Did Bulldozer Damage the AMD brand?

[bradley]

I cannot support a company that consistently misrepresents upcoming products and timelines. That was the final blow for me.

 

[ikachu]

Well, looking at the earnings it certainly didn't help...
http://allthingsd.com/20121011/amd-...ng-demand-for-chips/?reflink=ATD_yahoo_ticker

 

[Ventanni]

What about the Athlon 64 days?

You raise a good point. Back in those days, AMD definitely had the superior CPU, that's for sure. While I don't think AMD will ever return to that level of CPU performance that the A64 had over the P4, I don't think Intel will be able to ever catch up to AMD's expertise in the GPU department. I guess what I'm getting at is, while AMD was competitive with Intel in the A64 days, it was still in Intel's court. At least now, Intel can't really compete with AMD's GPU expertise.

But then again, all Intel needs to do is heavily optimize their iGPUs for OpenCL work, and AMD will be at a disadvantage again.

 

[Ben90]

It didnt change my opinion of the company. I expected failure, it delivered perfectly. AMD has been a mess since conroe showed up. Going on 7 years of garbage and false hopes.

The original Phenom and GeForce 9800 series were my first major disappointments with this hobby. I lost interest in hardware for quite a while because of those two.

 

[Mr. Pedantic]

The performance itself wasn't such a big factor, but the fact that they hyped it up so much spoiled it a bit. Making a crappy product is something that all companies do from time to time. Totally mis-representing it and trying to claim wins where there really weren't any is a completely different kettle of fish.

 

[Piano Man]

Bulldozer was being hyped as this generations K7, which now is laughable. It definitely changed my attitude towards AMD in the sense that I just don't really trust what they say we can expect from a product until I get to see an independent review. Their new tablet chip is supposedly the most awesome thing since forever but I don't really believe it because this is the same company who hyped the biggest failure in recent memory. Doesn't mean I wont' buy it though if it turns out to actually be good.

 

[Ferzerp]

Yes. More due to their marketing attitudes than the performance of the product. You can market to strengths without being deceptive, dihonest, etc.

Well, I assume most of us can. AMD's marketing department apparently cannot.

They've made large missteps, but instead of presenting an attitude of "Hey, we f'ed up, but we're working on it", the attitude is "You're just wrong, this is really better, even though only our fanboys (who there is no reason for us to market to) agree". Instead of admiting that at this point, they're mostly a low-end company, they try to flip that in to pretending that the areas where they are less weak are really the *only* areas that matter, and the areas that they are laughable are inconsequential (e.g. only integrated gpus matter, the cpu is pointless). This is pure head in the sand syndrome.

So... I guess, it wasn't bulldozer, per se. It's more the insane attitude that they present in official communications.

 

[eternalone]

AMD is no longer and option in any way or form when it comes to building a PC. If they were smart they would have made a deal with ecs or foxconn and given away free motherboards with a BD purchase like fry's used to and that would have kept them alive in the value section. But now AMD has a really bad reputation and seems like a company badly manged, deceptive, and shortly going out of business.

 

[frozentundra123456]

Anytime you hype-up a product or service....and then it fails to live up to the consumer expectation (for whatever reason), you end up doing damage to the company and the brand.

So yes, Bulldozer really hurt AMD.

Yea, to me the worst was the hype that led up to the launch. Namely the cherry picking isolated highly threaded benchmarks to mask the poor single threaded performance, the GPU bound benchmarks that were supposed to demonstrate CPU performance, and the price comparisons of BD to extreme edition Intel CPUs which no one in their right mind would use for a gaming rig when SB offered silmilar or better performance in games at 1/3 the price or less. I take any marketing claims from Intel, AMD, or any other company with a grain of salt, but I really dont trust AMDs marketing anymore, or their claims of future performance until proven by independent testers.

It is kind of sad really. I was just looking back on my favorite computers, and occasionally I still fire up probably my favorite, and AMD XP2600+ with an AGP video card. Far inferior to my current comp in performance, but for its time it was great, a huge improvement over the celerons and early P4 that I had. Those were the days for AMD.

 

[bunnyfubbles]

I never expected Bulldozer to compete with intel's massive IPC advantage, however I did expect it to do much better when it came to multithreaded scenarios, however it failed even in that regard as it struggled to compete with their own X6.

So what promised to have the potential to be a less expensive option to Intel's higher end platforms (6+core on s1366 and eventually s2011) turned out to be through and through inconsequential for those who already had a quadcore i7.

 

[LoneNinja]

I honesty don't care, I was a big amd supporter but I just don't care much about pc anymore, I'm much more interested in mobile products now.

 

[Maximilian]

AMD is no longer and option in any way or form when it comes to building a PC. If they were smart they would have made a deal with ecs or foxconn and given away free motherboards with a BD purchase like fry's used to and that would have kept them alive in the value section. But now AMD has a really bad reputation and seems like a company badly manged, deceptive, and shortly going out of business.

