Slashdot Mirror
Is Qualcomm the New AMD?
colinneagle writes "It's a darned shame, but the writing is on the wall for AMD. The ATI graphics business is the only thing keeping it afloat right now as sales shrivel up and the company faces yet another round of staffing cuts. You can only cut so many times before there's no one left to innovate you out of the mess you're in. Qualcomm, on the other hand, dominates this space, and it has the chips to back it up. The Snapdragon line of ARM-based processors alone is found in a ridiculous number of prominent devices, including Samsung Galaxy S II and S III, Nokia Lumia 900 and 920, Asus Transformer Pad Infinity and the Samsung Galaxy Note. Mind you, Samsung is also in the ARM processor business, yet it is licensing Qualcomm's parts. That's quite a statement."
Samsung is licensing the SoCs for the US market only. The flagship products (Galaxy S II,III and Note) are all using Exynos for every other market.
...can be answered with a "no"
Intel is already years ahead of AMD. They have well over 80% market share in the PC market and over 90% in the server and workstation market. There's a large performance spread between AMDs processors and Intels processors in both single threaded performance and overall performance per watt. If Intel wants to bend consumers over, they are already in a position to do so. However, they seem to be sticking to their roadmap despite the fact that AMD has been falling farther and farther behind.
Qualcomm manufacture ARM chips, like a dozen other companies, there is nothing special about them.
This is explicitly false. Qualcomm designed their own cores that implement the ARM instruction set. They did not license the Cortex A-x designs and glue them together (like every other ARM SoC vendor, including Samsung.) That also ignores the fact that they are the only ones making usable LTE basebands right now. Qualcomm right now is so dominant that if anything, they're the Intel of the mobile world.
If Intel wants to bend consumers over, they are already in a position to do so. However, they seem to be sticking to their roadmap despite the fact that AMD has been falling farther and farther behind.
Have you looked at Intel CPU prices lately? It hasn't been this bad since the Pentium II times. I would also point out that there are no Ivy Bridge server processors available, nor is their 6 core processor based on Ivy Bridge despite the first Ivy Bridge processors coming out a long time ago.
Have you looked at Intel CPU prices lately?
Yes. A high-end i7 costs less than my Pentium-4 did last time I built a Windows PC.
What are you talking about? In performance per US dollar, Intel has been winning the race hands down. I can get a Core i-5 3570k for about $220 USD. With decent memory, motherboard and cooling I can clock it up to 4.4 ghz and it's stable and not running too hot. To get that kind of performance at that price from AMD... oh, wait.. I can't.
AMD management made some bad decisions, then got rid of all the people who argued against those decisions. Now they are going to cut costs by firing the engineers who could develop new products. It is now inevitable: AMD is doomed.
"Unless the entire board and their puppets are removed in the next week or two, the little chance AMD has now will vanish. There is no up side here."
AMD's layoffs target engineering -- Board incompetence dooms the company
"AMD senior management, or (mis)management, as we are now calling them, have delayed the roadmap past the critical point. Project Win was survivable, barely. The churn of technical talent made things worse, far worse, and put the company at the breaking point. Layoffs sapped confidence, and senior management was negligent in not messaging a damn thing to those who mattered internally and externally. The cuts that will follow ensure that the plans in place are not achievable, and SemiAccurate can not see AMD surviving at this point."
AMD is imploding because management doesn't understand semiconductors -- Analysis: You can't Win by ignoring fundamentals
lf(1): it's like ls(1) but sorts filenames by extension, tersely
No, they do it in large part because Intel has engaged in some pretty obnoxious antitrust violations over the last decade or so and got what's barely a slap on the wrist. AMD for it's part did some really stupid stuff as well, but it's hard to make a profit when your competitor is paying systems integrators not to use your products.
Also, while most folks here seem to be on the AMD is walking dead meme, the fact of the matter is that Intel can't afford for AMD to go out of business any more than MS could have afforded Apple to go out of business during the '90s. If it really does get to that point, you'll see Intel laying off for a while to let AMD catch up.
