Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ Search Prices:

LOSTCIRCUITS	SHORTCUTS: Names and Confusion Benchmarking Issues Test Configurations Content Creation Office Productivity 3D Rendering Cinebench 2003 FutureMarks Gaming A-D Gaming M-X NX bit Tales More NX bit Tales Surprises and Conclusions Give Us Some Feedback to Help Us Improve our Reviews
AMD Athlon64 4000+
(Review by MS and Jerry Coffin, October 19, 2004)

Summary

The release of the Athlon64 4000+ and FX55 is about to extend AMD's leadership in the desktop domain but it also raises new questions about the validity of the performance rating and consequent naming convention. Multiple processors are branded with the same name on the basis of cache size vs. clock speed and memory interface and in the end, there is a wealth of confusion lingering on. Into this scenario are released the two new top contendors, that is the Athlon64 4000+ and the FX55. We were looking at the 4000+ in comparison to the 3800+ released a few months ago and, of course also in comparison to the competition in house and from Intel.

Among the different parameters we were looking at, that is system performance, gaming and 3DRendering we found some differences between the different models tested and then we dove headlong into the issue of memory protection and the NX bit. Well, at least, we tried.

Another day, another processor, another review. We could end it right here with the lapidary statement that it was faster, as expected. But wait, it is not that simple. We are looking at the latest version of AMD’s Athlon64 – 4000, using a Socket 939 interface, sporting a 1 MB L2 cache, running at a whopping 2400 MHz and manufactured on a 130 nm copper interconnect process . The specs are certainly impressive, on the other hand, the novelty factor is not. After all, the original FX53 had basically the same specifications, with the main difference of one additional pin and the requirement of running Registered DIMMs with the known drawback of about $20 per module and the additional latency cycle in the address and control decode pathway. Keep in mind, though that the term FX53 now applies to a different processor using the Socket 939 interface, which leads straight to the next issue at hand

      

The CPU and the Cooler

Impressions are one thing, confusion is what we were referring to above. There are Clawhammer and Newcastle cores out there with 1 MB or 512kB L2 cache, respectively, and the naming convention assigns different QuantiSpeed values to the same clock speed depending on the L2 cache size. Adding to the overall cluelessness is the issue that the same moniker can describe two totally different processors. Case in point is the already-mentioned FX53 with either 939 or else 940 pins, depending on the release date. Similar issues hold for quite a few other Athlon64 processors, cache is substituted for clock speed or vice versa and everything is based on some relatively obscure performance metric. Suffice it to say that AMD probably deserves an award for the most questionable processor numbering scheme. That aside, there is little doubt that the Athlon64 is a great processor, and AMD appears to be doing well, so there has to be something the company is doing right. Time to introduce the latest and second-greatest AMD processor of all times, the Athlon64 4000+

Names and Numbers

Along with the 4000+, AMD also released a new FX55 processor, running at 2600 MHz. Briefly, all FX-series CPUs are unlocked whereas the xx00+ processors are locked. To retract on our statement from above about the confusion, we need to grant AMD that the differences in processor migrating through the same model numbers are reflecting technological progress, that is, it is in the customers interest to incorporate changes as smoothly as possible. In many ways, the nomenclature is now on the way to a better understanding since all Socket940 CPUs are branded under the Opteron name, the FX are simply unlocked and the only issues left are the cache vs. clockspeed rating. In other words, it appears as if finally the Gordian knot has been unraveled to the point where the nomenclature makes some sort of sense – if one has the secret decoder pen - and it still doesn't really work.

A similar chart can be found on Tech-Report's A64-4000+ Coverage but it is not that pretty...

For the Benchmarks shown in this report, all “FX” references are pointing to Socket 940 processors rather than the latest versions of Socket 939 processor.

next page: => Benchmarking Issues -- Test Configurations =>