AMD has undergone a turnaround in the past two years, with its stock price rising from as low as $2 to $16. At the Deutsche Bank Technology Conference, AMD's Chief Financial Officer Devinder Kumar pointed out that this transformation was triggered by the company's reformed product strategy and target markets, a stable leadership team and a focus on product execution.

Kumar said the company is committed to growing revenue and improved its financials in 2017. AMD is launching multiple products in various markets, such as gaming and cloud computing, allowing the company to seek more market share.

In 2018, the company plans to shift its focus toward increasing profits by launching a complete suite of premium products across all its businesses.

AMD exploded onto the market in 2017 with new products across every market segment and price range. The company's product deluge began in March 2017 with the launch of its high-end desktop processor, the Ryzen 7. Then came the mid-range Ryzen 5 in April, and the low-end Ryzen 3 and entry-level Bristol Ridge in July.

In August, AMD shifted its focus to gaming. The company launched the high-performance gaming processor Ryzen Threadripper and Vega GPU. In September, it launched the commercial desktop processor Ryzen Pro.

The company has yet to launch a mobile version of its Ryzen series. The mobile versions will have Vega GPUs built in.

AMD has seen a lot of success with the launch of its server products. In June, the company launched the EPYC server processor. It is currently launching the Radeon Instinct GPU for cloud computing. The company is working with Baidu and Amazon in cloud computing.

With the launch of Ryzen and Vega, AMD skipped two manufacturing process generations and went directly from the 28nm node to 14nm. This shift helped AMD keep its products competitive with those of Intel and NVIDIA.

While the Ryzen desktop processors have been well received by the market, Vega faces some issues in the gaming space as it fails to offer any significant performance gains over NVIDIA GPUs. Strong GPU demand from crypto mining companies has increased the price of Vega GPUs, discouraging gamers from buying AMD’s GPUs at high prices. AMD has announced plans to move Ryzen and Vega to Global Foundries’ 12nm node in 2018 to stay competitive.

AMD is targeting the PC and traditional server markets, which Intel is trying to break away from, and is seeking to gain market share with its competing products.

At the Deutsche Bank Technology Conference, AMD CFO Devinder Kumar explained that the company is targeting markets with little or no market share. For example, it has almost no share in the server CPU market.

AMD will not target new emerging markets such as AI (artificial intelligence) or automotive electronics as they require large R&D expenditures, which the company lacks. Currently, the company is aiming to recover from years of losses and targeting markets that can achieve growth in the short and medium term.

Kumar said AMD is starting from a background of 28% gross margin in 2015. It gradually increased the margin to 31% in 2016 and plans to grow it to 34% in 2017, 36% in 2018 and 40% in 2020. The plan is to increase margins by entering the market with premium products and gaining market share.

AMD's 40% gross margin target is well below Intel and NVIDIA's current gross margins of 63% and 59%, respectively. AMD still has a long way to go before it reaches its competitors.

AMD's R&D productivity is higher than NVIDIA and Intel. NVIDIA focuses only on GPU, Intel on CPU, and AMD focuses on both CPU and GPU. AMD's operating expenses in 2016 were lower than those of Intel and NVIDIA, but AMD managed to introduce products that compete with its rivals and has higher R&D expenses.

Kumar said AMD needs to be "flexible in how it spends money." Like its competitors, AMD wants to spend money in emerging growth areas, but wants to limit its spending to maintain expenses. The company's operating expenses accounted for 34% in 2016 and plans to reduce it to 31% in 2017 and maintain it between 26% and 30% in the long term.

AMD is also in the process of clearing up all pending litigation. It recently agreed to settle a class action lawsuit with shareholders for a $29.5 million payment over the delayed launch of the Llam APU.

Shareholders filed the lawsuit in 2014, accusing AMD of misleading investors about the progress of its 32nm Llano chip, which was originally scheduled for launch in 2009 and was expected to compete with Intel's Sandy Bridge.

However, the product launch was pushed back to 2011. By this time, the chip was already obsolete and AMD had to write off $100 million in unsold inventory. This was one of the events that contributed to AMD's financial deficit.

AMD's long-awaited next-generation Vega GPUs launched in mid-August 2017. The company unveiled its first set of Vega cards -- the Radeon RX Vega 64 and RX Vega 56 -- to compete with NVIDIA's Pascal GTX 1080 and 1070.

