I hope this does not affect their future in building high performance GPUs. Without AMD, Nvidia has a hard lock on the dedicated graphics market. Of course with Nvidia trying to get developers locked into its technologies so it has better performance through more optimized code, its hard for AMD to compete.
It makes no sense to lose the dGPU side of their business, not unless APUs really sell like hotcakes and they need to shift production to satisfy demand. I'm not seeing any chance of that happening.
If it's their most profitable line and it's only marginally profitable, then it may be better to try and focus on things they can be insanely profitable at in the long run rather than emphasize the short term.
Not saying this is the case now, but just saying there is a universe where a company might sacrifice the sure thing now if that sure thing wasn't making bajillions for something they know the company must have in the future.
Given AMD named professional graphics as one of its new, high-growth business, why would it move away from it? Moreover, AMD has been building graphics IP that is scalable from ultra-low-power SOC to huge, 4K-gaming capable discrete GPUs. I don't see how the reorg can affect this capacity.
Generally high-end GPU's were low volume parts with high margins... However, if you look back over the last several years, those low volume parts have dramatically and I mean dramatically increased in volume, all due to compute, be it bitcoin mining or folding etc'.
They would be insane to throw all of that away and simply not take advantage of that potentially high-profit niche' in the future.
How is Mantle still considered an end product. The damned thing is still in development, nothing is final. If they released the specs now Nvidia would just pick the parts they liked and made their own version (CUDA 5.0 or something). Then we would have two new API's. The fact that you can try it out is just good will.
Also they did say that nothing (except not giving them the code yet) is stopping Nvidia from supporting Mantle on their hardware. We'll see how it turns out. Don't declare your hasty conclusions that they were purposefully vague about.
Mantle isn't going to be forever locked to AMD, it will eventually be opened up and someone like nVidia or Intel can build their Drivers to support it.
However, with that said... Even if Mantle flounders, it's done it's job as it prodded a stick over at Microsoft to bring some low-level API advantages to Direct X 12.
Converesly, TressFX and TrueAudio both AMD technologies are open enough that nVidia if it so wanted, could actually support it.
On the flip side... You just need to take a look at PhysX to see the kind of damaging "lock ins" nVidia can do to the market. I.E. Performance degradation for non-nVidia users.
The problem AMD has right now is that they're not financially healthy. The company has failed to produce an operating cash flow in three of the past four years and posted a negative $204 million this past quarter. The company couldn't generate internal cash even on a $400 million profit in 2010. AMD has lost 90% of its market capitalization since 2006 (a trend pre-dating the financial crisis so that isn't even an excuse), is worth just 1/3 of what Nvidia is and has lost over half of its entire balance sheet equity in four years.
Because of poor financials, AMD is unable to compete on two major fronts - CPU and GPU - effectively. Nvidia alone is outspending AMD by 50% in R&D and when Intel is thrown into the mix which is spending 10 times as much, AMD is just being overwhelmed.
Even if AMD focused purely on the GPU market, they would be overwhelmed by Nvidia's superior financial position. AMD is further cannibalizing its internal capital base to maintain cash for operations and is not investing in new capabilities. AMD is currently on path to be left behind in the GPU and CPU markets so badly they won't be able to catch up.
Another problem with AMD is that they're active in the console space. The key problems with the console space is that the technology required starts at two generations behind the current technology, has extremely low margins and is unable to take advantage of economics of scale and becomes a major burden as the console generation ages. In five years, AMD is going to have to maintain fabrication capability to produce extremely out of date hardware and produce it at margins that can fit within a final selling price of $200, which has to be split between CPU, GPU, mainboard, memory, storage, network controllers and input devices. With AMD locking itself into all three of the consoles as a primary supplier, they're setting themselves up to take large losses the company can barely handle as it is as the generation ages.
Further, AMD has a poorly developed customer development capability. The customer is not you, you're the end consumer, but the B2B component. The major reason why Nvidia is the market leader is a far superior business relationship capability. AMD is rather poor in maintaining key account customers, which allows Nvidia that can now and again sell more expensive but inferior chips without much worry of AMD supplanting them. AMD hasn't been able to convince buyers that their product is a superior value.
AMD is further harming itself by attempting to develop its own custom API, Mantle. These are resources the company can ill afford to produce since they do not have the necessary business connections to convince large volume buyers that Mantle and the AMD chips needed to run them are a good deal. One of the causes behind 3dfx's failure was attempting to push GLide and diverting resources away from producing a quality DirectX experience, which is what Nvidia focused on and ended up driving 3dfx out of the market.
Since AMD is no longer in the financial position to effectively continue competing in the CPU or GPU markets, they'll have to find an avenue where their deteriorated financial position and weak market reputation aren't going to be a hindrance. Since AMD still has a competent and capable engineering staff, they can refocus their competencies into the mobile space and cut ties with GPU and CPU production which the company lacks the resources to continue competing with. AMD is much better than Intel at producing inexpensive and efficient chips, so their competencies are better matched to mobile production.
Maybe that is because You cut out TIME FRAMES from Your forecasts.
But first You acknowledge AMD superior skills (they got ALL consoles for being inferior in design?), and then claim, that they do not have capability to create such skills (crisis started before GCN was made).
If AMD is capable of APU development, why should they chop that into pieces?
(However I do agree on the trend to finding better ways to conduct their R&D, bad ass simulators required for current day dGPU ain't cheap - and not always are enough)
Not sure what you mean about timeframe being off. The most recent quarter was also a large cash loss that was tempered with continued sales of internal assets.
Nvidia was rather public in the fact that they simply chose not to bid on the latest hardware generation of consoles. AMD didn't beat them out on the contract, Nvidia didn't want to play given the low sale price necessary for the unit production. Nvidia couldn't justify the low margins.
AMD is already chopping its capability. With only 2/3 of the investment base of Nvidia, they're attempting to split that between a broader product base than the competitors are. The API might be good, but at the end of the day, they do not have the resources necessary to support API development along with GPU and CPU investment. Besides, AMD's business connection and key account management is atrocious. They'd have to convince hardware resellers to prioritize AMD hardware if Mantle is ever going to be a success. Since Nvidia and Intel are unlikely to license out Mantle on their hardware, AMD has to green field develop a larger customer base.
The financials don't lie. AMD is not healthy and they can't handle all the projects they're currently committed to. Companies can't bleed cash for years and remain in business. I'm still concerned that the company won't survive past 2016 at this rate.
To deflect fanboy nonsense, especially since I doubt you bothered to even look at their quarterly reports (or could understand it for that matter since you find the analysis "strange"), I've been exclusive AMD for my personal built systems since the Winchester chip.
