It's an optimistic time for AMD, but in fairness; this still isn't nearly as good as things looked in the athlon (xp, x2, 64 etc) days. Back then AMD generally outperformed intel in most workloads across most of the market; now - they're competitive, but still generally slower in the still very critical single threaded stuff.
I don't know if ryzen is making an impact here. Amd has the habit of shuffling results around, over-reporting loss a few quarters so it can every once in a while report more optimistic results.
Amd finally has high end chips that are competitive and people want to buy, and they are selling at fairly good prices, for both Amd and consumers.
So I expect to see margins increasing, not decreasing by 1%. Come on Amd - let's see margins at 50%!
Higher ASP's on Ryzen and some effect on GPU - keep in mind the big OEM's just in the past few weeks have put Ryzen on their websites - the first 3 months of a CPU ramp are always slow but Ryzen is selling well - next quarter with OEM help and Threadripper/Vega i think they will provide a bigger beat although GM will be flattish due to all the console chips peaking in Q3.
They sold out their inventory, but they can't sell anymore than that. Lisa said that they ordered extra wafers for the next quarter to replenish the inventory. So, only a small amount of the extra was from GPU. It was mostly Ryzen. Next quarter we're also going to see a knock on effect as the GPU inventory has to be replenished.
Athlon 64 & 64 X2 (2003-2006) was the only time period where AMD had undeniably the performance crown - the other processors you mentioned were "competitive"
* Athlon vs P3 was usually a wash in terms of IPC and mostly a clockspeed war - (athlon having the grunt and P3 using compiler tricks and "cheats" to gain the upper hand in benchmarks). Reread this for a trip down memory lane http://www.anandtech.com/show/557 * Athlon XP was underperforming P4 after it matured, but represented a better value.
Back in the 80486 days, AMD had the x486-100, and later the x486-133, which outperformed the Intel counterparts. I believe Intel capped at the 486-66, but maybe had a 75MHz part.
We are now at a stage where performance is far more often "good enough", so being a bit behind doesn't matter as much. The main weakness is with regard to mobile products where AMD needs to have a presence.
"Back then AMD generally outperformed intel in most workloads across most of the market; now - they're competitive" back then AMD didnt destroy Intel's cost structure by gluing small high yielding dies together - this is the biggest longest lasting advantage AMD has ever had. If you throw in the manufacturing gap slowing and its the perfect storm for AMD to be around and do well for many years.
This guy gets it. While on the small, single chip scale, AMD's 8-core Zeppelin die might not be as lustrous compared to Intel's latest & greatest (Kaby-Lake, Skylake-X) for a number of reasons; lower IPC, lower clock-ceiling, increased cache (spec. L3) latency (more so vs KL; SL-X's new mesh topology brings a few of it's own cache latency issues to the table), etc.. This all changes though when you zoom out just a little bit; then the tables turn completely. This table turner isn't Zen's basic CPU layout, and it was never supposed to be; Zen's performance just needed to be competitive because Zen's basic functional CPU blocks (ALU's, FPU's, etc...) weren't the revolutionary part of the design. That otoh, would be it's modularity & scalability with Infinity Fabric; which is where Jim Keller and his team RIGHTLY choose to focus their efforts (IF is Keller's brainchild, just like it's ancestor, Hyper-Transport which he also designed). AMD needed a single, high performance design that could scale from laptops to servers. They simply couldn't afford to design/maintain 2-separate CPU arch's aimed at the consumer and HEDT/server markets like Intel can to deal with this problem (let's Intel's small 4-core chips continue to use a simple/fast ring bus for superior single-threaded perf, vs the game perf & cache latency hitting mesh interconnect required on Skylake-X/SP to be able to scale high enough in core count), and seeing as how it's REALLY hard to climb out of near bankruptcy selling MASSIVE monolithic chips; they set out to blaze a new path for CPU scalability. They succeeded gloriously; probably better than anyone could have predicted honestly. With Infinity Fabric they've managed to near single-handedly conquer Moore's Law, and have absolutely SHATTERED the cost margins on producing, selling, and buying high-core count CPU's. Everyone's going to be going this direction in the future. Intel knows this better than anyone (see their "EMIB" work; the "glued together desktop dies" press-kit comments were practically written in jealousy green). And they pulled this off not just way before anyone else, but on a "edge of bankruptcy" R&D budget.... It boggles the mind that that they managed to get the Zen CPU arch itself up to matching Broadwell-E (aka only 1 gen behind in IPC (SL==KL), though w/ a notably lower clock ceiling) while being almost infinitely scalable (aka "Infinity Fabric", as it can connect a near infinite # of devices over nearly any kind of bus fast enough. I.e. a 256bit bi-directional crossbar connecting the 2 CCX's in a Zeppelin die, or over 64 PCIe 3.0 lanes connecting a 2-socket EPYC), all while upending the entire economics of CPU production (and potentially any kind of large chip like GPU's).
@Cooe very true, "This table turner isn't Zen's basic CPU layout"
I also had heard one of the AMD execs talk about a Basic or Generic layout being helpful when working with the foundry partners. To me it seems that they not only hit a home run with Infinity fabric and and scaling so well but also had to manage coordination with the foundry partners to make a smooth transition to the new node. Making the design more generic made it easier for their partners and that seems to be what is happening at 7nm. (so far anyway)
The single threaded advantage is largely overblown. In real world usage I can't tell any difference between the platforms. We have 5820's and Ryzen 1600 systems at work and the biggest difference is the enormous savings with the latter. Responsiveness and overall performance is a wash. Look at the article on this site where the conclusion is stated that office professionals will notice the web and PDF responsiveness and then look at the measured differences. They're usually a measured in milliseconds or in the PDF test a whole second. It's clearly trying to show some reason why you would spend $500+ more on the Intel platform. Sure there are still some applications that show a measurable gain for Intel, but I'm guessing as both companies push more and more cores into the mainstream those applications will see patches that reduce this considerably. We've already seen that happen with games.
The next argument will be that we've had multi-core processors for years and applications still aren't using them. While that's true, we haven't had 6 cores / 12 threads at a $200 price point with an affordable mainboard ecosystem to back it. That is a game changer.
I doubt it will last too long. Intel is moving fast and we haven't seen Zen APU yet. Intel leak has already pointing 6 cores desktop and quadcore 15W class mobile, pretty sure the price will be competitive. People will realize that Intel has been milking the masses for so long but that number of people realizing that should be smaller than those who don't care at all.