The big problem that AMD has right now is old debt and an inability to produce enough chips to satisfy demand. That's not something that's generally true of chips that are being sold for the maximum price people will pay.
It's a lot more complicated than that. photolithography is a very complex process. As dies shrink due to a smaller nodal size it becomes increasingly more difficult to fabricate a single chip until that process matures.
All 150+ 4/6/8 core Sandybridge processors were sourced from only 5 different chips with 2/4/8 cores a piece and varying amounts of cache. The yield on the 8 cores is low even on the mature 32nm process so they demand a huge price premium. Those with defective cores have some disabled and are sold as 6 core variants.
Since defects are fairly consistent per wafer, yields on a 200mm^2 Sandybridge are exponentially higher than they are on a ~400mm^2 Sandybridge. The same is true for Ivybridge. I'm not sure if Intels 22nm process has matured enough to make 8 core Ivybridge processors economically feasible quite yet. Thus, 220mm^2 yields on Intel's 32nm process may be comparable or even higher than 160mm^2 yields on Intel's 22nm process.
TSMC's 28nm process was backlogged for quite some time due to low yields. The GTX680 was unavailable for the longest time because it required that a large chip be fabricated with no defects, the GTX670 which came later allowed for part of the chip to be disabled, thus increasing yields. AMD had the same problem with their HD 7000 series, low yields on the top end processors reduced their ability to ship those processors. Fortunately for them they had a stripped down version (HD 7950) ready to go at the same time rather than months later.
Intel is a remarkably conservative company. They're not known for announcing a product unless they know that they can make it available and thus it doesn't make sense to introduce an 8 core Ivybridge processor unless they know that they can actually deliver it. This is why the Sandybridge-E processors came around much later, and the same will be true for Ivybridge-E
Contrary to multiple postings on this thread, Qualcomm designs it's own ARM compatible CPUs (they call the latest version Krait) via an architecture licence from ARM. That's pretty make Qualcomm somewhat like the old-AMD designed it's own x86 compatible CPUs via an architecture license from Intel. However, the new-AMD licences their x86-64 architecture to Intel which designs it's own CPUs (arguably better than AMD).
On the other hand, Qualcomm acts alot like Intel in the cell-phone space. They use their patents on CDMA and other wireless communications and their first generation 4G-LTE modem/radio to bully cell-phone manufactuers into using Qualcomm SOC chipsets very similar to they way that Intel uses their CPUs to bully computer manufacturers into using Intel chipsets. They have been known to threaten to use bundling, bulk pricing, and even limited availability tricks on other low-end high-volume phone product lines to convince cell phone designers to use their chipsets. Thus you see even Samsung is forced to use Qualcomm SOC chipsets even though they make their own Exynos SOC. This makes them definitey not-like the new AMD in this sense.
Superior now. But when amd were ahead intel bribed the major pc makers not to use amd chips. During that time most of dell's income came from intel payments, for example. This is what destroyed amd since they could and can no longer afford r&d.
http://www.businessweek.com/technology/content/nov2009/tc2009114_975298.htm
The solution would have been for them to pay amd at least 10 billion in damages instead of 1, but that ship has sailed.
Well, I'm not entirely sure of that. For example: I do audio work, and video work, and like gaming, and compile my own software. All of those things take a robust desktop architecture to do well. You're not really suggesting that I'd switch to a tablet running BOINC in the background 24/7 while I process high-def audio files, are you?
So let's discuss alternatives. Say AMD goes down. What are my options as a consumer in, say, five years if I want to avoid Intel, but want all the horsepower I can get my hands on for a desktop workstation? I really don't thing it's going to be Qualcomm, if they're targeting low-wattage mobile devices. Are there any other CPU manufacturers who are positioned to step into that market?
That's bullshit. AMD couldn't afford to build a new fab because Hector Ruiz blew up AMD's cash reserves buying ATI lock stock and barrel over the stock market price just before the 2008 market crash. In fact this particular little deal smelled so bad a lot of people went to court and Hector was forced to quit his post.