They fall every year with market share, revenues and headcount, so they must constantly restructure itself to match smaller size of business. Soon they reach start-up or garage level business.
So since when was a market cap of 3.26 Billion anywhere near start up business level? They've only restructured once so far, and this would be it technically.
this^^^ Most here seem to be missing the point of your posts and others. They are forgetting what they WERE and how they keep shrinking what they ARE today.
The 3dfx point was spot on. AMD didn't learn from watching 3dfx go down I guess. AMD is forgetting DX/OpenGL for their own API which has ZERO dollars to push it on anyone. It is difficult and COSTLY to woo people to your API while others exist that serve everyone. IF you have the cash to blow, sure its great to add features your competition doesn't have to make you stand out. But when you're broke, and everyone around you is 3B+ rich, you should stop this nonsense. Most of the vidcard/driver problems of the last 2yrs were due to basically the same thing happening with diverted funds to make console custom chips. That money, had it been used on GPU/CPU might have put out something to compete with INTEL again, and also avoid GPU/driver problems on the gpu side. People don't seem to understand what is really going on here.
Also no matter how you slice up the data above, you should be on SOC rev4-5 by now like NV. Again funds from the diversion crap (console/mantle) could have been spent here, getting a modem company bought so they could at some point get into a phone etc...
Quick question: Why would a tech-site report on this kind of management-mutation?
Please don't get me wrong, I highly appreciate the AT special coverage on AMD progress, I believe in progress by competition, and I understand that communicating the technological considerations of all competitors even in the face of a commercial domination of a single competitor is in the best interest of the overall tech community and the average customer.
I mean it, I enjoy the equal opportunity communications going on at AT, even when it seems asymmetric sometimes at pairing Intel tests versus AMD ideas. But thats all for the better, we all have decades to live and should be interested in tomorrows products almost as much as we are in todays. So go on with all of that, I like it.
But now you go and babble about Business Units? Honestly, how and where will that ever turn up in your benchmarks? Does anybody really believe that an army of thousands of highly educated engineers and scientists will suddenly come up with ground-breaking new ideas because they are now grouped into a new business unit with a new shiny Mastermind?
Maybe it is time for AT to stop being paid by advertisers, and start being paid by the readers? I for one would not mind to give my money directly to an independent source of tech-information, rather than have it trickle down through the commercial departments of the companies I happen to buy stuff from.
Honestly, guys, please think about who your customer is, what your customer cares about. Me, I'd rather be your customer than your product to be sold.
This hardly reads like a press release. It's not like AMD really wants to announce to the world that we won't be able to find out how their individual units are doing, which is a major change; I remember reading Anandtech's reporting on how Nvidia and AMD were doing market share wise. Changing market share isn't going to create innovations, but it's still nice to note.
I agree that an independent tech review website supported by its readers would be a theoretically ideal business model for AT. I just don't see how not having partnerships with vendors would somehow result in more product coverage. AMD's sponsorship with this site CAN lead some to be suspicious of the selection of content like this article, however it isn't like this is a top tier story. Its a pipeline story after all, not a major news article. The title is clearly defined, so I knew what I was getting when I clicked the link. I'm not at all worried about AT thinking that its readers are just a marketing opportunity like some of the other major "tech" websites (looking at you big C.)
To answer your question we report on it because it's important. I believe that it's impossible to fully understand the decisions technology companies make without understanding the financial situations and people behind them. A change in management doesn't mean that AMD is going to suddenly figure out how to change the laws of physics, but it does signal what AMD considers important and what technologies they're going to be investing in for the future.
For example, to understand AMD's current product lineup and why they aren't falling over themselves to try to build a Intel-killer high-end CPU, you have to understand AMD's intentions to transition towards a smaller company that is focused on markets that AMD considers high growth, such as semi-custom silicon and dense servers. That is a management directed change that has directly impacted the kind of products AMD chooses to develop, and knowing something about Rory and Lisa along with AMD's financial situation helps to understand why AMD has gone that direction.
Similarly, AMD's decision on how to organize their business units indicates just what AMD considers important and how AMD expects their revenue to be aligned. The fact that they've gone from dividing the business between the GPU and CPU to dividing things up between core technologies and fringe technologies indicates that AMD expects a significant portion of their revenue to come from the latter. Which is to say that direct sales of products like Radeon and Athlon are no longer going to be the vast majority of AMD's business as they once were. This is immaterial to how many frames per second AMD's next GPU gets at 2160p, but it's material to understanding why AMD builds a GPU so powerful and charges that price, etc.
At the end of the day just looking at benchmarks and dissecting pipelines isn't enough; it tells us what happened but it doesn't tell us why. Understanding the business, the people in it, and the goals is the final piece of the puzzle that makes our understanding of these products complete. And in turn is what allows us to write better articles and offer better advice (do we suggest a CPU if we know the platform is at a dead end? etc).
As for advertising, that's above my pay grade. Advertising is handled by a 3rd party and whether to rely upon it or not for funding is something that's up to Anand. I just write about the things I'm interested in and collect a paycheck for it.
So why not report on something like Grid and K1 winning awards at computex (highly sought after gold award for K1)? With grid probably because over the last 3 quarters they went from 100 testing grid, then 200 testing grid, and this last Q now having 600 major companies testing grid. Interest like that doesn't make the news here...No need for an Nvidia portal and understanding NV products that are gaining some massive interest (especially grid, and car tech for K1)?. Where is the Nvidia portal? Where is the shield update coverage which added a LOT of features, causing tomshardware etc to re-review it and test the new features. They have the largest gpu market share in discrete/workstations buy far but you ignore them, grid info etc. I just went through the 3 pages of computex stories, and nothing on Nvidia. Nothing on Supermicro pitching grid. etc. Made it to tons of other booths, nothing on NV. IT's like they don't exist on anandtech...LOL. http://www.brightsideofnews.com/2014/06/02/nvidia-... NVIDIA Gamers day at computex...Nothing from anandtech...WOW. They gave away a Titan-Z, Shields, Tegra Note7's. http://blogs.nvidia.com/blog/2014/06/05/supermicro... Supermicro grid info from computex I digress...
Honestly, the K1 isn't interesting because everyone got burned by Tegra 4. AMD has been far more open about failures than Nvidia has.
Nvidia ticks the most boxes... but that doesn't mean they're the most innovative or interesting. Maxwell is interesting, but it's not like we got to see anything. If Maxwell had been showing we'd have seen something. Titan-Z is overpriced, Shields are a flop, and the TNote7 loses to Snap 800 devices.
