AMD manager speaks about Bulldozer, admits failure

[inf64]

Bulldozer is the first installment only. He doesn't say anything about completely abandoning the concept of Bulldozer or that there is no hope for improvement.

Exactly. The follow on products will definitely be better and we saw AMD executing in that direction with PD core. We now know that one of the two future iterations brings full 256bit SIMD pipeline ability and some other major updates to the design. AMD would be foolish to just scrap the design which has so much room to grow and start from "scratch" again. BD was definitely not up to the task but AMD is obviously not sitting still. Mark Papermaster gave an interview one year ago to Techreport in which he stated that engineering is doing a lot of work on BD follow up cores. Soon we will see how good (or bad) Kaveri is.

 

[mrmt]

Admitting Bulldozer was a failure doesn't mean that its successors are viewed as dead-ends within the company, or that Steamroller is now cancelled. We don't know that he's referring to anything more than the product launched late 2011 launch.

Just look at his talk on servers. Nothing about Bulldozer, nothing about Piledriver, nothing about Steamroller, only the mention of unmitigated failure. He even pointed out ARM as the future for servers, relegating Jaguar as nothing more than a stop gap.

When your own management calls a given product an unmitigated failure, and the current product heavily resembles the unmitigated failure, this is not really an endorsement, but an announcement that things will change soon.

IMO, it's just a matter of waiting for the coroner.

 

[MisterMac]

Just look at his talk on servers. Nothing about Bulldozer, nothing about Piledriver, nothing about Steamroller, only the mention of unmitigated failure. He even pointed out ARM as the future for servers, relegating Jaguar as nothing more than a stop gap.

When your own management calls a given product an unmitigated failure, and the current product heavily resembles the unmitigated failure, this is not really an endorsement, but an announcement that things will change soon.

IMO, it's just a matter of waiting for the coroner.

Normally i find everything you say very insightful - and just the right amount of cynicism\skepticism.

But your acting crazy about this.
Bulldozer per it's design goals(read, goals) is horrible - it falls short, nothing new here.
(Irrespectable of it might being a decent chip in certain niche areas).


...but that doens't mean they can't fix something and turn it around to greater than before.
It doesn't mean a new fresh team CAN'T come in and say:
"Well this is not optimal - but.... we could do this and this this etc" and end up with a decent chip.


Either you in this instance throw some statements on the table - such as % gains in key areas, or outright say it's dead.


Screaming "OMFG LOOK HIGHER UPPER MANAGEMENT SAYS IT SUCKED IN PUBLIC" - and then insinuating things without saying them is kinda weak man.


It speaks to the future of where AMD wants to go - but jesus have you seen haswell's marketing?

EVERYONE who's not already there - is rushing to get there.
Why is that different when AMD publicly says they want to go there too - and that a portion of their line-up simply isn't good enough for that.


PS:

Does every American here realize - admitting corporate fault and mistakes is fairly common in EU?
It's not really special at all - nor does it signal internal turmoil or create shockwaves in the market.

 

[mrmt]

...but that doens't mean they can't fix something and turn it around to greater than before.
It doesn't mean a new fresh team CAN'T come in and say:
"Well this is not optimal - but.... we could do this and this this etc" and end up with a decent chip.

Maybe it's just me or "unmitigated failure" and the ARM statements don't fit together with "Well this is not optimal - but.... we could do this and this this etc"?

It's just that this kind of statement isn't what you say when you have plans for something in the future. I wouldn't just say "it's an unmitigated failure. We know it" without laying a clear road map to fix my unmitigated failure if I had plans for something. What we have here a senior manager calling my product an unmitigated failure, touting other products as better for the future market and no word about fixing the unmitigated failure.

 

[Eeqmcsq]

Maybe it's just me or "unmitigated failure" and the ARM statements don't fit together with "Well this is not optimal - but.... we could do this and this this etc"?

It's just that this kind of statement isn't what you say when you have plans for something in the future. I wouldn't just say "it's an unmitigated failure. We know it" without laying a clear road map to fix my unmitigated failure if I had plans for something. What we have here a senior manager calling my product an unmitigated failure, touting other products as better for the future market and no word about fixing the unmitigated failure.

To me, the topic of the article was about AMD's low powered Kyoto chips, not their server business as a whole. That would explain why they didn't mention Steamroller or Excavator. So I'm not ready to conclude that AMD is about to scrap Steamroller and Excavator development yet, especially when Kaveri looks to be on track.

 

[Phynaz]

“Bulldozer was without doubt an unmitigated failure. We know it,” Feldman said.

Wow, sig worthy.

Anybody know if JF has landed a new job?

Edit, yes, he's a marketing VP at NextIO.

