iXBT Labs - Computer Hardware in Detail






Processors for Common People
Part 4: Sempron vs. Athlon 64
on the Socket 754 Platform

This part actually deserves to be called Part 3.5 instead of Part 4, as it should supplement the test results and conclusions, made in previous Part III. This time we decided (not without help from our readers, who pointed out some insufficiency of information in Part III) finally to clear up the issue of L2 Cache size and its influence on CPU performance on AMD Socket 754 platform, having compiled all possible L2 variants with the same core clock: 128 KB, 256, 512, and 1024. Besides, it was interesting to compare two neighbors in model numbers: Sempron 3100+ and 3300+, their clocks competing with the same L2 size: the latter is larger in the junior model, while the clock is higher in the senior model.

Testbed configurations

Sempron 3000+
Sempron 3100+
Sempron 3300+
CPU Socket
Socket 754
Socket 754
Socket 754
Core frequency, MHz
L2 Cache, KB

Sempron 3400+
Athlon 64 3000+
Athlon 64 3200+
CPU Socket
Socket 754
Socket 754
Socket 754
Core frequency, MHz
L2 Cache, KB

  • ASUS K8V Deluxe motherboard (Socket 754, VIA K8T800 chipset)
  • RAM: 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

References: AMD CPU designations

Unlike Intel, AMD has long given up the idea to write the model name on a processor. Instead, it uses quite a complex designation, resembling a cipher. :). We decided that we hadn't touched upon the deciphering issue for a long time. Besides, we have a nice occasion here: the article is devoted to AMD processors only. So, a technically correct name for AMD designation is OPN (Ordering Part Number). It's engraved on a packaging right under the CPU trademark (AMD Athlon 64, AMD Sempron, AMD Opteron, etc). It looks like this:


Or like this:


Or like this:


We can single out seven independent segments in OPN. Approximately like this (by the example of three OPNs, provided above):

AXDA | 2200 | D | K | V | 3 | C

ADA | 3200 | A | E | P | 5 | AP

SDA | 3100 | A | I | P | 3 | AX

What do they mean? It's quite simple.

  1. The first section contains 2-4 letters with "encoded" CPU family and core architecture names. For example, "AXDA" means Athlon XP manufactured by 0.13-micron process technology, "AX" 0.18-micron Athlon XP, "ADA" Athlon 64, "SDA" Sempron, and so on. Of course, there is some logic in the choice of letters. However, it's not always evident, so it's much simpler to remember designations of desktop CPUs for the most actual AMD64 platform (there are only two of them):
    1. SDA Sempron
    2. ADA Athlon 64, Athlon 64 FX, Athlon 64 X2

  2. The second section is Model Number. Most often it's a number, which (from AMD's point of view) characterizes performance of a given CPU (within its product line) in terms of conventional points. The same number can be seen in the model name in price lists: "AMD Athlon 34 3200+", "AMD Sempron 3000+", and so forth. "3200" and "3000" Model Numbers.
    Note: in case of Athlon 64 FX processors, the second section is not a number, as Model Number of these CPUs includes not only numbers, but also letters: "FX53", "FX55", and so an. This peculiarity gave rise to a popular delusion that the first section of Athlon 64 FX designation looks like "ADAFX". In fact, all Athlon 64 processors have the same first section: "ADA", "FX" in case of Athlon 64 FX refers to the second section — Model Number.
  3. The third section is just one letter denoting a packaging type of a processor.
    Note: some users are mistaken that this letter in the OPN means a CPU socket, but that's not the general case — it means exactly the packaging type.

    On the whole, this information in interesting only to specialists, so let's limit ourselves to technical abbreviations.

    1. A CPGA
    2. B OBGA
    3. D OPGA
    4. E uPGA
    5. F OPGA
    6. G uPGA

  4. The fourth segment is a single letter denoting the standard CPU voltage.
    1. A 1.351.40 V
    2. C 1.55 V
    3. E 1.50 V
    4. I 1.40 V
    5. K 1.35 V
    6. M1.30 V
    7. Q1.20 V
    8. S 1.15 V

  5. The fifth section is a single letter denoting the maximum permissible CPU temperature.
    1. A 71° C
    2. K — 65° C
    3. M — 67° C
    4. O — 69° C
    5. P — 70° C
    6. X — 95° C

  6. The sixth section is a single figure denoting the L2 Cache size in a given CPU.
    1. 2 128 KB
    2. 3 256 KB
    3. 4 512 KB
    4. 5 1024 KB (1 MB)

  7. And finally, the seventh section is one or two letters that provide some additional information on characteristics of the product. In the most general case, we can say that they denote a core revision. It's hardly reasonable to publish here all possible combinations, as they are too many (besides, revision comments will be even longer than their list, this information being of little use without comments).

On the whole, we can see only one drawback in AMD OPN: such purely consumer-aimed parameter as Model Number is included into the "ciphered" section instead of being engraved in its simple from. The following engraving seems much more logical:

AMD Athlon 64 3200+

Thus, a regular user would easily see a model number on the packaging, while technical specialists would be able to learn all they need from the code below. However, the above said is just "a petty objection"...

