We finally resume our series of articles about processors for common people. This time the heroes of our review are AMD processors. However, each platform requires individual approach. Hunting after the standardization will lead only to sheer formalism in our case, so we decided to change our criteria for selecting contenders. Indeed, first of all, fortunately, AMD allows to compare all the platforms nearly on identical hardware: even if the chipsets are different (however, a chipset makes almost no difference to AMD64 in terms of performance), but at least both Socket A and Socket 754/939 have AGP! That's why all the platforms can be compared almost correctly.

Secondly, AMD has a mysterious Sempron, which has identical (let's say "in spelling"...) model designations with Athlon XP and Athlon 64, but that's all they have in common. Moreover: one and the same Sempron model (for example, Sempron 3000+) may be found for Socket A as well as for Socket 754, but these are absolutely different processors!

In view of such situation, it was decided to sacrifice profundity to breadth in the third part of our article. The range of AMD model numbers under review is rather narrow: from 2800+ to 3200+. But on the other hand, all their existing modifications take part in our tests: Sempron for Socket A, "old" Athlon XP, Sempron for Socket 754, and Athlon 64 for Socket 754. Thus, we shall review "decent" (in terms of their model numbers) CPUs for a modern high-performance system (2800+ — 3200+). And on the other hand we shall preserve the general orientation to inexpensive solutions "for common people".

Of course, this approach covers well the middle part, while the edges "hang poised in mid-air", but... There were two articles about Intel processors, after all; so it's quite logical to devote the same number of articles to CPUs from AMD. That was why we decided to review the "tops" and "bottoms" of the popular AMD line in the fourth part. The third part will focus on the most interesting offers among inexpensive models: the golden mean.

Testbed configurations

• Processors
  • AMD Sempron 2800+ (Socket A, 2.0 GHz core, 256 KB L2)
  • AMD Sempron 3000+ (Socket A, 2.0 GHz core, 512 KB L2)
  • AMD Athlon XP 3000+ (2.167 GHz core, 512 KB L2, 333 MHz FSB)
  • AMD Athlon XP 3200+ (2.2 GHz core, 512 KB L2, 400 MHz FSB)
  • AMD Sempron 3000+ (Socket 754, 1.8 GHz core, 128 KB L2)
  • AMD Athlon 64 3000+ (Socket 754, 2.0 GHz core, 512 KB L2)
  • AMD Athlon 64 3200+ (Socket 754, 2.0 GHz core, 1 MB L2)
  • Intel Celeron D 335 (Socket 478, 2.8 GHz core, 256 KB L2, 533 MHz FSB)
  • Intel Pentium 4 2.8E GHz (Socket 478, 1 MB L2, 800 MHz FSB)

• Motherboards
  • Albatron KX18D Pro II (Socket A, nForce2 Ultra)
  • Gigabyte GA-8IPE1000 Pro2 (Socket 478, i865PE)
  • ASUS K8V Deluxe (Socket 754, VIA K8T800)

• Memory
  • 2x512 MB PC3200 (DDR400) DDR SDRAM DIMM 2-2-2-5 (Corsair TwinX)

• Video card: ATI Radeon 9800 Pro 256 MB (Manli)
• HDD: Samsung SP1614C (SATA), 7200 rpm, 8 MB Cache
• Windows XP Professional SP2, DirectX 9.0c
• ATI CATALYST 5.4 (Display Driver 6.14.10.6525)

AMD offers the richest model numbering pastures to be analyzed. We can...

1. Compare Sempron and Athlon XP for Socket A with the same model number (3000+).
2. Evaluate the effect of reduced cache size (256 KB) in Sempron Socket A (2800+ vs. 3000+).
3. Compare Sempron for Socket A and Socket 754 with the same model number (3000+).
4. Compare Sempron and Athlon XP for Socket A with Sempron and Athlon 64 for Socket 754 — again with the same model numbers in all processors!
5. Compare the performance of the top processor for Socket A (Athlon XP 3200+) with the equally-numbered Athlon 64, and lower-numbered Sempron and Athlon 64 for Socket 754.
6. Besides, the diagrams include top (for the previous tests in this series of articles) Intel processors for a reference point: Pentium 4 and Celeron D 2.8 GHz.

Test results

Remember that diagrams with all test results (61 items!) are published on a separate page — without comments, just as is. The article provides only summary diagrams that calculate the results of entire test groups into average scores. This approach appeases curiosity of the most inquisitive readers, who are against cutting down the number of test results published in our articles, and still makes the article less motley and graphics-intense. What concerns our comments, real professionals (who are interested in details) are expected to need none of them.

SPECapc for 3ds max 6 (3ds max 7.0)

Detailed results

Performance of processors for Socket A grows continuously, according to the model number and positioning: Sempron 2800+/3000+ processors bring up the rear, Athlon XP 3000+ and 3200+ processors are above them. However, the difference between these CPUs is not large — even if we compare Sempron 2800+ and Athlon XP 3200+. These are the results of just one test, mind it...

Athlon 64 3000+/3200+ processors, on the contrary, shoot forward... leaving Sempron 3000+ for Socket 754 (the same platform) far behind. Moreover: Sempron 3000+ / Socket 754 is outperformed even by Athlon XP 3200+! Yep, the reduced cache makes itself felt... However, all Semprons look excellent versus their direct competitor, Intel Celeron D. Even Sempron 2800+ for Socket A outperforms this processor. But Pentium 4 2.8 GHz easily outperforms all processors for Socket A, even the top Athlon XP 3200+.

