Check out this review to see how AMD’s latest socket 754 processor – the Athlon64 3400+ – performs versus the Athlon64 3200+ and the Athlon64 FX-51. At 2.2GHz, the 3400+ throws down some awesome performance numbers. Be sure to check it out!
The 64-bit revolution will not be televised… the revolution will not be televised… but it will be benchmarked to death, right? Of course! I’m going to contribute a little to that death-by-benchmarking fate with a quick look at the latest Socket 754 processor from AMD — the Athlon64 3400+. So what’s up with this new processor? It’s nothing surprising like the Athlon64 3000+, which interestingly got its L2 cache chopped to 512KB. It’s just the normal slight speed bump, from 2.0GHz (3200+) to 2.2GHz. Everything else is the same as the Athlon64 3200+, like the inclusion of 1MB of L2 cache, single-channel integrated memory controller, etc.
If you want a really technical article, this isn’t it (for technical information related to AMD64 technology, go here). I’m just going to show you several common benchmarks, comparing the Athlon 64 3400+ to the Athlon 64 3200+ and the Athlon 64 FX-51. It’s quite obvious that the 3400+ will outperform the 3200+ (hopefully by 10% or more), but how will it compare to the FX-51? Considering that both the FX-51 and the 3400+ are clocked at 2.2GHz, this should be interesting to see them go head to head. Read on to see which one wins this little bout and by how much.
Quick Processor Comparison
|Processor||Operating Frequency||L1 Cache||L2 Cache||Integrated Memory Controller||Socket||Retail Price|
|Athlon 64 3200+||2.0 GHz||64KB + 64KB||1024KB||72-bit||Socket 754||$279.99|
|Athlon 64 3400+||2.2 GHz||64KB + 64KB||1024KB||72-bit||Socket 754||$418.00|
|Athlon 64 FX-51||2.2 GHz||64KB + 64KB||1024KB||144-bit||Socket 940||$745.00|
Scott recently benchmarked his new Athlon 64 FX-51 system, so I’m using his numbers to compare to recent benchmarks I did with the 3200+ and 3400+. Refer to the table above if you want to quickly see the difference among these three processors. Also, note that all testing was done on an nForce3 platform (ASUS SK8N in Scott’s case with the FX-51 and Chaintech ZNF3-150 Zenith in my case with the 3200+ and 3400+). Here’s a quick look at pertinent test system info.
NOTE: Scott did something a little unusual when running his benchmarks. He actually kept his e-mail application and several browser windows open during the benchmarks (yes, on purpose). The performance hit in the benchmarks from this is negligible (not even 1%). So please keep that in mind as you compare scores.
|Athlon 64 System||Athlon 64 FX-51 System|
|CPU:||AMD Athlon 64 3400+/3200+||AMD Athlon 64 FX-51|
|Motherboard:||Chaintech ZNF3-150 Zenith (nForce3)||Asus SK8N (nForce3)|
|Memory:||2 x 256MB XMS3500 Corsair DDR||2 x Corsair TWINX1024RE-3200LL (Total 2GB)|
|Video Card:||Leadtek GeForce FX 5950 Ultra||NVIDIA 5950U Reference Board|
|Hard Drive:||Maxtor DiamondMax Plus 9 200GB 8MB Buffer||2 x WD 36GB Raptors (RAID 0)|
|OS:||Windows XP Professional SP1||Windows XP Professional SP1|
For testing, we used the following benchmarks: Futuremark’s PCMark2002, SiSoftware Sandra 2004, Futuremark’s 3DMark2001SE, AquaMark3 and Unreal Tournament 2003. Here are the results.
SiSoftware Sandra 2004 – CPU
SiSoftware Sandra 2004 – CPU Multimedia
SiSoftware Sandra 2004 – Memory Bandwidth
For the system benchmarks, with the exception of memory benchmarks, we see that the 3400+ performs about 10% better than the 3200+, which makes sense considering that it is 10% faster. While the 3400+ scores about 2% better than the 3200+ in the PCMark test, Sandra Memory Bandwidth scores are about the same.
When compared to the FX-51 processor, the 3400+ really shines, putting up slightly better scores in every test except in the memory benchmarks. The FX-51’s superior memory performance can be attributed to its 144-bit memory controller (the 3200+ and 3400+ feature a 72-bit memory controller).
3DMark2001SE (Build 330)
AquaMark3 – Default Benchmark
Unreal Tournament 2003 – Inferno
Unreal Tournament 2003 – Antalus
The FX-51 dominates all of the gaming benchmarks. Overall, the 3400+ puts up slightly higher numbers than the 3200+, but it’s only by about 2% in 3DMark2001SE.
Well, I was plenty happy with the 3400+’s scores, especially considering how it essentially went toe-to-toe with the much more expensive FX-51 in every benchmark except memory related ones, but I decided to see if I could overclock it a little. I tried a FSB (front side bus) of 220MHz at first, but not surprisingly, that was pushing something in my system too hard. I gradually lowered down to 210MHz FSB, where I was finally happy with the stability. A FSB of 210 results in a operating frequency of 2.31GHz, not much higher than the default 2.2GHz I know, but what the heck. Here are some of the test results at 2.31GHz.
3DMark2001SE (Build 330)
The 5% processor overclock resulted in about a 5% performance increase in PCMark2002 but only a 2% boost in 3DMark2001SE. Not bad and not too surprising really.
After seeing how well the 3400+ performs compared to the FX-51, I really can’t wait to see what AMD comes out with next. The FX-51 is still the top performing processor overall, but for about $330 less you can get most of the FX-51’s performance out of the 3400+ (with the major exception being memory performance). If you really don’t need the greater memory bandwidth but can’t do without topnotch performance, then I wholeheartedly recommend the 3400+ if you can afford it (priced at about $418.00 according to Pricegrabber).
If you’re like me, that $418 is a little (actually a lot) hard to swallow. At $279.99, the 3200+ is good investment if you just want to get your feet wet in the 64-bit revolution. The 3000+ can be had for $50-60 less even, but you only get 512KB of L2 cache.