Execution _老虎社区_美港股上老虎

2021-10-07

Execution is key

免责声明：上述内容仅代表发帖人个人观点，不构成本平台的任何投资建议。

精彩评论

我们需要你的真知灼见来填补这片空白

发表看法

{"i18n":{"language":"zh_CN"},"detailType":1,"isChannel":false,"data":{"magic":2,"id":823302744,"tweetId":"823302744","gmtCreate":1633578109458,"gmtModify":1633584394399,"author":{"id":4095222842422150,"idStr":"4095222842422150","authorId":4095222842422150,"authorIdStr":"4095222842422150","name":"aiyoh79","avatar":"https://static.tigerbbs.com/8b4d6ba5b13f7e724af3472b596ca6ac","vip":1,"userType":1,"introduction":"","boolIsFan":false,"boolIsHead":false,"crmLevel":2,"crmLevelSwitch":0,"individualDisplayBadges":[],"fanSize":5,"starInvestorFlag":false},"themes":[],"images":[],"coverImages":[],"extraTitle":"","html":"<html><head></head><body>Execution is key</body></html>","htmlText":"<html><head></head><body>Execution is key</body></html>","text":"Execution is key","highlighted":1,"essential":1,"paper":1,"likeSize":2,"commentSize":0,"repostSize":0,"favoriteSize":0,"link":"https://laohu8.com/post/823302744","repostId":1146744889,"repostType":4,"repost":{"id":"1146744889","kind":"news","pubTimestamp":1633577132,"share":"https://www.laohu8.com/m/news/1146744889?lang=&edition=full","pubTime":"2021-10-07 11:25","market":"us","language":"en","title":"Intel Vs. AMD: Competitive Prospects","url":"https://stock-news.laohu8.com/highlight/detail?id=1146744889","media":"Seeking Alpha","summary":"Summary\n\nThis is a (very) long look into the competitive positioning between Intel and AMD, on the C","content":"Summary\n<ul>\n <li>This is a (very) long look into the competitive positioning between Intel and AMD, on the CPU/SoC side, over the next 2+ years.</li>\n <li>It covers the desktop, laptop and server segments.</li>\n <li>There are changes happening, and more will be coming, to this competitive positioning.</li>\n</ul>\n<img src=\"https://static.tigerbbs.com/ad7be89aa0540e7ed8c51debe3aa0392\" tg-width=\"1536\" tg-height=\"1024\" width=\"100%\" height=\"auto\">Vedat OGUZCAN/iStock via Getty Images\nThis article will cover the present, and near-term (up to more than two years out), competitive prospects between Advanced Micro Devices(NASDAQ:AMD)and Intel(NASDAQ:INTC)on the CPU market. Namely, on its three main segments: desktop, laptop and server.\nThe article was built through the careful aggregation of information from dozens of sources, taking into account the most credible reports backed by the most recent benchmarks and statements from Intel and AMD.\nThe Context\nAMD has sprang a surprise. Since the first Zen 3-based chips reached the market, AMD has built up a significant advantage over Intel. This happened in the:\n<ul>\n <li>Desktop segment, where Zen 3 chips got to achieve equivalent single-threaded performance, while besting Intel on multithreaded performance due to much higher core counts.</li>\n <li>Laptop segment, though to a lesser extent, mostly on the strength of its multithreaded performance as well.</li>\n <li>Server, with its EPYC CPUs, on account of very competitive multithreaded abilities as a result of much higher core counts, as well as better performance / energy consumption metrics. Moreover, since Intel had been unchallenged in the server space for so long, AMD found it easy to price its CPUs at much better price / performance and TCO (Total Cost of Ownership) / performance ratios.</li>\n</ul>\nAMD achieved this incredible feat through three developments:\n<ul>\n <li>It finally built a competitive CPU architecture, enabling it to level the field in terms of IPC (Instructions Per Cycle).</li>\n <li>It got access to a process advantage due to using an external supplier, TSMC(NYSE:TSM). TSMC surpassed Intel's process abilities, and with Intel fabricating its chips internally, this created a semi-permanent disadvantage for Intel CPUs. This happened on account of many Intel missteps and delays advancing its process technology.</li>\n <li>And AMD's bet on a MCM (Multi-Chip Module) architecture, which enabled it to divide a monolithic CPU die (which is expensive to manufacture due to manufacturing yields being lower for larger dies) into many small, less expensive dies. This brought down the cost of the chip and allowed AMD to sell chips with many more cores than Intel's at similar price points.</li>\n</ul>\nIntel took a long time to respond to AMD’s challenge, giving AMD the possibility of catching and surpassing Intel.\nThings move slowly when it comes to designing chips or evolving chipmaking processes, so the initial reaction would always have to be slow. In the meantime, AMD has been slowly but surely eroding Intel’s market position, including on the very profitable server market. Any increase in AMD competitivity would lead to this, since Intel started from very high market shares on account of AMD’s prior near-absolute lack of competitiveness.\nThis is the starting point for my analysis today. At this point in time, AMD has been beating Intel like a redheaded stepchild across the market (by gaining market share), though slower on servers and laptops.\nServers are a conservative market and move slower, while on laptops Intel’s 11thgeneration chips still managed to keep AMD somewhat at bay due to a large GPU performance gain while keeping decent performance on the CPU side.\nAnyway, let’s check how things are set to evolve over the next 2-3 years.\nDesktopPresent\nIn terms of desktop CPUs, currently the battle is fought between Intel’s 11thgeneration and AMD’s Ryzen 5000 series. Intel has a small lead in single threaded performance, but AMD has a very large lead on multitasking performance, due to AMD fielding CPUs with a higher core number. For instance, at the top of the range, AMD’s Ryzen 5950x has 16 cores 32 threads) to Intel’s 11900K 8 (16 threads).\nIntel’s single threaded performance cannot compensate for the smaller number of cores, so presently AMD’s desktop offerings are more competitive.\nNext Month\nThe first shot on the renewed Intel/AMD war will be fired by Intel next month. That’s when Intel’s 12thgeneration, Alder Lake, will launch. This generation has one particularity: it will be the first time the x86 world will see the arrival of a big.little architecture.