Yeah that is one of the worst things for them right now, there's no niche for their chips anymore.

- High end performance... wont even go there
- Mid/low end performance, beaten by the i3's
- Power efficiency, beaten by everything
- HTPC/basic web pc, beaten by the pentiums/celerons
- Price, they have the best bang/buck in some situations but they are few

The only thing they have is the APU's which arent good enough to game and arent cheap or efficient enough to be a viable basic alternative to the pentium/celeron line.

 

[Shakabutt]

Personally i am content with my FX 6100 wich i purchased 2 months ago.

I never really catch the "good ol' amd days" and honestly i always go shopping for parts with budget/perfomance/longevity (im not brand biased) in mind.

After i got my processor (for like 100 bucks minus the cash i made from my old one) i was expecting the sky to fall down on me when i started using it....but nah it runs pretty smooth and i got a nice 4 GHZ OC on stock.

Thats not to say that my mouth doesnt drool when i see some Sandy Bridge action...but i drool at exotic cars to so...

And to be brutally honest this will probably be the last time i will buy AMD, not because they lied or some shit, in part due to the fact that intel is getting smart and is putting out cheap/beastly procs and AMD seems kinda dissoriented on the desktop front (they are moving away from the classic powerhouse/gaming procs imo).

Also i never had a intel cpu so yes, no more AMD for me

 

[OBLAMA2009]

i would say that i made me think of them negatively, but since i already thought of them negatively

 

[mikeymikec]

Given that AMD absolutely had the benefit of hindsight after witnessing Intel's failure with the the "more gigahertz, less IPC" mentality, yeah the fact they went ahead and built bulldozer really diminished my esteem for their prowess in creating products that customers are going to want to buy.

I think you're probably incorrect about AMD's approach with Bulldozer - they designed a CPU (allegedly using software rather than an intelligent human-designed approach), then IIRC had "manufacturing problems" (read: the CPUs didn't really compete at the intended frequencies, and needed to be massively overclocked). Management says "well, crap. We can't just abandon years of R&D effort and start again, so how do we market this?". Marketing says "More MHz, more cores!".

- edit - reading the article, now I'm not so sure about the timing with regard to SOI and the A64. I could have sworn that articles were coming out on tech news sites about this new idea shortly before the A64 was released.

Ok, reading that article I'm not so sure about the timing of SOI and the A64, but I could have sworn that the tech press was talking about SOI shortly before the A64 was released.

Someone mentioned the Athlon 64 era - AMD teamed up with IBM to come up with the SOI (http://en.wikipedia.org/wiki/Silicon_on_insulator) manufacturing technique, and the Athlon 64 was the first to use it. Bulldozer used something newer, but it didn't get any press attention at all. How much of a role did SOI play in AMD's success with the A64? I don't know because I don't know enough about that sort of thing.

 

[Ancalagon44]

Yes. More due to their marketing attitudes than the performance of the product. You can market to strengths without being deceptive, dihonest, etc.

Well, I assume most of us can. AMD's marketing department apparently cannot.

They've made large missteps, but instead of presenting an attitude of "Hey, we f'ed up, but we're working on it", the attitude is "You're just wrong, this is really better, even though only our fanboys (who there is no reason for us to market to) agree". Instead of admiting that at this point, they're mostly a low-end company, they try to flip that in to pretending that the areas where they are less weak are really the *only* areas that matter, and the areas that they are laughable are inconsequential (e.g. only integrated gpus matter, the cpu is pointless). This is pure head in the sand syndrome.

So... I guess, it wasn't bulldozer, per se. It's more the insane attitude that they present in official communications.

Yeah I think the arrogance was definitely part of it. JFAMD's refusal that IPC was in fact. He was quite arrogant about it. Now, I'm sure he was fed his info from elsewhere, but its hardly the only case.

What gets me too is that AMD dont really seem to care about this PR nightmare. Sure, they will probably be forgiven if Steamroller turns out well, I dont know, I just think they should like apologize for being so dishonest or something. Not that they ever will.

I think you're probably incorrect about AMD's approach with Bulldozer - they designed a CPU (allegedly using software rather than an intelligent human-designed approach), then IIRC had "manufacturing problems" (read: the CPUs didn't really compete at the intended frequencies, and needed to be massively overclocked). Management says "well, crap. We can't just abandon years of R&D effort and start again, so how do we market this?". Marketing says "More MHz, more cores!".

As far as I understand, the quest for high MHz was a deliberate design decision to offset the fact that a single BD core was going to have lower IPC than any other core except Atom and Bobcat. They knew that they had to clock the CPU higher to compete with both Intel but also their older products, and this was because they stripped execution resources to make leaner cores.

