The last fifteen months have seen a dramatic shift in the prospects of the AthlonXP core architecture when compared to the Pentium 4.  Throughout the first half of 2002 the AthlonXP fell farther and farther behind as Intel’s P4 effortlessly scaled upwards, appeared on new and higher-performing chipsets, and jumped to a higher front-side bus.  By comparison, the AMD’s flagship was wheezing, having gone nearly as far as even its excellently-tuned .18 micron process was capable of taking it.  Sunnyvale’s first foray into .13 micron technology on the desktop wasn’t exactly encouraging, as the first revision of the 2200+ ran far hotter than even an equivalently-clocked Palomino would’ve done.

Things began to turn in the fall.  AMD surprise-launched their 2600+ on a much-improved and cooler-running .13 micron process and NVIDIA’s nForce2 chipset made an impressive debut in late Nov, quickly sucking the high-end market away from VIA (who, nine months later, has yet to effectively answer the nForce2?but that’s another story?).  Now, armed with Barton’s larger L2 cache and yet another front-side bus jump courtesy of the nForce2 architecture, AMD is betting its latest CPU is powerful enough to award Sunnyvale the performance crown for the first time in over a year.  Does the rating fit the chip?  We’ll find out.

Playing Coy with Clock Speeds:

One interesting factor in AMD’s current level of performance is just how little actual clockspeed has changed over the last nine months.  Although AMD’s highest-end CPU has moved from a 2600+ to a 3200+ model number (an increase of 23%) the actual clockspeed of the CPU in question has increased by a whopping 62MHz.  There’s nothing wrong with this (AMD has, after all, increased their front-side bus speeds by 50% in the same period (266 times 50% is 133?266+133=400), but it’s interesting to note just how little the AthlonXP’s pure clockspeed has risen.    We’ll be discussing possible explanations for this behavior later in the review.

The Platform:

A year ago the motherboard market was dominated entirely by VIA KT333 boards while NVIDIA’s nForce had the markings of a promising idea that simply hadn’t delivered.  SiS was effectively out of the Socket A market, ALI hadn’t been heard from ever since the resounding splat of mediocrity that was MAGICK 1, and AMD’s own AMD760 platform had been retired from the single-CPU market.  With no real competition on the radar VIA seemed to be firmly at the head of Socket A chipset development?but looks can be deceiving.

Only twelve months later, VIA has been entirely de-throned from the Socket A high-end market by NVIDIA’s second-generation nForce2 chipset.  The Taiwanese company has even managed to surpass its own record for product re-introductions by introducing a KT400A chipset that still couldn’t run a 400 MHz FSB.  That facet of support won’t come until KT600, aka KT400AA. 

This has left AMD entirely dependent on NVIDIA for a high-performance, high-end chipset?and after some initial troubles and quirks, NVIDIA is proving themselves worthy of the task.  The latest version of the nForce2 chipset (nForce2 Ultra 400) is the revision that officially supports AMD’s 400 MHz FSB, but earlier revisions may be capable of reaching this speed as well.  Although all versions of the nForce2 chipset have the proper dividers in place to run at a 200 MHz FSB, the "Ultra 400”  revision is the only one fully optimized for synchronous operation with AMD’s latest speeds.

next > next > next > next >

---

• Test Setup (0 of 4)
• Benchmarks (0 of 4)
• Benchmarks, (Cont) (0 of 4)
• Analysis, Conclusion (0 of 4)