Sorry about that.
Oh, and I'm waiting for Maxwell coverage so don't count me biased. I'm in the Nvidia boat.
No only announced. You can't SHIP anything without having the chips in your hands for 4.5-6 months. So I don't understand your statement. Nothing will really be announced until AFTER shield2, and they know how many they can start shipping to others. If you were making ANY products, would you announce them before finding out if you can even get enough for a product launch? NO. That just ends up pissing off customers stuck on backorder. Everyone works like this.
You do realize they won the shows awards right? You may not be interested but people like google etc are. Car makers are etc. We will see how K1 does, but many devices are coming. They are just shipping chips now probably if that as most will go into shield 2 first as they did last year to allow ramp before giving them to bigger customers. Shield is a great way to advertise the chip's qualities during ramp. Shield R1 was july 31st device so not sure why you draw conclusions when likely nobody is getting them. Tell me this again at xmas when the OEM's have had the 4-6 months required to put out a product. I expect them to tell us more at Google I/O shortly and maybe even a gaming tablet announcement by google/nv. I'm sure shield 2 will be all over the floor just like last year front and center (only Anandtech ignored this...as usual).
Not sure how anyone can call shield a flop, they only have to sell 100K to profit, and they don't need to profit for a few years just not lose on it. IT was $10mil to dev, just like grid. The best way to spend money IMHO being so cheap to dev. They just drop in a new chip each year and keep adding more games. Have you been to tegrazone or seen all the console-like stuff on google's play store recently? They are adding massively and quality is on the up. It's not angry birds today, far from it. Have you seen the reviews at amazon etc. 4-4.5 stars, no whiners. Is there any other chip company pushing android so much? This is the first rev that will get deskop gpu's into mobile and people are starting to notice as everyone knows what a kepler is if they're involved in gaming at all. It will keep growing.
They rev the Tegra tabs yearly now. K1 already shown to "thrash" S800,801 and at worst tie or better S805. http://www.evolife.cn/html/2014/77075.html Just one of the articles. They've already benchmarked denver (not out until xmas), benchmarks are out for shield 2, mipad, thinkvision aio etc. As the gpu gains as the #1 reason you buy mobile instead of the last decade where it was the modem (even anand said it's not about the modem not long ago), we'll see NV take over some sales. We spend ~70% of our time on mobile playing games, and 40-65% of sales at the 4 major stores (google, apple, MS, amazon) are games. When you're pushing games and already have 20yrs+ experience with devs, drivers for games etc. who wins? :) If AMD gets in this race soon I think they can claw their way back up also, but they better hurry. The rest are starting to wake up to the fact that it's all about games (samsung ships gamepads now with top devices, their first demos were with GAMES...LOL), qcom hired 30 devs to show what they can do with their gpus (nothing from these guys yet, but they just hired them early last year IIRC) etc. GDC2013 shows devs massively left consoles for games on mobile/PC (both ~51%, everything else FAR behind as the survey showed from 2013 and 2014 shows).
T4 had no products out there getting benched or announced before shield. But we know K1 is already in devices getting benched. Denver will be inhouse also, unlike Qcom S810 so a gain on battery should be expected that at the least not to mention we already see the dual 2.5ghz matching the quad A15 K1 (2.3ghz). Definitely showing tablet to tablet K1 quad vs. the K1 dual is a very impressive showing for the NV custom in house cpu design. Qcom will be a year late to Custom party this time since they screwed up. So NV joins the custom chips this year (was apple swift, qcom krait), while Qcom drops the ball and goes off the shelf. oops.. I don't see how it's shaping up to be a bad next 12 months for K1 and expect more announcements of products once shield hits next month. As long as they're in xmas products I don't care about much else ;) It should do fairly well and denver will add to that.
Nvidia basically quitting the consumer SOC market with Tegra and taking their ball to embedded devices should tell you all about the state of that particular part of their business.
Well they did say in the slide bleeding edge hurt them and that was on the bad side replaced by language that sure sounds like ZERO bleeding edge chips will come.
Morawka was referring to AMD giving up on performance per watt, but the slide clearly indicates that they're not.
They're admitting that they cannot hope to be on the same process node as Intel, so they have to focus on getting the best out of what they have. The real telling statement was moving away from high performance products with an emphasis on cores, which - to me - would mean the more agile tactic of creating products with a required number of compute units for a given purpose. AMD have already said that's one of their goals. If you can make more effective use of the hardware on offer, it's cheaper - not to mention far less complex - to reach a specific performance level, meaning if HSA can get traction, it can help mask their performance deficit and help to reduce power usage over more conventional designs.
You'd rather bet on software tricks, while the other guys have the HARDWARE & software tricks. You'll lose as they have been showing year after year. R&D dropping, 6Billion in losses in the last 10yrs, be lucky to break even for the year this year etc. HSA will take years before it's of any financial value. It took Cuda years too (and billions of funding to get it to where it is today) and cuda6+maxwell gets much of the same benefits. Not to mention pascal etc coming with everything on the PCB with super highspeed links for sharing crap. To me moving away from high perf products just means you're shooting for 2nd/3rd because you can't compete for #1 in anything. That's not something I like seeing.
Process node? So they're admitting something that we've known for a decade? Wow, news. I don't think that's what they meant. Everyone isn't on Intel's node (yet). That's news?
I get what you're saying, but it reads like a company looking in every direction trying to figure out how to compete. I smell desperation all over this.
I guess AMD really hasn't been competitive in the CPU market for some time, so getting out of a race that they were losing makes sense.
The unfortunate side effect to that though is that Intel will continue to set their own pace unchallenged, and so far they've shown that sometimes the pace they select is even too slow for their own good.
yep, current intel cpu's are barely even much faster than sandybridge. without competition they just sit on their hands and slowly iterate, while keeping prices high.
This claim is so silky in the face of the facts, one has to wonder how much attention you have paid to the CPU market of the past decade.
Intel just introduced a 500 MHz bump for free.
Prices have been stable since the introduction of Conroe 8 years ago, when Intel obliterated AMD by introducing a $320 CPU that outperformed any AMD part, and used a lot less energy to do it, too.
Which is in shark contrast to what AMD did shortly after gaining the CPU crown: they raised prices from $700 to $800 to over $1000 while holding performance back with only 600 extra MHz over the course of 2 2/3rds years.
Interesting points, but with regard to your last statement i'd argue that the reason Intel prices aren't higher is simply that they've reached a state of nash equilibrium, in which raising them would not only create a net loss in revenue, but also damage their brand and cause users to start thinking about and researching alternatives.
