Examination II
Externally, the Athlon 64 3400+ is no different from the incumbent 3200+. Let's take a look.
Not listed in the enhancements discussion on the previous page, the use of an integrated heatspreader makes implicit sense. Expensive CPUs, large die sizes, and awkward cooler mounting is a sure-fire method of unnecessarily shortening the life span of a CPU. The metal jacket is reminiscent of the Pentium 4 processor. Other than the 3400 reference in the first line and 0341 manufacture date, there's little to be gleaned from the surrounding codes. Unlike, for example, certain Barton XP2500s with the desirable AQXEA stepping, there's not been enough research carried out to determine if one Athlon 64 3xxx+ stepping is markedly better than another.
If you have the patience of a saint and try to count the number of pins, there should be 754. That's a significant increase over the Barton's 462, but it is still less than the dual-channel SledgeHammer's 940 pins. The on-die memory controller accounts for part of the substantial increase. That's perhaps one of the side-effects of a general change in architecture - the forced change motherboards and logics. There's just no 'slocket' that's going to allow the 3400+ to run on an nForce2 motherboard. Even if there was, you'd realise that this CPU is a shift in AMD's performance thinking, evidenced by an on-die memory controller.
AMD's drive to educate the masses that pure clock speed isn't the single defining performance factor, which goes under the banner of True Performance Initiative, is the reason how and why AMD arrives at its model number speeds. The Model 3400+ doesn't infer that it runs at 3.4GHz. Rather, it's a performance figure based on the overall work done by the CPU (and motherboard) in a wide-ranging series of tests. The Model 3400+ runs at a nominal 2.2GHz, which is the Barton 3200+'s native speed. The factors listed on the previous page, AMD reckon, allow it to have a performance rating upgrade. Our tests with the Model 3200+ have shown that, if anything, AMD is being conservative in its estimates. Our glut of memory intensive tests, especially gaming, make it stand out as an excellent all-rounder. There's little reason for the 3400+ to be anything but faster.
A table should help us highlight the attributes that should see the Model 3400+ do well in our benchmarking section.
| CPU | AMD Athlon 64 Model 3400+ | AMD Athlon 64 FX-51 | XP3200+ Barton | Pentium 4 3.2GHz |
| Clock speed | 2200MHz | 2200MHz | 2200MHz | 3200MHz |
| L1 cache | 128kb | 128kb | 128kb | 20kb |
| L2 cache | 1024kb | 1024kb | 512kb | 512kb |
| Memory bandwidth | 3.2GB/s (DDR400) | 6.4GB/s (DDR400) | 6.4GB/s (nForce2) | 6.4GB/s (i875P) |
| FSB | 2200MHz (core speed) | 2200MHz (core speed) | 200MHz double pumped | 200MHz quad pumped |
| Integer pipeline length | 12 | 12 | 10 | 20 |
| CPU Die Size | 193mm² | 193mm² ?? | 101mm² | 131mm² |
| Transistor count | 105.9 million | 105.9mm² ?? | 54 million | 55 million |
| Manufacturing process | 0.13-mircon SOI | 0.13-micron SOI | 0.13-micron | 0.13-micron |
| Memory support | DDR400 Single Channel | DDR400 Dual Channel | DDR400 DC | DDR400 DC |
| OS Support | 32/64-bit | 32/64-bit | 32-bit | 32-bit |
| Voltage | 1.5v | 1.55v | 1.65v | 1.525 - 1.55v |
| Form Factor | 754 | 940 | 462 | 478 |
Precious little to separate it from the FX-51. The latter is potent in design, apparently. It requires registered memory and uses a dual-channel 64-bit controllers for a possible 6.4GB/s bandwidth at DDR400 speeds. The inherent trouble with registered memory is the difficulty in balancing protection against latency. It's only recently that low-latency registered modules have become widely available. And there's the stupendous price tag to consider. The ClawHammer 3400+ won't be cheap, but it should still be significantly less wallet-busting than the FX-51. Enough conjecture. Let's now see how it stacks up against the competition.
