AMD Ryzen, Epyc, Vega

Since everyone does not have the time to watch a 4-hour AMD investors’ event, here’s a brief summary of what went down at the AMD Financial Analyst Day presentation.

What was the event about?

As expected, AMD concentrated on products aimed at high-profit market segments, hoping to ramp up investor confidence. Also, it revealed several roadmaps – detailing timelines for future product releases and development.

AMD execs further stressed the company’s commitment to capturing untapped markets by launching products suitable for a variety of market segments. Enthusiast PCs, content creation, data centers, advanced AI and graphics computing PCs and servers, are some of the areas that AMD kept their focus on, during the event.

“We are the only company in the world to have high performance graphics and high performance compute” – Lisa Su, President and CEO, AMD

Understandably, gamers and mainstream consumers were disappointed due to the lack of information about products relevant to their use. However, being an investor event, the focus was always going to be the profit markets, not the mainstream consumer market.

First up – AMD Epyc for Datacenters

AMD showcased the true successor to their famed AMD Opteron lineup after over a decade, AMD Epyc. The company called this platform ‘Naples’ in a press release on 7th March 2017.

Epyc Infinity Fabric
Infinity Fabric creating a Naples (AMD Epyc) CPU [Source – AMD Official Slides FAD ’17]

Using the same Zen architecture that features in the successful Ryzen series of CPUs, each Epyc CPU houses an incredible amount of processing power. Four 8-core Ryzen dies are seamlessly interconnected using AMD’s Infinity Fabric technology.

With a mind-boggling 64-thread count and 32 physical cores, the Epyc flagship CPU will mark a significant shift in the datacenter market. Even though, pricing still remains a concern and is currently unknown, the industry hopes for a cheaper alternative to the Intel Xeon lineup.

Watch the official video here.

AMD Ryzen Mobile

This is perhaps the most exciting news from Financial Analyst Day for the mainstream consumer. The long-awaited APUs from AMD will house on-die Vega graphics, based on their latest GPU architecture. Moreover, these Ryzen mobile CPUs will use 50% less power than the previous generation of AMD APUs, while delivering 50% more compute performance and 40% more graphics performance.

Ryzen mobile release
Ryzen mobile release roadmap [Source – AMD Official Slides FAD ’17]

Aimed at OEMs and large notebook manufacturers, these Ryzen-based APUs will make high-performance mobile computing possible in smaller form factors. We should start hearing reports of Ryzen mobile based laptops hitting the market soon.

 

Furthermore, AMD also plans to launch a Ryzen Mobile Pro lineup. This will serve the mobile workstation market, with increased performance. However, the power efficiency and actual performance delivered by these products remain to be seen.

AMD Threadripper

Taking aim at the high-end enthusiast market for the first time, AMD showed off ‘Threadripper.’ This CPU boasts 16 cores and 32 threads, marking the first real competitor to the Intel enthusiast processor lineup after over a decade of no competition.

AMD Threadripper
AMD Threadripper Introduced at FAD ’17 [Source – Official AMD Slides from FAD ’17]

Performance benchmarks and details have not been showcased yet, but this promises to be a release to watch. With the upcoming Intel Core i9 line of enthusiast CPUs, the price to performance ratio will be of utmost importance.

Watch the official intro video here.

AMD Radeon Vega

Raja Koduri, Senior Vice President and Chief Architect, Radeon Technologies Group, came forward to highlight one of the most awaited AMD products.

GPUs built around AMD’s newest Vega architecture.

Deep Learning and AI

AMD continues embracing its open standards stand with the launch of Radeon Open Compute platform. This is designed to provide developers and engineers with necessary tools to take advantage of the Vega architecture’s processing power, for deep learning and AI.

AMD Radeon Vega Frontier
Radeon Vega Frontier Edition [Source – AMD Official Slides FAD ’17]

Dubbed the Radeon Vega Frontier Edition, this GPU is AMD’s flagship product for this market segment. With 64 compute units (4096 stream processors), the Radeon Vega Frontier Edition delivers an estimated 25 TFLOPS of FP16 and approximately 13 TFLOPS of FP32 peak performance

In DeepBench, a popular benchmark tool used to measure deep learning performance – the AMD Vega Frontier Edition outperforms its NVIDIA P100 competitor by a healthy 30% margin.

AMD Radeon Vega Frontier
Radeon Vega Frontier Edition Performance [Source – AMD Official Slides from FAD ’17]

4K Gaming

Gamers received a glimpse into Vega’s performance, albeit a short one, with a Sniper Elite 4 demo at 4K. The game held a consistent 60-70 FPS at 4K resolution. However, the graphics settings were not mentioned.

AMD Vega 4K
Vega 4K Gaming Performance [Source – AMD FAD ’17 Webcast]
Moreover, only a short stretch of the game was showcased, and we should definitely wait for more performance insight from reviewers once the RX Vega, the gaming Vega GPU is launched.

Radeon Pro SSG

This GPU is aimed at content creators who work with a massive amount of data every day. AMD announced the Radeon Pro SSG with a whopping 2 Terabytes of onboard memory. The SSG or Solid State Graphics card is AMD’s attempt at providing content creators ample blazing fast memory to enhance their productivity and reduce the time taken to complete mundane tasks.

In addition to the 16 GB of high bandwidth memory bundled inside the Vega Frontier Edition, the Radeon Pro SSG features two terabytes of non-volatile memory. This is crafted into the GPU with an internal NVMe device.

People who work with large datasets should see a significant performance bump with this AMD product. However, the actual performance gains, price, and power draw still remain a mystery.

Watch the launch video here.

Future Roadmap

According to AMD, the journey toward recapturing markets has just commenced. Their roadmap slides detail their intentions for the future.

x86 CPUs

AMD Ryzen CPU Roadmap
AMD Ryzen CPU Roadmap [Source – AMD Official Slides FAD ’17]
 As AMD previously announced, they will be leapfrogging the 10nm architecture, to the 7nm Zen 2. The Zen 2 release is expected in late 2018 or early 2019. Moreover, the Zen+ CPUs, which will be refined versions of the Ryzen CPUs available today, should be on shelves by early 2018.

GPUs

AMD Radeon GPU Roadmap
Radeon GPU Roadmap [Source – Official AMD Slides FAD ’17]
Vega is just the beginning. AMD assured investors that Radeon™ Technologies had their focus set on the discrete GPU segment, starting with the high-performance Vega architecture in 2017. Moreover, after a refined Vega release based on 14nm+, the next release will be ‘Navi.’ This will be a 7nm architecture for GPUs, providing increased performance and delivering higher performance per watt.

Datacenters

AMD Datacenter CPUs Epyc Naples
AMD Datacenter CPUs Roadmap [Source – Official AMD Slides from FAD ’17
Even the AMD datacenter CPU roadmap tells a similar story. The Zen-based Epyc CPUs will be followed by ‘Rome’ and ‘Milan,’ based on 7nm Zen 2 and 7nm+ Zen 3 architectures respectively.

Gaming Vega?

AMD is still keeping details about the RX Vega close to its chest. However, many expect the high-performance gaming GPUs to be unveiled at Computex 2017, on May 30th.