\nIn the case of Alder Lake, Intel will provide CPUs with up to 8 performance cores and another 8 efficiency cores. The performance cores have hyperthreading (two threads per core), the efficiency cores do not, hence, this 16-core CPU can run 24 threads, versus 32 on a top-of-the-line Ryzen. The top of the Intel Alder Lake line will be the i9-12900K CPU.\nThis chip will represent a huge step for Intel. From early tests, the 12900K is showing up to 2x higher performance in multitasking, when compared to the 11900K.\nThe jump in multitasking performance comes from two sources:\n<ul>\n <li>The small efficiency cores are actually quite performant. They perform&nbsp;above the level of an SkyLake core&nbsp;(thus, at slightly less than 1/2 the performance on the new Performance cores), but do so while demanding significantly less power. Hence, although they are efficiency cores, they can still provide significant computing capacity which impacts the overall multitasking ability of the chip. Roughly, about 33% of the chip’s multitasking power can come from these cores.</li>\n <li>Alder Lake also represents a large IPC jump, due to it using entirely new and much wider Performance core. This IPC gain is on the order of 19-20%, and a larger overall gain still might come from higher frequencies on top.</li>\n</ul>\nEfficiency Cores\nOne of the big surprises with Intel’s Alder Lake generation, is the performance the E-Cores (Efficiency Cores, Gracemont) are said to attain. These cores, schematically, seem to occupy 1/4ththe area taken by the new Golden Cove P-Cores (Performance Cores). Yet, in Intel’s&nbsp;Architecture Day 2021 presentation, these cores are able to attain levels of performance over those hit by SkyLake cores.\nNotice the relative sizes:\n<img src=\"https://static.tigerbbs.com/404ab04fb76822e023833b08316cccec\" tg-width=\"640\" tg-height=\"352\" width=\"100%\" height=\"auto\"><img src=\"https://static.tigerbbs.com/97305d665518f38620f00eedc89841f7\" tg-width=\"640\" tg-height=\"357\" width=\"100%\" height=\"auto\">\nAnd notice the power/performance charts vs SkyLake\n<img src=\"https://static.tigerbbs.com/4e65c15620b40f9cb4e5282a096bc8b4\" tg-width=\"640\" tg-height=\"348\" width=\"100%\" height=\"auto\">\nThe 80% performance jump doesn’t mean the core is 80% faster on single-threaded jobs. This is a multithreaded comparison, with four single-threaded efficiency cores vs twohyperthreadedSkyLake cores. For single-threaded performance:\n<img src=\"https://static.tigerbbs.com/714ffd1802677273d9b059fd270334c6\" tg-width=\"640\" tg-height=\"342\" width=\"100%\" height=\"auto\">\nPerformance Cores\nAt the same time, Intel’s big cores have made a large jump in performance. This jump is cumulative with what happened with Rocket Lake, and thus it has been enough to propel Intel well ahead of AMD’s current offerings when it comes to single threaded performance (~25%+).\nThe performance jump on the Performance cores, together with actually performant efficiency cores, also means that Alder Lake is able to match AMD's multithreaded performance even though it can only process 24 threads (16 threads on the 8 Performance cores, 8 threads on efficiency cores) vs AMD's 32 (on high-end chips).\nThis performance jump by Intel arguably will put Intel at an advantage in the desktop market next month, as soon as Alder Lake launches.\nQ1 2022\nIn Q1 2022, AMD’s 6000 series (Warhol) arrives. It will still be based on the Zen 3 architecture, and promises small (up to 5%) IPC gains, plus gains in other areas including frequency, aggregating to a~9% performance gain.\nThis will clearly fall short of what Alder Lake is showing now, though it can create a small advantage on multithreaded uses. The desktop core count will remain at 16 cores, hence there’s no large leapfrogging of Intel’s effort.\nAs a result, even with the AMD 6000 series shipping, Intel will likely retain the desktop performance crown.\nQ3 2022\nCome Q3 2022, Intel will likely release its 13thgeneration CPU family, Raptor Lake. This will be a refresh of Alder Lake, with minor IPC and memory improvements. The IPC and performance gains on this generation will likely mirror those attained by AMD on its AMD 6000 series.\nHowever, Intel will add another 8 efficiency cores to Raptor Lake on its high-end chip, upping the core count to 24 and the thread count to 32. What this means is:\n<ul>\n <li>Given the estimated contribution by the efficiency cores to Alder Lake’s multithreaded performance (~33%) and 8 additional efficiency cores (doubles their count), these additional cores can thus add a further 33% multithreaded improvement;</li>\n <li>And there will also be some single-threaded performance improvements.</li>\n</ul>\nThis seems enough to put the 13900K CPU ahead of the high-end Ryzen 6000 CPU even on multithreaded scenarios. Thus, Raptor Lake will solidify Intel’s position at the top of the desktop market.\nQ4 2022\nAMD is set to launch the Zen 4-based Ryzen 7000 series (Raphael) at this time, likely close to year-end.\nIt’s hard to estimate how large the performance uplift will be, though it seems&nbsp;up to 29% IPC&nbsp;and 40%-45% overall performance gains over the current generation (Zen 3) might be attainable. The core count for Desktop chips will remain at 16 cores (32 threads). Zen 4 will be built on a new TSMC process node.\nThese are very significant gains. If they’re realized, they are probably enough to put Zen 4 slightly ahead of Intel’s Raptor Lake, or match it.\nIn my view, even if an advantage is gained, it will likely be small (given the large jump achieved by Alder Lake, and then reinforced by Raptor Lake). Hence, at this point – Q4 2022 -- things will be too close to call on the Desktop market.\nQ2 2023\nIntel will launch Meteor Lake. This is an entirely new architecture on a new process node (Intel 4). Although there are no performance numbers at this point, a new architecture together with a new process node is bound to produce a large IPC gain. On top of this, Meteor Lake will also be a truly MCM design, allowing Intel to more easily ramp up the core count, which will favor multithreaded performance.