'We've an Unexpected Manufacturing Advantage For the First Time Ever': Intel's Manufacturing Glitch Opens Door For AMD (theinformation.com)

Over at The Information (paywalled), reporter Aaron Tilley has a splendid interview of Forrest Norrod, a senior executive who joined AMD four years ago. Mr. Norrod describes the challenge AMD has faced over the years and how, for the first time ever, it sees a real shot at making a significant dent in the desktop market. From the report: Advanced Micro Devices' battle with chip giant Intel has often seemed like a gnat fighting an elephant, with AMD struggling in recent years to gain even a tenth of the market for the chips that power PCs and data center servers. Forrest Norrod, a senior executive who joined AMD four years ago, says the company suffered from "little brother syndrome" where it tried and failed to compete with Intel on lots of different chips. Now, though, AMD may have a shot at coming out with a faster, more powerful chip than Intel for the first time. Intel in April said it was delaying the release of a more advanced chip manufacturing process until sometime in 2019. AMD has its own new, advanced chip, which it will now be able to release earlier than Intel, potentially giving it an edge in the market for high-performance chips for PCs and data center computers.

It's a market opportunity worth around $50 billion. That's what Intel makes from selling chips for PCs and data center servers, and it dominates both markets. The data center market is particularly important because of the growth of new technologies like artificial intelligence-related applications, much of which is handled in the cloud. Companies that buy chips for data centers or PCs could gravitate to AMD chips as a result of Intel's delay. "I think we have a year lead now," said Mr. Norrod, who oversees AMD's data center business. AMD now has "an unexpected [manufacturing] advantage for the first time ever," he added.

Comments

Intel lost their edge

[Kohath]

This is why the rumor about Apple making their own Mac CPUs is believable. Intel lost their 18 month chip fabrication lead and they are now 9-12 months behind TSMC.

Re:Intel lost their edge

[dfghjk]

"This is why the rumor about Apple making their own Mac CPUs is believable "

No it isn't. Apple doesn't even bother updating its product lines with current Intel offerings so it's hard to believe that Intel lagging behind is even a small problem for Apple.

That's not to say Apple's interest in making their own CPU's isn't believable, just that Apple wouldn't do it for this reason.

Re:Intel lost their edge

[Anonymous Coward]

They don't update their product lines because Intel isn't delivering any fucking value or performance increases with minor revisions... LPDDR4 in configurations bigger than 16GB being the biggest complaint to date which is totally an Intel problem.

Re:Intel lost their edge

[marklark]

Maybe not 3 "changes", but they have used the Motorola 68000 series, the IBM PowerPCs, and Intel's Core/i3/i5/i7" series

Re:Intel lost their edge

[Anonymous Coward]

This is the real key.

Intel has managed to stay ahead in the scaling race. The die shrink lets them speed up and reduce the power and stay ahead of their x86 competition and within spitting distance of the low power ARMs.

If they can't stay ahead in that... they've lost their biggest advantage.

Everyone seems to be thinking this is about AMD... it's not just them. AMD is fabless. They don't care who fabs their designs. Intel is being attacked from the bottom too... by ARM and they don't care who fabs their designs either.

And now it's falling behind on its fab business - which used to be outstanding and leading the field.

Wherever you look, Intel is falling behind.

It's not going to fail immediately, but there's a stink of decay about Intel these days.

Re:Intel lost their edge

[jwhyche]

This was a long time ago, back when there several 680x0 computers. 680x0 being Motorola. The first Macs where based on the Motorola 680x0 chipset. They switched to powerpc when the motorola design couldn't go above 66mhz, I believe.

With that being said the 680x0 was a better design than the x86 chips at the time. To bad Motorola couldn't keep on the ball.

Re:Intel lost their edge

[jwhyche]

I don't know. Back when the rumor came out that Apple might pony up their own chips I thought it would be a mistake. Now, with all intels issues with bugs, I think it might be a smart move for Apple to look into this. I read a report that Apple is about to become the first Trillion dollar company, go apple, so its not like they don't have the cash to waddle in this direction.

Pre Intel Core Chips.