There are two versions of the RX Vega 64 - air-cooled and water-cooled. The RX Vega 64 is priced at $499, similar to the GTX 1080, while the RX Vega 56 is priced $100 less than the GTX 1070.

PCGamesN reviewed AMD's Vega 56 GPU and found that it outperformed NVIDIA's GTX 1070 in games built using the DirectX 12 API, but that the GTX 1070 fell short when it came to last-gen DX11.

PCGamesN said that DirectX 12 was only a year ago, and most games are still built on the old DX11. It is likely that most games launched in the next six months will be built on DX11. DirectX 12 will probably be widely adopted in 2018. Therefore, PCGamesN recommends gamers to buy Vega 56 in six months.

Third-party reviews show that AMD's Vega cards match or are close to NVIDIA's GTX 10 series in performance, but lag behind in power consumption, which will threaten NVIDIA in the high-performance GPU market.

AnandTech states that the Vega 56 is 8% faster than the GTX 1070. This is not a significant performance increase compared to NVIDIA's GPU. Therefore, AMD has adopted a fallback strategy of offering similar performance at a lower price. The Vega 56 is 80% of the price of its opponent and offers 90% of the performance of the GTX 1070.

NVIDIA is still ahead of AMD, having launched high-end versions of the GTX 1080, the GTX 1080 Ti and Titan XP, in March and April 2017, respectively. The GTX 1080 Ti is priced at $699 and is similar to the liquid-cooled version of the Vega 64. However, the Ti is 35% faster than the Vega 64 and consumes 30% more power.

Since there has been a lot of discussion about Vega's power efficiency, Raja Koduri, head of RTG (Radeon Technologies Group), spoke in a Twitter chat about the company's strategy behind Vega.

During the chat, Koduri said that most users prefer performance per dollar/watt. With Vega, AMD gives users performance flexibility by enabling three separate operating power modes: Power Save, Balanced, and Turbo.

Turbo mode pushes the GPU to its performance peak while consuming the most power. Power Saver mode, however, provides the highest power efficiency without compromising performance. Balanced mode is somewhere in the middle of the other two nodes.

Koduri said that in power saving mode, RX Vega 64 provides 35% power efficiency while sacrificing only 4% performance. Third-party reviews tested Vega in balanced mode. So the power efficiency gap is huge.

After testing the Vega 64 in power saving mode, the power efficiency gap narrowed but did not disappear. NVIDIA's GTX 1080 still consumed 18% more power, while the GTX 1080 Ti consumed 24% more power than the Vega 64. " Tom's Hardware " measured the gaming power consumption of the Vega 64 in power saving mode and found that it consumed 196.7 watts, while the GTX 1080 consumed only 173 watts.

NVIDIA could extend its technological advantage over AMD by launching its new generation Volta GPUs. The company has already launched Volta GPUs for data centers , which offer significant performance improvements over Pascal GPUs. Some supply chain rumors suggest that NVIDIA could launch Volta-based GeForce GTX GPUs in early 2018.

AMD and NVIDIA are always hot topics. After launching the Vega GPUs, there was a rumor that NVIDIA would launch the GTX 1070 Ti , which would be in the middle of the GTX 1070 and 1080 in terms of features and pricing.

There is also a rumor that AMD will launch the RX Vega 32 and 28, which could replace its Radeon RX 500 series. The Radeon RX 500 series is a refresh of its Polaris-based RX 400 series. Rumor has it that these new Vega cards will compete with NVIDIA's GeForce GTX 1060.

Third-party reviewers have suggested that AMD 's Vega 56 GPUs are likely to be chosen by most gamers because they offer better value than NVIDIA's GTX 1070. However, AMD's most pressing issue is that the Vega GPUs are being promoted with this hidden risk.

In general GPUs, AMD and NVIDIA launch their reference designs based on their packaging as the first cards. After that, they offer cards to AIB (add-in-board) partners such as MSI, Gigabyte, EVGA, Asus, and PowerColor. AIB partners customize these GPUs in their custom designs by overclocking the GPU, adding better coolers, and other features.

However, AIB partners have been facing issues with Vega. It has been more than a month since AMD launched the Vega cards. Not a single AIB partner has released its custom GPUs yet, as most of the Vega cards have been mined by cryptominers.

Tom's Hardware Talked to several AIB partners and did several inquiries. XFX and Sapphire are still undecided on release dates, and PowerColor is facing a shortage of DRAM.

Furthermore, Gigabyte has yet to commit, MSI has dropped the ball on Vega, and Asus has delayed the launch of the card until mid-October 2017.

The custom Vega cards have been delayed due to two production issues - a design problem and a memory problem.

THG said that the problem with the Vega design was twofold. First, it was difficult for AIB partners to overclock Vega because of the variable design. Second, there were discrepancies in the temperatures reported by the board and the AIB.

The three variants of the Vega card had three different mounting methods, which made mass production of the cards difficult.

• As the first card to feature a molded bracket, the Samsung HBM2 (High Bandwidth Memory) is the same height as the GPU.

• Samsung HBM2 is 40 microns thinner than GPU.

  
  

• The third card has a variant with SK Hynix memory.

  
  

AMD's Vega launch time has extended from the normal 12 months to 16 months. HBM2 has been slower than expected to roll out. And the memory market has been experiencing supply shortages, which has increased memory prices.

There are reports that AMD lost money on Vega 64 due to high memory prices. This raises the question of whether AMD will abandon HBM2 and adopt GDDR6 in future GPUs.

While Vega has been experiencing manufacturing challenges, AMD's Ryzen has been doing very well. Next, we'll look at how Ryzen performs.

At the Deutsche Bank Technology Conference, AMD CFO Devinder Kumar said the company is seeing increasing adoption of Ryzen, with all major PC OEMs announcing premium Ryzen-based desktop products, including Acer, Dell, Asus, HP and Lenovo.

Kumar said that Ryzen Threadripper, Ryzen 7, Ryzen 5 and Ryzen 3 helped AMD to perform across the board. The rich product portfolio increased AMD's overall average selling price, thereby pulling the company's CG (computing and graphics) segment out of the quagmire of years of losses.

AMD has been rolling out its Ryzen desktop processors in phases. According to sales data from German retailer Mindfactory, 72.4% of CPUs sold in March 2017 were Intel processors and 27.6% were AMD's Ryzen processors.

Ryzen's contribution to retailers' overall CPU sales rose to 36.5% in April 2017 as AMD ramped up Ryzen 7 production and launched the Ryzen 5 series.

Ryzen sales started to rise in June 2017 as AMD accelerated Ryzen 5 production and launched the Ryzen 3 series. In July, Ryzen sales increased to 47.6%, up 48.7% year-over-year. The Ryzen 5 1600 processor is priced at $220.

In August, Ryzen surpassed Intel in CPU sales at Mindfactory, accounting for 56.1% of the retailer's overall CPU sales. This was due to the launch of AMD's high-end Ryzen Threadripper processors.

The chart above shows that in August 2017, the Ryzen Threadripper 1950X was priced at $999, outperforming Intel's i9 7900X.

This data is only available for one retailer. However, North American storefronts like Newegg and Amazon are showing similar trends, indicating that the market is increasingly adopting Ryzen.

Overall group sales revenue and average sales revenue increased 51% year-on-year, led by AMD's CG division. Another revenue driver was high graphics card demand from cryptocurrency miners.

In the second quarter, the CG division reported an operating profit for the first time in nearly four years. However, operating expenses also increased as the company continued to launch new products. The company is in its Ryzen. It plans to launch a mobile version of its Ryzen CPU for consumer applications in the second half of the year and commercial applications in the first quarter.

Next, we'll compare the performance of AMD's Ryzen to Intel's Core processors to see how the latter is responding to the growing competition.

AMD launched its first desktop Ryzen processor in March 2017. In the past six months, it has launched its entire range of Ryzen 7, 5, and 3 desktop processors for the high, mid, and low-end markets. Of all these processors, the Ryzen 5 1600X is the most popular offering. PCGamesN reviewed the Ryzen 5 1600X and found that the CPU offers value.

The site overclocked the 1600X and improved memory support, finding that the CPU offered a 13 percent performance boost at no extra cost. While the performance gain isn't significant, it offers value at a $220 price point.

