AMD versus Intel: Successes and pitfalls of their processor architectures – (1) Abstract and Introduction
AMD versus Intel: Successes and pitfalls of their processor architectures
Warning: This article is boring if you are not interested in computer architecture design. But if you are taking Computer Architecture class, this can be a very good study guide for you.
As we know, AMD and Intel are two leading companies in the processor market. Both of them has been existed for a while. There has been a lot of work posted online for identifying the differences of their architectures. However, those articles are either too general to capture the essence of computer architecture topics or too detailed to catch the overview of the their design differences. The goal of this paper is to compare the success and pitfalls between AMD and Intel in a proper-grained level of perspective. In order to get that perspective, I compare AMD and Intel processor architectures including the topics that are covered in the H&P textbook such as instruction set architectures, memory hierarchy, memory optimization, pipelining, instruction-level parallelism and thread-level parallelism. Instead of evaluating their performance, I mainly summarize their general designs. All information is collected from the textbook, online articles or research papers. The major contributions of this paper include: 1) This technical report can be served as a summary for understanding how processor architecture techniques are implemented in AMD and Intel. 2) It can also be used as a guide for selecting appropriate CPUs based on certain requirements. 3) It can be used as a brief manual to understand computer architecture concepts. 4) The comparison is an up-to-date version and therefore captures the latest technique.
Reference format: There are multiple places where information from the H&P textbook is used. The reference use a self-defined format like “P#”(# is page number).
I . Introduction
Processor architecture technology involves instruction set, memory hierarchy, memory optimization, pipelining, instruction level parallelism, and power and energy consumption. Each topic can be a very complex research area. AMD and Intel are two leading companies in the processor market. There have been a lot of questions regarding comparison between them, such as which one is cheaper, which one is better in graphics, etc. It may also be true that AMD is better in graphic computing. If we are able to compare AMD and Intel by following the major architecture topics, the comparison would bring more information which could be useful to entry-level users, architecture developers, or even architecture experts.
Hennessy and Patterson  have listed some examples of Intel and AMD in some chapters when they introduce each topic in computer architecture, but the examples are scattered and not in perspective of summary or overview. In addition, the data for Intel and AMD processors are not detailed enough to get a better understanding of how Intel and AMD differ to each other.
There are various kinds of online articles [8, 9] that illustrate different architectures such as Bulldozer, Sandy Bridge, etc. The pitfall of those articles is that they are either too general(they lack the deep perspective of architecture view) or too detailed to get the overview of the design. Also there is no comparison between those two vendors in a proper-grained level.
To demystify the differences of architectures between AMD and Intel, their general designs are frequently compared with each other. As the architecture design from each vendor changes with the time, I need to refer to a certain model for the processors compared. The following list introduces the major project from AMD and Intel; they are from their official websites [2, 3]. They are ordered by level of performance in an ascending order:
AMD for Desktop: Sempron, Athlon, Phenom, A-Series, FX
AMD for Server: Opteron
AMD x86 microarchitecture roadmap after K10: K10 -> Turion X2 Ultra -> Fusion -> Bobcat -> Bulldozer
Intel for Desktop: Celeron, Pentium, i3, i5, i7
Intel for Server: Itanium, Xeon E3, Xeon 5000, Xeon E5, Xeon E7
Intel x86 microarchitecture roadmap after P5: P5 -> P6 -> Core -> Nehalem -> Sandy Bridge -> Haswell -> Skylake
The main contributions of this technical report are:
- This technical report can be served as a summary for understanding how processor architecture techniques are implemented in AMD and Intel.
- It can also be used as a guide for selecting appropriate CPUs based on certain requirements.
- It can be used as a brief manual to understand computer architecture concepts.
- The comparison is an up-to-date version and therefore captures the latest technique.
<pre><code> String foo = "bar"; </code></pre>