Intel has nothing on the foreseeable roadmap that looks to be disruptive to AMD's Zen platform. We're seeing incremental updates in their moved up product lines. AMD is also working on an improved Zen (Zen+) and Zen 2. Which of the two future architectures pull ahead, is hard to say. Consumption wise Intel is not very competitive with Zen in their newest platform (Skylake - X). Since their next process jump is a ways away, it doesn't appear this will change anytime soon.
So unless Intel is holding something back from us and their shareholders, what we're seeing today is likely how the market will be for at least a few quarters.
OK so with all that RyZen with full availability on the market and a black-swan mining rush that saw AMD literally sell out of every single card it could make at elevated prices they.... still lost money but had a small operating profit when you discount the fact that they are deeply in debt.
When retailers started slashing the prices on RyZen CPUs it was hailed as some sort of miracle in AMD's manufacturing process with "magical" 80% yields (magical to fanboys, not to people who know the industry very well). Not one person stopped to consider that maybe the prices were being lowered because the chips weren't flying off the shelves at full price.
The opposite is true of AMD's old video cards that are just rebrands of last years ho-hum Rx 480. The miners scooped them all up.
So the real moral of the story is: AMD probably made its almost-but-not-really profit this quarter on the backs of miners who ate up the supply and are literally right now dumping the cards back on the secondary market as the coins start to crash. RyZen wasn't the real reason for that profit, and a 2% increase in margin from last year when AMD was literally selling bad retreads of its 2012 product lineup is no miracle.
its slightly more dramatic than that. intel has no counter in the server space in about 4 months due to epic cpu's. this will grow in due time. only question is how low will they go with the pricing. with a 16 core cpu for 1000 dollars, i bet its only time until intel is a little more shaken up. im not an amd guy either, i just know 32 cores + lower pricing + higher lane capacity > 28 cores + intel brand loyalty... i actually feel bad for the guys. it will be interesting.
You are overestimating the strength of the Epyc CPUs. During AMD's conference call today they said that they hoped to eventually win 10% of the server CPU market (Intel currently has well over 90% of the market), but that that would take time. They said inventory for Epyc is building much more slowly than for Ryzen because they don't expect significant Epyc sales in the coming quarter or two.
AMD is going to have to win server customers over, and that will take more than just iron. It will take a stable platform. Winning enterprise customers over may be a long, tough slog.
The fact that the hyperscalers are apparently not jumping at Epyc is not very good news, because they are the ones that have the power to switch architectures at the drop of a hat and move all their software over to it. It seems like they have already decided on Intel's Purley this generation. Probably AMD will have to wait for their second generation Ryzen to have a chance at winning real significant revenue from the hyperscalers.
And even besides platform stability and cost to change software, I myself don't know the competitiveness of Epyc processors in real world server workloads. Everybody has to do their own benchmarking of their actual use case and their own cost analysis to be sure.
"The fact that the hyperscalers are apparently not jumping at Epyc is not very good news": What's the source of this 'fact'? MS and Baidu have publicly announced deployment (the former was more emphatic on this). It's not unreasonable for them to wait for the platform to mature a bit: over at "serve the home(STH)", they've said that the latest firmware supporting DDR 2666 was only recently available, so I think it's still being tuned. It's a new architecture, new platform, and system software hasn't been tested on AMD for a while, so it's not unreasonable for them to be more cautious.
"And even besides platform stability and cost to change software, I myself don't know the competitiveness of Epyc processors in real world server workloads": Epyc's best use case is a 1P system with lots of I/O and memory capacity, which Intel has nothing to compete. What would've required a gold/silver 2P Skylake to get memory and I/O might get folded into a 1P Epyc. I don't believe AMD will beat intel across the board on all workloads; that's too much to expect. If it does well in certain areas like multi-VMs, storage, GPGPU for ML/HPC, web servers, certain distributed-system workloads. Anandtech's review shows that there are certain scenarios in which it excels. They are unlikely to do well -- based on prima facie evidence -- on traditional databases or HPC. It seems like a conscious choice. Time will tell.
"What's the source of this 'fact'?" The source of the fact is AMD, as well as the hyperscalers themselves.
"MS and Baidu have publicly announced deployment (the former was more emphatic on this)" I am not talking about the public cloud. It's easy for them to make some instances available, just as Microsoft has done with AMD Radeon. I'm talking about hyperscalers adopting Epyc for their own use.
"Epyc's best use case is a 1P system with lots of I/O and memory capacity, which Intel has nothing to compete." So you expect Epyc to be competing with ARM servers in the storage space? That's not very optimistic.
MS and others (like Baidu) ARE adopting Epyc for their own use. MS will be deploying them in Azure.
Also, while I think Epyc spans a wide range of scenarios: what 1P ARM solution are you referring to that supports as many PCIe lanes, memory channels, and capacity as 1P Epyc?
I'd also add, what ARM solution has a hope of getting adopted for storage? There are none at the moment, all pie-in-the-sky. AFAIK, software defined storage still uses intel machines, often 2P to get the PCIe lanes required.
Intel currently has 99% of the server market. Every percentage point means $200 million of revenue. If AMD can capture%5, it means a billion in fresh revenue.
You guys aren't really replying to what I am saying when you say it's reasonable for buyers of servers to change slowly or that if AMD can capture 5% of the server market it's 1 billion dollars in revenue. You're just creating ghosts to disbelieve.
My post was a direct reply to the statement, "intel has no counter in the server space in about 4 months due to epic cpu's." The user that made this statement is saying that AMD has the upper hand and Intel has no counter. He's saying it's gonna happen in 4 months. AMD eventually capturing 5% of the market or capturing the storage server market is in no way a confirmation of his statement. Not even close.
So just to clarify, Yojimbo-san: You believe that if Intel starts losing market share in servers to AMD that this somehow means Intel doesn't need to react? Of course they need to respond... but it will take them some time to do so.
"And even besides platform stability and cost to change software, I myself don't know the competitiveness of Epyc processors in real world server workloads"
Oh, so you purport to know that Epyc will not have a stable platform, and doesn't run the same software? Tell us more, Sensei.
EPYC is very good technology and its going to do well over time as AMD executes into the next year or so - but i dont think it happens quickly - link https://www.youtube.com/watch?v=osSMJRyxG0k&t=...
inserting that Link I am not implying that Intel is going to do what they did here but that there will probably be some of it. Intel is not going to let customers go - they will offer rebates and do what they can to hold onto share. Intel is not a pushover and will do what they can - also they have been the only player for so long that they have embedded relationships in advertising ect. AMD is different that last time though: they dont do their own manufacturing and have redundancy they have a GPU business that creates an opportunity for attach rate they also have an great CPU design with a mind blowing cost structure in workstations and servers. AMD is back but this will take some time IMHO.