\nHence, when Meteor Lake is launched, Intel is highly likely to put itself firmly ahead of Zen 4 CPUs in the desktop market. AMD’s answer will only come much later, in the form of Zen 5. At this point, it’s possible that Intel will have gained, and will keep, the performance leadership for a longer time.\nLaptopPresent\nCurrently, the laptop battle is fought between Intel’s Tiger Lake-H and -U families and AMD’s Ryzen 5000 H and U series.\nIntel and AMD are well matched in the H segment, with single-threaded and multithreaded performance broadly similar between the two.\nOn the U segment, Intel provides a lower core count at the top (4 vs 8), which gives AMD a performance lead on multithreaded tasks. However, Intel then has a large lead on single-threaded performance, as well as a large lead on GPU performance.\nGiven the above, I’ve considered that it’s too close to call this segment right now.\nQ1 2022\nDuring Q1 2023, Intel will launch new CPUs for the laptop market (H and U families) based on its very performant Alder Lake generation (as we saw from the desktop segment).\nThese CPUs will go up to 6+8 cores (14 cores, 20 threads, H family) and 2+8 cores (10 cores, 12 threads, U family), which will represent a large increase in core and thread counts from the present Tiger Lake generation (H goes from 8 to 14 cores, 16 to 20 threads; U goes from 4 to 10 cores, 8 to 12 threads). Simultaneously, the Performance cores will have a 20% IPC uplift (and likely a larger overall performance gain).\nHence, compute performance will be substantially increased, putting it well ahead of the existing AMD Ryzen 5000 series. However, the maximum EU (Execution Unit) count on the Alder Lake mobile CPUs remains capped at 96 EUs. Moreover, it seems these EUs will be based on the same GPU generation as the existing Tiger Lake GPUs. Hence, GPU performance will likely not increase a lot.\nAt about the same time, AMD will launch Ryzen 6000 APUs (Rembrandt), which will somewhat improve overall performance (~9%). But most importantly, AMD will mate these APUs with an entirely new GPU generation based on its RDNA 2 technology. This change will bring extreme GPU performance increases because the prior APUs were still based on AMD’s outdated and outclassed Vega technology. The performance increases will likely exceed 100%, and will put AMD’s mobile offerings well ahead of Intel’s when it comes to GPU performance.\nHence, as of Q1 2022 Intel will likely get the mobile CPU performance crown, but AMD will take away the integrated mobile GPU performance crown from Intel. As a result, I think the battle between these two products will be too close to call and will depend on each customer’s priorities.\nQ3 2022\nIntel’s mobile range will be refreshed with the launch of its 13thgeneration Raptor Lake CPUs. These will bring small CPU performance gains, which will build on the gains it attained with Alder Lake.\nHowever, the big news with the mobile Raptor Lake CPUs is that they’ll switch the GPU technology, going from Intel LP GPUs to Intel HPG GPUs. Moreover, the maximum number of EUs will be increased from 96 to 192 (doubled).\nThis is a significant change. Intel’s HPG EUs are supposed to bring a 50% performance uplift (on the same power budget), and the doubling of the maximum EUs will bring another 100% performance uplift. Conceivably, the maximum GPU performance on Raptor Cove high-end mobile CPUs might increase as much as 200%, though Intel might sacrifice some of the potential improvement for energy saving reasons.\n<img src=\"https://static.tigerbbs.com/799bea21a599b46badc1d99a507745ec\" tg-width=\"640\" tg-height=\"351\" width=\"100%\" height=\"auto\">\nWhen this is realized, Intel will attain performance superiority both on the CPU and GPU side in its mobile offerings. Hence, Intel will be at a clear advantage on the laptop market from Q3 2022 onwards.\nQ1 2023\nQ1 2023 marks the likely arrival of Zen 4-based laptop CPUs from AMD. These will likely enable AMD to regain the upper hand on CPU multitasking performance, and possibly level things on the single-threaded performance side. It remains to be seen what will happen at the GPU level.\nIn my view, like on the desktop side, AMD will regain the upper hand if the GPU side manages to surpass Intel’s big jump from using its HPG GPUs. Still, for me right now this segment will be too close to call once Zen 4 arrives.\nQ2 2023\nIn Q2 2023, Intel will likely start shipping its Meteor Lake 14thgeneration mobile CPUs. Meteor Lake is both a new architecture and it will be built on a new process node. Additionally, Meteor Lake will be Intel’s first truly MCM design of the modern era.\nAs a result of so many performance-enhancing developments, Meteor Lake is likely to bring significant performance gains (15-20% or more, depending on Intel choices), both on multitasking and single-threaded scenarios. Since Meteor Lake will arrive when things are more or less balanced between Intel and AMD, this will immediately push Intel towards a clear leadership in the laptop segment.\nServerPresent\nPresently, Intel’s Xeon Scalable 3rdGen tops out at 40 cores, while AMD’s EPYC Milan CPUs go all the way to 64 cores.\nThis large difference in number of cores means the Xeon is not very competitive in tasks involving maximum throughput. Cloud offerings often require high utilization rates, and thus the higher core count gives superiority to AMD and offers it a way to gain market share, which AMD has been taking advantage of.\nQ2 2022\nAround Q2 2022, Intel will launch its Sapphire Rapids Xeons. These will benefit from:\n<ul>\n <li>An early multi-die effort, allowing Intel to increase the core count to at least 56 cores. The CPU cores will be spread out in up to four dies, with 14 cores active in each (maybe more in the future).</li>\n <li>A very large jump in single-threaded performance, similar or larger than the one enjoyed by Alder Lake.</li>\n</ul>\nAt about the same time, though, AMD is likely to deploy its EPYC Milan-X server CPUs. The main change with these will be the use of 3D stacking to add extra cache on top of the execution units, making for much larger CPU caches. This will provide a performance improvement which will vary a lot across workloads (those more reliant on cache might benefit significantly, others not so much).