AMD: What happened?

[N4g4rok]

Only recently have i looked into the specifics of AMD's newer activities and product launches, and as an admittedly overzealous AMD fan, i have to say i'm a little bit confused.

When i built my first system a year ago, the Phenom II was an obvious choice. Price was great in comparison to the performance i expected from it. It kept up with all the games i played (always with vsync enabled), and managed to work with most of adobe's products and a few other software development tools at reasonably close to the same levels as Intel's somewhat expensive offerings at the time.

That was fine back then, but the end of the K10/10.5 production has forced me to look into BD, Interlagos, and the APU line. I saw the excessive market hype, and figured it may have paid off just a little bit. But, like most others, i just don't see it.

Even after remote success of their graphics devision, they seem to be cutting or otherwise loosing key players from that department. I hear rumors of a BD engineer expecting this somewhat poor performance. What ever happened with that? And has there been any news of what the new CEO plans to do about some of those problems?

Even though most of this is purely conjecture, i wouldn't mind laying out what's happening with what they've said they would do about it. It's difficult to support any company when it seems like they're actively making bad decisions.

 

[cantholdanymore]

Oh Boy! Here we go again

 

[N4g4rok]

Oh Boy! Here we go again

From the looks of things,i should have just browsed around the forums for a few hours. I've been out of the loop for a while. My apologies for continuing to beat a dead horse.

 

[formulav8]

Yeah there are so many threads already. Alot comes down to very bad managment. Ruins was a key player in that.

This new CEO is, well, very aggressive so far. Not sure what he has in mind by losing so many gpu engineers this past year and the one this year. I guess he knows what he's doing......

 

[Zor Prime]

Shareholders seem to approve of what AMD's new CEO is doing ... according to share performance, at least.

 

[alent1234]

the desktop/laptop market growth is over and even declining in some areas. ARM and SoC's is where it's at now so the most talented engineers will go to those jobs

 

[exar333]

Core architecture happened.

/thread

 

[tynopik]

Bulldozer

Instead of incrementally improving what they had, they took a huge gamble on a radical new architecture that required a ton of work. Then:
1. They couldn't execute. It turned into a much bigger job than they expected (complete rewrites usually do)
2. The design sucked anyways. Hard to know what exactly went wrong, but clearly the actual performance wasn't even close to what they were expecting

So they invested a ton of time and resources and got a chip that wasn't any better than what they already had.

 

[SlowSpyder]

Core architecture happened.

/thread

Yea... the A64 was a great chip, but it looked fantastic because it was competing against the P4. Intel used it's vast resources to make a much better chip than the A64 or P4, and AMD has been losing ground while simply trying to catch up ever since.

I bet even today, AMD's best parts probably aren't dramatically faster than C2D when looking at IPC.

 

[Rvenger]

IPC has never been an AMD strong point since the Athlon XP days really and this is why I think things have become a bit scewed with their 'more cores' strategy. The Athlon 64 being 64bit and it being priced significantly lower than the Intel Pentium 4/D is what kept AMD in competition. As soon as the dual core variants were released, it left Intel's Pentium 4 in the dust, and Intel spun wheels until they started getting clock gains with 'Core 2' series release. (Core duo clock speeds were too low to show significant gains.)

 

[sm625]

The problem with AMD is they fail to execute even on their "moar cores" strategy. They offer us a "8 thread" cpu that has a die size larger than their competitor, that consumes more power than their competitor for roughly the same level of multithreaded performance. That is a huge epic failure for them because given the die size of zambezi they should have been able to easily, easily fit sixteen K10 cores or even more bobcat cores. Instead they bloated the die with a bunch of garbage L3 that has such bad latency that it just makes me wonder if they arent diliberately failing, as if someone or multiple someones on their BoD has taken a payoff...

 

[IlllI]

Yeah there are so many threads already. Alot comes down to very bad managment. Ruins was a key player in that.

yeah, i'll have to agree with that. i remember when athlon 64 came out they were shooting for 25% market share. there server chips seemed to be doing well at the same time. everything seemed to be going in the right direction..

then they bought ATI for like 5 billion or something ridiculous.