>RyZen wasn't the real reason for that profit, and a 2% increase in margin from last year when AMD was literally selling bad retreads of its 2012 product lineup is no miracle.
You should consider reading the article.
"The year-over-year results may seem a bit skewed, since Q2 2016 was actually a profitable quarter for AMD, but that was due to a $150 million infusion of cash from a joint-venture with Nantong Fujitsu Microelectronics. "
That $150 million in Q2 2016 did not affect their Q2 2016 gross margin. It was recorded as "other income" and not part of net revenue. So AMD really did see a 2% y/y increase in gross margins.
And the point CajunArson made about the strong results this quarter possibly being mostly attributable to a temporary cryptocurrency craze is valid, as is his point that year over year this quarter is being compared to a Q2 2016 quarter with a weak product lineup.
However, I have to say that AMD's Q2 2016 computing and graphics only decreased 5% sequentially from Q1 2016, and at the time people were wondering if AMD's GPU business was turning around. But although they had picked up market share in GPUs, if I remember, their average selling price went down. My guess was that the revenue increase had to do with how AMD recorded sales for Polaris chips entering the channel or something. In any case, in Q3 2016 NVIDIA's GPU revenue went through the roof with Pascal and AMD's went nowhere (up only 9% sequentially over Q2 even though Q3 is seasonally a much stronger quarter). My point is that Q2 2016 was not really as bad for their GPU revenue as CajunArson made it out to be. They had only recently introduced Polaris so most of the cards sold in retail were probably pre-Polaris. But somehow I think they had counted Polaris revenue in Q2.
Miners did not affect amd income at all. The retal sellers did take the profit! Amd did sel normal amounth of graphic cards at normal price. The retalers did get all the extra money when the prices spiked! The Computer sales has declined all the time, so getting close to even is "ok" but all in all amd needs to get to green and that is not easy. They should increase margings (more expensive to consumer) but can not do it if it affect the sales too much. And people will buy Intel even if They cost considerably more because it is Intel, so the situation is still dire...
You have a point but also is it that surprising? Ryzen 5 and 7 might be great for DIY builds and some high priced gaming prebuilts. All in all a fairly small market. The real volume and hoepfully for AMD income should happen with the APU and EPYC. APU for volume and EPYC for margins. Both are not out yet.
EPYC's technically out; actually has been for a few weeks now (Anandtech here already had their hands on hardware for benchmarking). Rollouts at the very major partners (MS, Baidu, etc...) are likely underway, or will be in the very near future, but other than that, generally anything server/data center moves slower than molasses. It'll be a good while before parts start showing up in the off-channels where average consumers can buy them.
The yields are high and AMD wants more marketshare a mindshare. If the arch let's them sell for an even more attractive price to exceed meet the agenda, why not?
People are buying 1700s instead of 1800Xs, and 1600s instead of 1600Xs, plus the Ryzen 3s aren't out yet. That can't be helping AMD's bottom line. They still have Threadripper, EPYC and the Zen-based APUs to launch over the next six to nine months as well as Ryzen 3, and Ryzen Pro has just launched, but I don't imagine that Vega is going to help based off initial reports so I expect their graphics business to flounder for a bit.
What I'm saying is, they stand to do better than they are currently doing, financially speaking.
@CajunArson - ridiculous - the first quarter of a "Desktop" CPU launch - no server revs yet, no APU out yet, threadripper not out yet and you think AMD is peaking? its just beginning.
The way AMD segments their revenue drives me crazy. What do data centers have to do with consoles? Why aren't we allowed to resolve PC GPU revenue separately from PC CPU revenue? I don't think we really know how much of their revenue gain was from Ryzen, which, if it is responsible for the strong revenue numbers, would represent a fundamental strengthening of the company, and how much was because of cryptocurrency, which is most likely just a one or two quarter event.
I don't think it has anything to do with their competitors. AMD's competitors know how AMD are doing insofar that the competitors know how they themselves are doing in the market as a whole, which is what is important, anyway. It's entirely to do with investors, particularly smaller investors who can't afford the detailed research reports. But just a general obfuscation or delay in the resolution of the true state of their business seems to be AMD's goal.
That seems to be their tactic with the strength of their upcoming graphics architectures , too. When did we start hearing about Vega from AMD? It was before Polaris came out. And we heard about all the great improvements Vega would brings and how it would return AMD to a strong position in the high end of the GPU market and how it would compete in the data center. They put it on a graphic situated at the beginning of 2017. They hyped it up, even showing it off in careful demos in November or December of 2016. Then they finally said it would launch in Q2 of 2017. Now here it is, finally launching in August of 2017, smack in the middle of Q3, not looking like it's going to live up to their hard promises much less their hype. How long until they start talking about Navi?
The next step is some sort of Ryzen APU to go in laptops. AMD already has design wins for the last 4 MS and Sony consoles, and I expect that dominance to continue since they can produce the most powerful APU with their IP. So far, Ryzen, EPYC and Threadripper do nothing for the everyday consumer looking for a mobile offering.
""who can make this piece of crap cheapest" Why do you think people buy Intel servers instead of IBM? IBM has higher performance than Intel over a wide swathe of workloads. Intel makes them cheaper and more power efficient. That's largely why. That's why x86 dominates the server landscape and not RISC.
And, AMD does integrated GPUs way better than Intel. It's a no-contest there. The cores are now only marginally better than AMDs. Most of the PC TAM is in laptops, so the Ryzen APU is where the money is even if the margins aren't what they once were.
Marginally as in one core in my five year old 3570 is dangerously close to one core in R7? Okay sure, AMD was always a second-tier solution, so no surprises here, and it was _especially_ visible in laptops.
Oh, and people buy Intel instead of IBM because the latter is literally a shambling zombie led by a womyn. With entirely predictable results.