\nWithout the launch of AMD’s EPYC Milan-X, it would seem that the single- and multi-threaded performance afforded by the use of improved Alder Lake-derived cores (Golden Cove), along with a better (not as punishing, performance-wise) multi-die approach, would allow Sapphire Rapids to surpass AMD’s EPYC Milan. This would happen with Sapphire Rapids matching or beating EPYC Milan’s multitasking abilities, while also enjoying a significant single-threaded performance advantage.\nHowever, with the launch of EPYC Milan-X, I consider the situation too close to call, on account of Sapphire Rapids’ lower core count, along with some improvement in Milan’s overall performance.\nStill, on the server side, Intel being on equal terms with AMD is actually an improvement right now.\nQ4 2022\nIt all changes in AMD’s favor come Q4 2022, though. That’s when AMD will launch its EPYC Genoa generation, based on Zen 4 cores.\nThis generation will bring a new architecture, be built on a new process node and enjoy the benefits from the 3D cache stacking used on EPYC Milan-X. It will also bump the maximum number of cores to 96. This will result in:\n<ul>\n <li>EPYC Genoa matching or beating Sapphire Rapids' single-threaded performance.</li>\n <li>EPYC Genoa greatly outclassing Sapphire Rapids when it comes to the all-important (for server loads) multithreaded performance (96 vs 56 cores).</li>\n</ul>\nOn account of this, AMD will again regain the upper hand in the server market.\nQ1 2023\nOn Q1 2023, Intel will likely field a refresh of the Sapphire Rapids generation, code named Emerald Rapids. This generation will bring the benefits from the Raptor Lake desktop refresh to the server side. It will also bump the maximum number of cores to 64 from 56.\nStill, the advantage gained by the Zen 4-based EPYC Genoa seems large enough that Emerald Rapids won’t be enough to counter it. Emerald Rapids might get close on single-threaded use cases, but it will still remain far away in terms of multithreaded use cases – the most important for server CPUs most often handling virtualized workloads. Basically, the “number of cores” advantage enjoyed by EPYC Genoa will be unsurmountable by this iterative Intel effort.\nAMD will keep its advantage in the server space, and will thus continue to chip away at Intel’s market share throughout this period.\nQ3 2023\nAt this point, on the server side, there will be a truly significant event. Intel will bring its Granite Rapids server CPU generation to the market.\n<img src=\"https://static.tigerbbs.com/f7c77d5b18968ad5fdc85e0aff296903\" tg-width=\"1280\" tg-height=\"502\" width=\"100%\" height=\"auto\">\nThis generation enjoys benefits similar to those enjoyed by Meteor Lake on the desktop side, but to an even larger extent. Namely, this generation is set to bump the number of cores to 120, spread across two 60-core dies. This is significant in several ways:\n<ul>\n <li>These cores, on a standalone basis, will likely beat Zen 4 cores by a wide single-threaded margin. This is so because at the starting point Zen 4 will be roughly matched to the previous Intel cores (Raptor Cove on Raptor Lake). Meteor Lake (and Granite Rapids), however, will be built with a new core architecture (Redwood Cove) and on a new process step. Thus, the performance gains are likely to be significant on a single-threaded basis.</li>\n <li>By putting 120 cores on a single server CPU, Intel will for the first time in a long while, be having more cores per CPU than AMD. With each core being more powerful, this already ensures a significantly higher multithreaded performance.</li>\n <li>But worse still, since the core dies are much larger on Intel’s Granite Rapids (two dies, 60 cores each) than on AMD’s EPYC Genoa (12 CCDs with 8 cores each), this will lead to another performance benefit for Granite Rapids: The latencies between its cores will be much lower, leading to a large relative performance gain (other things being equal).</li>\n</ul>\nFrom these effects, it seems a certainty that Intel will regain its performance lead, both on single-threaded and multithreaded use cases, on the server segment. And Intel's advantage will, at that point, be large. This is set to happen in Q3 2023, nearly two years from now.\nAfter Granite Rapids, Intel is also promising another very large jump in performance, but this will already fall outside the timeframe for this article. AMD will also bring Zen 5, but its arrival schedule is still very uncertain.\nA Charted Summary\nThe above translates into the following color-coded schematic for all the segments:\n<img src=\"https://static.tigerbbs.com/19775dfb6f19634f9d4f428ef9b27d13\" tg-width=\"640\" tg-height=\"264\" width=\"100%\" height=\"auto\">\nConclusion\nTaking into account the near-term roadmaps, benchmarks, leaks and guidance, it seems Intel is on the verge of taking away the performance crown from AMD. This will happen by Intel first regaining the single-threaded crown by a visible margin, while more or less matching AMD on multitasking, followed by a period where Intel will get a lead on both counts, interrupted only briefly by AMD's Zen 4.\nHowever, perhaps in the most important market segment, servers, AMD looks to be ahead of Intel until late 2023 – mostly on account of much higher core counts. This is two years away from today, and that’s a long time for the market to look forward to. Until then, AMD might well continue to take server share away from Intel. Also, it seems possible that near-term, AMD will also increase its threat to the laptop market, due to an incoming, very large, GPU processing power improvement.\nWithin two years, it seems likely that Intel will again regain a sustainable performance advantage over AMD across all market segments. In the near term, Alder Lake will already imply pressure on AMD’s desktop margins (for a while, AMD has been commanding premium pricing in that market, when compared to Intel). Still, one wonders if the market will look that far forward right now.\nAMD stock is rendered dangerous in light of these developments. Intel stock might become more attractive, since Intel also seems to be improving its process technology at a higher cadence now, and its push into the foundry business might also be favorable.\nHowever, the current situation remains hard for Intel on the server market, and thus makes an immediate long bet somewhat hard as well even if the potential is visible. Considering this, I’m leaning slightly bearish on AMD and leaning slightly bullish on Intel, but it’s hard to commit to the position fully given the short-term server headwinds for Intel and tailwinds for AMD resulting from Zen 4 expectations.","collect":0,"html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>Intel Vs. AMD: Competitive Prospects</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\nIntel Vs. AMD: Competitive Prospects\n</h2>\n\n<h4 class=\"meta\">\n\n\n2021-10-07 11:25 GMT+8&nbsp;&nbsp;&nbsp;<a href=https://seekingalpha.com/article/4458744-intel-vs-amd-competitive-prospects>Seeking Alpha</a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\nSummary\n\nThis is a (very) long look into the competitive positioning between Intel and AMD, on the CPU/SoC side, over the next 2+ years.\nIt covers the desktop, laptop and server segments.\nThere are ...\n\n<a href=\"https://seekingalpha.com/article/4458744-intel-vs-amd-competitive-prospects\">Web Link</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","type":0,"thumbnail":"","relate_stocks":{"AMD":"美国超微公司","INTC":"英特尔"},"source_url":"https://seekingalpha.com/article/4458744-intel-vs-amd-competitive-prospects","is_english":true,"share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","article_id":"1146744889","content_text":"Summary\n\nThis is a (very) long look into the competitive positioning between Intel and AMD, on the CPU/SoC side, over the next 2+ years.\nIt covers the desktop, laptop and server segments.\nThere are changes happening, and more will be coming, to this competitive positioning.\n\nVedat OGUZCAN/iStock via Getty Images\nThis article will cover the present, and near-term (up to more than two years out), competitive prospects between Advanced Micro Devices(NASDAQ:AMD)and Intel(NASDAQ:INTC)on the CPU market. Namely, on its three main segments: desktop, laptop and server.\nThe article was built through the careful aggregation of information from dozens of sources, taking into account the most credible reports backed by the most recent benchmarks and statements from Intel and AMD.\nThe Context\nAMD has sprang a surprise. Since the first Zen 3-based chips reached the market, AMD has built up a significant advantage over Intel. This happened in the:\n\nDesktop segment, where Zen 3 chips got to achieve equivalent single-threaded performance, while besting Intel on multithreaded performance due to much higher core counts.\nLaptop segment, though to a lesser extent, mostly on the strength of its multithreaded performance as well.\nServer, with its EPYC CPUs, on account of very competitive multithreaded abilities as a result of much higher core counts, as well as better performance / energy consumption metrics. Moreover, since Intel had been unchallenged in the server space for so long, AMD found it easy to price its CPUs at much better price / performance and TCO (Total Cost of Ownership) / performance ratios.\n\nAMD achieved this incredible feat through three developments:\n\nIt finally built a competitive CPU architecture, enabling it to level the field in terms of IPC (Instructions Per Cycle).\nIt got access to a process advantage due to using an external supplier, TSMC(NYSE:TSM). TSMC surpassed Intel's process abilities, and with Intel fabricating its chips internally, this created a semi-permanent disadvantage for Intel CPUs. This happened on account of many Intel missteps and delays advancing its process technology.\nAnd AMD's bet on a MCM (Multi-Chip Module) architecture, which enabled it to divide a monolithic CPU die (which is expensive to manufacture due to manufacturing yields being lower for larger dies) into many small, less expensive dies. This brought down the cost of the chip and allowed AMD to sell chips with many more cores than Intel's at similar price points.\n\nIntel took a long time to respond to AMD’s challenge, giving AMD the possibility of catching and surpassing Intel.\nThings move slowly when it comes to designing chips or evolving chipmaking processes, so the initial reaction would always have to be slow. In the meantime, AMD has been slowly but surely eroding Intel’s market position, including on the very profitable server market. Any increase in AMD competitivity would lead to this, since Intel started from very high market shares on account of AMD’s prior near-absolute lack of competitiveness.\nThis is the starting point for my analysis today. At this point in time, AMD has been beating Intel like a redheaded stepchild across the market (by gaining market share), though slower on servers and laptops.\nServers are a conservative market and move slower, while on laptops Intel’s 11thgeneration chips still managed to keep AMD somewhat at bay due to a large GPU performance gain while keeping decent performance on the CPU side.\nAnyway, let’s check how things are set to evolve over the next 2-3 years.\nDesktopPresent\nIn terms of desktop CPUs, currently the battle is fought between Intel’s 11thgeneration and AMD’s Ryzen 5000 series. Intel has a small lead in single threaded performance, but AMD has a very large lead on multitasking performance, due to AMD fielding CPUs with a higher core number. For instance, at the top of the range, AMD’s Ryzen 5950x has 16 cores 32 threads) to Intel’s 11900K 8 (16 threads).\nIntel’s single threaded performance cannot compensate for the smaller number of cores, so presently AMD’s desktop offerings are more competitive.\nNext Month\nThe first shot on the renewed Intel/AMD war will be fired by Intel next month. That’s when Intel’s 12thgeneration, Alder Lake, will launch. This generation has one particularity: it will be the first time the x86 world will see the arrival of a big.little architecture.