Of course, most people at AT knew that the high MHz option was a bad idea, but anyway, benchmarks have spoken and they agree that the high MHz option was not a good idea.

 

[lifeblood]

BD really soured me on AMD. Prior to BD, AMD was an inovator. A64 with its 64bit extension, followed by A64 X2 with its dual core architecture, truly made AMD the leader. With BD and all the talk of fusion I expected AMD to again out-inovate Intel. Add to that the selective benchmarks and JF's insistance that IPC would stay the same while clocks would greatly increase and it was clear that BD would be a winner.

I guess AMD is still an inovator. To be fair, when AMD came out with 64bit and dual cores, few operating systems or apps could take advantage of it. It took time for the rest of the world to catch up. So now BD has capabilities that most apps don't support yet. The problem is that A64 still had excellent performance compared to P4 while BD's performance compared to Core sucks.

I also must throw some mud at GloFlo. BD wouldn't have been so horrible if they had been able to hold up their end of the bargain and deliver higher clock rates at acceptable heat/power levels. Unfortunitly they did what they seem to always do and delivered less than they promised, later than they promised it.

The fact that JF so soured things is annoying to me. I accept he was told by others that IPC would stay the same. My problem is that when he did learn IPC dropped he said nothing. It was good to have an "insider" on Anandtech giving us information, and I really wish we had that again. Unfortunitly because of the whole JF affair I don't think that will happen anytime soon. No company is going to allow it given the backlash that resulted from BD.

 

[BenchPress]

But then again, all Intel needs to do is heavily optimize their iGPUs for OpenCL work, and AMD will be at a disadvantage again.

Not at all. Intel will soon have AVX2, while AMD hasn't even added it to the roadmaps yet. Haswell will easily more than double the OpenCL performance thanks to AVX2, FMA, twice the cache bandwidth, three AGUs, execution port 6 to offload port 0, etc.

Heterogeneous computing using the iGPU is a dead end. It doesn't have more computing power, it's much harder to program, and suffers from the CPU-GPU data transfer bottleneck. AVX2 is far more versatile and the data doesn't have to leave the CPU cores and it's just as powerful.

 

[Ferzerp]

Not at all. Intel will soon have AVX2, while AMD hasn't even added it to the roadmaps yet. Haswell will easily more than double the OpenCL performance thanks to AVX2, FMA, twice the cache bandwidth, three AGUs, execution port 6 to offload port 0, etc.

Heterogeneous computing using the iGPU is a dead end. It doesn't have more computing power, it's much harder to program, and suffers from the CPU-GPU data transfer bottleneck. AVX2 is far more versatile and the data doesn't have to leave the CPU cores and it's just as powerful.

On an igpu, yes. On a real GPU though, I'd say it will make perfect sense for a long time to come, but then again, only on the specific workloads that it already makes sense for.

 

[pelov]

I also must throw some mud at GloFlo. BD wouldn't have been so horrible if they had been able to hold up their end of the bargain and deliver higher clock rates at acceptable heat/power levels. Unfortunitly they did what they seem to always do and delivered less than they promised, later than they promised it.

This wasn't GloFo. Had clock speeds not been attained due to poor yields we'd have heard about poor yields. The poor yields stemmed from Llano and it's really dense and really big on-die GPU.

AMD and GloFo didn't release a stepping with higher clocks and lower power consumption either. That's because it wasn't a GloFo problem but an AMD one. Just take a look at the idle power consumption numbers and you'll see that AMD's BD chips are very good at idle and sip power. The issue stems from architectural problems and that's noted by the huge issues with power consumption under load. Had it been a GloFo problem we'd have seen both, not just one.

AMD really left a sour taste in my mouth. The stupid marketing gimmicks they pulled before and after the release of BD were horrible. Internally they knew it was a dud, yet they held that stupid overclocking record, that fixed "blind gaming test" that featured a single GPU and told nothing about the CPU's performance. Then there was JF spreading FUD on the forums and even that [H] question/response session they had where the engineers clearly said one thing and then had marketing say the exact opposite in a separate question.

If you've got a dud, fine. There's no reason to straight up lie to the enthusiast community, thinking that the sales numbers might go up if you go into damage control mode. There are quite a few of us who waited for a good AMD chip and were disappointed by the benchmarks but were flat out disgusted with the way it was represented.

 

[Idontcare]

This wasn't GloFo. Had clock speeds not been attained due to poor yields we'd have heard about poor yields. The poor yields stemmed from Llano and it's really dense and really big on-die GPU.

AMD and GloFo didn't release a stepping with higher clocks and lower power consumption either. That's because it wasn't a GloFo problem but an AMD one. Just take a look at the idle power consumption numbers and you'll see that AMD's BD chips are very good at idle and sip power. The issue stems from architectural problems and that's noted by the huge issues with power consumption under load. Had it been a GloFo problem we'd have seen both, not just one.

