Even though Intel is an incontestable leader in the segment of top-end desktop x86 processors, and even though single-socket quad-core solutions for this price range are offered only by Intel so far, this manufacturer continues to please users who are ready to pay any money for ultimate performance with newer and faster CPUs. Today we are going to review another champion: Intel Core 2 eXtreme QX6850. It's even faster than the previous Intel's favourite, to say nothing of the last but one. It has a higher core clock, of course, and a higher FSB clock of 1333 MHz (QP). As a result, our testbed went euphoric and its results rose as high as the sky. :) Jokes aside, it's a top solution, what else can we say. It will hardly be affordable to many users. They can just admire the test results and feel the inexorable pace of the technical progress that leads us to bright future, when such processors are called Celerons. :)
Hardware and Software
* - "2 x ..." means per core;
Essential foreword to charts
Our test method has two peculiarities of data representation: (1) all data types are reduced to one - integer relative score (performance of a given processor relative to that of Intel Core 2 Duo E4300, given its performance is 100 points), and (2) detailed results are published in this Microsoft Excel table, while the article contains only summary charts by benchmark classes. We will nevertheless focus your attention on detailed results, when needed.
3D Modeling and Rendering
If we have a look at detailed test results, we'll see that the four cores do not help in all operations except rendering, they are even interfering with performance (both QX processors are outperformed by the X6800). But owing to the huge advantage in rendering speed (this process can be distributed between cores almost perfectly) the total score of quad-core processors is still much higher.
To all appearances, QX6850 gains an advantage owing to a higher core clock rate plus a faster bus (you cannot explain this victory only by the clock rate), not because of additional two cores (or QX6700 would have outperformed X6800).
Digital Photo Processing
The most popular operations and filters in Adobe Photoshop cope well with distributing the load between two or four cores. It's a rare example of a popular program that is ready for quad-core processors even now.
Despite the official support for multithreaded operation, MS Visual Studio 2005 compiler failed to use four cores effectively: the advantage of QX6850 over X6800 fits in their frequency difference.
We have already noted that Apache Benchmark leads Intel quad-core processors into some strange stupor. Here is another proof: four cores at 3 GHz are outperformed (even if nominally) by two cores at 2.93 GHz. How could it happen? We've come up with only one hypothesis so far: a lot of threads jump from core to core with a non-shared L2 cache and pollute it so much that caching becomes inefficient.
The rendering module of CPU RightMark can use up to 32 processors. So there is nothing surprising about the results of the "4 vs 2 cores" competition - quad-core processors take top lines in the diagram.
The archivers somehow use the two additional cores, or the difference between X6800 and QX6850 wouldn't have been so noticeable. But the authors still have some issues to improve.
QX6850 outperforms its closest competitor solely due to its higher clock rate, twice as many cores give absolutely no advantage here. However, we already wrote that FineReader did not enable SMP optimization in the batch recognition mode.
It's the old group of tests, which have lost their relevance because of high predictability of their results. No comments.
Most video codecs used in our tests support two or four cores, so the advantage of quad-core processors in this test does not come as a surprise.
A purely nominal advantage of QX6850 over X6800 suggests two ideas: firstly, four cores will give no real advantages in games, that's for certain (many game developers still don't know what to do with two cores...); secondly, performance seems to be limited by a graphics card here.
Funny, the total score actually puts on a par four cores at 2.66 GHz and two cores at 2.93 GHz (Core 2 eXtreme QX6700 and X6800). It's a demonstration for enthusiasts, who are raving about the forthcoming advent of quad-core processors to the mass market: It was enough to increase the clock rate of two cores by 0.27 GHz instead of using four cores. :) QX6850 breaks away from a couple of its closest competitors by 12% owing to three advantages: firstly, it's a quad-core processor (unlike X6800); secondly, it has the highest core clock; and thirdly, it has the fastest bus. Unfortunately, we cannot determine a contribution of each advantage to the final result, because we don't have other Intel Core processors with the same bus clock, or the same core clock.
Estimated power consumption
The difference in idle power consumption between QX6700 and QX6850 can hardly be explained solely by higher frequencies of the bus and cores - it's too big. We can only assume that BIOS of our motherboard is not finetuned enough, and some power saving features are disabled (for example, we failed to register a core clock reduction, although the testbed was idle for more than two hours.) It makes no sense to blame the motherboard manufacturer either - this CPU did not officially exist when we ran our tests. What concerns power consumption under 100% load, there are no surprises here. 3 GHz for this core manufactured by a given process technology is close to the limit, so it would have been naive to expect power saving miracles. To crown it all, it's a quad-core processor...
We can actually only repeat the title of this article: store is no sore. Intel hardly designed Core 2 eXtreme QX6850 to prove anything. It's crystal clear that the company is beyond competition in the top segment of x86 processors until AMD launches a cardinally new Athlon core. That's why we are not in raptures over the QX6850, despite its super-high performance results. Another top solution. Fast. Relatively hot (no wonder - the clock rate is higher, while the process technology has not been changed). The bar has been raised by another several centimeters. We are waiting for the competitor to jump...
Memory modules for our testbeds are kindly provided by
Corsair Memory Russia
Stanislav Garmatiuk (firstname.lastname@example.org)
July 16, 2007
Write a comment below. No registration needed!