So far, Intel has squeezed profits out of the declining PC market by segmenting its products and raising prices. This strategy has given it a monopoly in the market. However, increasing competition from AMD will reduce Intel's pricing power.

AMD's Ryzen poses a threat to Intel as the latter earns nearly 80% of its operating income from CCG (Client Computing Group). Competition from AMD will reduce Intel's revenue and operating profit.

Technology gurus debate whether AMD's Ryzen products are superior to Intel's products because the former offer better power efficiency and similar performance at a lower price. In fact, Ryzen's products outperform Intel's products in multi-threaded tasks.

AMD's Ryzen 5 1600X offers 6 cores and 12 threads to the mainstream market for the first time. In response to the growing competition, Intel will launch the 8th generation Coffee Lake Core i7-8700K processor. Intel claims that the new processor will have 6 cores and 12 threads and surpass the Ryzen 5 1600X. However, some leaks do not support Intel's claims.

According to an ArsTechnica article, Intel's i7-8700K Coffee Lake processor outperforms its predecessor, the four-core/eight-thread Core i7-7700K, by 30 percent in the multi-threaded benchmark Cinebench R15. The i7-8700K is competitive with AMD's Ryzen 1600X, but is inferior to the eight-core/12-thread Ryzen 7 1700 and 1800X.

Even if Intel's "Coffee Lake" isn't a dramatic improvement over Ryzen, the latter will benefit greatly. That's because Intel has launched the first Coffee Lake chips for laptops, while AMD's Ryzen Mobile CPUs won't be available until the fourth quarter.

Several PC OEMs, such as Lenovo and HP Inc., will feature more than 145 laptop designs with Coffee Lake processors, with the first devices launching in September 2017. Coffee Lake could benefit from holiday sales, while AMD would miss out.

Next, we look at Ryzen Threadripper.

AMD 's high-end Ryzen Threadripper CPU has been well received by the market, and consumers prefer Threadripper to Intel's i9 processor.

AMD's Threadripper lineup ranges in price from $549 to $999, with the latter being the most popular among customers, Devinder Kumar said.

TechRadar lists the eight-core/16-thread Ryzen Threadripper 1900X at $549, versus the eight-core/16-thread Ryzen 7 1800X at $499. When it comes to gaming, both CPUs deliver similar frames per second. In fact, the 1800X performs better in some cases.

The 1900X, on the other hand, offers better multitasking performance as it offers quad-channel DDR4 2667 memory support compared to the Ryzen 7's two channels, as well as 64 PCIe lanes.

Ryzen Threadripper is comparable to AMD's EPYC server processors as they share the same architecture. This was revealed by popular overclocker der8auer, who acquired an engineering sample of Threadripper in August 2017 and discovered that the CPU has four dies, just like the EPYC processors.

In September 2017, der8auer acquired a retail Threadripper 1950X processor and discovered two dummy dies with integrated circuits and transistors. This discovery led to speculation that there might be a way to activate the dummy dies. The likelihood that this speculation is true is reduced because AMD will not go through the pain of attaching fully functional dies to Threadripper for free.

In response, it was previously explained that the two non-functional dies "have no path to operation". Therefore, the company uses the term virtual die instead of inactive die. However, one thing is clear - Threadripper uses the same design as EPYC.

AMD announced its enterprise-class Ryzen Pro desktop CPUs at IBC 2017. The Pro series is similar to the Ryzen 7, 5, and 3 series, but is targeted at workstations. The Ryzen 7 Pro offers 8 cores and 16 threads, twice as many as an Intel Core i7.

With clock speeds up to 3.7 GHz, the Ryzen 7 Pro is particularly suited for power-hungry enterprise users and content creation. In Adobe Premier Pro video production, the Ryzen 7 Pro is 30% faster than the Core i7.

The Ryzen 7 Pro's 16 threads make it ideal for multitasking, but regular productivity software may not be able to take advantage of its 16 threads.

In 2017, AMD launched Zen CPUs and Vega GPUs on the 14nm node, marking its process transition from being built on the 28nm node.

While the Ryzen desktop CPUs were well received by customers, the Vega GPUs faced manufacturing issues due to HBM2 and were unable to deliver significant performance gains over NVIDIA's offerings.