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

We can easily assume that our readers will have two main questions concerning the comparison of CPU performance in this article: does the Sempron model number correspond to its real performance and is Sempron with a higher model number preferable than an Athlon 64 with a lower model number. In case of 3ds max, the answers will be yes and no correspondingly. Sempron performance grows steadily with its Model Number, but even a Sempron 3400+ failed to outperform an Athlon 64 3000+. Comparing technical characteristics of the above couple, we inevitably come to a conclusion that all that matters is L2 Cache size. It's confirmed by the Athlon 64 3200+, running at the same clock as the Athlon 64 3000+, but with a larger cache — this processor outperformed it as well.

SPECapc for Maya 6 (Maya 6.5)

Detailed results

This situation is more controversial, as the advantage of Athlon 64 3000+ over Sempron 3400+ is minimal. To all appearances, the point is in cache again, but with a slightly different accent: 128 KB (Sempron 3300+) are too insufficient, while the program makes no difference between 256 and 512. However, increasing L2 Cache to 1 MB (Athlon 64 3200+) again yields a positive effect.

Lightwave 8.2, rendering

Strange as it may seem, the situation is similar to that in Maya. Even though in SPECapc for Maya we tested interactive operations, here we deal with the final rendering of a complex scene. We can again see the three "performance steps": low (128 KB L2), middle (256 and 512 KB L2) and high (1 MB L2).

SPECapc for SolidWorks 2003

Detailed results

In SolidWorks all processors line up evenly. But the conclusion on Sempron is as sad as in case of 3ds max: even the Sempron 3400+ failed to outperform the Athlon 64 3000+. So, the table with technical characteristics of the processors will help you understand the reasons — but regular users trust the "magic of numbers" (3400 vs. 3000) more than technical characteristics...

Adobe Photoshop CS (8)

Detailed results

Considering the good performance of Intel Celeron D in Adobe Photoshop, we could assume that this program was no less critical to clock than to L2 Cache size. That's why the similar performance demonstrated by Sempron 3300+/3400+ and Athlon 64 3000+ is not surprising: perhaps, AMD processors do fine even with 128 KB L2 cache. However, 1 MB cache still allows Athlon 34 3200+ to outperform all other processors. That is we again witness some "L2 Cache size threshold": cache growth has little effect on performance up to a certain size; but when a threshold is reached, performance starts growing again.

Adobe Acrobat 6.0

We can establish a fact that Sempron model numbers are approved: performance grows together with Model Number.

All-purpose data compression (archiving)

Detailed results

Quite an unexpected result, as Sempron 3300+ managed to outperform Sempron 3100+ even in archiving (archivers' sensitivity to cache size is well known) due to a higher clock (the 3300+ model is faster by 200 MHz), even though its L2 cache is twice as small. But if we have a look at the detailed results, it becomes clear that one of the applications (RAR) actually expressed its scorn of Sempron 3300+ for its small cache size. But the results of the other archiver (7-Zip) counterbalanced this small defeat.

Multimedia lossy compression (MP3/MPEG2-4)

Detailed results

It's an excellent illustration to our hypothesis that performance growth, when 128 KB L2 grows to 256, does not mean that it will grow, when 256 KB L2 grows to 512. And vice versa — no performance growth in case of 256 —> 512 change does not mean that there will be no performance growth when 512 KB L2 is increased to 1024. On the whole, it's a classical case of "everything is clear, but still obscure". Unfortunately, there is nothing we can add to comfort you: it's practically impossible to predict behavior of a given program with a given L2 cache size without testing. Except for thorough analysis of the source code... but no one will grant access to it.

CPU RightMark 2004B

There is a strange dip in the Athlon 64 3000+, there are no reasonable explanations to it. In other respects the situation is quite normal for a CPU RM 2004B: nothing matters, except for clock and architecture.

3D games and graphics visualization
in professional packages

Total score in games

Detailed results

On the whole, there is a tendency towards justified model numbers, that is all Sempron models operate according to their Model Number: the higher, the better.

SPEC viewperf

Detailed results

Perhaps, it's the only case when the mythic "improved memory controller" in new Sempron processors has [presumably] shown its worth (Sempron 3400+ outperformed Athlon 64 3000+). It illustrates well the worth of hunting after new revisions...


Our conclusion will be brief, as the test results are neither new nor offering global discoveries or exposures.

  1. You must not trust the magic of numbers when you choose between Sempron and Athlon 64. Model Numbers of these processors provide an idea of their performance rating within their own series. But comparing Model Numbers of processors with different names and trying to draw conclusions from this comparison makes absolutely no sense. On the other hand, comparing technical characteristics of a CPU (in this case clocks and L2 Cache sizes) sometimes dots one's “i's” and crosses one's “t's”.
  2. But still there are more cases when AMD model numbers are true, that is they give a real idea of the performance relations between them (within the same product line!) than the cases when they are wrong. So even though this approach is controversial, we can draw a conclusion that, in general, the policy to use Model Numbers to rate processors justifies itself and really informs customers rather than misleads them.
  3. I repeat: if we make another step from comparing Model Numbers and compare technical characteristics of Socket 754 CPUs from AMD (at least two simplest ones — clock and L2 Cache size), predicting performance is not that difficult. Especially after a preliminary analysis of test results, such as above.

Stanislav Garmatiuk (nawhi@ixbt.com)
October 25, 2005.

Write a comment below. No registration needed!

Article navigation:

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.