Intel finds itself in a situation in which most users already have a rig that's fast enough for their needs, and the only way you could get them to maybe, possibly upgrade is if the costs are reasonable (as in more performance for not more than what they're used to paying). Intel can't sell chips if no one wants them, keeping prices stable each generation is a way to keep sales in some sort of equilibrium with their release schedule.
Agreed, FABs need to be used to be profitable, but I doubt there would have been a meaningful impact on Intel's utilization if they had offered the 4790K at $5xx rather than $3xx.
That they did not, and have not after years without meaningful competition in that space, should be plenty of answer to the 'Intel's keeping the prices high' crowd.
If they released the 4790k at $5xx, nobody would buy it because they could just grab a 4770k for $330, and have performance just a little ways behind, yet they'd save $200 for other parts and peripherals.
If they had released the 4770k at $5xx, people would have bought the 3770k instead, at $330, same situation as the 4790k.
Intel actually does keep its price higher. How do you think they get a profit margin so high? They charge more for individual units, as well as charge more for system building companies like HP and Dell.
Intel knows just how to perform this fine balancing act of keeping prices high enough to make massive profits, yet not so high as to discourage current customers from upgrading and potential customers from taking some of their pricier, more high-end products.
It costs more money for AMD to make their FX processors, yet intel charges much more for their processors.
The around-the-corner i7-5960x is to be sold at $1000. The FX-9590 is sold at $350ish.
A big problem for AMD is brand recognition. Many of my clients have no idea who AMD is, but can clearly identify Intel. When i hear about "Intel Marketing" I think of the Blue Man Group commercials, when I hear AMD Marketing, all i can think of it tech news releases. Intel made themselves a household name and AMD was relegated to being known as a cheaper alternative to Intel, like an off-brand. Even in the days where they had superior CPU's, they were only really known as an Intel alternative. They really need to find a way to make themselves better known. I have never seen a AMD commercial of any kind. I was hoping AMD was just "playing possum" and would release a great CPU to send Intel reeling, but with this, I can't see that ever happening.
no one cares. AMD are too far behind intel on CPU tech they may as well just become a GPU only company. AMD did a PR thing about going to space, and instead of getting good positive thoughts about it. every comment and post was telling them to get back to work! AMD has a cult following because its early days they really did hurt big CPU and ppl loved them for it. now they do nothing 4 the investors but cost us money
AMD was ahead when they used Simultaneous Multi-Threading (SMT) architectures. (Intel still does). Both dealt with the same problems and advantages, but AMD was just much better at manipulating SMT to get maximum power. They designed their architectures to be able to run circles around intel. Then, with Bulldozer, they tried to run even bigger circles around intel, by creating a Clustered Multi-Threading (CMT, or modular) architecture. 1 module was 2 out-of-order cores that shared early pipeline stages. Each one had different independent resources (cache, etc.) Basically, with Sandybridge coming up AMD needed to fight the emerging hyper-threaded threat. So they did the opposite, hoping it would speed up processing. They made it so that each core in each module was single-threaded, and together, 1 module = 2 threads. It was horribly inefficient. The modules took up a lot of die space, and needed a lot of power to satisfy their need to supply each module. They were absolutely crushed by multithreaded performance, and single because a single-threaded app would be split apart in the early pipeline, but a multi-threaded app would be condensed in the later pipeline stages.
AMD recently announced that excavator (2016) would be SMT. Fingers crossed for another big one from the comeback king.
On top of that, looking at intel's release patterns, we see this:
Sandybridge (Q1 2011) Ivybridge (Q3 2011)
Haswell (Q3 2013) Haswell refresh (Q2 2014)
Broadwell (Q1 2015) Skylake (2016 or 2017???)
Their gap is widening, and while tick-tock says Skylake should be 2016, intel hasn't used tick-tock for long enough for us to be able to predict everything, and if you look at the widening gap, it started half a year apart, now it's a year apart, so it could be 2 years apart.
If this was a generic situation you would be correct; however, AMD doesn't own the x86 instruction set, they license it from Intel. The cross-licensing agreement doesn't permit AMD to license the ISA to anyone else thereby requiring them to sell the chips to Microsoft and Sony. AMD makes a profit selling the chips and not royalities as is the norm.
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
51 Comments
Back to Article
FITCamaro - Friday, June 13, 2014 - link
I hope this does not affect their future in building high performance GPUs. Without AMD, Nvidia has a hard lock on the dedicated graphics market. Of course with Nvidia trying to get developers locked into its technologies so it has better performance through more optimized code, its hard for AMD to compete.jimjamjamie - Friday, June 13, 2014 - link
This all seems to be focused on CPUs so I can only presume the GPU roadmap is business as usual.Wreckage - Friday, June 13, 2014 - link
It is possible that AMD is slowly moving away from the discrete GPU market.Death666Angel - Friday, June 13, 2014 - link
Wasn't that one of the most profitable parts of their last business year, though?silverblue - Friday, June 13, 2014 - link
It makes no sense to lose the dGPU side of their business, not unless APUs really sell like hotcakes and they need to shift production to satisfy demand. I'm not seeing any chance of that happening.HisDivineOrder - Monday, June 16, 2014 - link
If it's their most profitable line and it's only marginally profitable, then it may be better to try and focus on things they can be insanely profitable at in the long run rather than emphasize the short term.Not saying this is the case now, but just saying there is a universe where a company might sacrifice the sure thing now if that sure thing wasn't making bajillions for something they know the company must have in the future.
silverblue - Monday, June 16, 2014 - link
It's just as well that NVIDIA didn't bet the farm on Tegra.tipoo - Friday, June 13, 2014 - link
Move away from one of their biggest, most profitable cash flows? Why? This would imply just the opposite. GPUs are already high revenue for them.pTmdfx - Saturday, June 14, 2014 - link
Given AMD named professional graphics as one of its new, high-growth business, why would it move away from it? Moreover, AMD has been building graphics IP that is scalable from ultra-low-power SOC to huge, 4K-gaming capable discrete GPUs. I don't see how the reorg can affect this capacity.StevoLincolnite - Wednesday, June 18, 2014 - link
Generally high-end GPU's were low volume parts with high margins...However, if you look back over the last several years, those low volume parts have dramatically and I mean dramatically increased in volume, all due to compute, be it bitcoin mining or folding etc'.
They would be insane to throw all of that away and simply not take advantage of that potentially high-profit niche' in the future.
