Ripping apart the AMD Ryzen Threadripper - [Part 1] Introduction and Micro-Architecture

Introduction


It has been more almost two months since the launch of the new AMD Ryzen Threadripper platform. This high-end platform officially signifies the return of AMD into the consumer CPU space, completing the full range of desktop processors from the entry-level to High-End Desktop Segments. Ryzen was hugely welcomed by consumers which sent a very strong message to competitors in the same market that AMD is back in the game.

In this month’s special, we are going to take a deep dive into the AMD Ryzen Threadripper platform, to fully understand the strategies that AMD is taking to make their CPUs great again. *Make AMD Great Again* The whole series will be segmented into 5 parts. In each article, we will focus on certain topics of the entire platform. For a start, let’s study AMD’s strategy with the Ryzen Threadripper, as well as the whole Ryzen Micro-Architecture to find out why AMD is able to scale their Ryzen series so well.


AMD had quite a journey for the past 10 years

I built my very first desktop computer back when I was 16, with a AMD Athlon x2 5000+. Quite frankly speaking, I have always been a AMD fanboy (You could also probably guess my age from that). AMD always focused on providing the best value to consumers as well as better performance at a lower price than the competition. 

While Intel continued to innovate with their CPU architectures with their tick-tock strategy, AMD’s CPU Micro-Architectures seemed to have reached a peak. Performance increase was achieved largely by putting in more compute modules/cores into a single CPU chip, while ramping up core clocks. Since then, the gap in performance between AMD and Intel was slowly widening. As such, AMD simply lost their competitive edge in the consumer desktop CPU space. That was the time when Intel was able to leverage on their strength in advancements in the CPU Micro-Architecture, and dominated the market for many product generations.


The rise of the Zen Micro-Architecture

It wasn’t easy to come out with a whole new modern Micro-Architecture for a new CPU. After a long 5 years of conceptualization, design, development and further improvements, AMD’s Ryzen series CPUs were introduced earlier this year in March. The market was extremely receptive when the new Ryzen CPUs were launched and AMD’s desktop CPU market share rose significantly over the past few months. AMD’s target was to have the Zen based CPUs be 40% faster in terms of Instruction Per Cycle than their previous micro-architecture. With an aging Micro-Architecture design, AMD knew that they had to start from the ground up, and re-imagined the whole Zen CPU. It seems that their hard work over the past years in development of the new micro-architecture had definitely paid off, as AMD has announced that they had actually achieved a 52% increase in IPC performance, which is much higher than the targeted 40%.


The Ryzen Threadripper Strategy


The AMD Ryzen Threadripper is a monstrous CPU. It is the first ever desktop class CPU that features a whopping 16 cores and 32 threads, redefining the idea of a High-End Desktop (HEDT) computer. AMD has put their bet largely on bringing increased performance with multi-processing in order to compete with their competition. As Moore’s law reaches its limitations, it seems that the only way to improve performance is to increase parallelism by having more compute cores. This will also probably change the way how many other software applications will be developed and executed in the future. Of course, AMD’s strategy is to be the first out in the market to prepare users for the future of highly threaded software applications and show off the raw performance capabilities of their CPU. All these are done while still making their products a lot more affordable than competition.

The AMD Ryzen Threadripper is a product that is for a very niche market segment. In today’s context, I would rather label the AMD Ryzen Threadripper as a more application specific processor, rather than a HEDT processor. Not many applications will be able to utilize the huge number of computing cores, but this may change in the very near future. Of course, highly threaded applications such as 3D Modelling, Audio Engineering, Video Editing, Software Development, Data Analysis and Data Processing work will perform extremely well and are mostly optimized for the AMD Ryzen Threadripper. The keyword here is “optimized”. Of course, the Ryzen Threadripper will still be able to play games, or compute single threaded applications really well as compared to other processors. But as shown in the initial reviews of the Threadripper, the blue camp members seem to still do better.

The AMD Ryzen Threadripper sets itself aside as a serious multi-tasker. In complex compute applications, this could simply mean that the completion of computations cuts down from days to hours or even minutes. However, we would again like to state that the highest end Threadripper is not meant for mainstream users, as you would likely not to fully utilize all the CPU resources. However, let’s not forget the other processors in the Threadripper lineup. Luckily, AMD has also clearly thought about this, and introduced the Threadripper in 8 cores, 12 cores and 16 cores variants. In fact, the 8 core 1900X Threadripper will be a great CPU for HEDT. With the high clocks, large PCI-E bandwidth and many other features of the TR4 chipset, it will entice many gamers to their Threadripper platform as well.

To sum up the introduction of the whole AMD Ryzen Threadripper platform, we would say that the whole system is made and designed for creative professionals and scientists, but also works well in competitive gaming and streaming (particularly the 1900X). It is on an extremely scalable platform has great expandability options, and supports overclocking which can even bring it to a next level of performance.


Quick Look at the Zen Architecture in the Ryzen Threadripper

The whole Zen Micro-Architecture is able to reach its performance thanks to the focus on four areas.
  1. Increased Instruction-Level Parallelism (ILP) – better pipeline designs and enhanced branch prediction
  2. Enhanced Throughput – Introduction of Simultaneous Multi-Threading to reduce idle-time in pipelines
  3. Better Efficiencies – Low-power design methodologies that manages active and idle power for better performance per watt under different usage scenarios
  4. Increased Scalability – A modular 4C core complex (CCX) attached to a new interconnect called Infinity Fabric (previously known as Hyper Transport) that allows the whole Zen platform to scale, integrate and communication with other AMD IP in the core. 



  1. [Part 1] Introduction and Micro-Architecture
  2. [Part 2] The X399 Chipset and TR4 Socket (Featuring ASUS Zenith Extreme Motherboard)
  3. [Part 3] Optimized for Highly Multi-Threaded Creative and Professional Applications
  4. [Part 4] Overclocking, Temperatures and Energy Consumption
  5. [Part 5] Final Words and Conclusion


For the LATEST tech updates,
FOLLOW us on our Twitter
LIKE us on our FaceBook
SUBSCRIBE to us on our YouTube Channel!
SHARE
    Blogger Comment
    Facebook Comment

0 comments: