https://seekingalpha.com/article/4152240-amds-growing-cpu-advantage-intel?page=1 submitted by
AMD's Growing CPU Advantage Over Intel Mar. 1.18 | About: Advanced Micro (AMD)
Raymond Caron, Ph.D. Tech, solar, natural resources, energy (315 followers) Summary AMD's past and economic hazards. AMD's Current market conditions. AMD Zen CPU advantage over Intel. AMD is primarily a CPU fabrication company with much experience and a great history in that respect. They hold patents for 64-bit processing, as well as ARM based processing patents, and GPU architecture patents. AMD built a name for itself in the mid-to-late 90’s when they introduced the K-series CPU’s to good reviews followed by the Athlon series in ‘99. AMD was profitable, they bought the companies NexGen, Alchemy Semiconductor, and ATI. Past Economic Hazards If AMD has such a great history, then what happened? Before I go over the technical advantage that AMD has over Intel, it’s worth looking to see how AMD failed in the past, and to see if those hazards still present a risk to AMD. As for investment purposes we’re more interested in AMD’s turning a profit. AMD suffered from intermittent CPU fabrication problems, and was also the victim of sustained anti-competitive behaviour from Intel who interfered with AMD’s attempts to sell its CPU’s to the market through Sony, Hitachi, Toshiba, Fujitsu, NEC, Dell, Gateway, HP, Acer, and Lenovo. Intel was investigated and/or fined by multiple countries including Japan, Korea, USA, and EU. These hazard needs to be examined to see if history will repeat itself. There have been some rather large changes in the market since then.
1) The EU has shown they are not averse to leveling large fines, and Intel is still fighting the guilty verdict from the last EU fine levied against them; they’ve already lost one appeal. It’s conceivable to expect that the EU, and other countries, would prosecute Intel again. This is compounded by the recent security problems with Intel CPU’s and the fact that Intel sold these CPU’s under false advertising as secure when Intel knew they were not. Here are some of the largest fines dished out by the EU
2) The Internet has evolved from Web 1.0 to 2.0. Consumers are increasing their online presence each year. This reduces the clout that Intel can wield over the market as AMD can more easily sell to consumers through smaller Internet based companies.
3) Traditional distributors (HP, Dell, Lenovo, etc.) are struggling. All of these companies have had recent issues with declining revenue due to Internet competition, and ARM competition. These companies are struggling for sales and this reduces the clout that Intel has over them, as Intel is no longer able to ensure their future. It no longer pays to be in the club. These points are summarized in the graph below, from Statista, which shows “ODM Direct” sales and “other sales” increasing their market share from 2009 to Q3 2017. 4) AMD spun off Global Foundries as a separate company. AMD has a fabrication agreement with Global Foundries, but is also free to fabricate at another foundry such as TSMC, where AMD has recently announced they will be printing Vega at 7nm.
5) Global Foundries developed the capability to fabricate at 16nm, 14nm, and 12nm alongside Samsung, and IBM, and bought the process from IBM to fabricate at 7nm. These three companies have been cooperating to develop new fabrication nodes.
6) The computer market has grown much larger since the mid-90’s – 2006 when AMD last had a significant tangible advantage over Intel, as computer sales rose steadily until 2011 before starting a slow decline, see Statista graph below. The decline corresponds directly to the loss of competition in the marketplace between AMD and Intel, when AMD released the Bulldozer CPU in 2011. Tablets also became available starting in 2010 and contributed to the fall in computer sales which started falling in 2012. It’s important to note that computer shipments did not fall in 2017, they remained static, and AMD’s GPU market share rose in Q4 2017 at the expense of Nvidia and Intel.
7) In terms of fabrication, AMD has access to 7nm on Global Foundries as well as through TSMC. It’s unlikely that AMD will experience CPU fabrication problems in the future. This is something of a reversal of fortunes as Intel is now experiencing issues with its 10nm fabrication facilities which are behind schedule by more than 2 years, and maybe longer. It would be costly for Intel to use another foundry to print their CPU’s due to the overhead that their current foundries have on their bottom line. If Intel is unable to get the 10nm process working, they’re going to have difficulty competing with AMD. AMD: Current market conditions In 2011 AMD released its Bulldozer line of CPU’s to poor reviews and was relegated to selling on the discount market where sales margins are low. Since that time AMD’s profits have been largely determined by the performance of its GPU and Semi-Custom business. Analysts have become accustomed to looking at AMD’s revenue from a GPU perspective, which isn’t currently being seen in a positive light due to the relation between AMD GPU’s and cryptocurrency mining.