 

[BUnit1701]

Maybe it's just me or "unmitigated failure" and the ARM statements don't fit together with "Well this is not optimal - but.... we could do this and this this etc"?

It's just that this kind of statement isn't what you say when you have plans for something in the future. I wouldn't just say "it's an unmitigated failure. We know it" without laying a clear road map to fix my unmitigated failure if I had plans for something. What we have here a senior manager calling my product an unmitigated failure, touting other products as better for the future market and no word about fixing the unmitigated failure.

The original quote is 'was' not 'is'. I.e. Bulldozer failed, but we are fixing it with PD and SR. To say they are going to scrap the entire family based on the flop of the first iteration is foolish. Design cycles run 5-7 years, you dont just walk in to the office on Monday and say 'This design didn't work, build me a new one for next quarter'. It literally takes years to change the design. The Bulldozer family IS the future of AMD for at least 2-3 more years, there is no escaping it.

 

[sushiwarrior]

Just look at his talk on servers. Nothing about Bulldozer, nothing about Piledriver, nothing about Steamroller, only the mention of unmitigated failure. He even pointed out ARM as the future for servers, relegating Jaguar as nothing more than a stop gap.

He didn't say ARM is the future for servers, he just said ARM works better in some applications/uses. There is no "future for servers", but rather some uses which ARM (or low power x86 like Jaguar and Atom) are better at than HPC modules. Jaguar isn't a "stop gap", Jaguar is a transition into what some servers are starting to need, but are not willing to take the plunge to ARM.

 

[krumme]

Oh dear, what's left for the fanbois now?

Maybe Mr. Feldman is just an Intel shill.

I think a good management culture is where a failure is called a failure, and a victory called a victory, in plain language. I have no other experience saying otherwise.

I simply dont understand your emotional reaction to what i think is quite good - and common - practice? - especially perhaps in silicon valley where success is build after faillure - the dynamics and learning from bitter experience and knowing how to get knowledge in failure.

 

[SiliconWars]

Just look at his talk on servers. Nothing about Bulldozer, nothing about Piledriver, nothing about Steamroller, only the mention of unmitigated failure. He even pointed out ARM as the future for servers, relegating Jaguar as nothing more than a stop gap.

When your own management calls a given product an unmitigated failure, and the current product heavily resembles the unmitigated failure, this is not really an endorsement, but an announcement that things will change soon.

IMO, it's just a matter of waiting for the coroner.

Is this your new forlorn hope for AMD's death now that your constant prediction of WSA doom is ended due to the massive console wins? :awe:

When something isn't working you change it. The SeaMicro aquisition was designed to head AMD down this route, now that Jaguar kills Atom in perf/Watt they are the clear current option for their servers.

If ARM turns out to offer better perf/Watt then AMD will go down that route, unlike Intel who will stick with what they think is best even when everybody else recognises their failures.

 

[StinkyPinky]

I really don't understand the glee some people seem to portray at the failure of bulldozer. Don't they realize it is in everyones best interest for AMD to be competitive? Do you think Intel would still be releasing 4 core cpu's with limited IPC increases if AMD were suddenly offering highly competitive CPU's?

I really want steamroller to be much improved although I have my doubts. The marketing talk of "greater parallelism" doesn't exactly seem to suggest anything amazing.

 

[Ajay]

To me, the topic of the article was about AMD's low powered Kyoto chips, not their server business as a whole. That would explain why they didn't mention Steamroller or Excavator. So I'm not ready to conclude that AMD is about to scrap Steamroller and Excavator development yet, especially when Kaveri looks to be on track.

That how I read it. For big core x86, Bulldozer was an unmitigated disaster (saying this publicly is like tossing the skeletons out of the closet and telling your customers they will see a fresh start from a new management team). I'd really be surprised if we don't see an SR Opteron. XV, I'm not so sure about; seems like AMD really needs 20nm FD-SOI* to pull of a good jump in power/performance. With the current state of things, it would be _amazing_ to see 20nm CPUs from GFL by the end of 2015. GFL is the biggest threat to AMD's big core x86 line and SR could be the end of the line unless GFL gets it's act together.


*FD-SOI would be needed to push the performance high enough by using the buried gate. The bulk 20nm process would allow for more xtors, but not likely to achieve higher clocks typical of modern CPUs at the same time - leaving AMD with the task of getting a huge bump in IPC out of XV. Kaveri will give us a clue as to what to expect from GFL bulk, IMHO.