Dribble - Friday, June 13, 2014 - link
Other then by AMD getting developers locked into there technologies (mantle) so they have better performance through optimized code.SleepyFE - Friday, June 13, 2014 - link
How is Mantle still considered an end product. The damned thing is still in development, nothing is final. If they released the specs now Nvidia would just pick the parts they liked and made their own version (CUDA 5.0 or something). Then we would have two new API's. The fact that you can try it out is just good will.Also they did say that nothing (except not giving them the code yet) is stopping Nvidia from supporting Mantle on their hardware. We'll see how it turns out. Don't declare your hasty conclusions that they were purposefully vague about.
StevoLincolnite - Wednesday, June 18, 2014 - link
Mantle isn't going to be forever locked to AMD, it will eventually be opened up and someone like nVidia or Intel can build their Drivers to support it.However, with that said... Even if Mantle flounders, it's done it's job as it prodded a stick over at Microsoft to bring some low-level API advantages to Direct X 12.
Converesly, TressFX and TrueAudio both AMD technologies are open enough that nVidia if it so wanted, could actually support it.
On the flip side... You just need to take a look at PhysX to see the kind of damaging "lock ins" nVidia can do to the market. I.E. Performance degradation for non-nVidia users.
LaughingTarget - Saturday, June 14, 2014 - link
The problem AMD has right now is that they're not financially healthy. The company has failed to produce an operating cash flow in three of the past four years and posted a negative $204 million this past quarter. The company couldn't generate internal cash even on a $400 million profit in 2010. AMD has lost 90% of its market capitalization since 2006 (a trend pre-dating the financial crisis so that isn't even an excuse), is worth just 1/3 of what Nvidia is and has lost over half of its entire balance sheet equity in four years.Because of poor financials, AMD is unable to compete on two major fronts - CPU and GPU - effectively. Nvidia alone is outspending AMD by 50% in R&D and when Intel is thrown into the mix which is spending 10 times as much, AMD is just being overwhelmed.
Even if AMD focused purely on the GPU market, they would be overwhelmed by Nvidia's superior financial position. AMD is further cannibalizing its internal capital base to maintain cash for operations and is not investing in new capabilities. AMD is currently on path to be left behind in the GPU and CPU markets so badly they won't be able to catch up.
Another problem with AMD is that they're active in the console space. The key problems with the console space is that the technology required starts at two generations behind the current technology, has extremely low margins and is unable to take advantage of economics of scale and becomes a major burden as the console generation ages. In five years, AMD is going to have to maintain fabrication capability to produce extremely out of date hardware and produce it at margins that can fit within a final selling price of $200, which has to be split between CPU, GPU, mainboard, memory, storage, network controllers and input devices. With AMD locking itself into all three of the consoles as a primary supplier, they're setting themselves up to take large losses the company can barely handle as it is as the generation ages.
Further, AMD has a poorly developed customer development capability. The customer is not you, you're the end consumer, but the B2B component. The major reason why Nvidia is the market leader is a far superior business relationship capability. AMD is rather poor in maintaining key account customers, which allows Nvidia that can now and again sell more expensive but inferior chips without much worry of AMD supplanting them. AMD hasn't been able to convince buyers that their product is a superior value.
AMD is further harming itself by attempting to develop its own custom API, Mantle. These are resources the company can ill afford to produce since they do not have the necessary business connections to convince large volume buyers that Mantle and the AMD chips needed to run them are a good deal. One of the causes behind 3dfx's failure was attempting to push GLide and diverting resources away from producing a quality DirectX experience, which is what Nvidia focused on and ended up driving 3dfx out of the market.
Since AMD is no longer in the financial position to effectively continue competing in the CPU or GPU markets, they'll have to find an avenue where their deteriorated financial position and weak market reputation aren't going to be a hindrance. Since AMD still has a competent and capable engineering staff, they can refocus their competencies into the mobile space and cut ties with GPU and CPU production which the company lacks the resources to continue competing with. AMD is much better than Intel at producing inexpensive and efficient chips, so their competencies are better matched to mobile production.
przemo_li - Tuesday, June 17, 2014 - link
Really, really, really strange post.Maybe that is because You cut out TIME FRAMES from Your forecasts.
But first You acknowledge AMD superior skills (they got ALL consoles for being inferior in design?), and then claim, that they do not have capability to create such skills (crisis started before GCN was made).
If AMD is capable of APU development, why should they chop that into pieces?
(However I do agree on the trend to finding better ways to conduct their R&D, bad ass simulators required for current day dGPU ain't cheap - and not always are enough)
LaughingTarget - Friday, June 27, 2014 - link
Not sure what you mean about timeframe being off. The most recent quarter was also a large cash loss that was tempered with continued sales of internal assets.Nvidia was rather public in the fact that they simply chose not to bid on the latest hardware generation of consoles. AMD didn't beat them out on the contract, Nvidia didn't want to play given the low sale price necessary for the unit production. Nvidia couldn't justify the low margins.
AMD is already chopping its capability. With only 2/3 of the investment base of Nvidia, they're attempting to split that between a broader product base than the competitors are. The API might be good, but at the end of the day, they do not have the resources necessary to support API development along with GPU and CPU investment. Besides, AMD's business connection and key account management is atrocious. They'd have to convince hardware resellers to prioritize AMD hardware if Mantle is ever going to be a success. Since Nvidia and Intel are unlikely to license out Mantle on their hardware, AMD has to green field develop a larger customer base.
The financials don't lie. AMD is not healthy and they can't handle all the projects they're currently committed to. Companies can't bleed cash for years and remain in business. I'm still concerned that the company won't survive past 2016 at this rate.
To deflect fanboy nonsense, especially since I doubt you bothered to even look at their quarterly reports (or could understand it for that matter since you find the analysis "strange"), I've been exclusive AMD for my personal built systems since the Winchester chip.
Zoomer - Tuesday, June 17, 2014 - link
Didn't realize AMD owns all the console manufacturers.-> Console manufacturers usually license the design and then make the part themselves.
