The benchmarking procedure (the list of software and test conditions) is described in this article. To make the diagrams easier to read, results are represented in percents (100% stands for the result of Intel Core 2 Quad Q9300 in each test). The detailed results in absolute values are available in this Microsoft Excel spreadsheet.
This type of tasks is still mostly indifferent to multi-core processors, memory bandwidth, and other improvements in modern processors. We've got obvious results -- the top desktop processor from Intel outperforms a Mid-End model from the previous series just by 40%, while the price difference is great. And the cheapest Core i7 is outperformed only by 10% (however, results in all tests, not just this one, suggest some problems with our 920 processor from the first presale shipment, about 5%). Performance gains are actually proportional to the base clock rate of the core, no matter what core. I also want to note that memory in the nominal mode does not yield any advantages to the eXtreme editions. On the contrary, they lost some tests. Thus, even if you have a choice, it always makes sense to think what to choose -- higher clock rates or lower relative timings.
We've separated interactive and computing parts of our tests for a reason. Now it's clear what the new processors are good for. Even the Core i7 920 outperforms the Mid-End model from the previous series in computing tasks by half. And the Extreme 975 is almost twice as fast as the latter! It goes without saying that the Core 2 Q9650 or Q9770 would have made this gain smaller, but we haven't tested these processors using our new test procedure so far. Besides, it wouldn't have changed the situation on the principal level -- win 50% back within the same family is impossible even if the clock rate grows by 50%. And these processors aren't that fast. So, the Core i7 is an excellent computing CPU in any case, be it a single computer or a cluster. Dual-processor systems with Xeon processors will be even better, but they are ore expensive.
Scientific and engineering analysis
These applications are more conservative, but the Core i7 family still demonstrates good results. It was to be expected.
Bitmap image processing
These three groups show similar behavior. What do they have in common? Their applications often cannot use multiprocessing (running in two or often even one thread), they sometimes have low memory requirements, etc. As a result, Core i7 processors barely justify their top status. The vow factor does not work when you look at their results.
The situation changes, when an applications loads the maximum number of cores. In other words, let the processor thrash data, and it will prove its worth. This talent reveals best in those applications that provide much data for processing. And don't forget that video codecs update very often. So the situation is simple and clear in all tests.
Testing processors with games is a complex task even at the planning stage. You are always tempted to create a situation that depends on a processor under review, so that its results look very effectively. Alas, it's attractive, but absolutely useless. If we use realistic settings and graphics systems, top processors will be almost no different from each other. But it's the least evil. That's why we have such a straight picture. It's another proof of the fact that performance in modern games depends mostly on a graphics card. It depends on a processor as well, especially with a good graphics card. And we use one of the most powerful single-GPU solutions in our testbed. However, it's evident that to get significant performance gains, gamers should upgrade to multi-GPU systems with all their pros and cons.
We tried to make our tests at least a tad more interesting by using different memory modes -- and we found out that this parameter does not affect results. The only result that stands out against the common background is demonstrated by Core i7 920, but it's not that different. The other processors line up by their base clock rates. They control it in the same way, and our tests cannot overheat any of them. At least none of processors from this series -- they possess a great reserve, so even the first threshold can be reached only with special utilities. In other respects, it's a crystal clear situation. The only new processor in our tests was Extreme 975, which differs from its predecessor by just one step of the multiplier.
In our next articles about Intel processors we'll dig deeper and expand our knowledge about Xeon (in single- and dual-processor configurations) as well as Core 2, which are still very popular. That should be enough to fill the gap until the LGA1156 debut. So our today's test results will come in handy many times. Just remember that the Core i7 performance covers 130-170 points in our new test procedure. It will be interesting to see how this domain is covered by the other CPU families.
Write a comment below. No registration needed!