Lol I upgraded from a 3570K cranked to the ceiling (4.6GHz) to a Ryzen 7 1700 (4.0GHz) and you are completely talking out your ass. Ryzen is WAAAY faster than Ivy Bridge in just about every way measurable, single-threaded performance per clock (aka IPC) is VASTLY superior (win some, loss some with Broadwell/-E) and with more than 4x the threads and 3x the cache, is more than 2x as fast (!!!!) multi-threaded. For example, despite being clocked a whopping 600MHz faster than my Ryzen, it's single-threaded performance is no better (pretty close to equal), but loses by a factor of around 2.5x in total, multi-threaded performance. That's MASSIVE. My performance in every single CPU intensive thing I do has absolutely skyrocketed, I can multi-task like you're system couldn't even believe (I witnessed my 3570K and it's 4-measly threads get overloaded CONSTANTLY). Plus, I'm on a way more modern platform that will be supported for like 3-4+ years for cheap, swap in CPU upgrades, not 1 then ditched. Heck, even if the only thing you care about is games, even in games where they do preform similarly (which is shrinking and shrinking as 4-threads is quickly becoming a SERIOUS bottleneck), the actual gaming experience on Ryzen is SOOOO far superior it's kinda crazy. This effect has been studying extensively, 4-thread CPU's are rapidly becoming thread constrained in an 8-core console world. This leads to "thread-hitching" (small frame-rate hitchs whenever a games idle-threads and active-threads have to be swapped to/from a CPU core), and much lower fps minimum's, 1% averages, and worse frame-times, even if the frame-rate counter is on average showing a similar number. You would get the same effect going to an i5 to an i7, but the point remains, 4-thread CPU's are a dying breed for gaming, just as they're a dying bread for anyone who does real work with their computers. 4-threads, no matter how fast you clock them, just aren't enough to cut the mustard for any kind of significant, modern workload. But, hey, whatever you have to tell yourself to make your old hardware look better :). (And I'm not hating the hardware, in it's day the 3570K was a great chip, and it's had a remarkable lifespan (along with the 2500K), but the older i5's are seriously starting to become thread-constrained. Unlike the newest i5's can't make up some of that loss with like 15-20% faster IPC (IB->KL), and like 15% faster clocks, though honestly, I wouldn't recommend a 4-thread chip to anyone at this point (for a primary use desktop) no matter how fast the threads are.
Synthetic benchmarks clearly show that even an i5 2500K, over 6 years old at this point, is faster at single and 4-threaded workloads than the top of the line Ryzen SKUs if both are overclocked to the max. AMD made an impressive IPC jump, but they barely reach 3.9/4.0Ghz. Sandy bridge at 4.8 to 5.0Ghz is not a rarity with proper cooling, and 4.5Ghz is possible even with a stock cooler.
It's funny how the people most vocal about Ryzen, the gamer fanboys, profit the least from its strengths. They would be much better off buying old i5s on the cheap and overclocking them, saves you from having to spend a fortune on DDR4 as well. Ryzen is much better suited for programmers, CAD artists and other professionals.
Hurr Durr vomited this crap: "Oh, and people buy Intel instead of IBM because the latter is literally a shambling zombie led by a womyn. With entirely predictable results."
What an amazingly thoughtful and unbiased, technically literate critique! He must've gone to Trump University....
Your 3570 is marginally worse than whatever equivalent 6/7 gen intel CPU too. And the graphics as as third-tier as it always was; third, because it can't go any lower. AMD was bad on the bulldozer architecture and a bulk 28nm process that leaked far more than Finfet. Both of those are history. Zen is especially efficient at the 2-3 GHz range (power starts going off the rails over 3.7GHz). So, yeah, as a package, Ryzen Mobile as a package is *better* than anything Intel has, if they can match or beat power, which seems likely.
"It`s hardly the dominance worth pursuing when deciding factor is "who can make this piece of crap cheapest"." in a way that is a deciding factor - Intels and AMD's products are commodities - they are not brands - their chips go into products and are not products themself. AMD is taking a step into stopping the myth that is the "Intel Brand" - the commercials and slogans wont stop the fact that they make chips and that is it. Nobody cares what kind of ARM chip is in an Ipad - they just want to enjoy their "Juicy fruit" device.
Yep. AMD's APUs couldn't come fast enough. AMD's products has slight edge now over Intel's probably for more than a year in desktops. I wonder when they will release mobile parts as Intel might release Cannonlake at the end of the year.
It's a long shot but I'm hoping for an Atom-class Zen APU to show up someday. The market for small, low to mid-end Windows tablet still exists but Intel dropped out of it by killing most Apollo Lake SKUs.
The 2% increase in margins from last year is great news, shame they're still losing money. I expected them to dominate the market in DIYs/desktops since the Ryzen series seems to be great value at mid and high tier segments.
The real MVPs are the accountants that are able to keep them afloat while they've bled money the last 5/10? years. The one good thing about cheap credit and 0% interest rates are these borderline companies still being able to stay in business.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
60 Comments
Back to Article
austinsguitar - Tuesday, July 25, 2017 - link
for the first time since athlon x2..... welcome back amd, we missed you <3emn13 - Wednesday, July 26, 2017 - link
It's an optimistic time for AMD, but in fairness; this still isn't nearly as good as things looked in the athlon (xp, x2, 64 etc) days. Back then AMD generally outperformed intel in most workloads across most of the market; now - they're competitive, but still generally slower in the still very critical single threaded stuff.ddriver - Wednesday, July 26, 2017 - link
I don't know if ryzen is making an impact here. Amd has the habit of shuffling results around, over-reporting loss a few quarters so it can every once in a while report more optimistic results.Amd finally has high end chips that are competitive and people want to buy, and they are selling at fairly good prices, for both Amd and consumers.
So I expect to see margins increasing, not decreasing by 1%. Come on Amd - let's see margins at 50%!
ddriver - Wednesday, July 26, 2017 - link
My guess is the extra revenue came from the GPUs which are essentially sold out everywhere thanks to the mining craze.stockolicious - Wednesday, July 26, 2017 - link
Higher ASP's on Ryzen and some effect on GPU - keep in mind the big OEM's just in the past few weeks have put Ryzen on their websites - the first 3 months of a CPU ramp are always slow but Ryzen is selling well - next quarter with OEM help and Threadripper/Vega i think they will provide a bigger beat although GM will be flattish due to all the console chips peaking in Q3.lefty2 - Thursday, July 27, 2017 - link
They sold out their inventory, but they can't sell anymore than that. Lisa said that they ordered extra wafers for the next quarter to replenish the inventory. So, only a small amount of the extra was from GPU. It was mostly Ryzen. Next quarter we're also going to see a knock on effect as the GPU inventory has to be replenished.Spoelie - Wednesday, July 26, 2017 - link
Athlon 64 & 64 X2 (2003-2006) was the only time period where AMD had undeniably the performance crown - the other processors you mentioned were "competitive"* Athlon vs P3 was usually a wash in terms of IPC and mostly a clockspeed war - (athlon having the grunt and P3 using compiler tricks and "cheats" to gain the upper hand in benchmarks). Reread this for a trip down memory lane http://www.anandtech.com/show/557
* Athlon XP was underperforming P4 after it matured, but represented a better value.