\nIn the case of Alder Lake, Intel will provide CPUs with up to 8 performance cores and another 8 efficiency cores. The performance cores have hyperthreading (two threads per core), the efficiency cores do not, hence, this 16-core CPU can run 24 threads, versus 32 on a top-of-the-line Ryzen. The top of the Intel Alder Lake line will be the i9-12900K CPU.\nThis chip will represent a huge step for Intel. From early tests, the 12900K is showing up to 2x higher performance in multitasking, when compared to the 11900K.\nThe jump in multitasking performance comes from two sources:\n\nThe small efficiency cores are actually quite performant. They perform above the level of an SkyLake core (thus, at slightly less than 1/2 the performance on the new Performance cores), but do so while demanding significantly less power. Hence, although they are efficiency cores, they can still provide significant computing capacity which impacts the overall multitasking ability of the chip. Roughly, about 33% of the chip’s multitasking power can come from these cores.\nAlder Lake also represents a large IPC jump, due to it using entirely new and much wider Performance core. This IPC gain is on the order of 19-20%, and a larger overall gain still might come from higher frequencies on top.\n\nEfficiency Cores\nOne of the big surprises with Intel’s Alder Lake generation, is the performance the E-Cores (Efficiency Cores, Gracemont) are said to attain. These cores, schematically, seem to occupy 1/4ththe area taken by the new Golden Cove P-Cores (Performance Cores). Yet, in Intel’s Architecture Day 2021 presentation, these cores are able to attain levels of performance over those hit by SkyLake cores.\nNotice the relative sizes:\n\nAnd notice the power/performance charts vs SkyLake\n\nThe 80% performance jump doesn’t mean the core is 80% faster on single-threaded jobs. This is a multithreaded comparison, with four single-threaded efficiency cores vs twohyperthreadedSkyLake cores. For single-threaded performance:\n\nPerformance Cores\nAt the same time, Intel’s big cores have made a large jump in performance. This jump is cumulative with what happened with Rocket Lake, and thus it has been enough to propel Intel well ahead of AMD’s current offerings when it comes to single threaded performance (~25%+).\nThe performance jump on the Performance cores, together with actually performant efficiency cores, also means that Alder Lake is able to match AMD's multithreaded performance even though it can only process 24 threads (16 threads on the 8 Performance cores, 8 threads on efficiency cores) vs AMD's 32 (on high-end chips).\nThis performance jump by Intel arguably will put Intel at an advantage in the desktop market next month, as soon as Alder Lake launches.\nQ1 2022\nIn Q1 2022, AMD’s 6000 series (Warhol) arrives. It will still be based on the Zen 3 architecture, and promises small (up to 5%) IPC gains, plus gains in other areas including frequency, aggregating to a~9% performance gain.\nThis will clearly fall short of what Alder Lake is showing now, though it can create a small advantage on multithreaded uses. The desktop core count will remain at 16 cores, hence there’s no large leapfrogging of Intel’s effort.\nAs a result, even with the AMD 6000 series shipping, Intel will likely retain the desktop performance crown.\nQ3 2022\nCome Q3 2022, Intel will likely release its 13thgeneration CPU family, Raptor Lake. This will be a refresh of Alder Lake, with minor IPC and memory improvements. The IPC and performance gains on this generation will likely mirror those attained by AMD on its AMD 6000 series.\nHowever, Intel will add another 8 efficiency cores to Raptor Lake on its high-end chip, upping the core count to 24 and the thread count to 32. What this means is:\n\nGiven the estimated contribution by the efficiency cores to Alder Lake’s multithreaded performance (~33%) and 8 additional efficiency cores (doubles their count), these additional cores can thus add a further 33% multithreaded improvement;\nAnd there will also be some single-threaded performance improvements.\n\nThis seems enough to put the 13900K CPU ahead of the high-end Ryzen 6000 CPU even on multithreaded scenarios. Thus, Raptor Lake will solidify Intel’s position at the top of the desktop market.\nQ4 2022\nAMD is set to launch the Zen 4-based Ryzen 7000 series (Raphael) at this time, likely close to year-end.\nIt’s hard to estimate how large the performance uplift will be, though it seems up to 29% IPC and 40%-45% overall performance gains over the current generation (Zen 3) might be attainable. The core count for Desktop chips will remain at 16 cores (32 threads). Zen 4 will be built on a new TSMC process node.\nThese are very significant gains. If they’re realized, they are probably enough to put Zen 4 slightly ahead of Intel’s Raptor Lake, or match it.\nIn my view, even if an advantage is gained, it will likely be small (given the large jump achieved by Alder Lake, and then reinforced by Raptor Lake). Hence, at this point – Q4 2022 -- things will be too close to call on the Desktop market.\nQ2 2023\nIntel will launch Meteor Lake. This is an entirely new architecture on a new process node (Intel 4). Although there are no performance numbers at this point, a new architecture together with a new process node is bound to produce a large IPC gain. On top of this, Meteor Lake will also be a truly MCM design, allowing Intel to more easily ramp up the core count, which will favor multithreaded performance.\nHence, when Meteor Lake is launched, Intel is highly likely to put itself firmly ahead of Zen 4 CPUs in the desktop market. AMD’s answer will only come much later, in the form of Zen 5. At this point, it’s possible that Intel will have gained, and will keep, the performance leadership for a longer time.\nLaptopPresent\nCurrently, the laptop battle is fought between Intel’s Tiger Lake-H and -U families and AMD’s Ryzen 5000 H and U series.\nIntel and AMD are well matched in the H segment, with single-threaded and multithreaded performance broadly similar between the two.