Tests

Our test method is briefly described here). The scores on diagrams are relative to that of our reference testbed that always scores 100 points. As of 2011, it's based on the AMD Athlon II X4 620 CPU, 8GB of RAM and Palit's NVIDIA GeForce GTX 570 1280MB. Detailed (absolute) results are traditionally provided in this summary.

3D Modeling

HT still takes a toll on tasks that cannot reap its benefits, but it doesn't really matter much for these benchmarks. You can easily see the difference between the first and second iterations of the Core architecture: the former boosts performance by 8%, while the latter, by only a couple. Frankly, this is nearly the worst result in today's testing, not completely unexpected though.

Final 3D Rendering

Hyper-Threading is very efficient in these benchmarks, so processors do differently with and without it. With it, the performance boosts provided by each architecture generation are about the same. Without it, the boost (and efficiency) deteriorates with each generation.

Data Compression/Decompression

The noticeable growth provided by the transition from LGA1156 to the first version of LGA1155 continued in the move to the 3rd generation. Though 3% might not seem much, but it's still better than what we saw on the first diagram.

Audio Encoding

Here, the performance boost is even bigger. This is interesting, because audio codecs are rather 'conservative' software. As soon as performance in these tasks had become satisfactory for practical use even on cheap CPUs, further optimizations were abandoned. Here, even multitasking is just encoding several files in parallel. And mind you, a couple of threads is what developers had in mind. Still, the performance continues to grow with each CPU generation: both in equal conditions and as a result of methods like increasing clock rates, for example.

Compiling

This case might be the most interesting: the boost provided by the 3rd generation is bigger than that provided by the 2nd. We don't know for sure what helped in this case, but Sandy Bridge has always shown minor improvements over rivals in these tasks. In turn, Ivy Bridge has surely got some upgrades it seems.

Mathematical and Engineering Computations

The performance boosts from the 2nd and 3rd generations are comparable. You can also see that while Hyper-Threading has provided a small but stable performance boost in the 1st Gen Core CPUs, it doesn't do so in the 2nd and 3rd generations. Well, it does in some cases if you look at the detailed results, but it also reduces performance in other benchmarks, so it balances out. This wasn't the case with the 1st generation.

Raster Graphics Processing

The raster graphics performance continues to improve. Unlike compilers, the dash forward has been provided by Sandy Bridge while Ivy Bridge just continues the progress (which is still very nice).

Vector Graphics Processing

As opposed to Nehalem, the Sandy and Ivy Bridge processors do not lose performance with HT on. In general, we have a good microarchitecture progress in both cases.

Another dash forward here albeit shorter. Unlike Nehalem, Ivy Bridge doesn't lose performance in these few-threaded benchmarks when you enable Hyper-Threading. Besides, the 3rd generation provides a boost comparable to that of the 2nd. So you see, there's progress even when the number of threads is low.

Write a comment below. No registration needed!