PGFan - Wednesday, July 26, 2017 - link
Back in the 80486 days, AMD had the x486-100, and later the x486-133, which outperformed the Intel counterparts. I believe Intel capped at the 486-66, but maybe had a 75MHz part.Kvaern1 - Thursday, July 27, 2017 - link
Intel made a 100mhz 486 DX4 and by the time the 100+ mhz AMD 486 CPUs came out Intel's highend had moved on to Pentiums.Lonyo - Wednesday, July 26, 2017 - link
We are now at a stage where performance is far more often "good enough", so being a bit behind doesn't matter as much. The main weakness is with regard to mobile products where AMD needs to have a presence.stockolicious - Wednesday, July 26, 2017 - link
"Back then AMD generally outperformed intel in most workloads across most of the market; now - they're competitive"back then AMD didnt destroy Intel's cost structure by gluing small high yielding dies together - this is the biggest longest lasting advantage AMD has ever had. If you throw in the manufacturing gap slowing and its the perfect storm for AMD to be around and do well for many years.
Cooe - Wednesday, July 26, 2017 - link
This guy gets it. While on the small, single chip scale, AMD's 8-core Zeppelin die might not be as lustrous compared to Intel's latest & greatest (Kaby-Lake, Skylake-X) for a number of reasons; lower IPC, lower clock-ceiling, increased cache (spec. L3) latency (more so vs KL; SL-X's new mesh topology brings a few of it's own cache latency issues to the table), etc.. This all changes though when you zoom out just a little bit; then the tables turn completely. This table turner isn't Zen's basic CPU layout, and it was never supposed to be; Zen's performance just needed to be competitive because Zen's basic functional CPU blocks (ALU's, FPU's, etc...) weren't the revolutionary part of the design. That otoh, would be it's modularity & scalability with Infinity Fabric; which is where Jim Keller and his team RIGHTLY choose to focus their efforts (IF is Keller's brainchild, just like it's ancestor, Hyper-Transport which he also designed). AMD needed a single, high performance design that could scale from laptops to servers. They simply couldn't afford to design/maintain 2-separate CPU arch's aimed at the consumer and HEDT/server markets like Intel can to deal with this problem (let's Intel's small 4-core chips continue to use a simple/fast ring bus for superior single-threaded perf, vs the game perf & cache latency hitting mesh interconnect required on Skylake-X/SP to be able to scale high enough in core count), and seeing as how it's REALLY hard to climb out of near bankruptcy selling MASSIVE monolithic chips; they set out to blaze a new path for CPU scalability. They succeeded gloriously; probably better than anyone could have predicted honestly. With Infinity Fabric they've managed to near single-handedly conquer Moore's Law, and have absolutely SHATTERED the cost margins on producing, selling, and buying high-core count CPU's. Everyone's going to be going this direction in the future. Intel knows this better than anyone (see their "EMIB" work; the "glued together desktop dies" press-kit comments were practically written in jealousy green). And they pulled this off not just way before anyone else, but on a "edge of bankruptcy" R&D budget.... It boggles the mind that that they managed to get the Zen CPU arch itself up to matching Broadwell-E (aka only 1 gen behind in IPC (SL==KL), though w/ a notably lower clock ceiling) while being almost infinitely scalable (aka "Infinity Fabric", as it can connect a near infinite # of devices over nearly any kind of bus fast enough. I.e. a 256bit bi-directional crossbar connecting the 2 CCX's in a Zeppelin die, or over 64 PCIe 3.0 lanes connecting a 2-socket EPYC), all while upending the entire economics of CPU production (and potentially any kind of large chip like GPU's).stockolicious - Saturday, July 29, 2017 - link
@Cooevery true,
"This table turner isn't Zen's basic CPU layout"
I also had heard one of the AMD execs talk about a Basic or Generic layout being helpful when working with the foundry partners. To me it seems that they not only hit a home run with Infinity fabric and and scaling so well but also had to manage coordination with the foundry partners to make a smooth transition to the new node. Making the design more generic made it easier for their partners and that seems to be what is happening at 7nm. (so far anyway)
Threska - Sunday, July 30, 2017 - link
Wonder if their interposer tech might make an appearance later?DrKlahn - Wednesday, July 26, 2017 - link
The single threaded advantage is largely overblown. In real world usage I can't tell any difference between the platforms. We have 5820's and Ryzen 1600 systems at work and the biggest difference is the enormous savings with the latter. Responsiveness and overall performance is a wash. Look at the article on this site where the conclusion is stated that office professionals will notice the web and PDF responsiveness and then look at the measured differences. They're usually a measured in milliseconds or in the PDF test a whole second. It's clearly trying to show some reason why you would spend $500+ more on the Intel platform. Sure there are still some applications that show a measurable gain for Intel, but I'm guessing as both companies push more and more cores into the mainstream those applications will see patches that reduce this considerably. We've already seen that happen with games.The next argument will be that we've had multi-core processors for years and applications still aren't using them. While that's true, we haven't had 6 cores / 12 threads at a $200 price point with an affordable mainboard ecosystem to back it. That is a game changer.
cryosx - Thursday, July 27, 2017 - link
I think this time around Intel's dirty anti competitive tactics of locking in OEMs isn't going to work because of YouTube and the internet.WorldWithoutMadness - Wednesday, July 26, 2017 - link
I doubt it will last too long. Intel is moving fast and we haven't seen Zen APU yet. Intel leak has already pointing 6 cores desktop and quadcore 15W class mobile, pretty sure the price will be competitive. People will realize that Intel has been milking the masses for so long but that number of people realizing that should be smaller than those who don't care at all.DrKlahn - Wednesday, July 26, 2017 - link
Intel has nothing on the foreseeable roadmap that looks to be disruptive to AMD's Zen platform. We're seeing incremental updates in their moved up product lines. AMD is also working on an improved Zen (Zen+) and Zen 2. Which of the two future architectures pull ahead, is hard to say. Consumption wise Intel is not very competitive with Zen in their newest platform (Skylake - X). Since their next process jump is a ways away, it doesn't appear this will change anytime soon.So unless Intel is holding something back from us and their shareholders, what we're seeing today is likely how the market will be for at least a few quarters.
CajunArson - Tuesday, July 25, 2017 - link
OK so with all that RyZen with full availability on the market and a black-swan mining rush that saw AMD literally sell out of every single card it could make at elevated prices they.... still lost money but had a small operating profit when you discount the fact that they are deeply in debt.When retailers started slashing the prices on RyZen CPUs it was hailed as some sort of miracle in AMD's manufacturing process with "magical" 80% yields (magical to fanboys, not to people who know the industry very well). Not one person stopped to consider that maybe the prices were being lowered because the chips weren't flying off the shelves at full price.
The opposite is true of AMD's old video cards that are just rebrands of last years ho-hum Rx 480. The miners scooped them all up.