[jellomizer]

Back in 2005 time AMD was making significant headway in becoming the Chip for your PC right before Intel released the Core duo chip. The Pentium Line was getting aging and the Pentium-5 wasn't that popular and AMD was the chip for your PC. AMD had about a year or two of popularity.

Then Intel made the Intel Core Duo and the Core 2 Duo chip (64 bit) which put AMD back. But right before then, Intel was seen as the dying giant.

Re:Pre Intel Core Chips.

[ITRambo]

That's correct. At one point AMD briefly passed 50% of retail desktop sales, thanks to the Athlon 64. Intel came back strong and almost crushed AMD who screwed up with the A-series and other pre-Ryzen processors.

Re:Pre Intel Core Chips.

[Anonymous Coward]

There's a lot more to it than that. AMD had far better chips for just over 6 years but Intel had a huge head start in terms of brand recognition and market share, as well as far more fabrication capacity. While their market share dropped over the years as people discovered that AMD CPUs were faster, more reliable, ran cooler and cost less all at the same time, Intel used every dirty trick they could to keep AMD from growing. Despite that, AMD continued to become more popular, and Intel decided to throw their market share weight while they still could. When there were around 4 Intel customers for every AMD one they started offering a better price for CPU stockists who refused to sell AMD, effectively forcing sellers to choose between selling either only Intel or only AMD, resulting in AMD being extremely difficult to buy even if you wanted to. This cut off AMD's revenue, limiting their R&D ability. It was very illegal, and Intel was found guilty in court, but it was worth it to completely hamstring their competitor.

Intel had several lucky windfalls around this time too. The unplanned development of the Core architecture was a big one, as was the court's decision to give them an incredibly small penalty for getting caught. Another minor one was that AMD's architecture capitalised on low memory latency, so subsequent DDR versions having improved throughput at the expense of worse latency worked slightly in their favour. The main advantage was that AMD no longer had enough money to pay their top talent, so Intel had time to catch up to AMD's tech advantage,

Intel isn't pure evil, of course - there's a lot to like about the company - but it was dirty tricks that allowed them to cement their dominate position in the mid 2000's.

Re:Pre Intel Core Chips.

[Kjella]

At one point AMD briefly passed 50% of retail desktop sales, thanks to the Athlon 64.

If I remember correctly it was retail CPU sales excluding pre-built systems from OEMs, Intel was still by far the biggest by total volume.

Intel came back strong and almost crushed AMD who screwed up with the A-series and other pre-Ryzen processors.

The actual screw-up was earlier, when AMD stretched waaaaaaaay too far to buy ATI for $5.4 billion where $4.2 billion was cash. That's the war chest AMD should have had to counter Intel Core 2, instead they were stretched super thin. To cut development cost they replaced manually designed circuits with inferior designs created by automation and they couldn't afford to invest as much as they should in process technology so that even then they managed an equivalent design they were behind on cost, performance and power consumption. Meanwhile ATI was under siege by nVidia and couldn't really contribute much and there wasn't really all that much gained by APUs over discrete/integrated graphics because it took special code paths to take advantage and the niche was too small.

Strategically it was also a huge mistake because sure ATI would be fully aligned with AMD (the CPU side). But it meant nVidia had little choice but to deal with Intel on their terms, which Intel used to kick nVidia out of the integrated chipset business and then took all integrated graphics on Intel chips for themselves. AMD opened that door and Intel said "look, we're just doing what the competition is doing". Even if worst case Intel had bought ATI as was rumored they'd have gotten nVidia's full support for free in the fight against Chipzilla instead of paying billions. They should have seen that the battle wouldn't be that easily won instead of thinking CPUs was in the box, on to GPUs...

Re:Pre Intel Core Chips.

[epine]

But right before then, Intel was seen as the dying giant.

No serious observer regarded Intel as a dying giant, though you'd have my vote for a Napoleonic (Itanium) psychopath (RDRAM), sleeping off a boozy bender (Prescott, Caminogate).

One humble phone call to their Israeli design center ("maybe let's just put the engineers back in charge for a short while"), and Intel bounced right back off the matt again, big time, rocking those giant abulous fabs we all knew they were still packing under their delirious anti-competitive power-grab.