The market views cryptocurrency as further risk to AMD. When Bitcoin was introduced it was also mined with GPU’s. When the currency switched to ASIC circuits (a basic inexpensive and simple circuit) for increased profitability (ASIC’s are cheaper because they’re simple), the GPU’s purchased for mining were resold on the market and ended up competing with and hurting new AMD GPU sales. There is also perceived risk to AMD from Nvidia which has favorable reviews for its Pascal GPU offerings. While AMD has been selling GPU’s they haven’t increased GPU supply due to cryptocurrency demand, while Nvidia has. This resulted in a very high cost for AMD GPU’s relative to Nvidia’s. There are strategic reasons for AMD’s current position:
1) While the AMD GPU’s are profitable and greatly desired for cryptocurrency mining, AMD’s market access is through 3rd party resellers whom enjoy the revenue from marked-up GPU sales. AMD most likely makes lower margins on GPU sales relative to the Zen CPU sales due to higher fabrication costs associated with the fabrication of larger size dies and the corresponding lower yield. For reference I’ve included the size of AMD’s and Nvidia’s GPU’s as well as AMD’s Ryzen CPU and Intel’s Coffee lake 8th generation CPU. This suggests that if AMD had to pick and choose between products, they’d focus on Zen due higher yield and revenue from sales and an increase in margin.
2) If AMD maintained historical levels of GPU production in the face of cryptocurrency demand, while increasing production for Zen products, they would maximize potential income for highest margin products (EPYC), while reducing future vulnerability to second-hand GPU sales being resold on the market. 3) AMD was burned in the past from second hand GPU’s and want to avoid repeating that experience. AMD stated several times that the cryptocurrency boom was not factored into forward looking statements, meaning they haven’t produced more GPU’s to expect more GPU sales.
In contrast, Nvidia increased its production of GPU’s due to cryptocurrency demand, as AMD did in the past. Since their Pascal GPU has entered its 2nd year on the market and is capable of running video games for years to come (1080p and 4k gaming), Nvidia will be entering a position where they will be competing directly with older GPU’s used for mining, that are as capable as the cards Nvidia is currently selling. Second-hand GPU’s from mining are known to function very well, with only a need to replace the fan. This is because semiconductors work best in a steady state, as opposed to being turned on and off, so it will endure less wear when used 24/7.
The market is also pessimistic regarding AMD’s P/E ratio. The market is accustomed to evaluating stocks using the P/E ratio. This statistical test is not actually accurate in evaluating new companies, or companies going into or coming out of bankruptcy. It is more accurate in evaluating companies that have a consistent business operating trend over time.
“Similarly, a company with very low earnings now may command a very high P/E ratio even though it isn’t necessarily overvalued. The company may have just IPO’d and growth expectations are very high, or expectations remain high since the company dominates the technology in its space.” P/E Ratio: Problems With The P/E I regard the pessimism surrounding AMD stock due to GPU’s and past history as a positive trait, because the threat is minor. While AMD is experiencing competitive problems with its GPU’s in gaming AMD holds an advantage in Blockchain processing which stands to be a larger and more lucrative market. I also believe that AMD’s progress with Zen, particularly with EPYC and the recent Meltdown related security and performance issues with all Intel CPU offerings far outweigh any GPU turbulence. This turns the pessimism surrounding AMD regarding its GPU’s into a stock benefit. 1) A pessimistic group prevents the stock from becoming a bubble. -It provides a counter argument against hype relating to product launches that are not proven by earnings. Which is unfortunately a historical trend for AMD as they have had difficulty selling server CPU’s, and consumer CPU’s in the past due to market interference by Intel. 2) It creates predictable daily, weekly, monthly, quarterly fluctuations in the stock price that can be used, to generate income. 3) Due to recent product launches and market conditions (Zen architecture advantage, 12nm node launching, Meltdown performance flaw affecting all Intel CPU’s, Intel’s problems with 10nm) and the fact that AMD is once again selling a competitive product, AMD is making more money each quarter. Therefore the base price of AMD’s stock will rise with earnings, as we’re seeing. This is also a form of investment security, where perceived losses are returned over time, due to a stock that is in a long-term upward trajectory due to new products reaching a responsive market.
4) AMD remains a cheap stock. While it’s volatile it’s stuck in a long-term upward trend due to market conditions and new product launches. An investor can buy more stock (with a limited budget) to maximize earnings. This is advantage also means that the stock is more easily manipulated, as seen during the Q3 2017 ER.
5) The pessimism is unfounded. The cryptocurrency craze hasn’t died, it increased – fell – and recovered. The second hand market did not see an influx of mining GPU’s as mining remains profitable.