So the real moral of the story is: AMD probably made its almost-but-not-really profit this quarter on the backs of miners who ate up the supply and are literally right now dumping the cards back on the secondary market as the coins start to crash. RyZen wasn't the real reason for that profit, and a 2% increase in margin from last year when AMD was literally selling bad retreads of its 2012 product lineup is no miracle.
austinsguitar - Tuesday, July 25, 2017 - link
its slightly more dramatic than that. intel has no counter in the server space in about 4 months due to epic cpu's. this will grow in due time. only question is how low will they go with the pricing. with a 16 core cpu for 1000 dollars, i bet its only time until intel is a little more shaken up. im not an amd guy either, i just know 32 cores + lower pricing + higher lane capacity > 28 cores + intel brand loyalty... i actually feel bad for the guys. it will be interesting.Yojimbo - Tuesday, July 25, 2017 - link
You are overestimating the strength of the Epyc CPUs. During AMD's conference call today they said that they hoped to eventually win 10% of the server CPU market (Intel currently has well over 90% of the market), but that that would take time. They said inventory for Epyc is building much more slowly than for Ryzen because they don't expect significant Epyc sales in the coming quarter or two.AMD is going to have to win server customers over, and that will take more than just iron. It will take a stable platform. Winning enterprise customers over may be a long, tough slog.
The fact that the hyperscalers are apparently not jumping at Epyc is not very good news, because they are the ones that have the power to switch architectures at the drop of a hat and move all their software over to it. It seems like they have already decided on Intel's Purley this generation. Probably AMD will have to wait for their second generation Ryzen to have a chance at winning real significant revenue from the hyperscalers.
And even besides platform stability and cost to change software, I myself don't know the competitiveness of Epyc processors in real world server workloads. Everybody has to do their own benchmarking of their actual use case and their own cost analysis to be sure.
deltaFx2 - Wednesday, July 26, 2017 - link
"The fact that the hyperscalers are apparently not jumping at Epyc is not very good news": What's the source of this 'fact'? MS and Baidu have publicly announced deployment (the former was more emphatic on this). It's not unreasonable for them to wait for the platform to mature a bit: over at "serve the home(STH)", they've said that the latest firmware supporting DDR 2666 was only recently available, so I think it's still being tuned. It's a new architecture, new platform, and system software hasn't been tested on AMD for a while, so it's not unreasonable for them to be more cautious."And even besides platform stability and cost to change software, I myself don't know the competitiveness of Epyc processors in real world server workloads": Epyc's best use case is a 1P system with lots of I/O and memory capacity, which Intel has nothing to compete. What would've required a gold/silver 2P Skylake to get memory and I/O might get folded into a 1P Epyc. I don't believe AMD will beat intel across the board on all workloads; that's too much to expect. If it does well in certain areas like multi-VMs, storage, GPGPU for ML/HPC, web servers, certain distributed-system workloads. Anandtech's review shows that there are certain scenarios in which it excels. They are unlikely to do well -- based on prima facie evidence -- on traditional databases or HPC. It seems like a conscious choice. Time will tell.
Yojimbo - Wednesday, July 26, 2017 - link
"What's the source of this 'fact'?"The source of the fact is AMD, as well as the hyperscalers themselves.
"MS and Baidu have publicly announced deployment (the former was more emphatic on this)"
I am not talking about the public cloud. It's easy for them to make some instances available, just as Microsoft has done with AMD Radeon. I'm talking about hyperscalers adopting Epyc for their own use.
"Epyc's best use case is a 1P system with lots of I/O and memory capacity, which Intel has nothing to compete."
So you expect Epyc to be competing with ARM servers in the storage space? That's not very optimistic.
Alexvrb - Wednesday, July 26, 2017 - link
MS and others (like Baidu) ARE adopting Epyc for their own use. MS will be deploying them in Azure.Also, while I think Epyc spans a wide range of scenarios: what 1P ARM solution are you referring to that supports as many PCIe lanes, memory channels, and capacity as 1P Epyc?
deltaFx2 - Wednesday, July 26, 2017 - link
I'd also add, what ARM solution has a hope of getting adopted for storage? There are none at the moment, all pie-in-the-sky. AFAIK, software defined storage still uses intel machines, often 2P to get the PCIe lanes required.STH has done a good pros-and-cons of Epyc for the interested: https://www.servethehome.com/amd-epyc-v-intel-xeon...
andychow - Wednesday, July 26, 2017 - link
Intel currently has 99% of the server market. Every percentage point means $200 million of revenue. If AMD can capture%5, it means a billion in fresh revenue.Yojimbo - Wednesday, July 26, 2017 - link
You guys aren't really replying to what I am saying when you say it's reasonable for buyers of servers to change slowly or that if AMD can capture 5% of the server market it's 1 billion dollars in revenue. You're just creating ghosts to disbelieve.My post was a direct reply to the statement, "intel has no counter in the server space in about 4 months due to epic cpu's." The user that made this statement is saying that AMD has the upper hand and Intel has no counter. He's saying it's gonna happen in 4 months. AMD eventually capturing 5% of the market or capturing the storage server market is in no way a confirmation of his statement. Not even close.
Alexvrb - Wednesday, July 26, 2017 - link
So just to clarify, Yojimbo-san: You believe that if Intel starts losing market share in servers to AMD that this somehow means Intel doesn't need to react? Of course they need to respond... but it will take them some time to do so."And even besides platform stability and cost to change software, I myself don't know the competitiveness of Epyc processors in real world server workloads"
Oh, so you purport to know that Epyc will not have a stable platform, and doesn't run the same software? Tell us more, Sensei.
stockolicious - Saturday, July 29, 2017 - link
EPYC is very good technology and its going to do well over time as AMD executes into the next year or so - but i dont think it happens quickly - link https://www.youtube.com/watch?v=osSMJRyxG0k&t=...inserting that Link I am not implying that Intel is going to do what they did here but that there will probably be some of it. Intel is not going to let customers go - they will offer rebates and do what they can to hold onto share. Intel is not a pushover and will do what they can - also they have been the only player for so long that they have embedded relationships in advertising ect. AMD is different that last time though:
they dont do their own manufacturing and have redundancy
they have a GPU business that creates an opportunity for attach rate
they also have an great CPU design with a mind blowing cost structure in workstations and servers.
AMD is back but this will take some time IMHO.
vanilla_gorilla - Tuesday, July 25, 2017 - link
>RyZen wasn't the real reason for that profit, and a 2% increase in margin from last year when AMD was literally selling bad retreads of its 2012 product lineup is no miracle.You should consider reading the article.