For about a five year period, during their Hewlett Packard joint venture, that must have been one hell of dysfunctional board room, perhaps even arcing as high as 100 mFi (milli-Fiorinas).

Itanium Sales Forecasts edit.png

I can never review that chart without hearing Julie Andrews in my inner ear chirruping gaily away about kettles of kittens and mittens of string.

Re:Pre Intel Core Chips.

[LinuxIsGarbage]

Back in 2005 time AMD was making significant headway in becoming the Chip for your PC right before Intel released the Core duo chip. The Pentium Line was getting aging and the Pentium-5 wasn't that popular and AMD was the chip for your PC. AMD had about a year or two of popularity.

Then Intel made the Intel Core Duo and the Core 2 Duo chip (64 bit) which put AMD back. But right before then, Intel was seen as the dying giant.

What actually happened was at the turn of the century Pentium III was replaced with Pentium 4, which had a worse per clock and per watt performance than Pentium III. But Intel kept pushing it because Mhz. They were also thinking Itanium was the way of the future.

AMD managed to make cheaper, lower MHz, faster chips. They also invented x86-64 with full backwards capability.

Intel kept pushing the Spaceheater 4. That thing was such an energy hog it was completely unsuitable for laptops. Intel in Israel created the Pentium M chip in 2003 as a new mobile chip. It was based on the Pentium III. It was usually marketed as a "Centrino" platform: Intel chipset, Pentium M, Intel Wireless.

The Pentium 4 was such a piece of shit Intel eventually threw the whole Netburst Architecture in the garbage, and based all future processors on the Pentium M: Core, Core 2, Core i3/5/7.

They also had to adopt AMD's 64 bit platform.

Only because of their huge missteps of Netburst and Itanium was AMD able to grab a sizable chunk of the market. Even when Core / Core 2 were on the market (~2007) it was easy to get better performance / dollar with AMD.

Re:Pre Intel Core Chips.

[Kjella]

That is just false. AMDs APU graphics could/can actually play games worth a salt.

Absolutely, but the division of labor between the CPU and GPU remains pretty much the same. The theory with APUs was that you'd mix and match CPU and GPU resources, calling the GPU for parallelism more because of the tight interconnect compared to going over PCIe. In reality AMDs APUs provided competitive value to Intel's CPU + nVidia GPU but they didn't really add any extra value. The gamer market didn't care because they used dGPUs anyway, in fact it's only when AMD released Zen processors with no graphics that competition returned. Integrating graphics didn't provide any value or advantage at all, they'd do just as fine integrating ATI/nVidia graphics chipset as they did before the buyout.

Re:Pre Intel Core Chips.

[TheRaven64]

NetBurst wasn't a failure of microarchitecture, it was a failure of process technology. NetBurst was designed on the assumption that Intel's fab people would be able to produce 10GHz versions in 2005, in the same power budget as the 2GHz version in 2000. This failed dismally and was one of the first indications that Intel's dominance over fabrication techniques was not going to last forever. If Intel in 2005 had been able to produce the process technology that they were forecasting in 1996, when the NetBurst design was started, then AMD would be long dead. They had absolutely nothing that could have competed with a 10GHz Pentium 4. Unfortunately for Intel, AMD had quite a lot that could compete with a 3GHz Pentium 4.

You are only as good as your last invention

[bobbied]

Isn't capitalism grand! You stubble and you failed!

Intel isn't in any danger here. AMD may gain market share and Intel may make less money, but this isn't the beginning of the end of Intel. Not by a long shot. It may mean that AMD finds it easier to be competitive, but Intel will get it's manufacturing back on track eventually and recover.

It's going to take more than a couple of stumbles for Intel to fall to second place to AMD.

Intel will fight dirty.

[Gravis Zero]

If I've learned anything about Intel from their past behavior then it's that they will lie, cheat and steal if that's what it takes to suppress AMD. I wouldn't be surprised if they paid off a bunch of companies to have a supply disruption occur, bad firmware updates bricking machines or creating a shell company to make purposefully shitty AMD machines.