TristanSDX - Friday, June 13, 2014 - link
They fall every year with market share, revenues and headcount, so they must constantly restructure itself to match smaller size of business. Soon they reach start-up or garage level business.ZandalarZandali - Friday, June 13, 2014 - link
So since when was a market cap of 3.26 Billion anywhere near start up business level? They've only restructured once so far, and this would be it technically.Wreckage - Friday, June 13, 2014 - link
That's about what a headphone company just sold for.lmcd - Friday, June 13, 2014 - link
Yeah and the selling price was over their value, since all deals end up like that. AMD if purchased would probably cost closer to 3.5 Billion.LaughingTarget - Saturday, June 14, 2014 - link
AMD had a market cap of $30 billion in 2006.TheJian - Sunday, June 15, 2014 - link
this^^^ Most here seem to be missing the point of your posts and others. They are forgetting what they WERE and how they keep shrinking what they ARE today.The 3dfx point was spot on. AMD didn't learn from watching 3dfx go down I guess. AMD is forgetting DX/OpenGL for their own API which has ZERO dollars to push it on anyone. It is difficult and COSTLY to woo people to your API while others exist that serve everyone. IF you have the cash to blow, sure its great to add features your competition doesn't have to make you stand out. But when you're broke, and everyone around you is 3B+ rich, you should stop this nonsense. Most of the vidcard/driver problems of the last 2yrs were due to basically the same thing happening with diverted funds to make console custom chips. That money, had it been used on GPU/CPU might have put out something to compete with INTEL again, and also avoid GPU/driver problems on the gpu side. People don't seem to understand what is really going on here.
Also no matter how you slice up the data above, you should be on SOC rev4-5 by now like NV. Again funds from the diversion crap (console/mantle) could have been spent here, getting a modem company bought so they could at some point get into a phone etc...
ShieTar - Friday, June 13, 2014 - link
Quick question: Why would a tech-site report on this kind of management-mutation?Please don't get me wrong, I highly appreciate the AT special coverage on AMD progress, I believe in progress by competition, and I understand that communicating the technological considerations of all competitors even in the face of a commercial domination of a single competitor is in the best interest of the overall tech community and the average customer.
I mean it, I enjoy the equal opportunity communications going on at AT, even when it seems asymmetric sometimes at pairing Intel tests versus AMD ideas. But thats all for the better, we all have decades to live and should be interested in tomorrows products almost as much as we are in todays. So go on with all of that, I like it.
But now you go and babble about Business Units? Honestly, how and where will that ever turn up in your benchmarks? Does anybody really believe that an army of thousands of highly educated engineers and scientists will suddenly come up with ground-breaking new ideas because they are now grouped into a new business unit with a new shiny Mastermind?
Maybe it is time for AT to stop being paid by advertisers, and start being paid by the readers? I for one would not mind to give my money directly to an independent source of tech-information, rather than have it trickle down through the commercial departments of the companies I happen to buy stuff from.
Honestly, guys, please think about who your customer is, what your customer cares about. Me, I'd rather be your customer than your product to be sold.
Egg - Friday, June 13, 2014 - link
This hardly reads like a press release. It's not like AMD really wants to announce to the world that we won't be able to find out how their individual units are doing, which is a major change; I remember reading Anandtech's reporting on how Nvidia and AMD were doing market share wise. Changing market share isn't going to create innovations, but it's still nice to note.HigherState - Friday, June 13, 2014 - link
I agree that an independent tech review website supported by its readers would be a theoretically ideal business model for AT. I just don't see how not having partnerships with vendors would somehow result in more product coverage. AMD's sponsorship with this site CAN lead some to be suspicious of the selection of content like this article, however it isn't like this is a top tier story. Its a pipeline story after all, not a major news article. The title is clearly defined, so I knew what I was getting when I clicked the link. I'm not at all worried about AT thinking that its readers are just a marketing opportunity like some of the other major "tech" websites (looking at you big C.)Ryan Smith - Friday, June 13, 2014 - link
Hi ShieTar;To answer your question we report on it because it's important. I believe that it's impossible to fully understand the decisions technology companies make without understanding the financial situations and people behind them. A change in management doesn't mean that AMD is going to suddenly figure out how to change the laws of physics, but it does signal what AMD considers important and what technologies they're going to be investing in for the future.
For example, to understand AMD's current product lineup and why they aren't falling over themselves to try to build a Intel-killer high-end CPU, you have to understand AMD's intentions to transition towards a smaller company that is focused on markets that AMD considers high growth, such as semi-custom silicon and dense servers. That is a management directed change that has directly impacted the kind of products AMD chooses to develop, and knowing something about Rory and Lisa along with AMD's financial situation helps to understand why AMD has gone that direction.
http://www.anandtech.com/show/5503/understanding-a...
http://www.anandtech.com/show/7281/understanding-a...
Similarly, AMD's decision on how to organize their business units indicates just what AMD considers important and how AMD expects their revenue to be aligned. The fact that they've gone from dividing the business between the GPU and CPU to dividing things up between core technologies and fringe technologies indicates that AMD expects a significant portion of their revenue to come from the latter. Which is to say that direct sales of products like Radeon and Athlon are no longer going to be the vast majority of AMD's business as they once were. This is immaterial to how many frames per second AMD's next GPU gets at 2160p, but it's material to understanding why AMD builds a GPU so powerful and charges that price, etc.
At the end of the day just looking at benchmarks and dissecting pipelines isn't enough; it tells us what happened but it doesn't tell us why. Understanding the business, the people in it, and the goals is the final piece of the puzzle that makes our understanding of these products complete. And in turn is what allows us to write better articles and offer better advice (do we suggest a CPU if we know the platform is at a dead end? etc).
As for advertising, that's above my pay grade. Advertising is handled by a 3rd party and whether to rely upon it or not for funding is something that's up to Anand. I just write about the things I'm interested in and collect a paycheck for it.
TheJian - Friday, June 13, 2014 - link
So why not report on something like Grid and K1 winning awards at computex (highly sought after gold award for K1)? With grid probably because over the last 3 quarters they went from 100 testing grid, then 200 testing grid, and this last Q now having 600 major companies testing grid. Interest like that doesn't make the news here...No need for an Nvidia portal and understanding NV products that are gaining some massive interest (especially grid, and car tech for K1)?. Where is the Nvidia portal? Where is the shield update coverage which added a LOT of features, causing tomshardware etc to re-review it and test the new features. They have the largest gpu market share in discrete/workstations buy far but you ignore them, grid info etc. I just went through the 3 pages of computex stories, and nothing on Nvidia. Nothing on Supermicro pitching grid. etc. Made it to tons of other booths, nothing on NV. IT's like they don't exist on anandtech...LOL.http://www.brightsideofnews.com/2014/06/02/nvidia-...
NVIDIA Gamers day at computex...Nothing from anandtech...WOW. They gave away a Titan-Z, Shields, Tegra Note7's.
http://blogs.nvidia.com/blog/2014/06/05/supermicro...