"The year-over-year results may seem a bit skewed, since Q2 2016 was actually a profitable quarter for AMD, but that was due to a $150 million infusion of cash from a joint-venture with Nantong Fujitsu Microelectronics. "
Yojimbo - Tuesday, July 25, 2017 - link
That $150 million in Q2 2016 did not affect their Q2 2016 gross margin. It was recorded as "other income" and not part of net revenue. So AMD really did see a 2% y/y increase in gross margins.And the point CajunArson made about the strong results this quarter possibly being mostly attributable to a temporary cryptocurrency craze is valid, as is his point that year over year this quarter is being compared to a Q2 2016 quarter with a weak product lineup.
However, I have to say that AMD's Q2 2016 computing and graphics only decreased 5% sequentially from Q1 2016, and at the time people were wondering if AMD's GPU business was turning around. But although they had picked up market share in GPUs, if I remember, their average selling price went down. My guess was that the revenue increase had to do with how AMD recorded sales for Polaris chips entering the channel or something. In any case, in Q3 2016 NVIDIA's GPU revenue went through the roof with Pascal and AMD's went nowhere (up only 9% sequentially over Q2 even though Q3 is seasonally a much stronger quarter). My point is that Q2 2016 was not really as bad for their GPU revenue as CajunArson made it out to be. They had only recently introduced Polaris so most of the cards sold in retail were probably pre-Polaris. But somehow I think they had counted Polaris revenue in Q2.
haukionkannel - Wednesday, July 26, 2017 - link
Miners did not affect amd income at all. The retal sellers did take the profit! Amd did sel normal amounth of graphic cards at normal price. The retalers did get all the extra money when the prices spiked!The Computer sales has declined all the time, so getting close to even is "ok" but all in all amd needs to get to green and that is not easy. They should increase margings (more expensive to consumer) but can not do it if it affect the sales too much. And people will buy Intel even if They cost considerably more because it is Intel, so the situation is still dire...
yankeeDDL - Wednesday, July 26, 2017 - link
Miners did not affect margins, as AMD sold at their MSRP, but it did affect teh income, as AMD sold "everything" they had.Threska - Sunday, July 30, 2017 - link
"And people will buy Intel even if They cost considerably more because it is Intel, so the situation is still dire...""Nobody ever got fired for buying IBM." ;-)
beginner99 - Wednesday, July 26, 2017 - link
You have a point but also is it that surprising? Ryzen 5 and 7 might be great for DIY builds and some high priced gaming prebuilts. All in all a fairly small market. The real volume and hoepfully for AMD income should happen with the APU and EPYC. APU for volume and EPYC for margins. Both are not out yet.Cooe - Wednesday, July 26, 2017 - link
EPYC's technically out; actually has been for a few weeks now (Anandtech here already had their hands on hardware for benchmarking). Rollouts at the very major partners (MS, Baidu, etc...) are likely underway, or will be in the very near future, but other than that, generally anything server/data center moves slower than molasses. It'll be a good while before parts start showing up in the off-channels where average consumers can buy them.Lolimaster - Wednesday, July 26, 2017 - link
The yields are high and AMD wants more marketshare a mindshare. If the arch let's them sell for an even more attractive price to exceed meet the agenda, why not?silverblue - Wednesday, July 26, 2017 - link
People are buying 1700s instead of 1800Xs, and 1600s instead of 1600Xs, plus the Ryzen 3s aren't out yet. That can't be helping AMD's bottom line. They still have Threadripper, EPYC and the Zen-based APUs to launch over the next six to nine months as well as Ryzen 3, and Ryzen Pro has just launched, but I don't imagine that Vega is going to help based off initial reports so I expect their graphics business to flounder for a bit.What I'm saying is, they stand to do better than they are currently doing, financially speaking.
Kvaern1 - Thursday, July 27, 2017 - link
I'm not sure why anyone would expect the X models to sell better than their much better price/performance siblings?stockolicious - Wednesday, July 26, 2017 - link
@CajunArson - ridiculous - the first quarter of a "Desktop" CPU launch - no server revs yet, no APU out yet, threadripper not out yet and you think AMD is peaking? its just beginning.Yojimbo - Tuesday, July 25, 2017 - link
The way AMD segments their revenue drives me crazy. What do data centers have to do with consoles? Why aren't we allowed to resolve PC GPU revenue separately from PC CPU revenue? I don't think we really know how much of their revenue gain was from Ryzen, which, if it is responsible for the strong revenue numbers, would represent a fundamental strengthening of the company, and how much was because of cryptocurrency, which is most likely just a one or two quarter event.zepi - Wednesday, July 26, 2017 - link
I think it is on purpose to make it difficult for competitors to know how well they are doing and to guess their margins and costs of production.Yojimbo - Wednesday, July 26, 2017 - link
I don't think it has anything to do with their competitors. AMD's competitors know how AMD are doing insofar that the competitors know how they themselves are doing in the market as a whole, which is what is important, anyway. It's entirely to do with investors, particularly smaller investors who can't afford the detailed research reports. But just a general obfuscation or delay in the resolution of the true state of their business seems to be AMD's goal.That seems to be their tactic with the strength of their upcoming graphics architectures , too. When did we start hearing about Vega from AMD? It was before Polaris came out. And we heard about all the great improvements Vega would brings and how it would return AMD to a strong position in the high end of the GPU market and how it would compete in the data center. They put it on a graphic situated at the beginning of 2017. They hyped it up, even showing it off in careful demos in November or December of 2016. Then they finally said it would launch in Q2 of 2017. Now here it is, finally launching in August of 2017, smack in the middle of Q3, not looking like it's going to live up to their hard promises much less their hype. How long until they start talking about Navi?
MonkeyPaw - Tuesday, July 25, 2017 - link
The next step is some sort of Ryzen APU to go in laptops. AMD already has design wins for the last 4 MS and Sony consoles, and I expect that dominance to continue since they can produce the most powerful APU with their IP. So far, Ryzen, EPYC and Threadripper do nothing for the everyday consumer looking for a mobile offering.Droblesa - Tuesday, July 25, 2017 - link
The next APU is code Named Raven Ridge. It will have 1 Ryzen CCX and 11 Vega Cores. It will be interesting for Laptops come in the 2H of the year.http://wccftech.com/amd-pinnacle-ridge-raven-ridge...
Hurr Durr - Wednesday, July 26, 2017 - link
It`s hardly the dominance worth pursuing when deciding factor is "who can make this piece of crap cheapest".deltaFx2 - Wednesday, July 26, 2017 - link
""who can make this piece of crap cheapest" Why do you think people buy Intel servers instead of IBM? IBM has higher performance than Intel over a wide swathe of workloads. Intel makes them cheaper and more power efficient. That's largely why. That's why x86 dominates the server landscape and not RISC.And, AMD does integrated GPUs way better than Intel. It's a no-contest there. The cores are now only marginally better than AMDs. Most of the PC TAM is in laptops, so the Ryzen APU is where the money is even if the margins aren't what they once were.
