Articles: CPU
 

Bookmark and Share

(19) 
Pages: [ 1 | 2 ]

Performance

Office and Content Creation Applications

First of all, we would like to follow our good tradition and to measure the processor performance in office and digital content creation applications. For this purpose we used Winstone benchmark sets.

In Business Winstone including top office applications the laurels belong to Athlon XP processors, which appear considerably faster than their competitors. Barton core manages to show its highs here. Due to larger L2 cache, Barton performance turns faster than that of Thoroughbred even though the latter worked at higher clock rates.

This benchmark set includes mostly serious applications working mostly with streaming multi-media data. Pentium 4 manages to outperform Athlon XP here. No wonder, as NetBurst architecture implemented in Pentium 4 aims exactly at streaming data processing.

However, the results of Multimedia Content Creation Winstone 2003 reveal a much more interesting regularity. Athlon XP 2800+ processor on Thoroughbred core turns faster than Athlon XP 3000+ on Barton core. It means that in this case the rating system offered by AMD doesn't at all reflect the actual processor performance. Why does this happen? The answer is very simple: Athlon XP 2800+ on Thoroughbred core works at higher core frequency than Athlon XP 3000+ on Barton core. And the performance in applications included into Multimedia Content Creation Winstone appears more dependent on the processor working frequency rather than on the L2 cache size.

Streaming Data Processing

In this section we would like to start with the synthetic PCMark2002. We decided on this test package because the algorithms used in PCMark2002 to evaluate the system performance include JPEG decompression, compression and decompression according to LZ77 algorithm text search and audio stream transformation.


The CPU performance test included into PCMark2002 showed that the top Pentium 4 processors were faster than the top Athlon XP ones. Moreover, Athlon XP 3000+ again failed to outpace Athlon XP 2800+ with the older Thoroughbred core. The matter is that PCMark2002 doesn't use any huge amounts of data and the results of this test are not dependent that much on the L2 cache size.

And during the data compression with the WinRAR utility, L2 cache size has a very serious influence on the final result. Note that the results in this test are beyond the limits of Athlon XP rating system. This way Athlon XP 2500+ based on Barton core proved as fast as Athlon XP 2800+ on Thoroughbred core, even though the latter works at nearly 0.5GHz higher clock frequency. However, even the larger L2 cache fails to help the fastest Athlon XP processors to outperform Pentium 4 3.06GHz with Hyper-Threading technology.

During the encoding of the sound stream into mp3 format Pentium 4 3.06GHz leaves all the rivals far behind. It is most likely to owe its victory in this test to Hyper-Threading technology. Moreover, the results obtained in this test indicate that the algorithm used in this test is much more dependent on the processor frequency than on L2 cache size. As a result, CPUs on Thoroughbred core with lower rating outpace Barton based ones throughout the entire test.

Video encoding into MPEG-4 format is another task where Hyper-Threading technology shows its best. Pentium 4 3.06GHz managed to leave its competitors far behind.

As for the Barton and Thoroughbred rivalry, the situation is very similar to what we have already seen before: the core clock frequency appears more important for the CPU performance in this case than the L2 cache size.

This way, we have to admit that the new Athlon XP 3000+ failed to outperform Intel Pentium 4 3.06GHz in all the tests measuring how fast the streaming data could be encoded.

Gaming Applications

Now let's see how well the new Athlon XP with larger L2 cache will perform in gaming applications.

Since 3DMark 2003 will be released only tomorrow, we have to work with the older version of this software set. In 3dMark2001 SE the situation is not so dramatic for the new Athlon XP processors. Athlon XP 3000+ is even a little bit faster than Intel Pentium 4 3.06GHz. And the performance of all other Athlon XP processors fits well into their rating system. Athlon XP 3000+ on Barton core outpaces Athlon XP 2800+ on Thoroughbred core, and Athlon XP 2800+ on Barton core is faster than Athlon XP 2700+ on Thoroughbred core.

Return to Castle Wolfenstein game built on Quake3 engine makes Pentium 4 3.06GHz the leader. Although its advantage over Athlon XP 3000+ is not that significant. And the relative performance of other Athlon XP processors is directly dependent on the CPU ratings, just as in the previous case.

Unreal Tournament 2003 is a game that loads the processor floating-point unit a lot. No wonder that in this case Athlon XP CPUs appear faster than Pentium 4. Moreover, Unreal Tournament 2003 makes real use of the larger L2 cache of the new Barton core, which gives us every right to consider Athlon XP 3000+ the fastest CPU for playing Unreal Tournament 2003.

Summing up a few already discussed things, we have to say that the new Athlon XP processors prove very worthy. They gained revenge for the streaming data encoding tests.

3D Rendering

Now let's find out how efficient the new AMD Athlon XP processors are for image rendering in popular applications. This time we have slightly enlarged the list of tests used.

Since 3ds max is very good at using multi-threading, Pentium 4 3.06GHz capable of processing two threads at a time managed to leave the rivals far behind. Athlon XP 3000+ failed to compete with it in this test. Moreover, rendering in 3ds max depends a lot on the CPU core frequency and not on the L2 cache size. We already know what it leads to: Athlon XP 2800+ on Thoroughbred core outperforms its counterpart on Barton core with 3000+ rating.


