QuantiSpeed Architecture

This content was originally featured on Amdmb.com and has been converted to PC Perspective’s website. Some color changes and flaws may appear.

Okay, you are probably wondering exactly what in the world is this QuantiSpeed architecture term that I have been throwing at you. First take a look at how AMD described the importance of this feature in their white paper document:

QuantiSpeed Architecture allows the AMD Athlon XP processor to accomplish more work per clock cycle (IPC) when compared to other competing processors and achieve relatively high operating frequencies. For these two reasons, QuantiSpeed Architecture is the key element that provides the AMD Athlon XP Processor with a “quantum leap” in real-world application performance.

Basically, the QuantiSpeed architecture is the term AMD has applied to the changes and enhancements made to the new Palomino core. As you probably already know, these same changes were also brought into the AMD line of processors in the mobile Athlon 4 chip, the Athlon MP chip as well as the Duron processors at or over 1.0 GHz. Why did AMD just now decide to apply this new marketing term? Simply because the home user market is vastly different than the mobile or server markets and thus the need for different tactics to sell their products in each corresponding market was needed.

What are the changes that QuantiSpeed Architecture brings about? There are four changes that AMD lists as being the heart and soul of the Palomino core.

1. Nine-issue, superscalar, fully pipelined micro-architecture

The newly pipelined Palomino core is what AMD refers to as “a balanced approach” to processor pipeline length. Additional pipelines simply open up more pathways for processor instructions to flow through to the core and this translates into more work done per clock cycle. AMD’s balance of both the frequency and pipelines (thus, IPC) produces the high levels of performance that the Athlon XP provides. The longer pipelines by themselves, such as in the Pentium 4 processor, translate into a lower IPC but higher clock frequencies, and similarly, shorter pipelines alone translate into increased IPC but slower clock frequencies. Finding a balance between the two is what AMD has accomplished in the Athlon XP processor.

2. Superscalar, fully pipelined Floating Point Unit (FPU)

This item that AMD mentions is basically just to emphasize the floating-point power of the Athlon processor. It by far completes more floating-point operations per clock cycle that any other x86 processor. Floating-point operations are used in scientific calculations such as those in 3D programs including games and rendering in professional programs.

3. Hardware data prefetch

As I described in my Athlon MP article a few months ago, hardware prefetching consists of retrieving instructions from the system memory and bringing them to the processors on chip L1 cache. This is a very common technique used to improve a processors IPC value. It reduces the time (in clock cycles) needed to retrieve/process/output data from applications and the operating system. This enhancement does not require any kind of software rewriting or changes, so ALL applications and games receive the benefits of this feature to varying degrees.

4. Exclusive and speculative Translation Look-aside Buffers (TLBs)

The final way that both the Athlon XP and Athlon MP improve their work completed per cycle is with better TLBs. TLB structures keep the maps to data in memory close to the processor for easy access and look ups of data to be retrieved from memory. Changed in three ways, the new Palomino core TLBs improves real-world application performance. First, the new TLB structure is larger and gives the Athlon XP processor access to additional data maps. Secondly, the exclusivity of the TLBs means there will be less duplication of mapped memory, again freeing up more space in the L2 cache for other useful data. Finally, the speculation of the TLBs, similar to hardware prefetch, attempt to predict what TLBs will be needed before they are called upon.

Also, if you would like to read more about all of these changes, start reading on page 6 of the Athlon MP article.

To help the customers better understand the MHz myth and how it pertains to AMD and Intel processors, AMD included a few interesting analogies in their QuantiSpeed Architecture white paper and I am going to share one of them with you here:

Tour de Performance
Two cyclists ride together on 10-speed bikes. One cyclist uses the 10th gear, pedaling slower but moving faster down the road and covering more distance with each stroke. The other cyclist uses 1st gear and has to pedal like a lunatic to achieve even close to the same speed on the road and cover the same ground.

Processors work like this as well. Some processors have to pedal faster (measured in Megahertz or Gigahertz) just to travel the same distance. The AMD Athlon XP processor is more efficient because it travels farther per stroke, therefore it can outperform a processor that simply spins it pedals faster.

« PreviousNext »