Honestly, the FTC should have had their boot on Intel's neck decades ago and kept it their.

Seen this before

[ghoul]

K8.
Intel tried to make its next chip in Bangalore and screwed up so the K8 Opteron was a better chip and for a year AMD was the darling of the markets.
Intel caught up and ate AMDs lunch. AMD instead of using the windfall from the Opteron to build a sustainable chip pipeline (3-4 chips in dev instead of 1-2) used the moeny to buy ATI.
People in the CPU div were pissed when the 40 dollar RSUs went to 3 dollar.
But with AI and computation shifting more towards GP-GPUs than CPUs the ATI purchase has now started to payoff.

Of course that's the measurement on desktops

[raymorris]

> If you want to factor cost then, maybe

Performance is measured as "X per Y", such as "miles per hour", "miles per gallon", etc.

In a phone, the most important measurement is "instructions per watt", how fast can you go for the amount of power you use. Per dollar is also important in a phone. If you didn't' care about power usage / heat, and didn't care about dollar cost, your phone might have four Core i7 CPUs. It would have a ten pound battery and cost $2,000, and it would be fast.

On the desktop, power usage isn't nearly so important - it's plugged in. Your budget isn't a power budget on the desktop, it's a dollar budget. The main measure is instructions per dollar. AMD gives you better performance. If you didn't care about dollar cost, if what you cared about was instructions per second, you'd have a $97 million Cray OLCF-3 at your home office. You don't choose the OLCF-3 because cost is the primary measurement of interest. You want the performance for the $500 you intend to spend.

Re:Of course that's the measurement on desktops

[valnar]

Well in your opinion. I think elegance is more important than brute force. AMD chips run hotter, have a higher TDP and worse chipsets overall than Intel. Honestly other than INITIAL price, there isn't anything they do better. Cost has nothing to do with the quality of the chips. And if you have multiple computers at your house (or business), electricity matters. I leave a couple low powered (Intel) servers running 24/7. I wouldn't want to do that without Intel CPU's and Intel graphics.

Eh, not for the first time...

[Ecuador]

Not for the first time. How old are you, 15? AMD's Athlon was faster than the P3, especially when the latter couldn't keep up with clock speeds (there was even a P3 that was unstable at the rated speed and had to be recalled), and then Athlon 64 was much faster than the P4 (esp. with 64 bit OSes) but most publications at the time were at Intel's pocket and were trying to pass off that absolute turd Netburst architecture as gold, while at the same time Intel was strong-arming or bribing system integrators into not using the superior AMD. So AMD has had better solutions for years in the past, but due to Intel's illegal tactics they did not gain a big enough market share. In the end, Intel was forced to pay a fine which was nothing compared to the revenue AMD lost over that time and that lost revenue when they had superior technology meant they eventually were not competitive which meant consumers lost.

AMD64 anyone?

[Attila Dimedici]

Has everyone forgotten AMD64?

Back when Intel was trying to sell Itanium as the 64 bit successor to the x86 instruction set, AMD came out with what is now known as x86-64. It was worlds better than Itanium and Intel was forced to license it from AMD in order to stay in business (Of course, AMD had no choice but to offer Intel such a license on reasonable terms because it was built on the x86 architecture which AMD licensed from Intel).

Re:AMD64 anyone?

[complete loony]

There's a huge amount of complexity in a modern x86 - x64 based CPU around decoding instructions and detecting where work can be done in parallel. Part of the Itanium design required shifting a bunch of complexity into the compiler.

I can imaging that in the HPC space, where you really care about how fast this single loop is running, it makes sense to invest the engineering effort to improve the compiler. But that work had not been done.

It's easy to sell a CPU that will run your existing binaries faster. It's hard to sell a CPU that will require massive investment on your part to recompile *everything*, with a compiler that doesn't exist yet, in order to see any benefit at all.

So we are going to just forget K7 existed?

[Luckyo]

>Mr. Norrod describes the challenge AMD has faced over the years and how, for the first time ever, it sees a real shot at making a significant dent in the desktop market.

And then we remember how the first Athlon wiped the table with Pentium 3 and Pentium 4.