\nOn the U segment, Intel provides a lower core count at the top (4 vs 8), which gives AMD a performance lead on multithreaded tasks. However, Intel then has a large lead on single-threaded performance, as well as a large lead on GPU performance.\nGiven the above, I’ve considered that it’s too close to call this segment right now.\nQ1 2022\nDuring Q1 2023, Intel will launch new CPUs for the laptop market (H and U families) based on its very performant Alder Lake generation (as we saw from the desktop segment).\nThese CPUs will go up to 6+8 cores (14 cores, 20 threads, H family) and 2+8 cores (10 cores, 12 threads, U family), which will represent a large increase in core and thread counts from the present Tiger Lake generation (H goes from 8 to 14 cores, 16 to 20 threads; U goes from 4 to 10 cores, 8 to 12 threads). Simultaneously, the Performance cores will have a 20% IPC uplift (and likely a larger overall performance gain).\nHence, compute performance will be substantially increased, putting it well ahead of the existing AMD Ryzen 5000 series. However, the maximum EU (Execution Unit) count on the Alder Lake mobile CPUs remains capped at 96 EUs. Moreover, it seems these EUs will be based on the same GPU generation as the existing Tiger Lake GPUs. Hence, GPU performance will likely not increase a lot.\nAt about the same time, AMD will launch Ryzen 6000 APUs (Rembrandt), which will somewhat improve overall performance (~9%). But most importantly, AMD will mate these APUs with an entirely new GPU generation based on its RDNA 2 technology. This change will bring extreme GPU performance increases because the prior APUs were still based on AMD’s outdated and outclassed Vega technology. The performance increases will likely exceed 100%, and will put AMD’s mobile offerings well ahead of Intel’s when it comes to GPU performance.\nHence, as of Q1 2022 Intel will likely get the mobile CPU performance crown, but AMD will take away the integrated mobile GPU performance crown from Intel. As a result, I think the battle between these two products will be too close to call and will depend on each customer’s priorities.\nQ3 2022\nIntel’s mobile range will be refreshed with the launch of its 13thgeneration Raptor Lake CPUs. These will bring small CPU performance gains, which will build on the gains it attained with Alder Lake.\nHowever, the big news with the mobile Raptor Lake CPUs is that they’ll switch the GPU technology, going from Intel LP GPUs to Intel HPG GPUs. Moreover, the maximum number of EUs will be increased from 96 to 192 (doubled).\nThis is a significant change. Intel’s HPG EUs are supposed to bring a 50% performance uplift (on the same power budget), and the doubling of the maximum EUs will bring another 100% performance uplift. Conceivably, the maximum GPU performance on Raptor Cove high-end mobile CPUs might increase as much as 200%, though Intel might sacrifice some of the potential improvement for energy saving reasons.\n\nWhen this is realized, Intel will attain performance superiority both on the CPU and GPU side in its mobile offerings. Hence, Intel will be at a clear advantage on the laptop market from Q3 2022 onwards.\nQ1 2023\nQ1 2023 marks the likely arrival of Zen 4-based laptop CPUs from AMD. These will likely enable AMD to regain the upper hand on CPU multitasking performance, and possibly level things on the single-threaded performance side. It remains to be seen what will happen at the GPU level.\nIn my view, like on the desktop side, AMD will regain the upper hand if the GPU side manages to surpass Intel’s big jump from using its HPG GPUs. Still, for me right now this segment will be too close to call once Zen 4 arrives.\nQ2 2023\nIn Q2 2023, Intel will likely start shipping its Meteor Lake 14thgeneration mobile CPUs. Meteor Lake is both a new architecture and it will be built on a new process node. Additionally, Meteor Lake will be Intel’s first truly MCM design of the modern era.\nAs a result of so many performance-enhancing developments, Meteor Lake is likely to bring significant performance gains (15-20% or more, depending on Intel choices), both on multitasking and single-threaded scenarios. Since Meteor Lake will arrive when things are more or less balanced between Intel and AMD, this will immediately push Intel towards a clear leadership in the laptop segment.\nServerPresent\nPresently, Intel’s Xeon Scalable 3rdGen tops out at 40 cores, while AMD’s EPYC Milan CPUs go all the way to 64 cores.\nThis large difference in number of cores means the Xeon is not very competitive in tasks involving maximum throughput. Cloud offerings often require high utilization rates, and thus the higher core count gives superiority to AMD and offers it a way to gain market share, which AMD has been taking advantage of.\nQ2 2022\nAround Q2 2022, Intel will launch its Sapphire Rapids Xeons. These will benefit from:\n\nAn early multi-die effort, allowing Intel to increase the core count to at least 56 cores. The CPU cores will be spread out in up to four dies, with 14 cores active in each (maybe more in the future).\nA very large jump in single-threaded performance, similar or larger than the one enjoyed by Alder Lake.\n\nAt about the same time, though, AMD is likely to deploy its EPYC Milan-X server CPUs. The main change with these will be the use of 3D stacking to add extra cache on top of the execution units, making for much larger CPU caches. This will provide a performance improvement which will vary a lot across workloads (those more reliant on cache might benefit significantly, others not so much).\nWithout the launch of AMD’s EPYC Milan-X, it would seem that the single- and multi-threaded performance afforded by the use of improved Alder Lake-derived cores (Golden Cove), along with a better (not as punishing, performance-wise) multi-die approach, would allow Sapphire Rapids to surpass AMD’s EPYC Milan. This would happen with Sapphire Rapids matching or beating EPYC Milan’s multitasking abilities, while also enjoying a significant single-threaded performance advantage.\nHowever, with the launch of EPYC Milan-X, I consider the situation too close to call, on account of Sapphire Rapids’ lower core count, along with some improvement in Milan’s overall performance.