6) Blockchain is an emerging market, that will eclipse the gaming market in size due to the wide breath of applications across various industries. Vega is a highly desired product for Blockchain applications as AMD has retained a processing and performance advantage over Nvidia. There are more and rapidly growing applications for Blockchain every day, all (or most) of which will require GPU’s. For instance Microsoft, The Golem supercomputer, IBM, HP, Oracle, Red Hat, and others. Long-term upwards trend AMD is at the beginning of a long-term upward trend supported by a comprehensive and competitive product portfolio that is still being delivered to the market, AMD referred to this as product ramping. AMD’s most effective products with Zen is EPYC, and the Raven Ridge APU. EPYC entered the market in mid-December and was completely sold out by mid-January, but has since been restocked. Intel remains uncompetitive in that industry as their CPU offerings are retarded by a 40% performance flaw due to Meltdown patches. Server CPU sales command the highest margins for both Intel and AMD.
The AMD Raven Ridge APU was recently released to excellent reviews. The APU is significant due to high GPU prices driven buy cryptocurrency, and the fact that the APU is a CPU/GPU hybrid which has the performance to play games available today at 1080p. The APU also supports the Vulcan API, which can call upon multiple GPU’s to increase performance, so a system can be upgraded with an AMD or Nvidia GPU that supports Vulcan API at a later date for increased performance for those games or workloads that been programmed to support it. Or the APU can be replaced when the prices of GPU’s fall.
AMD also stands to benefit as Intel confirmed that their new 10 nm fabrication node is behind in technical capability relative to the Samsung, TSMC, and Global Foundries 7 nm fabrication process. This brings into questions Intel’s competitiveness in 2019 and beyond. Take-Away • AMD was uncompetitive with respect to CPU’s from 2011 to 2017 • When AMD was competitive, from 1996 to 2011 they did record profit and bought 3 companies including ATI. • AMD CPU business suffered from: • Market manipulation from Intel. • Intel fined by EU, Japan, Korea, and settled with the USA • Foundry productivity and upgrade complications • AMD has changed • Global Foundries spun off as an independent business • Has developed 14nm &12nm, and is implementing 7nm fabrication • Intel late on 10nm, is less competitive than 7nm node • AMD to fabricate products using multiple foundries (TSMC, Global Foundries) • The market has changed • More AMD products are available on the Internet and both the adoption of the Internet and the size of the Internet retail market has exploded, thanks to the success of smartphones and tablets. • Consumer habits have changed, more people shop online each year. Traditional retailers have lost market share. • Computer market is larger (on-average), but has been declining. While Computer shipments declined in Q2 and Q3 2017, AMD sold more CPU’s. • AMD was uncompetitive with respect to CPU’s from 2011 to 2017. • Analysts look to GPU and Semi-Custom sales for revenue. • Cryptocurrency boom intensified, no crash occurred. • AMD did not increase GPU production to meet cryptocurrency demand. • Blockchain represents a new growth potential for AMD GPU’s. • Pessimism acts as security against a stock bubble & corresponding bust. • Creates cyclical volatility in the stock that can be used to generate profit. • P/E ratio is misleading when used to evaluate AMD. • AMD has long-term growth potential. • 2017 AMD releases competitive product portfolio. • Since Zen was released in March 2017 AMD has beat ER expectations. • AMD returns to profitability in 2017. • AMD taking measureable market share from Intel in OEM CPU Desktop and in CPU market. • High margin server product EPYC released in December 2017 before worst ever CPU security bug found in Intel CPU’s that are hit with detrimental 40% performance patch. • Ryzen APU (Raven Ridge) announced in February 2018, to meet gaming GPU shortage created by high GPU demand for cryptocurrency mining. • Blockchain is a long-term growth opportunity for AMD. • Intel is behind the competition for the next CPU fabrication node. AMD’s growing CPU advantage over Intel About AMD’s Zen Zen is a technical breakthrough in CPU architecture because it’s a modular design and because it is a small CPU while providing similar or better performance than the Intel competition.
Since Zen was released in March 2017, we’ve seen AMD go from 18% CPU market share in the OEM consumer desktops to essentially 50% market share, this was also supported by comments from Lisa Su during the Q3 2017 ER call, by MindFactory.de, and by Amazon sales of CPU’s. We also saw AMD increase its market share of total desktop CPU’s. We also started seeing market share flux between AMD and Intel as new CPU’s are released. Zen is a technical breakthrough supported by a few general guidelines relating to electronics. This provides AMD with an across the board CPU market advantage over Intel for every CPU market addressed.