Hurr Durr - Wednesday, July 26, 2017 - link
Marginally as in one core in my five year old 3570 is dangerously close to one core in R7? Okay sure, AMD was always a second-tier solution, so no surprises here, and it was _especially_ visible in laptops.Oh, and people buy Intel instead of IBM because the latter is literally a shambling zombie led by a womyn. With entirely predictable results.
Cooe - Wednesday, July 26, 2017 - link
Lol I upgraded from a 3570K cranked to the ceiling (4.6GHz) to a Ryzen 7 1700 (4.0GHz) and you are completely talking out your ass. Ryzen is WAAAY faster than Ivy Bridge in just about every way measurable, single-threaded performance per clock (aka IPC) is VASTLY superior (win some, loss some with Broadwell/-E) and with more than 4x the threads and 3x the cache, is more than 2x as fast (!!!!) multi-threaded. For example, despite being clocked a whopping 600MHz faster than my Ryzen, it's single-threaded performance is no better (pretty close to equal), but loses by a factor of around 2.5x in total, multi-threaded performance. That's MASSIVE. My performance in every single CPU intensive thing I do has absolutely skyrocketed, I can multi-task like you're system couldn't even believe (I witnessed my 3570K and it's 4-measly threads get overloaded CONSTANTLY). Plus, I'm on a way more modern platform that will be supported for like 3-4+ years for cheap, swap in CPU upgrades, not 1 then ditched. Heck, even if the only thing you care about is games, even in games where they do preform similarly (which is shrinking and shrinking as 4-threads is quickly becoming a SERIOUS bottleneck), the actual gaming experience on Ryzen is SOOOO far superior it's kinda crazy. This effect has been studying extensively, 4-thread CPU's are rapidly becoming thread constrained in an 8-core console world. This leads to "thread-hitching" (small frame-rate hitchs whenever a games idle-threads and active-threads have to be swapped to/from a CPU core), and much lower fps minimum's, 1% averages, and worse frame-times, even if the frame-rate counter is on average showing a similar number. You would get the same effect going to an i5 to an i7, but the point remains, 4-thread CPU's are a dying breed for gaming, just as they're a dying bread for anyone who does real work with their computers. 4-threads, no matter how fast you clock them, just aren't enough to cut the mustard for any kind of significant, modern workload. But, hey, whatever you have to tell yourself to make your old hardware look better :). (And I'm not hating the hardware, in it's day the 3570K was a great chip, and it's had a remarkable lifespan (along with the 2500K), but the older i5's are seriously starting to become thread-constrained. Unlike the newest i5's can't make up some of that loss with like 15-20% faster IPC (IB->KL), and like 15% faster clocks, though honestly, I wouldn't recommend a 4-thread chip to anyone at this point (for a primary use desktop) no matter how fast the threads are.grefot - Monday, August 7, 2017 - link
Synthetic benchmarks clearly show that even an i5 2500K, over 6 years old at this point, is faster at single and 4-threaded workloads than the top of the line Ryzen SKUs if both are overclocked to the max. AMD made an impressive IPC jump, but they barely reach 3.9/4.0Ghz. Sandy bridge at 4.8 to 5.0Ghz is not a rarity with proper cooling, and 4.5Ghz is possible even with a stock cooler.It's funny how the people most vocal about Ryzen, the gamer fanboys, profit the least from its strengths. They would be much better off buying old i5s on the cheap and overclocking them, saves you from having to spend a fortune on DDR4 as well. Ryzen is much better suited for programmers, CAD artists and other professionals.
TesseractOrion - Wednesday, July 26, 2017 - link
Hurr Durr vomited this crap: "Oh, and people buy Intel instead of IBM because the latter is literally a shambling zombie led by a womyn. With entirely predictable results."What an amazingly thoughtful and unbiased, technically literate critique! He must've gone to Trump University....
deltaFx2 - Wednesday, July 26, 2017 - link
Your 3570 is marginally worse than whatever equivalent 6/7 gen intel CPU too. And the graphics as as third-tier as it always was; third, because it can't go any lower. AMD was bad on the bulldozer architecture and a bulk 28nm process that leaked far more than Finfet. Both of those are history. Zen is especially efficient at the 2-3 GHz range (power starts going off the rails over 3.7GHz). So, yeah, as a package, Ryzen Mobile as a package is *better* than anything Intel has, if they can match or beat power, which seems likely.stockolicious - Saturday, July 29, 2017 - link
"It`s hardly the dominance worth pursuing when deciding factor is "who can make this piece of crap cheapest"."in a way that is a deciding factor - Intels and AMD's products are commodities - they are not brands - their chips go into products and are not products themself. AMD is taking a step into stopping the myth that is the "Intel Brand" - the commercials and slogans wont stop the fact that they make chips and that is it. Nobody cares what kind of ARM chip is in an Ipad - they just want to enjoy their "Juicy fruit" device.
zodiacfml - Wednesday, July 26, 2017 - link
Yep. AMD's APUs couldn't come fast enough. AMD's products has slight edge now over Intel's probably for more than a year in desktops. I wonder when they will release mobile parts as Intel might release Cannonlake at the end of the year.serendip - Wednesday, July 26, 2017 - link
It's a long shot but I'm hoping for an Atom-class Zen APU to show up someday. The market for small, low to mid-end Windows tablet still exists but Intel dropped out of it by killing most Apollo Lake SKUs.Threska - Sunday, July 30, 2017 - link
NASs are a market, especially with them going past just being for storage.Jumangi - Wednesday, July 26, 2017 - link
Hopefully Ryzen based server sales will start to show up as that's where they can make some serious cash.Dionisiaco - Wednesday, July 26, 2017 - link
at amd the accountants be like, go check the printer its seems we are out of red ink again.webdoctors - Wednesday, July 26, 2017 - link
The 2% increase in margins from last year is great news, shame they're still losing money. I expected them to dominate the market in DIYs/desktops since the Ryzen series seems to be great value at mid and high tier segments.The real MVPs are the accountants that are able to keep them afloat while they've bled money the last 5/10? years. The one good thing about cheap credit and 0% interest rates are these borderline companies still being able to stay in business.
HomeworldFound - Thursday, July 27, 2017 - link
It's wonderful to see a competitive alternative, but even better where you really don't have to worry about picking the wrong option.