 

[Rvenger]

The problem with AMD is they fail to execute even on their "moar cores" strategy. They offer us a "8 thread" cpu that has a die size larger than their competitor, that consumes more power than their competitor for roughly the same level of multithreaded performance. That is a huge epic failure for them because given the die size of zambezi they should have been able to easily, easily fit sixteen K10 cores or even more bobcat cores. Instead they bloated the die with a bunch of garbage L3 that has such bad latency that it just makes me wonder if they arent diliberately failing, as if someone or multiple someones on their BoD has taken a payoff...

People used to complain they didn't have enough cache, now they have too much cache! lol

I think its the L2 cache that is really pointless.

 

[WT]

Theoretical question: Would it have made any difference if they had NOT acquired ATI ??

 

[sm625]

Buying ATI was a really smart move. Unfortunately after all this time they really have nothing to show for it. Who knows what's really going on in their labs as far as true CPU/GPU integration, but from what I am seeing on the market they are nowhere close to realizing their potential. I think someone like apple would have gobbled them up if they actually truly fused a gpu with a cpu core.

 

[Kristijonas]

AMD is a stubborn, rebellous teenager. It is charming and dangerous at the same time. As most boys of this age, AMD is known to make rash, uncalculated decisions and then follow them stubbornly till the end when they have to face consequences. It all wouldn't be bad, but AMDs teenage-period happened late in it's life. While it is still rebelling, other boys, like Intel, is making well-thought, calculated step-by-step progress. Some could say that AMD should stop smoking weed and start making a career, but I like hippies and I hope AMD stays alive and continues to surprise us.

 

[Soulkeeper]

The old AMD:
http://media.bestofmicro.com/,T-Z-264167-3.jpg

The new AMD:
http://thegirlcanwrite.files.wordpress.com/2010/11/obeseladies.jpg

Let's not inline those since this is a tech forum
-ViRGE

 

[BD231]

AMD got narrow minded. They put all their efforts into multi-threaded performance then proceeded to slap a bunch of out dated Phenom CPU tech together and weaken per core performance hoping multi-threaded performance alone would sell their CPU's. Unfortunately they cut to many corners, the process tech they use lacks quality, memory performance is still in the stone age and performance per watt is abysmal.

Just a bad time for AMD cpu's right now, hopefully they pull something worthwhile out of their rear in the next few years.

 

[Dravic]

Looking at the consumer performance issues and Anand's recent Server (Opteron) review. My best GUESSES are..

They didn't hit clock speed target to offset lost in IPC from the shared resources design. Additionally at those clock speeds 8 cpu's eat a TON of power. You look at all that power being used and wonder where the performance went. I think the bottleneck is inter CPU and memory communications, and especially after seeing the Server review

HP (or Dell) have been adding a snoop filter CPU (traffic cop) to their 8+ core server SKU's specifically because they started seeing the natural reduced scaling effects of 8+ cpu workloads.

My completely wild and unconfirmed guess is that the additional logic for some kind of a snoop filter got CUT somewhere in the design, or they thought they could get away without it until > 8 cores.

 

[frozentundra123456]

Buying ATI was a really smart move. Unfortunately after all this time they really have nothing to show for it. Who knows what's really going on in their labs as far as true CPU/GPU integration, but from what I am seeing on the market they are nowhere close to realizing their potential. I think someone like apple would have gobbled them up if they actually truly fused a gpu with a cpu core.

For a long time I had a running arguement with a collegue at work who said buying ATI was a mistake, while I supported it. But now I am not so sure, or maybe the problem was that they paid way too much for the company.

They are only now beginning to get out of the debt incurred from the purchase and they have had to sell off their fabs. So it was a risk, and I think the jury is still out on whether or not it was a good move, considering all the problems they are having with their fabrications. I suppose we will never know how Bulldozer would have turned out if they had devoted some of the money spent for ATI into development of that chip and been able to produce it at their own fabs.