Supermicro grid info from computex
I digress...
lmcd - Friday, June 13, 2014 - link
Honestly, the K1 isn't interesting because everyone got burned by Tegra 4. AMD has been far more open about failures than Nvidia has.Nvidia ticks the most boxes... but that doesn't mean they're the most innovative or interesting. Maxwell is interesting, but it's not like we got to see anything. If Maxwell had been showing we'd have seen something. Titan-Z is overpriced, Shields are a flop, and the TNote7 loses to Snap 800 devices.
Sorry about that.
Oh, and I'm waiting for Maxwell coverage so don't count me biased. I'm in the Nvidia boat.
Egg - Saturday, June 14, 2014 - link
I'm with lmcd... what's the only shipping K1 product? http://www.anandtech.com/show/8022/xiaomi-announce... this?TheJian - Sunday, June 15, 2014 - link
No only announced. You can't SHIP anything without having the chips in your hands for 4.5-6 months. So I don't understand your statement. Nothing will really be announced until AFTER shield2, and they know how many they can start shipping to others. If you were making ANY products, would you announce them before finding out if you can even get enough for a product launch? NO. That just ends up pissing off customers stuck on backorder. Everyone works like this.TheJian - Sunday, June 15, 2014 - link
You do realize they won the shows awards right? You may not be interested but people like google etc are. Car makers are etc. We will see how K1 does, but many devices are coming. They are just shipping chips now probably if that as most will go into shield 2 first as they did last year to allow ramp before giving them to bigger customers. Shield is a great way to advertise the chip's qualities during ramp. Shield R1 was july 31st device so not sure why you draw conclusions when likely nobody is getting them. Tell me this again at xmas when the OEM's have had the 4-6 months required to put out a product. I expect them to tell us more at Google I/O shortly and maybe even a gaming tablet announcement by google/nv. I'm sure shield 2 will be all over the floor just like last year front and center (only Anandtech ignored this...as usual).Not sure how anyone can call shield a flop, they only have to sell 100K to profit, and they don't need to profit for a few years just not lose on it. IT was $10mil to dev, just like grid. The best way to spend money IMHO being so cheap to dev. They just drop in a new chip each year and keep adding more games. Have you been to tegrazone or seen all the console-like stuff on google's play store recently? They are adding massively and quality is on the up. It's not angry birds today, far from it. Have you seen the reviews at amazon etc. 4-4.5 stars, no whiners. Is there any other chip company pushing android so much? This is the first rev that will get deskop gpu's into mobile and people are starting to notice as everyone knows what a kepler is if they're involved in gaming at all. It will keep growing.
They rev the Tegra tabs yearly now. K1 already shown to "thrash" S800,801 and at worst tie or better S805.
http://www.evolife.cn/html/2014/77075.html
Just one of the articles. They've already benchmarked denver (not out until xmas), benchmarks are out for shield 2, mipad, thinkvision aio etc. As the gpu gains as the #1 reason you buy mobile instead of the last decade where it was the modem (even anand said it's not about the modem not long ago), we'll see NV take over some sales. We spend ~70% of our time on mobile playing games, and 40-65% of sales at the 4 major stores (google, apple, MS, amazon) are games. When you're pushing games and already have 20yrs+ experience with devs, drivers for games etc. who wins? :) If AMD gets in this race soon I think they can claw their way back up also, but they better hurry. The rest are starting to wake up to the fact that it's all about games (samsung ships gamepads now with top devices, their first demos were with GAMES...LOL), qcom hired 30 devs to show what they can do with their gpus (nothing from these guys yet, but they just hired them early last year IIRC) etc. GDC2013 shows devs massively left consoles for games on mobile/PC (both ~51%, everything else FAR behind as the survey showed from 2013 and 2014 shows).
T4 had no products out there getting benched or announced before shield. But we know K1 is already in devices getting benched. Denver will be inhouse also, unlike Qcom S810 so a gain on battery should be expected that at the least not to mention we already see the dual 2.5ghz matching the quad A15 K1 (2.3ghz). Definitely showing tablet to tablet K1 quad vs. the K1 dual is a very impressive showing for the NV custom in house cpu design. Qcom will be a year late to Custom party this time since they screwed up. So NV joins the custom chips this year (was apple swift, qcom krait), while Qcom drops the ball and goes off the shelf. oops.. I don't see how it's shaping up to be a bad next 12 months for K1 and expect more announcements of products once shield hits next month. As long as they're in xmas products I don't care about much else ;) It should do fairly well and denver will add to that.
MartinT - Monday, June 16, 2014 - link
Nvidia basically quitting the consumer SOC market with Tegra and taking their ball to embedded devices should tell you all about the state of that particular part of their business.Morawka - Friday, June 13, 2014 - link
So this is it, AMD admitting they are giving up on the race to Desktop CPU and Server CPU performance per watt.silverblue - Friday, June 13, 2014 - link
...except there was nothing of the sort to indicate either.TheJian - Friday, June 13, 2014 - link
Well they did say in the slide bleeding edge hurt them and that was on the bad side replaced by language that sure sounds like ZERO bleeding edge chips will come.silverblue - Saturday, June 14, 2014 - link
Morawka was referring to AMD giving up on performance per watt, but the slide clearly indicates that they're not.They're admitting that they cannot hope to be on the same process node as Intel, so they have to focus on getting the best out of what they have. The real telling statement was moving away from high performance products with an emphasis on cores, which - to me - would mean the more agile tactic of creating products with a required number of compute units for a given purpose. AMD have already said that's one of their goals. If you can make more effective use of the hardware on offer, it's cheaper - not to mention far less complex - to reach a specific performance level, meaning if HSA can get traction, it can help mask their performance deficit and help to reduce power usage over more conventional designs.
TheJian - Sunday, June 15, 2014 - link
You'd rather bet on software tricks, while the other guys have the HARDWARE & software tricks. You'll lose as they have been showing year after year. R&D dropping, 6Billion in losses in the last 10yrs, be lucky to break even for the year this year etc. HSA will take years before it's of any financial value. It took Cuda years too (and billions of funding to get it to where it is today) and cuda6+maxwell gets much of the same benefits. Not to mention pascal etc coming with everything on the PCB with super highspeed links for sharing crap. To me moving away from high perf products just means you're shooting for 2nd/3rd because you can't compete for #1 in anything. That's not something I like seeing.Process node? So they're admitting something that we've known for a decade? Wow, news. I don't think that's what they meant. Everyone isn't on Intel's node (yet). That's news?
I get what you're saying, but it reads like a company looking in every direction trying to figure out how to compete. I smell desperation all over this.
Phartindust - Monday, June 16, 2014 - link
I don't see how you could arrive at that conclusion. You do know that AMD has 14nm tech on the way right?http://www.theregister.co.uk/2014/04/17/samsung_gl...