1) The larger the CPU the lower the yield. - Zen architecture that makes up Ryzen, Threadripper, and EPYC is smaller (44 mm2 compared to 151 mm2 for Coffee Lake). A larger CPU means fewer CPU’s made during fabrication per wafer. AMD will have roughly 3x the fabrication yield for each Zen printed compared to each Coffee Lake printed, therefore each CPU has a much lower cost of manufacturing.
2) The larger the CPU the harder it is to fabricate without errors. - The chance that a CPU will be perfectly fabricated falls exponentially with increasing surface area. Intel will have fewer high quality CPU’s printed compared to AMD. This means that AMD will make a higher margin on each CPU sold. AMD’s supply of perfect printed Ryzen’s (1800X) are so high that the company had to give them away at a reduced cost in order to meet supply demands for the cheaper Ryzen 5 1600X. If you bought a 1600X in August/September, you probably ended up with an 1800X.
3) Larger CPU’s are harder to fabricate without errors on smaller nodes. -The technical capability to fabricate CPU’s at smaller nodes becomes more difficult due to the higher precision that is required to fabricate at a smaller node, and due to the corresponding increase in errors. “A second reason for the slowdown is that it’s simply getting harder to design, inspect and test chips at advanced nodes. Physical effects such as heat, electrostatic discharge and electromagnetic interference are more pronounced at 7nm than at 28nm. It also takes more power to drive signals through skinny wires, and circuits are more sensitive to test and inspection, as well as to thermal migration across a chip. All of that needs to be accounted for and simulated using multi-physics simulation, emulation and prototyping.“ Is 7nm The Last Major Node? “Simply put, the first generation of 10nm requires small processors to ensure high yields. Intel seems to be putting the smaller die sizes (i.e. anything under 15W for a laptop) into the 10nm Cannon Lake bucket, while the larger 35W+ chips will be on 14++ Coffee Lake, a tried and tested sub-node for larger CPUs. While the desktop sits on 14++ for a bit longer, it gives time for Intel to further develop their 10nm fabrication abilities, leading to their 10+ process for larger chips by working their other large chip segments (FPGA, MIC) first.” There are plenty of steps where errors can be created within a fabricated CPU. This is most likely the culprit behind Intel’s inability to launch its 10nm fabrication process. They’re simply unable to print such a large CPU on such a small node with high enough yields to make the process competitive. Intel thought they were ahead of the competition with respect to printing large CPU’s on a small node, until AMD avoided the issue completely by designing a smaller modular CPU. Intel avoided any mention of its 10nm node during its Q4 2017 ER, which I interpret as bad news for Intel shareholders. If you have nothing good to say, then you don’t say anything. Intel having nothing to say about something that is fundamentally critical to its success as a company can’t be good. Intel is on track however to deliver hybrid CPU’s where some small components are printed on 10nm. It’s recently also come to light that Intel’s 10nm node is less competitive than the Global Foundries, Samsung, and TSMC 7nm nodes, which means that Intel is now firmly behind in CPU fabrication. 4) AMD Zen is a new architecture built from the ground up. Intel’s CPU’s are built on-top of older architecture developed with 30-yr old strategies, some of which we’ve recently discovered are flawed. This resulted in the Meltdown flaw, the Spectre flaws, and also includes the ME, and AMT bugs in Intel CPU’s. While AMD is still affected by Spectre, AMD has only ever acknowledged that they’re completely susceptible to Spectre 1, as AMD considers Spectre 2 to be difficult to exploit on an AMD Zen CPU. “It is much more difficult on all AMD CPUs, because BTB entries are not aliased - the attacker must know (and be able to execute arbitrary code at) the exact address of the targeted branch instruction.” Technical Analysis of Spectre & Meltdown * Amd
Further reading Spectre and Meltdown: Linux creator Linus Torvalds criticises Intel's 'garbage' patches | ZDNet FYI: Processor bugs are everywhere - just ask Intel and AMD Meltdown and Spectre: Good news for AMD users, (more) bad news for Intel Cybersecurity agency: The only sure defense against huge chip flaw is a new chip Kernel-memory-leaking Intel processor design flaw forces Linux, Windows redesign Take-Away • AMD Zen enjoys a CPU fabrication yield advantage over Intel • AMD Zen enjoys higher yield of high quality CPU’s • Intel’s CPU’s are affected with 40% performance drop due to Meltdown flaw that affect server CPU sales.