AMD really left a sour taste in my mouth. The stupid marketing gimmicks they pulled before and after the release of BD were horrible. Internally they knew it was a dud, yet they held that stupid overclocking record, that fixed "blind gaming test" that featured a single GPU and told nothing about the CPU's performance. Then there was JF spreading FUD on the forums and even that [H] question/response session they had where the engineers clearly said one thing and then had marketing say the exact opposite in a separate question.

If you've got a dud, fine. There's no reason to straight up lie to the enthusiast community, thinking that the sales numbers might go up if you go into damage control mode. There are quite a few of us who waited for a good AMD chip and were disappointed by the benchmarks but were flat out disgusted with the way it was represented.

Exactly. And I'll point out this is exactly the same thing that happened with Prescott.

Everyone blamed Intel's 90nm process tech as being the culprit for Prescott's stupid high power consumption, but then Intel released their mobile dothan (precursor to merom/core2) on the same 90nm process node and it was awesome at low power consumption, not held back by Intel's 90nm at all.

GloFo hit their electrical parametrics, bulldozer performed exactly as it was designed to perform. The loss in confidence here is that according to AMD they didn't design bulldozer to perform this way, which means they really have no idea how to make their models and simulations be a good guide for basing design trade-offs on.

If you set out to grow bananas but you end up harvesting apples, something is fundamentally wrong.

This is what makes me all the more concerned with AMD's continued reliance on simulated performance and marketing their expectations of streamroller's performance and jaguar's performance without having any silicon in hand. The track record is not confidence building and yet they keep putting that same foot forward (see Phynaz's sig for example).

 

[pelov]

Everyone blamed Intel's 90nm process tech as being the culprit for Prescott's stupid high power consumption, but then Intel released their mobile dothan (precursor to merom/core2) on the same 90nm process node and it was awesome at low power consumption, not held back by Intel's 90nm at all.

GloFo hit their electrical parametrics, bulldozer performed exactly as it was designed to perform. The loss in confidence here is that according to AMD they didn't design bulldozer to perform this way, which means they really have no idea how to make their models and simulations be a good guide for basing design trade-offs on.

If you set out to grow bananas but you end up harvesting apples, something is fundamentally wrong.

This is what makes me all the more concerned with AMD's continued reliance on simulated performance and marketing their expectations of streamroller's performance and jaguar's performance without having any silicon in hand. The track record is not confidence building and yet they keep putting that same foot forward (see Phynaz's sig for example).

This whole debacle likely stems from the poor simulation in the labs rather than any foundry issue. There's a reason AMD initially planned it for 45nm, pulled the plug and waited for a mature 32nm SOI node before attempting it again, all the while the 45nm Thubans and Phenom IIs did just fine.

I'm hoping that whatever simulation issues they encountered with Bulldozer were dropped along the way and that they had to push the design through no matter what. The latter looks likely, considering the huge list of improvements in PD on the same node just a year later. Clearly they got the silicon in hand, went "Oh s**t, there's a lot wrong that we didn't see" and set out to fixing it by plucking at the lowest hanging fruit first, such that they can release a revision only a year later. If the new Jaguar cores under-perform as well then things aren't looking up for Steamroller either. I mean, just the fact that they're quoting a 7% IPC increase on the same process means there was a massive list of screwups with the initial Bulldozer design, and that's 7% without even addressing the major issues with the architecture.

I think they have to have silicon in hand, but they're just not showing it. At least for the Bobcat successor. I'm not sure they have anything for Steamroller yet

 

[chihlidog]

It had a significant impact on my view of AMD. It was overhyped. They didnt just market it, they were flat out deceptive. And the whole slower IPC than the previous gen was a huge issue for me. This is all coming from a former AMD fan.

I still love AMD's graphics, but I certainly do not see myself using any AMD CPUs in the foreseeable future - something I never thought I would say 2 or 3 years ago.

 

[Batmeat]

I see AMD compensating by slowing down cpu development for desktops and focussing more on mobile computing and gpu development. Last AMD chip I owned was an Athlon 64. In the end, it's all about profit, and if they can get into the mobile market with phones and tablets they stand to make a killing. Hope if works for them.
Linky

 

[Idontcare]

I think they have to have silicon in hand, but they're just not showing it. At least for the Bobcat successor. I'm not sure they have anything for Steamroller yet

If they do have silicon in hand but are falling back to discussing the theoretical performance on the basis of the simulations they had done before they had silicon in hand then that is even the less confidence building

That is, after all, exactly what JFAMD did...hyped the performance of bulldozer on the basis of simulations but didn't correct those expectation once AMD had silicon in hand.

In theory there is no difference between theory and practice, but in practice there is.