cache is small and clock rates are low.
3D rendering
Common sense tells us that processors should've separated into three groups according to the number of cores, because these benchmarks can use of all those. But different clock rates and Hyper-Threading amended the picture considerably. In particular, dual-core Core i3 (with HT) outperform the triple-core Phenom II X3 740. Core i5-680 gets as close to Athlon II X4 645 as possible. Core 2 Quad Q8200 with its four physical cores is nearly the last. And Phenom II X2 560 with two physical cores is the last. Core i3-2100, the today's hero, could only do so much and no more. But it did well nevertheless, outperforming 'older' Core i3 CPUs (including those working at higher clock rates). It would be nice, if it could also outperform Core i5-680, but without AVX supported by software that might be too much. Still, Core i3-2100 is much cheaper and consumes less power — quite a compensation for those 5% extra performance.
Scientific calculations
This one resembles the first group of tests with one exception: one of the applications can fully load even a Core i7, so the number of cores does matter (but not that much compared with the previous diagram). But that's what Core i3-2100 has, and it takes the second place.
Image editing
This group of tests has always favored Core most. It's also one of the few where higher-end dual-core CPUs look especially good. The reason is simple: multithreaded optimization is only supported by Photoshop (not completely, too). And the other three applications do fine with just a couple of cores (although ACDSee tries no spread load among more, but without any significant effect on performance). These three also 'disfavor' large cache and are sort of sceptical about AMD processors. Well, we have what we have. Anyway, Core i3-2100 performs very well, only yielding to Core i5-680 and even outperforming the quad-core Core i5-2300.
Data compression/decompression
Archivers have significantly different preferences: they all like large cache, and 7-Zip is also very efficient in load-parallelizing. So Core i3-2100 fails to catch up with Core i5-2300, but that means second place, which is great. The efficiency of the new architecture makes up for reduced cache and low clock rate (compared with high-end Clarkdales). In general, the only processor capable of demonstrating similar performance is the once-top-end Phenom II X4 965, the former $300 rival of higher-end Core 2 Quad and first Core i7 processors.
Compiling
Of course, as we move from partial to full multithreading, we have to compare Core i3-2100 with processors very different from Phenom II X4 965. But still the competition is quite favorable. Core i3-2100 yields to Core i5-680 due to clock rate and cache difference, but it outperforms older Core i3, Athlon II X4 and Phenom II X3. And it only has two cores and is quite cheap. Right now AMD can only answer with processors with four physical cores and full-fledged L3 cache. That's even worse than the competition of Phenom II X6 and Core i7.
Java
Unlike Visual Studio, Java Virtual Machine uses Hyper-Threading with less efficiency and also cares less about cache size. That's why this benchmark has always favored Athlon II X4 and processors with four physical cores in general. But HT does help, too, otherwise all Core i3 CPUs would've performed like Phenom II X2. And with HT they outdo Phenom II X3 and get close to low-end Core 2 Quad. Core i3-2100 performs like old dual-core Core i5 — even faster than the reference Athlon II X4 620. In other words, it's not a record, but still a decent result.
Write a comment below. No registration needed!