 

[Olikan]

well, at least now.....
it seems that the new manager team aren't lunatics, like the faildozer team

 

[CHADBOGA]

Well it is good to see AMD management finally admit what anyone with a third digit in their IQ knew from the moment Bulldozer was launched. :awe:

 

[os2wiz]

Excellent point Revenger. Though I wish it wasn't so, I doubt my AM3+ mbs will be the preferred chipsets for SteamRoller, if it does get released.

My hunch is the SteamRoller core will be released with a new socket and chipset much the same way as Intel is doing with Haswell.

Your hunch is dead wrong. It has been explicitly stated on more than one occasion that Steamroller FX will run on AM3+ socket. Where do you get these notions from??? Perhaps Steve Ballmer's snuff box?

 

[Maximilian]

At least they still have moar cores!

Its that time again! :awe:

 

[itsmydamnation]

Wow, sig worthy.

Anybody know if JF has landed a new job?

Edit, yes, he's a marketing VP at NextIO.

I dont get your peoples obsession, your like a partisan shill it politics, must be a sad life having so much loathing for something so unimportant in the grand scheme.

I don't think "unmitigated failure" leaves much hope for Steamroller, Excavator, Concrete Mixer or any other chip from this family.

then stop thinking, it obviously hurts.

Do you still expect future products from an "unmitigated failure"?

See case in point a nonsensical statement, bulldozer isn't producing the next iteration of cores people are.

 

[Phynaz]

Your hunch is dead wrong. It has been explicitly stated on more than one occasion that Steamroller FX will run on AM3+ socket. Where do you get these notions from??? Perhaps Steve Ballmer's snuff box?

Wow, a troll comes out of the woodwork.

 

[Phynaz]

I dont get your peoples obsession, your like a partisan shill it politics, must be a sad life having so much loathing for something so unimportant in the grand scheme.

Funny that you felt it's important enough that you felt the need to quote me.

 

[NTMBK]

Your hunch is dead wrong. It has been explicitly stated on more than one occasion that Steamroller FX will run on AM3+ socket.

Has it? Could you show me the links? (This isn't a trolling thing, I'd just like to see; I haven't seen anything confirming that, and it'd be good to know.)

 

[Sheep221]

Bulldozer was killed especially by its marketing rather than design.
The problem is the AMD advertised BD to be true 8 core experience, the 8 core CPU that performed slower than the competing quad cores. The CPU can have 100 cores but if it won't have computing power of 100 cores of competitor, you don't basically sell it as a 100 core, or you fail.
Just a bit better power optimizations and branding as dual cores with added threads and a bit lower prices would pwn the intels ass like nothing. The overclocking capabilities, turbo and greatly higher performance over intel's DC offerings would probably cause them to be selling like a hotcakes and greatly increasing revenues, they so badly needed at the time the Bulldozer was launched. Ideal for home computers and gamers but also enthusiasts, which was their target market for BD.

 

[Vesku]

"Advanced Micro Devices had no plans to release low-power x86 server chips until the release of its ARM-based servers in 2014"

What are they smoking at AMD? They really thought they could just sit out for a couple of years and then show up with their ARM designs having laid no many efficient core foundations?

Isn't there enough examples of companies making business acquisitions, in this case Seamicro, and squandering their potential by putting them on a shelf as if they could just be stored until needed?

I was kind of expecting to read about some lessons learned after the "yes your right it was a failure" admission. Instead it's "well we weren't really going to do much in this space until mid late next year but then someone with some brains pointed out how we could deliver this product line without spending much money..."

Now I have serious doubts they have properly focused their products. Seemed pretty obvious to me to pursue maximum efficiency through their low power x86 line (adding in ARM models when they are ready) and refocus for more performance and providing more transistors (useful ones ;p )/$ in their mainstream x86 line. Reinforces the zombie company jibes, imo.

 

[Ajay]

What, AMD has hired some key new players and one of them may well have said 'I think we are missing an opportunity here'. That's why you build a new management team.

 

[Vesku]

What, AMD has hired some key new players and one of them may well have said 'I think we are missing an opportunity here'. That's why you build a new management team.

That they wouldn't have half or more of the company telling them they have an opportunity is the WTF part. Before reading that article, and doing some more checking to make sure it wasn't just that writers spin, I just assumed they had plans to have jaguar in dense server plans. Wouldn't they have those plans going all the way back to the first canceled Brazos refresh? Apparently they didn't and that's sad.

 

[NTMBK]

Kind of surprised that the only parts they have ready for this are Kabini with various parts fused off. They should have had a quad (or even oct) core die with a tinytinytiny GPU (the smallest thing that would still support remote desktop), with all the video decode etc stripped out. I thought that rapidly producing very tailored parts was the whole idea behind Jaguar, and their semicustom part push in general? So why is a harvested tablet chip trying to be a server chip?