Related Articles

Introduction

It has been just over six months since AMD first rolled out its first 90nm processors with the Winchester core which promised cooler running chips and a greater number of processors on each wafer. The success of this transition is apparent today as AMD has not run into heat and power issues and they remain the enthusiast favorites for gaming and overclocking. AMD has not been resting on their laurels however and today, we take a look at an updated Athlon 64 core, code-named, Venice, which will have the higher end parts from AMD make the move to 90nm also. While the manufacturing process remains the same as the Winchester's there are several internal differences between the two cores, and the changes are noted here -

• SSE3 multimedia instruction sets
• Additions and enhancements to the Integrated Memory controller
• Mismatched DIMM support (ability to configure and use different size DIMMs on the same channel)
• Improved memory mapping (more efficient use of memory space)
• Improved memory loading (fully populating the memory with double-bank DIMMs no longer requires the memory controller to run @ DDR333)

For those that are not familiar with the Athlon 64 architecture, there are several significant differences between it and Intel's Pentium 4 line up, most notable of which is the on-die memory controller.

With the onboard memory controller on the Athlon 64, it scales in speed along with the processor which means that latency is minimized. The links to I/O and memory are also independent which means there is no bandwidth contention. This has also allowed AMD to keep the socket fairly consistent as there has already been two updates to the memory controller, none of which required a new board. Prior to the Athlon 64, memory controller duties were taken care of by the Northbridge thus the processor communicates with the memory through the front side bus. With the 925XE, Intel has tried to address bandwidth limitations by raising the FSB speed from 800Mhz to the 1066Mhz.

The Venice Core and the AMD clockspeed philosophy

It is important to note that until now, all processors above the 3500+ range are based on the 130nm process which means that power consumption will drop at the same clockspeeds. CPU-Z reported our core voltage to be 1.36 while the BIOS reported it as 1.38; interestingly enough, the AMD webpage lists the voltage as variable. We are working to find out whether this means that the voltage scales during operation or if it only varies between different chips. With the reduction in voltage, AMD should be able to kick the clockspeed of their processors up a couple notches, however AMD does realize that there is a ceiling to what they can do and they articulated this in no uncertain terms at WinHEC. The future is dual core but in the near term, but the faster single core solutions will still be aimed at gaming enthusiasts. Since AMD has favored the 200 Mhz increases in clockspeed and it seems likely that the next iteration of the FX will debut at 2.8 Ghz.