On the other hand, perhaps the graphics expertise obtained from ATI will enable them to gain back some market share and compete in the mobile and SOC areas. But there is a lot of competition in that area already and intel is trying to get into it too.

 

[Anarchist420]

Instead they bloated the die with a bunch of garbage L3 that has such bad latency that it just makes me wonder if they arent diliberately failing, as if someone or multiple someones on their BoD has taken a payoff...

Agreed, the biggest issue with BD is cache performance/efficiency. That was largely the issue with the PII as well. It would be nice if they could get away with 256KB of L2 cache/core, but (I think) that would require shortening the pipeline (or whatever the correct jargon is).

People used to complain they didn't have enough cache, now they have too much cache! lol I think its the L2 cache that is really pointless.

They've never had too much cache. The problem has been that's it's timings are too slow. They'd have done well with a shorter pipeline so they could've done with less cache. That way, the cache could've had less latency and it could've run at higher clocks.

 

[N4g4rok]

So, what's the plan of action if Piledriver fails as well? Dump the desktop CPU game entirely?

 

[Texashiker]

AMD: What happened?

They dropped the ball, that is what happened,,,, they dropped the ball.

 

[Idontcare]

IPC has never been an AMD strong point since the Athlon XP days really and this is why I think things have become a bit scewed with their 'more cores' strategy. The Athlon 64 being 64bit and it being priced significantly lower than the Intel Pentium 4/D is what kept AMD in competition. As soon as the dual core variants were released, it left Intel's Pentium 4 in the dust, and Intel spun wheels until they started getting clock gains with 'Core 2' series release. (Core duo clock speeds were too low to show significant gains.)

I thought IPC was AMD's strong suit since K7 came along, up through K10.

Intel was not able to beat their IPC until they had Core.

AMD's IPC did not get weaker at the time, it just didn't get strong enough quick enough to maintain the lead they had going back to 1999.

Now bulldozer changes that, IPC did indeed get weaker, which is the problem.

AMD is second to only Intel in terms of x86 IPC, a lot of companies have come and gone trying to get there themselves.

Buying ATI was a really smart move. Unfortunately after all this time they really have nothing to show for it. Who knows what's really going on in their labs as far as true CPU/GPU integration, but from what I am seeing on the market they are nowhere close to realizing their potential. I think someone like apple would have gobbled them up if they actually truly fused a gpu with a cpu core.

Yeah there is really no question that AMD needed to make the move they did if they were going to stand any chance of remaining relevant a decade later.

The only second-guessing I have is that I really wish Hector and the BoD had set aside their personal ego's and agreed to let JHH be the CEO of a combined AMD/Nvidia.

Not because the NV technology would have made all that much of a difference in comparison to ATI's for the fusion products, but because AMD really needed a passionate CEO who knew how to get stuff done. JHH is that person, like Steve Jobs. Hector was not, nor was Dirk.

 

[Rvenger]

I thought IPC was AMD's strong suit since K7 came along, up through K10.

Intel was not able to beat their IPC until they had Core.

AMD's IPC did not get weaker at the time, it just didn't get strong enough quick enough to maintain the lead they had going back to 1999.

Now bulldozer changes that, IPC did indeed get weaker, which is the problem.

AMD is second to only Intel in terms of x86 IPC, a lot of companies have come and gone trying to get there themselves.

Yeah there is really no question that AMD needed to make the move they did if they were going to stand any chance of remaining relevant a decade later.

The only second-guessing I have is that I really wish Hector and the BoD had set aside their personal ego's and agreed to let JHH be the CEO of a combined AMD/Nvidia.

Not because the NV technology would have made all that much of a difference in comparison to ATI's for the fusion products, but because AMD really needed a passionate CEO who knew how to get stuff done. JHH is that person, like Steve Jobs. Hector was not, nor was Dirk.

I never said the IPC got weaker. In the Athlon XP days, the IPC was very strong that they really didn't need much of an increase when they moved to A64, all they had to do is bump up the clock speeds and the processor scaled really well to speed bumps. Remember, Intels IPC with Pentium 4 was worse to begin with. Core Duo solved that problem.