Today we will compare the Athlon 64 3000+ to other top processors from Intel and AMD (last time it was Athlon 64 3400+). The processor itself is very interesting. Its main purpose is to bring closer the new 64-bit platform to more mortals who can't afford the right to discover 64-bit mysteries. Until the Athlon 64 3000+ was released we could watch the battle of Titans from afar because an average user can't afford a processor which is as expensive as a quite fast PC case.

AMD's approach to promote its new platform is not original: they just cut off half of the L2 cache of the Athlon 64 3200+. That was how the 3000+ was given a birth to. Sure, such models appear only from crippled samples of Athlon 64 3200+ with a faulty cache. Intel Celeron and AMD Duron were developed the same way, that is why such approach is well mastered by both companies. It's nice that the company didn't cut down the other specs: both 3200+ and 3000+ work at 2,000 MHz. That is why the Athlon 64 3000+ is a promising processor: if the L2 cache trimming from 1 MB to 512 KB didn't have such a great effect on performance, potential customers would get a system with pretty good performance at a moderate sum of money; plus it would have some prospects because 64-bit software might become very popular on the market if AMD64 based PCs got widely available. 

That is exactly why we decided to focus on comparison of processors within the Athlon 64 line. If you are eager to have the comparison diagrams for processors from both companies open the Athlon 64 3400+ Benchmarking to see in parallel how they perform relative to each other. All the scores can be compared because we changed neither the test technique nor the testbed. 

Hero: Athlon 64 3000+. Green.
Athlon 64 3200+ was brown.
Does AMD consider green to be a low-end color?
What about all Intel's CPUs then? :)

Testbed and software

Testbed

• CPUs:
• AMD Athlon 64 3000+ (2000 MHz, 512 KB L2) 
• AMD Athlon 64 3200+ (2000 MHz, 1 MB L2) 
• AMD Athlon 64 3400+ (2200 MHz, 1 MB L2) 
• AMD Athlon 64 FX-51 (2200 MHz, 1 MB L2, dual-channel memory controller)
• Mainboards:
• Gigabyte K8NNXP-940 (BIOS F3) on NVIDIA nForce3 Pro 150
• ABIT KV8-MAX3 (BIOS 17) on VIA K8T800
• Memory:
• 2x512 MB PC3200 DDR SDRAM DIMM TwinMOS (timings: 2-2-2-5) 
• 2x512 MB PC3200 Registered DDR SDRAM DIMM Corsair (timings: 2-2-2-5)
• Video card: Manli ATI Radeon 9800Pro 256 MB
• Hard drive: Western Digital WD360 (SATA), 10000 rpm

Software

• Windows XP Professional SP1
• DirectX 9.0b
• NVIDIA UDP 3.13
• VIA Hyperion 4.51
• VIA SATA Driver 2.10a
• Silicon Image Driver 1.1.0.52
• Terratec DMX 6fire Drivers 5.40
• ATI Catalyst 3.9
• Gray Matter Studios & Nerve Software Return To Castle Wolfenstein v1.1, demo checkpoint.dm_57
• Croteam/GodGames Serious Sam: The Second Encounter 1.07, demo Grand Cathedral
• Digital Extremes/Epic Games/Atari Unreal Tournament 2003 v2225, botmatch antalus
• Cacheburst 32 0.91.07

Low-level tests

CPU core performance

The L2 cache doesn't have a big effect, Athlon 64 3200+ and 3000+ have equal scores. 

The rendering module prefers a greater cache, though the gap is inconsiderable. 

Combining the two diagrams into one is very predictable.

Memory performance

The only really interesting fact is the increased memory write speed in case of the Athlon 64 3000+. The gap is small and it could be even considered a measurement error but remember that the low-level tests are measured very precisely, and all other components of Athlon 64 3000+/3200+/3400+ are identical. That is why we suppose that the cache trimming had a positive effect on the memory write speed. The Pentium 4 core (relative to the eXtreme Edition) demonstrates the opposite picture. 

Tests in real applications

Graphics

Well, Athlon 64 3000+ and 3200+ go on a par.

It's the ideal case to estimate how Adobe Photoshop depends on a cache. Remember that comparison of Pentium 4 and Pentium 4 eXtreme Edition can't be considered correct for estimating the cache effect in Photoshop - the cache was increased not at the expense of L2 but due to the added L3 cache (whose cache wasn't 256-bit). This time the situation is more fair: Photoshop likes a big L2 cache a lot: Athlon 64 3000+ falls behind its 1MB counterpart by 9%. On the other hand, the price difference is very noticeable... Athlon 64 3000+ can be a very good choice for an average designer. At least, it's faster than Athlon XP 3200+ . 

Audio encoding

The scores are almost identical, with the Athlon 64 3000+ being a little bit faster that the Athlon 64 3200+. Is it really because of the increased memory write speed? Well, it would be more correct to consider the victory of 3000+ over 3200+ as the proof of such difference. 

Video encoding

There is nothing interesting except the fact that Canopus ProCoder prefers larger caches. But it's not the first time (see this review). There is not difference between 3200+ and 3000+ in three tests out of four, which lets users not to spend a fortune on this CPU. 

Archiving

The archivers which can use big dictionaries must be sensible to caches, and the diagrams prove it. However, the Athlon 64 3000+ doesn't fall much behind its sibling. 

Games

The situation is similar to the previous ones: Athlon 64 3000+ is the slowest within the Athlon 64 line but the gap is not that big.

Conclusion

AMD has developed a very successful processor. Athlon 64 3000+ is almost twice cheaper than Athlon 64 3200+ (around $460 on our local market). But the price difference doesn't justify the performance gap, which is miserable in most cases and even zero in some applications. It looks like a collision inside AMD's line: those who need a speedy processor will go with Athlon 64 3400+ and Athlon 64 FX-51. Those who don't strive for megahertz and want to save on it will probably take Athlon 64 3000+. Athlon 64 3200+ doesn't seem to have potential buyers. Nevertheless, this fact underlines the advantages of the Athlon 64 3000+. 

P.S. 

This material proves one simple truth. No CPU characteristics reflected in its name can help you judge about its performance in a random program. It can be a clock speed (Intel is often blamed for "false" MHz), or AMD's model index, or cache size, or bus bandwidth... The times when you could judge about performance of one or another processor by its name in the price list are over. If you are interested in performance in your applications, the only way to find it out is to test a given processor. The parameters in CPU names like "3.2 GHz", "3400+", or even "FX-51" mean nothing :).