As we see, we can obtain different results depending on the rendering type in Lightwave. However, since the last version of this test set is optimized for SSE2 instructions, which are not supported by Athlon XP (SSE2 will be supported only in Athlon 64) and since the increased L2 cache of the new Athlon XP didn't have any noticeable effect on the performance, Pentium 4 3.06Ghz appeared the fastest CPU in all cases. As for the new Athlon XP with the 512KB L2 cache, it doesn't make much sense to use them for Lightwave rendering, because they appear almost as fast as their predecessors working on the same core clock and featuring Thoroughbred core.

The highest rendering speed in Cinema 4D application measured with the help of a special CINEBENCH 2000 test belongs to Intel Pentium 4 3.06GHz processor, which managed to become the leader due to Hyper-Threading technology support. As for the performance of different Athlon XP processors, the core frequency again appears of greater importance than the L2 cache size.

The situation is absolutely the same in POV-Ray 3.5.

As a result, we can say the following. Although AMD processors used to be just perfect in 3D rendering tasks, now their leadership era seems to have come to an end. Now Hyper-Threading technology speeds up 3D rendering quite a lot and allows Pentium 4 3.06GHz processor to become an indisputable leader in this type of applications. Moreover, the new Athlon XP 3000+ appeared nearly as fast in 3D rendering tasks as Athlon XP 2700+ with twice as small L2 cache but the same core clock frequency of 2.167GHz.

CAD Performance





In AutoCAD Athlon XP processors show much higher performance than Pentium 4. Hyper-Threading technology, which helped the fastest Pentium 4 processor quite a lot in many cases, doesn't do it any good in this application. Sometimes, even the contrary happens. Besides, the bigger L2 cache of Barton based CPUs tells positively on the performance in AutoCAD. As a result, the laurels here go over to Athlon XP 3000+.

Scientific Applications

To test the performance of the new AMD processors in scientific tasks we took ScienceMark 2.0 package (read more about this application here). This benchmark supports multi-threading, and all SIMD-instructions, including MMX, 3DNow!, SSE and SSE2.



It has been known for a long time now that Athlon XP processors are very efficient for math1ematic modeling or cryptography tasks. Here we see another proof of this case the performance depends mostly on the tasks type.

Besides ScienceMark, we decided to check the performance of our testing participants in the client of the Distributed Computing Environment TSC project (Tuberous Sclerosis Complex).This project has a very well arranged scientific basis, and its client models the interaction of chemical reactions with well-known AutoDock software set.

There are no more doubts: Athlon XP is beyond any competition in scientific tasks. Of course, the increase in L2 cache size is a pretty doubtful improvement in this case, but the three-pipeline FPU makes AMD processors faster than Intel Pentium 4 even though the new Athlon XP models do not work at higher core frequencies.

Professional OpenGL Applications





Athlon XP processors are quite strong in benchmarks included into SPECviewperf 7.0 test set. We have already discussed it several times in our previous articles: the algorithms used in this test package are quite old and do not involve SSE2 instructions. And the intensive calculations have always been a trump of Athlon XP processors.

As for Barton, its advantage over the older Thoroughbred is still quite arguable. Athlon XP 3000+ working at a lower core frequency than Athlon XP 2800+ on Thoroughbred core very often fails to outpace its counterpart, even though it boasts a largest L2 cache.

Conclusion

Well, the conclusions, which we tend to make, are quite ambiguous. It is true, AMD Company managed to improve its Athlon XP processors architecture by adding another 256KB of L2 cache memory. However, the manufacturing technology for these processors remained the same. As a result, the clock frequencies of Athlon XP based on the new Barton core cannot be increased over the clock frequencies of the Athlon XP processors with 256KB L2 cache. This way, we cannot state that the new core will be faster in all applications. Despite the fact that AMD assigned its new CPUs higher processor ratings.

Unfortunately, Athlon XP 3000+ on Barton core doesn't support higher clock frequency than Athlon XP 2700+ on Thoroughbred core. As a result, Athlon XP 3000+ very often appears just a little faster than Athlon XP 2700+. Moreover, in quite a bit of tests Athlon XP 2700+ performs better than Athlon XP 2800+ on Barton core. We suppose that this reshuffle may discredit the rating system used by AMD to mark its processors.

As for the rivalry between the top Athlon XP and top Pentium 4 models, the situation is not that favorable for AMD here as well. Bigger L2 cache doesn't allow Athlon XP to improve its performance significantly. This leads to even fewer applications where Athlon XP manages to defeat Intel Pentium 4 processors. At present Athlon XP can boast its superiority only in 3D games, CAD and scientific tasks. Hyper-Threading technology implemented in Pentium 4 processors proved highly efficient and improved the performance of Pentium 4 CPUs quite a lot. The announcement of the new Barton core can hardly be called an adequate response to Hyper-Threading.

However, the harder times for AMD are still ahead. In the end of April Intel Company will announce Pentium 4 3.2GHz with 800MHz processor bus and Hyper-Threading technology. The only thing AMD can actually respond with will be the launching of their new Athlon XP 3200+ based on Barton core. But, we really doubt that it will manage to withstand the new powerful rival.

So, we have to admit that it looks as if the situation in the high-end desktop processor market promises to be not in AMD's favor until their so long-awaited Athlon 64 is out.
 

 
Pages: [ 1 | 2 ]

Discussion

Comments currently: 19
Discussion started: 07/18/03 06:17:08 PM
Latest comment: 09/25/07 08:11:59 PM

View comments

Add your Comment




Latest materials in CPU section

Article Rating

Article Rating: 9.5745 out of 10
 
Rate this article:
Excellent
Average
Poor