AMD will launch Ryzen Mobile APUs that integrate the Ryzen CPU and Vega GPU on a single chip. These new APUs are based on the Raven Ridge architecture. The Raven Ridge APUs will first appear in low-end and mid-range laptops before the end of 2017. This means that the mobile versions of Ryzen 5 and 3 will launch before the mobile version of Ryzen 7.

AMD claims that the Raven Ridge APU will provide up to 50% more CPU performance and 40% more GPU performance, while consuming only 50% of the power, compared to the previous Bristol Ridge APU. The Raven Ridge APU was tested on Geekbench, and the results show that AMD's claims are true.

So far, AMD has targeted the PC market with its full range of Ryzen CPUs and Vega GPUs. The company is now looking to expand its CPUs and GPUs in the data center market.

With the launch of its EPYC server processors and Radeon Instinct GPUs, AMD is betting big on the data center market, where it has almost no share.

Devinder Kumar said the company is targeting the data center market because of the profit margins it offers. It started increasing production of EPYC processors in the second quarter, but the process takes time. Production in the server space is growing slower than in the PC space.

Kumar discussed the company's data center product strategy. By bringing more memory and more IO (input/output) to the EPYC processors, AMD is going after "a large portion of the data center market." The company hopes to reduce the total cost of ownership (TCO) of the data center by providing high performance, efficiency, security, and flexibility.

AMD expects enterprises to be the early adopters of EPYC processors, followed by hyperscale data centers and cloud computing companies, placing large orders for server processors.

Benchmark reports from third-party benchmarking site ServeTheHome show that AMD's EPYC 7351P single-socket CPU is in fierce competition with Intel's Xeon Silver CPU. Five hyperscale cloud computing companies, including Google, Baidu, and Amazon, have shown interest in EPYC processors.

At its 2017 investor day, AMD said its server CPUs represent a $16 billion opportunity, while GPUs represent another $5 billion. At the Deutsche Bank Technology Conference, Kumar said AMD aims to capture 10% of the server processor market share in terms of units. The server processor market is currently owned by Intel, which has a market share of more than 95%.

Jefferies analyst Mark Lipacis expects AMD to gain 2% of the server processor market in 2017, growing to 8% in 2018 as companies such as Microsoft, Baidu and Tencent announce plans to use AMD chips for cloud computing.

AMD is simultaneously launching its CPU and GPU products into data centers. The company began shipping its Radeon Instinct GPUs in the second quarter. The Instinct GPU portfolio includes the Vega-based MI25, the Fiji-based MI8, and the Polaris-based MI6 that supports heterogeneous computing.

AMD has invested in open source software ROCm (Radeon Open Compute), which supports MIOpen 1.0, an open source GPU acceleration library. Radeon GPUs also support popular open source machine learning frameworks such as Facebook Caffe, Google TensorFlow, and Torch.

In August 2017, AMD partnered with Baidu to provide high-performance GPUs and CPUs to handle next-generation data center workloads. NVIDIA partnered with Baidu in July 2017, but this partnership goes beyond cloud computing to include autonomous driving, open source deep learning optimization, and virtual assistants on the NVIDIA Shield TV.

According to an article in Forbes magazine, Chinese companies are the second largest investors in AI after American companies. In 2016, Baidu spent a whopping $30 billion on AI projects.

China is expected to increase spending in several AI projects such as smart home appliances, smart cars, drones, robots, and wearables. Partnering with Baidu will help AMD penetrate China and benefit from AI investments.

AMD has also received a design win from Amazon Web Services. AWS will use the AMD FirePro S7150x2 server GPU to power its new Appstream 2.0 graphics design instances. The graphics design instances allow users to stream graphics applications in the cloud. Amazon is using AMD Radeon Pro MxGPUs (multi-user GPUs) to allow up to 16 users to run graphics software in the cloud simultaneously.

With this service, Amazon is enabling the performance of a cloud-based design workstation for $2.40 per hour on 16 vGPU graphic design instances. This service will enable engineering companies with limited budgets to access high-end workstations without investing in hardware.

As more and more clouds adopt this service, it will drive the demand for more server GPUs among cloud customers, and CIC (cloud computing) presents strong growth potential in the market. According to Jon Peddie Research , the CiC market is in its early stages, with 23% of the involved industries implementing or planning to implement CiC in their workflows.