\nStill, on the server side, Intel being on equal terms with AMD is actually an improvement right now.\nQ4 2022\nIt all changes in AMD’s favor come Q4 2022, though. That’s when AMD will launch its EPYC Genoa generation, based on Zen 4 cores.\nThis generation will bring a new architecture, be built on a new process node and enjoy the benefits from the 3D cache stacking used on EPYC Milan-X. It will also bump the maximum number of cores to 96. This will result in:\n\nEPYC Genoa matching or beating Sapphire Rapids' single-threaded performance.\nEPYC Genoa greatly outclassing Sapphire Rapids when it comes to the all-important (for server loads) multithreaded performance (96 vs 56 cores).\n\nOn account of this, AMD will again regain the upper hand in the server market.\nQ1 2023\nOn Q1 2023, Intel will likely field a refresh of the Sapphire Rapids generation, code named Emerald Rapids. This generation will bring the benefits from the Raptor Lake desktop refresh to the server side. It will also bump the maximum number of cores to 64 from 56.\nStill, the advantage gained by the Zen 4-based EPYC Genoa seems large enough that Emerald Rapids won’t be enough to counter it. Emerald Rapids might get close on single-threaded use cases, but it will still remain far away in terms of multithreaded use cases – the most important for server CPUs most often handling virtualized workloads. Basically, the “number of cores” advantage enjoyed by EPYC Genoa will be unsurmountable by this iterative Intel effort.\nAMD will keep its advantage in the server space, and will thus continue to chip away at Intel’s market share throughout this period.\nQ3 2023\nAt this point, on the server side, there will be a truly significant event. Intel will bring its Granite Rapids server CPU generation to the market.\n\nThis generation enjoys benefits similar to those enjoyed by Meteor Lake on the desktop side, but to an even larger extent. Namely, this generation is set to bump the number of cores to 120, spread across two 60-core dies. This is significant in several ways:\n\nThese cores, on a standalone basis, will likely beat Zen 4 cores by a wide single-threaded margin. This is so because at the starting point Zen 4 will be roughly matched to the previous Intel cores (Raptor Cove on Raptor Lake). Meteor Lake (and Granite Rapids), however, will be built with a new core architecture (Redwood Cove) and on a new process step. Thus, the performance gains are likely to be significant on a single-threaded basis.\nBy putting 120 cores on a single server CPU, Intel will for the first time in a long while, be having more cores per CPU than AMD. With each core being more powerful, this already ensures a significantly higher multithreaded performance.\nBut worse still, since the core dies are much larger on Intel’s Granite Rapids (two dies, 60 cores each) than on AMD’s EPYC Genoa (12 CCDs with 8 cores each), this will lead to another performance benefit for Granite Rapids: The latencies between its cores will be much lower, leading to a large relative performance gain (other things being equal).\n\nFrom these effects, it seems a certainty that Intel will regain its performance lead, both on single-threaded and multithreaded use cases, on the server segment. And Intel's advantage will, at that point, be large. This is set to happen in Q3 2023, nearly two years from now.\nAfter Granite Rapids, Intel is also promising another very large jump in performance, but this will already fall outside the timeframe for this article. AMD will also bring Zen 5, but its arrival schedule is still very uncertain.\nA Charted Summary\nThe above translates into the following color-coded schematic for all the segments:\n\nConclusion\nTaking into account the near-term roadmaps, benchmarks, leaks and guidance, it seems Intel is on the verge of taking away the performance crown from AMD. This will happen by Intel first regaining the single-threaded crown by a visible margin, while more or less matching AMD on multitasking, followed by a period where Intel will get a lead on both counts, interrupted only briefly by AMD's Zen 4.\nHowever, perhaps in the most important market segment, servers, AMD looks to be ahead of Intel until late 2023 – mostly on account of much higher core counts. This is two years away from today, and that’s a long time for the market to look forward to. Until then, AMD might well continue to take server share away from Intel. Also, it seems possible that near-term, AMD will also increase its threat to the laptop market, due to an incoming, very large, GPU processing power improvement.\nWithin two years, it seems likely that Intel will again regain a sustainable performance advantage over AMD across all market segments. In the near term, Alder Lake will already imply pressure on AMD’s desktop margins (for a while, AMD has been commanding premium pricing in that market, when compared to Intel). Still, one wonders if the market will look that far forward right now.\nAMD stock is rendered dangerous in light of these developments. Intel stock might become more attractive, since Intel also seems to be improving its process technology at a higher cadence now, and its push into the foundry business might also be favorable.\nHowever, the current situation remains hard for Intel on the server market, and thus makes an immediate long bet somewhat hard as well even if the potential is visible. Considering this, I’m leaning slightly bearish on AMD and leaning slightly bullish on Intel, but it’s hard to commit to the position fully given the short-term server headwinds for Intel and tailwinds for AMD resulting from Zen 4 expectations.","news_type":1},"isVote":1,"tweetType":1,"viewCount":317,"commentLimit":10,"likeStatus":false,"favoriteStatus":false,"reportStatus":false,"symbols":[],"verified":2,"subType":0,"readableState":1,"langContent":"CN","currentLanguage":"CN","warmUpFlag":false,"orderFlag":false,"shareable":true,"causeOfNotShareable":"","featuresForAnalytics":[],"commentAndTweetFlag":false,"andRepostAutoSelectedFlag":false,"upFlag":false,"length":14,"xxTargetLangEnum":"ZH_CN"},"commentList":[],"isCommentEnd":true,"isTiger":false,"isWeiXinMini":false,"url":"/m/post/823302744"}