SPECapc for Maya 6 (Maya 6.5)

Detailed results

On the whole, the situation is similar to the previous test in many aspects: continuous performance growth in CPUs for Socket A, not very impressive results of the Sempron 3000+ / Socket 754; both Athlon 64 processors demonstrate a good breakaway. But Celeron D shows a smaller gap from the group. Pentium 4 still confines itself to a small victory over the old AMD platform.

Lightwave 8.2, rendering

The first case, when the progressive architecture bested the frequency: Sempron 3000+ / Socket 754 outperforms all processors for Socket A, even though its clock is just 1.8 GHz (Athlon XP 3200+ — 2.2 GHz). Even the small cache was not a problem... However, renderers usually adjust themselves to the cache size on their own. Interestingly, Pentium 4 2.8E is approximately on a par with Athlon 64 3000+. Most likely due to the Hyper-Threading support: Lightwave render engine can generate more than one thread.

SPECapc for SolidWorks 2003

Detailed results

The situation is practically the same as in case of 3ds max, so there is no point in describing it in detail for the second time.

Adobe Photoshop CS (8)

Detailed results

That's what we call the Socket A defeat. To all appearances, Socket 754 processors have won thanks to their built-in memory controller. However, Adobe Photoshop has always been behindhand in loving Athlon XP... But in this case the dislike looks too deliberate: we can understand the defeat of the top Athlon XP 3200+ to the processors of the same class (the old core, the old platform...), but we are in stupor over the comparison results with Celeron D...

Adobe Acrobat 6.0

There is nothing interesting here: the worst results are demonstrated by Celeron D (well, it's supposed to, considering its rank) and Sempron 3000+ for Socket 754. Well, sometimes the frequency matters...

All-purpose data compression (archiving)

Detailed results

It's a triumph of Socket 754 again. Even if Celeron D does not outperform Athlon XP 3200+, they are again nearly on a par. Funny: that's the second such situation already.

Multimedia lossy compression (MP3/MPEG2-4)

Detailed results

As always, Intel processors look generally up to the mark as far as encoding media is concerned. However, have a look at the detailed results: if you always read our articles, you already know "the secret": all AMD processors without exception score very low in LAME with the highest encoding quality.

CPU RightMark 2004B

And again there is nothing surprising about it: Pentium 4 wins due to Hyper-Threading, everything else is up to the frequency.

3D games and graphics visualization
in professional packages

DOOM 3

With high quality graphics settings, nearly all processors were limited by the video card. Quite a natural result.

Far Cry

...Which recurred in the next game...

Painkiller

But Painkiller depends much on a processor. It demonstrates phenomenal dislike of "everything non AMD K8". However, it dislikes Intel platform more: even Sempron 2800+ for Socket A catches up with Pentium 4 2.8E.

Unreal Tournament 2004

Strange as it may seem, this situation is similar to the previous one. It seems we haven't tested processors for Socket A for a long time. How many interesting things can be revealed with relatively new software! Well, interesting they are, but not very consoling for this platform: even the 1.8 GHz Sempron 3000+ for Socket 754 (with a tiny L2!) easily outperforms Athlon XP 3200+.

Total score in games

Detailed results

To keep it brief: the games obviously prefer the new AMD platform.

SPEC viewperf

Detailed results

SPEC viewperf agrees with the games, but with one amendment: Intel Pentium 4 looks much better in professional OpenGL than in games. Driver optimizations? Perhaps... But the strange thing is why it is not felt in games.

Conclusion

It would be logical to resort to the run-down system in our conclusions: brief points — what new things we have learnt from the diagrams?

1. On the whole, the differences between Socket A Sempron and Athlon XP with identical model numbers are not significant. However, their technical characteristics are not significant either. So if you choose between Sempron / Socket A and Athlon XP, you have to take into account prices in the first place.
2. The differences between identically-numbered Athlon XP and Athlon 64 / 754 are often immense. Athlon XP is inferior here. Sometimes "the old guys" profit by their clocks, but on the whole Athlon 64 (often with a lower model number) looks obviously more promising.
3. On the other hand, Sempron-754 is sometimes outperformed by processors for Socket A. The reason is on the surface: even a more progressive architecture cannot make up for the low clock and small L2 Cache in some tasks. Thus, when you choose between Sempron-754 and Sempron/Athlon Socket A, you'd better pay attention to the test results in the applications you are interested in.
4. Compared to Pentium 4, processors for Socket A are defeated in the majority of cases. There is actually nothing surprising about it: the old platform, no SSE2/3 support, low FSB bandwidth. But the Celeron D vs. Socket A Sempron/Athlon comparison is not that straightforward, you have to look through the tests in applications.
5. Processors for Socket 754 are quite decent-looking compared to Pentium 4 / Socket 478 in the majority of cases. They can be certainly recommended for gaming as the best price/performance solution (just for gaming — even Sempron).
6. By the way, considering that Pentium 4 performance (similar in clock and cache size) differs very little for Socket 478 and LGA775, we can draw a conclusion that the Socket 754 platform in the low and middle price range can easily compete even with the latest platform from Intel. Especially its Socket 754 + PCI-Express combo (such motherboards are available).

In the fourth part of the article we shall try to demonstrate the performance difference between "the AMD golden mean", reviewed today, and the low end (Sempron 2200+ and maybe even Duron — if we find them in stores ;).

Write a comment below. No registration needed!