All you need to know about your products!
 |
 |
|
3DNews Vendor Reference English Resource - All you need to know about your products! |
|
||||
|
|
|
||||
|
|
|
||||
|
|
|
||||
|
||||||
AMD Athlon 64 X2 5000+. Diagonal racesAuthor: Vladimir RomanchenkoDate: 26/12/2006
Typical roundup materials devoted to CPU tests are normally biased towards two extremes. As a rule, on the moment of release of new chips they test and compare top-end models, while putting an emphasis at value chips in the "off-season". As regards the processors of the pricing range in which the performance is no longer at the peak and the price is affordable, the tests are normally not conducted because we normally do them once respective specimen arrives at the test lab. To be honest, AMD Athlon 64 X2 5000+ arrived at our test lab over a month ago, and we've always been pressed for time to complete the research to the end. That does not at all mean that the test platform was idle in the dusty corner – on the contrary, it's just the AMD Athlon 64 X2 5000+ in combination with the ASUS M2N32-SLI DELUXE motherboard based on which we ran the first express roundup test of performance of the platform powered by Windows XP SP2 and Windows Vista RC2 (read the article Using Windows Vista. Part III: installation, performance, licensing).  Some might think of publishing an article on testing the capabilities of CPU AMD Athlon 64 X2 5000+ right on the eve of year 2007 a bit belated, especially because we recently published a research into dual-core Athlon 64 X2 4800+. Perhaps, it is a bit late, but not late for the urgency of the topic – CPUs like other components turn cheaper with time, and thus become more affordable to the wide groups of buyers. Moreover, the recent statement of AMD's top management that DirectX 10 won't be topical until the second half of 2007 (made regarding the graphic cards), as well as lack of interest towards support for the new generation DDR3 memory in desktop PCs based on AMD chips at least no earlier than year 2008, suggests that if Athlon 64 X2 5000+ eventually becomes a middling solution, in any case the Socket AM2 platform its current form will be topical long enough. Anyway, let's leave the analytical forecasts and reasoning and move on to the practical side. And the first question which those who wish to compare the performance of Athlon 64 X2 5000+ versus some competitor by Intel will come up against is of course the choice of reasonable criteria for comparison. Let's be frank - things are really bad at that for manufacturers of modern processors because there is almost no technical parameter to be used for comparison of chips. Is it the L2 cache size, its organization, the system bus speed (HT), and the principle of interaction with memory, or finally the clock speed? No, no, and no, because architectures have left us no chance for an adequate comparison. Then what is left is the good old and proven method - the price. All in all, there does exist something used for reference at Intel and AMD who present comparable prices for CPUs and keep them at a certain level for some time. So let's see what return each dollar invested into purchase of a processor at an approximately comparable price will give, which is a bit more than 300 USD these days. Proceeding from the price criterion, the closest "opponent" to Athlon 64 X2 5000+ is Intel Core 2 Duo E6600. That's about all the methodology, and for the purity of experiment let's mention how the Intel Core 2 Duo E6600 came to the test lab. All is very simple - for that we used a preliminarily prepared test platform based on Core 2 Extreme X6800 – just that one which took part in our first test of the new architecture named Core 2 Duo processors: shock and awe. Well, what do we get? Both Core 2 Extreme X6800 and Core 2 Duo E6600 are built on the Conroe core, both offer 4 MB L2 cache, support for FSB=1066 MHz and are based on Socket LGA775. The major distinction between them, of course if we disregard the price, TDP and other details which don't matter now, is the multiplier. Yes, it is just the multiplier õ11 in Core 2 Extreme X6800 (2,93 GHz) which we changed to õ9 (the chip allows for that) in order to emulate quite an adequate model of the CPU Core 2 Duo E6600 (2.4 GHz). I suspect that such method might not be to the liking of some, but let's not forget that this platform was to a larger extent used for "test-driving" the new methodology of tests in which the possible insignificant deviation from statistical error is not critical. Now let's move on to describing the test benches and test conditions. The testing procedure  CPU Athlon 64 X2 5000+ built on the version of working name Windsor (Revision F), offers the core clock speed 2.60 GHz, 2 õ 128 K L1 cache and 2 õ 512 K L2 cache. The chip supports the HyperTransport bus with the clock speed 2 GHz and is equipped with an integrated memory controller with support for 2-channel DDR2-800/667/533/400 memory modules.   Among the key specifications is the comparatively new 940-pin Socket AM2, production at the lines following the 90-nm process technology and using SOI, support for the AMD Virtualization (AMD-V, formerly - Pacifica) and reduced power consumption as compared to the predecessors: the TDP of just the Athlon 64 X2 5000+ chip is rated at 89 W. In the specifications of the chip, we should also mention 153.8 mln transistors, core area 183 mm, T.Case (Max) 55-70°, the core supply voltage 1.30-1.35 V, ICC (Max) 66.2 A.   We tested Athlon 64 X2 5000+ on the base of a system with the following components:
The comparable system for "emulation" of the Core 2 Duo E6600 chip had the following specifications:
I must admit, the major task of that test was not only to try the CPU in all possible ways at extreme settings and through the maximum number of tests, but also testing with various benchmarks including the debut "test-driving" of test suites new to our test lab. As a result, some least interesting results - e.g. those related to games, are not included into the review as having little visual experience, although some were indeed added to the "armory" of our test lab.  