The above products are built on the 14nm node. AMD now plans to transition to the 12nm node to remain competitive. We will explore this in the next section.

AMD has returned to the competitive landscape with the launch of its 14nm Ryzen CPUs. Prior to Ryzen, AMD had a huge technological disadvantage as it was stuck with the outdated 28nm node while Intel transitioned to 14nm in 2014.

In September 2017, AMD's foundry partner GlobalFoundries announced its 12nm LP (low power) node, which will begin production in 2018, and AMD announced that it would build Ryzen CPUs and Vega GPUs on the 12nm node. The company did not mention EPYC, which leaves a gray area if server processors can also get a process upgrade.

NVIDIA is ahead of AMD as the former is already using TSMC 12nm node to build its Volta GPU. Intel will also surpass AMD in 2018 as it starts production on its 10nm node. This forces AMD to transition to the 12nm node to stay competitive.

GlobalFoundries said 12nm is not an actual node, but just an iterative improvement on the 14nm node. The company claims that the 12nm node will increase transistor density by 15% and performance by 10% compared to the 16nm node.

GlobalFoundries is developing 12nm in an effort to deliver the latest performance enhancements before moving to the 7nm node. The company expects 7nm to deliver a 40% performance increase and 60% power reduction over 14nm.

At the Canaccord Genuity Growth Conference, AMD Chief Technology Officer Mark Papermaster said the company will continue to adhere to Moore's Law even if performance and costs are below historical levels.

However, the company will not rely solely on Moore's Law for performance improvements. It will use its multi-chip module approach, in which various combinations of CPUs, GPUs, memory, and other accelerators will be used to improve performance. It will also use software to optimize the chips for end-use applications. AMD calls this strategy Moore's Law .

AMD 's new product launches have made the stock one of the most highly traded semiconductor stocks in 2017. Wall Street analysts have a "hold" recommendation on AMD, even as they remain optimistic about the company's new products.

The stock has already priced in the estimated revenue from new products. On the other hand, AMD's rivals Intel and NVIDIA also have "buy" recommendations.

Wells Fargo analyst Weiwei Wang set an $18 price target for AMD as he believes the company’s new chips have a competitive advantage. He concluded:

• AMD's Vega GPUs are comparable to NVIDIA's GPU offerings in the $379-$599 range.

• AMD's EPYC server processors offer comparable performance to Intel's latest Skylake server processors in the two-way server space, hitting a $4,000 price point.

• Ryzen Threadripper performs well against Intel's i9 desktop chips, and in some cases beats the latter, at a comparably priced price.

Susquehanna analyst Christopher Rolland set a $15 price target for AMD. Susquehanna's PC-SIGnals report tracks semiconductor industry trends by analyzing data from multiple desktops, notebooks, GPUs, and SSDs. According to reports, AMD Vega's GPU average selling price increased by 25% sequentially, increasing the company's GPU market share by 4% in the third quarter of 2017.

The report also expects AMD's CPU market share to rise in the third quarter. Rolland expects Ryzen to increase AMD's CPU average selling price by as much as 70%. The report expects ODM notebook sales to grow 7.2% sequentially in the third quarter of 2017, up from 2.6% previously. High ODM sales will benefit both AMD and Intel.

Jefferies analyst Mark Lipacis is bullish on AMD and has set a $19 price target on the stock. He expects GPU demand for stealth miners to remain strong on July 17, a trend that could generate long-term growth.

Lipacis is also optimistic about EPYC server processors. He pointed to third-party benchmarks of EPYC from ServeTheHome. The site said AMD's "mainstream offerings are competing well against the emerging [Intel] Xeon Silver CPUs."

A recent report from Digitimes states that Intel has delayed the launch of its 10nm Cannon Lake laptop processors from mid-2018 to late 2018, which Lipacis said is good news for AMD as it will have a greater chance of selling Ryzen mobile processors.

Overall, AMD's slew of new product launches could put the company back on track for growth.

Source | MarketRealist

Reply " report " to view all semiconductor industry research reports

Highlights of previous issues ( click on the blue words to read directly ) :

It seems that Apple is determined to break free from the constraints of the two major chip giants!

RF-SOI is far ahead, and the battle for RF front-end technology is over!

How to pretend to know a lot about Industrial Internet of Things?