AMD stock drivers 1) EPYC • -A critically acclaimed CPU that is sold at a discount compared to Intel. • -Is not affected by 40% software slow-downs due to Meltdown. 2) Raven Ridge desktop APU • - Targets unfed GPU market which has been stifled due to cryptocurrency demand - Customers can upgrade to a new CPU or add a GPU at a later date without changing the motherboard. • - AM4 motherboard supported until 2020. 3) Vega GPU sales to Intel for 8th generation CPU’s with integrated graphics. • - AMD gains access to the complete desktop and mobile market through Intel.
4) Mobile Ryzen APU sales • -Providing gaming capability in a compact power envelope.
5) Ryzen and Threadripper sales • -Fabricated on 12nm in April. • -May eliminate Intel’s last remaining CPU advantage in IPC single core processing. • -AM4 motherboard supported until 2020. • -7nm Ryzen on track for early 2019. 6) Others: Vega, Polaris, Semi-custom, etc. • -I consider any positive developments here to be gravy. Conclusion While in the past Intel interfered with AMD's ability to bring it's products to market, the market has changed. The internet has grown significantly and is now a large market that dominates when in computer sales. It's questionable if Intel still has the influence to affect this new market, and doing so would most certainly result in fines and further bad press.
AMD's foundry problems were turned into an advantage over Intel.
AMD's more recent past was heavily influenced by the failure of the Bulldozer line of CPU's that dragged on AMD's bottom line from 2011 to 2017.
AMD's Zen line of CPU's is a breakthrough that exploits an alternative, superior strategy, in chip design which results in a smaller CPU. A smaller CPU enjoys compounded yield and quality advantages over Intel's CPU architecture. Intel's lead in CPU performance will at the very least be challenged and will more likely come to an end in 2018, until they release a redesigned CPU.
I previously targeted AMD to be worth $20 by the end of Q4 2017 ER. This was based on the speed that Intel was able to get products to market, in comparison AMD is much slower. I believe the stock should be there, but the GPU related story was prominent due to cryptocurrency craze. Financial analysts need more time to catch on to what’s happening with AMD, they need an ER that is driven by CPU sales. I believe that the Q1 2018 is the ER to do that. AMD had EPYC stock in stores when the Meltdown and Spectre flaws hit the news. These CPU’s were sold out by mid-January and are large margin sales.
There are many variables at play within the market, however barring any disruptions I’d expect that AMD will be worth $20 at some point in 2018 due these market drivers. If AMD sold enough EPYC CPU’s due to Intel’s ongoing CPU security problems, then it may occur following the ER in Q1 2018. However, if anything is customary with AMD, it’s that these things always take longer than expected.
Better yet, your FPGA miner won’t turn into a paperweight if the algorithm for the cryptocurrency you are mining is updated or if you decide to start mining something else. And because UltraMiner is designed to be a standalone device, it does not require that you invest in a desk full of peripheral components. Its small form factor and flexible cooling options allow you to set it up and plug ... A selection of 28nm graphic cards and FPGA devices are analysed and used for comparison purposes. GPUs performance is derived from commercial graphic cards characteristics. Table 2 lists a selection of the best cards for the money as representative samples for the 2016’s state-of-the-art technology combining older models with newer flagship graphic cards. GPUs price ranges from less than 100 ... Einmal investieren, fortan Geld im Schlaf verdienen: Das Bitcoin-Prinzip klingt zu schön, um wahr zu sein. Wir haben einige der Goldgräber-Maschinen im c’t-Labor schürfen lassen und gezählt ... 21.10.2020 - Lattice Semiconductor Corporation (NASDAQ: LSCC), the low power programmable leader, announced the Lattice CrossLink-NX FPGA for embedded vision and AI applications won the Best FPGA ... The whole Bitcoin mining system using the two-level pipeline SHA256 Core on FPGA Virtex-7 VC707 reaches the frequency of 514.92 MHz, the hash rate of 514.92 MH/s, and the power consumption of 1.74 W. Table 4 is the comparison between our proposed mining system and the related systems.
Quick demo of the Butterfly Labs 5 GH/s miner. KnCMiner Jupiter - Bitcoin Miner 500GH/s+ 28nm ASIC chips - unboxing and setup 1080p - Duration: 4:18. Florian Uhlemann 131,038 views World's first 28nm Bitcoin Mining ASIC. Worlds'a fastest 28nm concept to silicon ever. 722 GH/s - HashFast's Golden Nonce has met it's match! See our GN 28nm ASIC on the new Yoli Evo Mining Board!!! The setup of my Avalon Bitcoin Miner. Setup and in action. The day is here! Nathan from Syn Shop, Las Vegas' local hackerspace, helps me disassemble BFL's fancy new ASIC miner. I see a bunch of chips and capacitors, he sees a lot more. Of course the one ...