Test ResultsOur readers must have already noticed that for the recent months we have substantially renewed the list of benchmarking suites used at our test lab. That is related to a number of reasons, but apart from the natural "ageing" regarding the measurement of CPU performance there are two more key reasons worth mentioning. That is, first, the emergence of multicore chips which need parallelized tasks to adequately estimate their performance. Secondly, the emergence of truly interesting test applications which skillfully emulate such real processes like scientific computations, rendering, multimedia content operations, etc. In future, additionally to the benchmarking suites presented today we'll add a number of other no less important and visual. Today, taking the opportunity, let me briefly describe the capabilities of some popular benchmarking suites. PCMark 2005 – a synthetic benchmark that allows deeply enough to investigate the performance of specific PC subsystems – CPU, memory, storage. Due to the feature for auto detection of the CPU type installed, it is possible to load dynamic libraries with the code optimized for AMD or Intel. Of special interest are the tests which fit within the voluminous L2 cache of modern chips.       In fact, no surprise has been brought by testing under PCMark 2005 - Core 2 Duo E6600 takes a lead everywhere except the memory latency test. 3DMark 2006 – a good synthetics to emulate gaming applications with support for modern 3D technologies (less the DX10 capabilities) at various screen resolutions. This benchmarking suite is also interesting in that it allows putting adequate load upon multicore chips. The leadership of Core 2 Duo E6600 at this suite is evident, although we should be always aware that 3DMark 2006 is still synthetics, and while testing in real games the result may be not so evident.   ScienceMark 2.0 – also a synthetic benchmark, however, it is aimed at imitation of scientific and engineering computations with a heavy load upon the memory bus.  Therefore, it is quite logical to observe some superiority of Athlon 64 X2 5000+ related to the integrated memory controller that offers low latency.  POV-Ray 3.70 - a very interesting and free rendering utility. The current version of the suite for Windows is under development and beta testing, but already supports multithreading.   Cinebench 9.5 – a fresh version of benchmark that imitates creation of 3D content based on Maxon Cinema 4D engine.  Final wordsWell, is the CPU Athlon 64 X2 5000+ worth the money it is asked for today? The answer is not that evident as one would like to. Of course, this CPU demonstrates a superb performance, but in most cases – especially where the load upon memory bus is less intensive, the new architecture Intel Core wins back, and Athlon 64 X2 5000+ anyway gives in to the conditionally "equivalent" Core 2 Duo E6600. At the same time, if we view Athlon 64 X2 5000+ as an option for migration to the Socket AM2 platform (e.g., from Socket 939) keeping a future upgrade in mind, the purchase of such a chip can be quite reasonable - new 65 nm processors for desktop PCs are already at hand and they are also aimed at Socket AM2. In fact, transition to the Socket AM2 platform should be effected from other chips – say, something like AMD Athlon 64 X2 4200+ and lower. But that is already a different story...
|
|
|
| |||||