I can't speak for other industries, but for the semiconductor industry gross margin is measured as revenue from a chip minus the immediate production costs. For AMD this would be how much they paid GloFo for the chip (or rather averaged across the wafer), plus the costs of testing, assembly/packaging, boxing, and shipping. It does not include advertising, R&D, taxes, etc. And as I stated earlier, R&D is a massive expense. All of those engineers designing the next chip are a huge cost that have to be paid.
You can take a look at AMD's finances first-hand and see how this plays out; AMD has never made a profit with gross margins below 44% or so. Intel would be an even better example: 13.5B in revenue, 3B in net income, and a gross margin of 63.3%. That would put Intel's profit margin at 22% versus their gross margin of 63.3%. Where did all the money go? R&D and fab upgrades. Gross margin only covers your immediate expenses in the semiconductor industry.
I disagree. Without Intel's backroom dealings AMD would have made enough money to weather the idiocy of Hector the Sector Wrecker (as muh ex-Motorolla buddies call him).
The whole reason that GloFo had to be spun off is because AMD invested in huge new fabs because they were fab-limited, but then found out they were Intel-limited, their marketshare didn't increase and their fab capacity was unusued. That's crippling for a silicon manufacturing company, so AMD had to stop being one.
Hector was no help, that's for sure, but I really think it was Intel that crippled AMD at the worst/best time.
The enemies of Democracy are
I guess you're just unaware of when AMD had the superior product, but couldn't get OEMs to sell products at the volumes such price and performance superiority would have suggested, because Intel, still the dominant player, had made deals with them not to sell AMD parts. Their market share was growing, necessitating a new fab, but then they hit the artificial limits defined by Intel, a crippling blow after investing billions in a new fab.
There's only been several verdicts against them by the regulatory authorities of multiple governments, and a lawsuit settled between Intel and AMD in AMD's favor with a 1.75 (iirc) billion payout. A pittance compared to what was lost, of course, but still heavily in the news.
I suppose it would have been easy to miss if you only just started following the CPU industry.
The enemies of Democracy are
They had a faster processor, but that is only one part of the equation. They had two major problems:
1) They didn't offer a CPU/chipset/mobo solution. Intel does it all for customers, they make the entire core if you want. This is useful to OEMs because there's no finger pointing when there's problems. Doesn't matter which of those components is broken, same company is responsible, they need to find and implement the fix. With the Athlons you could have a 3 way pointing match between AMD, VIA, and whoever made the board all claiming the other guy was responsible for a problem.
2) No good chipset. The processor was all kinds of fast but woe betide you if you wanted to use it with, say a GeForce DDR. The VIA chipset that was the "premier" solution for it implemented the AGP spec improperly and wouldn't work with the GeForce card since the AGP slot wasn't really AGP, basically just a fast PCI slot. This wasn't the only problem, just one of the most major ones.
So it is no surprise that some OEMs shied away from them. I built an Athlon system and it was a couple weeks of hell trying to make it work before I found out that no, there was just no way my GeForce would work with it. Back the parts went and in came Intel parts that functioned without error.
Likewise at work we did have some Athlon systems, Gateway I believe, and they were far more trouble than the Intel systems as a whole.
Intel isn't just popular because of the power, but their stability. It matters in business. AMD never really had a competitive solution in that regard.
I'm not saying Intel didn't also try to squash AMD (IA64 was another attempt, since there is no cross licensing for that instruction set) but AMD did little to help themselves. They produced a good processor without the hardware to support it.
Then they caught another break, with the fuckup that was the P4, but they rested on their laurels and didn't really do much in the way of architecture updates. Intel hit back with the Core 2, then Core i, then Sandy Bridge all of which are stellar performers per clock and there was just nothing new from AMD, until now Bulldozer which is pathetic, worse than their old chips at times.
Intel is not blameless, but AMD has done themselves few favours.