What went wrong with AMD?

[MyWorld]

I have often wondered what went wrong with AMD. I was a die hard AMD fan for many a year, until Intel released their Core 2 Duo chips, from there on AMD never stood a chance.

What went wrong?
In the early x86_64 days they had Intel on the ropes, they produced brilliant processors and then all of a sudden it all went pear shaped. Right now they are battling for their very survival!

I came across this article that explains it quite nicely. A bit lengthy, but worth the read.

http://www.pcpro.co.uk/features/372859/amd-what-went-wrong

 

[Ancalagon]

Some of that article is inaccurate or incorrect.

Such as:

Barcelona was officially finalised in the summer of 2006, with president Dirk Meyer promising a summer 2007 launch for the new chips, branded “Phenom”.

But the August launch was postponed until September and, when AMD finally put its chips on the table, they received a frosty critical reception – our tests concluded we were “underwhelmed by its performance levels” – thanks to a flaw that required a key internal buffer to be disabled. Fixed chips didn’t arrive until April 2008.

Even after that bug was fixed, performance of Phenom I still sucked. Heck, although Phenom II was a great product, it was never competitive with even the Core 2 Duo on a clock for clock basis. I dont think it could overclock as well, either.

Also they mention the lawsuit against Intel. As far as I'm aware, Intel ceased doing most of that stuff sometime between 2005 and 2007. So although it contributed (and probably still does) to AMD's problems, it wasnt a factor in the Phenom I/II era.

It also lacks concrete info on what actually went wrong. Supply side issues are not a cause, they are an affect. Something caused those supply problems to appear - was it poor planning, poor design, a rush to market, poor management, etc etc? To say that supply issues are causing AMD woes is to say a lack of runs caused us to lose the cricket. Yes, but what caused us to not get the runs?

It does point out two problems that AMD has though, both quite interesting:

Light was shed on Bulldozer’s problems when ex-AMD engineer Cliff Maier spoke out about manufacturing issues during the earliest stages of design. “Management decided there should be cross-engineering [between AMD and ATI], which meant we had to stop hand-crafting CPU designs,” he said.

Thats pretty bad! It could be the reason BD stinks.

AMD’s battles with Intel have overshadowed a more insidious problem: a chronic lack of leadership. Eight senior executives have departed in the past four years: Henri Richard, the chief sales officer, left shortly before Barcelona’s debut, and Phil Hester, AMD’s chief technology officer, left soon after.

 

[Archer]

It also lacks concrete info on what actually went wrong. Supply side issues are not a cause, they are an affect. Something caused those supply problems to appear - was it poor planning, poor design, a rush to market, poor management, etc etc? To say that supply issues are causing AMD woes is to say a lack of runs caused us to lose the cricket. Yes, but what caused us to not get the runs?

Thats not always true. Sometimes the fabs just plain suck (or the raw materials are contaminated), and no matter how perfect your product, yields are bad and there is sweet nothing that the customer can then do about it.

 

[Mike Hoxbig]

The same thing that goes wrong with any company that Rouxenator touts - they have him as a fanboy trolling the interwebs.

 

[Ancalagon]

Thats not always true. Sometimes the fabs just plain suck (or the raw materials are contaminated), and no matter how perfect your product, yields are bad and there is sweet nothing that the customer can then do about it.

yeah, but why do they suck? What causes the poor yield issues? Did they not give the engineers time to mature the process? Were the dies simply too large to get decent yields? Were the wafers too small? Were the engineers constrained by bad management?

Supply issues dont just happen, something causes them.

 

[Archer]

yeah, but why do they suck? What causes the poor yield issues? Did they not give the engineers time to mature the process? Were the dies simply too large to get decent yields? Were the wafers too small? Were the engineers constrained by bad management?

Supply issues dont just happen, something causes them.

Yes, sometimes the die is to big to get good yields. But sometimes the manufacturer simply cant get good yields on a node. For example, there was nothing wrong with HD6xxx series, but it had to be a 40nm part, instead of 32nm because none of the fabs were ready with 32nm. Hence, absolutely nothing AMD could do about it except fallback to the old 40nm node. And remember the lack of HD5xxx parts at launch? Again, good design by AMD, but TSMC had contamination problems. Nothing you can do about that. But because AMD had a good design (vs Fermi) they got quite a few parts out despite all that, whereas Nvidia struggled their rear end off.

What you must realise is AMD does not own any fabs, so some things will be out of their control. This does not mean to say I think their CPU department is any good though

 

[Ancalagon]

Perhaps the next question one could ask is, was going fabless a good idea for AMD?

As you say, owning no fabs means a lot of these issues are completely out of their control, whereas Intel has complete control over their fabs. Has owning no fabs translated to supply issues? Or would they still have had supply issues even if they did own fabs?

I guess ARM is also fabless, but then, ARM doesnt really design chips. They just produce ISA specifications (as far as I know?) and leave it up to licensees to actually design chips. I guess all of those companies (Qualcomm, Samsung etc) use TSMC?