And that the 295x2 put a smack down on Titan Z:
http://www.pcper.com/reviews/Graphics-Cards/NVIDIA...
Oh, and Richard Huddy left Intel for AMD:
http://www.maximumpc.com/graphics_guru_richard_hud...
C'mon man!
twtech - Saturday, June 14, 2014 - link
I guess AMD really hasn't been competitive in the CPU market for some time, so getting out of a race that they were losing makes sense.The unfortunate side effect to that though is that Intel will continue to set their own pace unchallenged, and so far they've shown that sometimes the pace they select is even too slow for their own good.
bwat47 - Saturday, June 14, 2014 - link
yep, current intel cpu's are barely even much faster than sandybridge. without competition they just sit on their hands and slowly iterate, while keeping prices high.MartinT - Saturday, June 14, 2014 - link
This claim is so silky in the face of the facts, one has to wonder how much attention you have paid to the CPU market of the past decade.Intel just introduced a 500 MHz bump for free.
Prices have been stable since the introduction of Conroe 8 years ago, when Intel obliterated AMD by introducing a $320 CPU that outperformed any AMD part, and used a lot less energy to do it, too.
Which is in shark contrast to what AMD did shortly after gaining the CPU crown: they raised prices from $700 to $800 to over $1000 while holding performance back with only 600 extra MHz over the course of 2 2/3rds years.
alacard - Sunday, June 15, 2014 - link
Interesting points, but with regard to your last statement i'd argue that the reason Intel prices aren't higher is simply that they've reached a state of nash equilibrium, in which raising them would not only create a net loss in revenue, but also damage their brand and cause users to start thinking about and researching alternatives.Intel finds itself in a situation in which most users already have a rig that's fast enough for their needs, and the only way you could get them to maybe, possibly upgrade is if the costs are reasonable (as in more performance for not more than what they're used to paying). Intel can't sell chips if no one wants them, keeping prices stable each generation is a way to keep sales in some sort of equilibrium with their release schedule.
MartinT - Monday, June 16, 2014 - link
Agreed, FABs need to be used to be profitable, but I doubt there would have been a meaningful impact on Intel's utilization if they had offered the 4790K at $5xx rather than $3xx.That they did not, and have not after years without meaningful competition in that space, should be plenty of answer to the 'Intel's keeping the prices high' crowd.
Antronman - Sunday, June 22, 2014 - link
Intel has to compete with it's own products.If they released the 4790k at $5xx, nobody would buy it because they could just grab a 4770k for $330, and have performance just a little ways behind, yet they'd save $200 for other parts and peripherals.
If they had released the 4770k at $5xx, people would have bought the 3770k instead, at $330, same situation as the 4790k.
Intel actually does keep its price higher. How do you think they get a profit margin so high? They charge more for individual units, as well as charge more for system building companies like HP and Dell.
Intel knows just how to perform this fine balancing act of keeping prices high enough to make massive profits, yet not so high as to discourage current customers from upgrading and potential customers from taking some of their pricier, more high-end products.
It costs more money for AMD to make their FX processors, yet intel charges much more for their processors.
The around-the-corner i7-5960x is to be sold at $1000. The FX-9590 is sold at $350ish.
Morg72 - Saturday, June 14, 2014 - link
A big problem for AMD is brand recognition. Many of my clients have no idea who AMD is, but can clearly identify Intel. When i hear about "Intel Marketing" I think of the Blue Man Group commercials, when I hear AMD Marketing, all i can think of it tech news releases. Intel made themselves a household name and AMD was relegated to being known as a cheaper alternative to Intel, like an off-brand. Even in the days where they had superior CPU's, they were only really known as an Intel alternative. They really need to find a way to make themselves better known. I have never seen a AMD commercial of any kind.I was hoping AMD was just "playing possum" and would release a great CPU to send Intel reeling, but with this, I can't see that ever happening.
nunya112 - Sunday, June 15, 2014 - link
no one cares. AMD are too far behind intel on CPU tech they may as well just become a GPU only company.AMD did a PR thing about going to space, and instead of getting good positive thoughts about it. every comment and post was telling them to get back to work! AMD has a cult following because its early days they really did hurt big CPU and ppl loved them for it. now they do nothing 4 the investors but cost us money
Antronman - Sunday, June 22, 2014 - link
Wrong, wrong wrong oh so wrong.Sorry, but you're wrong.
AMD was ahead when they used Simultaneous Multi-Threading (SMT) architectures. (Intel still does). Both dealt with the same problems and advantages, but AMD was just much better at manipulating SMT to get maximum power. They designed their architectures to be able to run circles around intel. Then, with Bulldozer, they tried to run even bigger circles around intel, by creating a Clustered Multi-Threading (CMT, or modular) architecture. 1 module was 2 out-of-order cores that shared early pipeline stages. Each one had different independent resources (cache, etc.) Basically, with Sandybridge coming up AMD needed to fight the emerging hyper-threaded threat. So they did the opposite, hoping it would speed up processing. They made it so that each core in each module was single-threaded, and together, 1 module = 2 threads. It was horribly inefficient. The modules took up a lot of die space, and needed a lot of power to satisfy their need to supply each module. They were absolutely crushed by multithreaded performance, and single because a single-threaded app would be split apart in the early pipeline, but a multi-threaded app would be condensed in the later pipeline stages.
AMD recently announced that excavator (2016) would be SMT. Fingers crossed for another big one from the comeback king.
On top of that, looking at intel's release patterns, we see this:
Sandybridge (Q1 2011)
Ivybridge (Q3 2011)
Haswell (Q3 2013)
Haswell refresh (Q2 2014)
Broadwell (Q1 2015)
Skylake (2016 or 2017???)
Their gap is widening, and while tick-tock says Skylake should be 2016, intel hasn't used tick-tock for long enough for us to be able to predict everything, and if you look at the widening gap, it started half a year apart, now it's a year apart, so it could be 2 years apart.
WC20 - Wednesday, June 25, 2014 - link
If this was a generic situation you would be correct; however, AMD doesn't own the x86 instruction set, they license it from Intel. The cross-licensing agreement doesn't permit AMD to license the ISA to anyone else thereby requiring them to sell the chips to Microsoft and Sony. AMD makes a profit selling the chips and not royalities as is the norm.WC20 - Wednesday, June 25, 2014 - link
Oops...that reply didn't "reply" to the correct post.arturp - Saturday, February 20, 2016 - link
Going away from High Performance Computing - it really sounds like a plan how to kill company.