CPU: AMD's new Athlon XP CPU line :: Features

10-18-2001 · Category: Hardware - Processors

By Doc Overclock

The XP processors from AMD boast many new features and instruction sets that make it a more performance oriented and versatile product for both the desktop and the server market. One of the key ingredients is their new QuantiSpeed architecture which allows the AMD Athlon XP processor to accomplish more instructions per clock cycle (IPC).

Some of these enhanced features include the new nine-issue, superscalar, fully pipelined microarchitecture which provides more pathways to feed application instructions into the execution engines of the CPU core. This allows the processor to complete more work in a given clock cycle making for much quicker response times and faster data transfers.

A Superscalar, fully pipelined Floating Point Unit (FPU) completes more floating point operations per clock cycle than the competitors x86 processors and permits higher operating frequencies. The end result of this process is a processor with the computing power to tackle the most computation-intensive software applications available today.

Another one of the new instructions is the hardware data prefetch, a feature which prefetches data from system memory to the processor's Level 1 cache. This reduces the overall time it takes to feed the processor critical data resulting in increased work throughput, improving the overall processing performance.

Exclusive and speculative Translation Look-aside Buffers (TLBs) are a feature that keeps the maps to critical data close to the CPU. What this does is help prevent the processor from stalling or waiting when future data is being requested. These TLB structures on the Athlon XP are now larger and offer better exclusivity between caches, and speculative processing.

Larger TLB's give the AMD Athlon XP processor access to additional data maps for faster access times to your data. Exclusivity removes the duplication of information, freeing up more space in the Level 2 cache for other useful data to be used by the processor. And the speculative nature of these structures allows the processor to generate future maps of critical data quickly and without error.

These four key advances are what allow this new QuantiSpeed architecture to perform more calculations per second which should boost overall productivity and enable a more robust computing experience allowing better performance gains across the board for the user. These are the main feature benefits that pull this chip ahead of the pack and in our tests we will see if it lives up to the image it has presented for itself on paper.

The physical differences between the Athlon Thunderbird and the XP Processors are among the biggest and most important. The newer "organic" PCB, very similar to what Intel has been using for years, should allow more "give" when installing heatsinks as well as force even more of the heat through the primary heat pathway compared to the old ceramic PCB.

Other physical changes include a slight change in DIE shape (contrary to popular belief, the DIE solely consists of the portion of the CPU lifted above the PCB) as well as moving some resistors to the backside of the CPU. Intel has been using this method for quite sometime now and it is a good move on AMD's part to change the design in this fashion.

Among the most important of these physical changes is the change in PCB material. In theory, the new PCB will reduce the amount of heat "leakage" from the core, forcing more through the primary heat pathway, further increasing heat transfer and efficiency. However, a noteworthy item would be that current socket-thermistor based temperature measurements, already inaccurate with Thunderbird processors, become even more inaccurate with XP processors.

Already we are seeing reports on the internet of people switching from Thunderbird to Athlon XP chips and seeing massive "drops" in temperature despite CPU wattage staying the same or even increasing. This is due to both the new PCB allowing even less heat through the backside, in addition to the new resistors on the backside preventing certain motherboard/socket-thermistor combinations from contacting the CPU backside PCB.

Therefore, in my opinion, it is imperative that sites and people do not compare old T-bird temps with newer Athlon XP temps. They are not the same CPU, and the changes are significant enough to invalidate a direct comparison. Also worth noting is, when comparing CPU wattage output the Athlon XP does not deliver on AMD's promised 20% heat reduction.

While the slightly larger contact surface does help (128mm squared versus 120mm squared), and any heat reduction is great but it seems likely AMD's PR department just overestimated how much the engineering department could achieve in terms of heat reduction. The XP does however run cooler than its 1.4GHz brother but the margin is more in the 10% range and since heat is still the enemy it is better than nothing.

The actual internal differences of the Athlon XP processor are already well documented. They should certainly increase the performance of the chip, which in addition to the physical changes, should allow AMD an "interim" ramping process until they start mass-producing the .13micron Thoroughbred CPUs next year.

All of these changes should keep AMD very competitive with whatever processors Intel might release in the near future and that always is good for the people as price wars will soon ensure. I have listed in bulleted form the complete technical features of the Athlon XP as listed on the AMD site. In our performance section we will compare the Athlon XP against an Intel Pentium 4 and see how well the CPU performs in our series of tests. Check out the performance section and see if the XP can tackle the P4, read on.

Athlon XP 1500+ QuantiSpeed architecture operates at 1333 MHz
Athlon XP 1600+ QuantiSpeed architecture operates at 1400 MHz
Athlon XP 1700+ QuantiSpeed architecture operates at 1467 MHz
Athlon XP 1800+ QuantiSpeed architecture operates at 1533 MHz
  • QuantiSpeed™ Architecture
  • 384K Total High-Performance, Full-Speed Cache
  • Advanced 266MHz Front-Side Bus with ECC (Error Correcting Code) Support
  • 3DNow!™ Professional Technology (72 instructions, full SSE compatibility)
  • Support for Double Data Rate (DDR) Memory
  • Multiple parallel x86 instruction decoders
  • Three out-of-order, superscalar, fully pipelined floating point execution units, which execute x87 (floating point), MMX™ and 3DNow!™ instructions
  • Three out-of-order, superscalar, pipelined integer units
  • Three out-of-order, superscalar, pipelined address calculation units
  • 72-entry instruction control unit
  • Advanced hardware data prefetch
  • Exclusive and speculative Translation Look-aside Buffers
  • Advanced dynamic branch prediction
  • 21 original 3DNow!™ instructions-the first technology enabling superscalar SIMD
  • 19 additional instructions to enable improved integer math calculations for speech or video encoding and improved data movement for Internet plug-ins and other streaming applications
  • 5 DSP instructions to improve soft modem, soft ADSL, Dolby Digital surround sound, and MP3 applications
  • 52 SSE instructions with SIMD integer and floating point additions offer excellent compatibility with Intel's SSE technology
  • Compatible with Windows® XP, Windows 98, Windows 95, and Windows NT® 4.x operating systems
  • Established Socket A Infrastructure
  • Source synchronous clocking (clock forwarding) technology
  • Support for 8-bit ECC for data bus integrity
  • Peak data rate of 2.1GB/s
  • Multiprocessing support: point-to-point topology, with number of processors in SMP systems determined by chipset implementation
  • Support for 24 outstanding transactions per processor

Feature comparison between the AMD Athlon XP and the Intel Pentium 4 processor

Feature AMD Athlon XP Pentium 4
QuantiSpeed™ Architecture Yes No
Operations per clock cycle 9 6
Integer pipelines 3 4
Floating point pipelines 3 2
Full x86 decoders 3 1
L1 cache size 128K 12k µop (Trace Cache) + 8KB (Data Cache)
L2 cache size 256KB (on-chip) 256KB (on-chip)
Total on-chip full-speed cache 384KB 64KB + 12k µop
Total effective on-chip full-speed cache 384KB (exclusive) 256KB - 12k µop (inclusive)
System bus speed 266MHz 400MHz
3D Enhancement instructions 3DNow!™ Professional SSE2
Cache/prefetch controls Yes Yes
Streaming controls Yes Yes
DSP/comm extensions Yes Yes