 

[killadoob]

The same thing that goes wrong with any company that Rouxenator touts - they have him as a fanboy trolling the interwebs.

LOL, i wonder what he is going to do now that amd have said the desktop is over for them. He may be their only client for cpu's soon haha.

I loved amd, i only bought amd. Recommended to my clients and then the core2 bomb dropped and it was game over for amd. Personally i think amd fluffed their lines because they got arrogant and thought people would buy rubbish cpu's and many did, they had die hard fans who would buy any crap they released but those die hard fans are few and far between now.

You look at amd's last release, they release a cpu so pathetic it could barely beat their old line and then they use an excuse that it will only work in windows 8 and windows 7 is to blame. Now what rocket scientist thought it would be good to release hardware that cannot work well on current systems? Did they think people would buy the cpu's a year before windows 8 came out? They could have taken intel but they sat back when they were on top. Ivy bridge and sandy bridge have all but killed amd in the desktop arena. Even the most die hard fan boys cannot justify to us nor themselves that amd is a good choice now.

 

[The_Unnamed]

The only reason the Athlon 64 series of processors were good is because the Prescott processors sucked so bad.

 

[Ancalagon]

@killadoob

Agreed. I have a hunch that management is to blame.

Manager: "Guys what we need is a really high clockspeed part - because clockspeed sells - with lots of cores, because cores sell too! Lets reduce the number of functional units and increase the pipeline length to make it cheaper and let us hit higher clockspeeds."
Engineer: "Uh did nobody here watch that disaster movie Pentium 4 Prescott: The Heatening, in which a giant processor overheats and destroys the planet? there is a reason Intel abandoned Netburst - it sucked!"
Manager: "You are paid to design what we tell you to design, not think. Design faster!"

 

[killadoob]

@killadoob

Agreed. I have a hunch that management is to blame.

Manager: "Guys what we need is a really high clockspeed part - because clockspeed sells - with lots of cores, because cores sell too! Lets reduce the number of functional units and increase the pipeline length to make it cheaper and let us hit higher clockspeeds."
Engineer: "Uh did nobody here watch that disaster movie Pentium 4 Prescott: The Heatening, in which a giant processor overheats and destroys the planet? there is a reason Intel abandoned Netburst - it sucked!"
Manager: "You are paid to design what we tell you to design, not think. Design faster!"

Without a doubt. Also amd had a bit of an issue where people would buy intel regardless of what amd offered. Even when they had a better platform many of my clients would say we want intel. So they never really got on top of intel but still they went backwards from the time intel dropped the core 2 bomb, perhaps their arrogance never allowed them to see intel coming but core2 killed amd in my opinion, the nail in the coffin was releasing a cpu line made for an OS that was 12-18 months away from being launched.

it's all good and well to push the boundaries but rather push them when people can make use of the new amazing technology .

 

[Ancalagon]

DAMNIT! I typed a whole long post, we had an internet whoopsie, and now its gone.

Anyway, I think the entire Bulldozer project is the reason why AMD is in such dire straits. Had they released a decent follow up to Phenom II 12-18 months after its release, they would have been fine. All they needed to is to have started such a project around the days of Phenom I, and had the management team not filled by idiots. Then they could have done it. Instead they let monkeys be in charge.

The combination of the massive delay between Phenom II, and the suckage that is Bulldozer, I think are the nails in the coffin. And this time they have only themselves to blame.

 

[Gtx Gaming]

The tech did not work out so great, im sure it can work with some tweaks etc.

Just like vista tried something new and it failed, and windows 7 fixed it.

 

[Ancalagon]

Tweaks? In my opinion, they need to abandon the idea of raising clock speed and core count at the expense of IPC. Thats not a tweak, thats a redesign. But they likely wont do it. If they tweak it, it could be suitable for server usage. Also lower power usage would be nice, but with clock speeds so high, not likely to happen.

Also when you release your next major architecture many years late, we dont want to hear excuses about how it will be better next time. We want actual performance!

 

[ponder]

It's a pity though as we need AMD to keep Intel on it's toes and ensure prices don't go up due to a market monopoly.

My first Intel cpu is my current C2Q, I'm not a fanboy of any brand and will buy what makes sense. AMD served me well before.

 

[droid]

With Windows 8 the improvements to run multithreaded applications will push most benchmarks to maybe look like this.

http://www.amdzone.com/phpbb3/viewtopic.php?f=532&t=138864


Also AMD's is still very good bang for buck.
http://paulisageek.com/compare/cpu/

The x86 need competition. Intel has already postpone some releases because the lacklustre competition.
Intel fans realise that AMD bad fortunes influence the speed of releases and might even influence the tempo of innovation at Intel. Love AMD for the total benefit they bring to the CPU market.

AMD and IBM is also starting to work closer together. It will help push up the yield.
Wonder if IBM might eventually take over AMD. Then we'll see fireworks.