AMD Appoints Dr. Lisa Su as President and Chief Executive Officer

[HumanSmoke]

I wholeheartedly disagree.
We have one purpose. As high of performance within a reasonable TDP as possible.
That is of course a huge oversimplification, but to say they don't understand that isn't true, what they have done is to allow the design teams to run in multiple directions and end up getting to the end of their leash still running full tilt.

The problem there is that the design goal was an architecture aimed squarely at the high ASP enterprise sector. The APU design as originally envisioned (10h based) wasn't actually too bad, and had AMD's foundries (later GloFo) delivered Llano might have had more impact. It was a management decision to starve the company of R&D funds, it was a management decision to deprive their own foundry business of income from third party fabbing, it was also a management decision to pursue a high overhead CPU architecture* when Intel bet on a mobile-centric high IPC arch, and it was also a management decision to sell off AMD's baseband and mobile GPU IP, which kind of indicates where their thinking was in 2006-07 when Bulldozer was being drawn up.

* “Bulldozer” is actually very well aligned with server workloads now and on into the future. In fact, a great deal of the trade-offs in Bulldozer were made on behalf of server, and not just one type of workload, but a diversity of workloads. It's very important that a server not just do one thing well; it needs to do a range of things well - AMD Chief Architect Chuck Moore

Whoever keeps dreaming up the shitty ideas for "modules" when we still have so many single threaded applications needs to be beaten.

Good luck with that. Mike Butler, Bulldozers architect, has been at Samsung for almost two years.

AMD had a great idea, dual core, then quad core, then its like they made that a god, always make more cores, never waiver on the making of more cores, instead of better and more logic throw in two or four more cores.

Quite simply, AMD blinked in a stare-down with Intel. AMD had a vision of what they thought was competitive, Intel talked up Nehalem, and AMD scuttled back to the drawing board to make an architecture better than Intel's vapourware (remember Nehalem as originally touted by Intel was supposed to be closer to Sandy Bridge with integrated graphics). Development basically stopped and was repurposed while AMD dithered over what they needed to include to counter Nehalem - which is what I meant by AMD's management lacking a clear end goal. AMD's goal wasn't a definitive product, it was some nebulous accretion of features added depending upon what Intel was doing allied with a lot of crossed fingers that their foundry could produce on time.

When performance failed to materialize lets throw hotdogs down this hallway in the form of PR spins and pure Ghz to fill up the gaps.
AMD FX 85Ghz, now with higher TDP!! (Not for use below the arctic circle, may cause permafrost melt, may kill polar bears, may cause heat stroke)

Probably not a lot else that can be done TBH. A CPU architecture takes upwards of five years to develop and bring to market. Bulldozer as originally envisioned would have been competing with Intel's Bloomfield and Lynnfield desktop processors - all on 45nm. Halting development and repurposing the design meant that the roadmap slipped and things got pretty unfavourable pretty fast - behind in architecture, behind in IPC, behind in process node.

AMD's problem is that management have almost always been reactive rather than proactive. When they've grabbed the bull by the horns it has paid off handsomely (adoption of MIPS features within x86 architecture, AMD64), but too often they tend to be content to play follow-the-leader. It's hard enough to develop product on your own timetable without changing horses midstream to also keep up with a competitors cadence.

 

[cadaveca]

AMD's problem is that management have almost always been reactive rather than proactive. When they've grabbed the bull by the horns it has paid off handsomely (adoption of MIPS features within x86 architecture, AMD64), but too often they tend to be content to play follow-the-leader. It's hard enough to develop product on your own timetable without changing horses midstream to also keep up with a competitors cadence.

That's "Asset Light" for ya, can't do much else since any IP is an asset, and any un-used IP is a waste of R&D. So you build IP to meet market demands... not push the market towards your IP.

 

[HumanSmoke]

That's "Asset Light" for ya, can't do much else since any IP is an asset, and any un-used IP is a waste of R&D. So you build IP to meet market demands... not push the market towards your IP.

Probably AMD's problems in a nutshell. With limited R&D, prioritization and picking the right direction become paramount. It probably doesn't help that AMD and ATI have historically overachieved in developing IP and any subsequent IP sales that eventually bear massive fruit for the new owners looks doubly bad for AMD ( the mobile IP sale to Qualcomm for example). The scattergun approach to development from such a creative environment also works against the company in other ways - so many new ideas that focus gets lost, and the incorporation of IP that never gets proper recognition or is marketed to the extent it should be - kind of like when Nvidia brought out G-Sync and someone at AMD belatedly realised that AMD GPUs had supported variable VBLANKing for a few years.

As for the not pushing the market towards your IP....I'd agree in AMD's case. They aren't a software company so they concentrate on providing the tools while the software coders develop the work they'll be used for. The strategy of developing and shaping a market based on your IP only works through force of will with a definite goal in mind. Nvidia's CUDA would be a prime example. They created the enterprise GPGPU sector - and likely remain a viable company because of it's development. Six years (until the advent of Xeon Phi) with virtually no competitor in the professional graphics/ math co-processor segments certainly allowed the company to shape many markets (medical imaging, scientific/economic forecasting and simulation, CG rendering etc.) - basically a continuation of the ethos that SGI's pro graphics division had already been employing up until it became part of Nvidia back in 1999

 

[cadaveca]

I don't think NVidia would have been so successful with CUDA if that hadn't taken over @ Stanford when AMD bought ATI. UP till that point, GPGPU was primarily ATI's forte, and even Gabe Newell complained about NVidia's lack of math capabilities. Yet that stands as the perfect example of how years of investment on both hardware, and software paid off for NVidia, while when it was ATI that had the strong footing, there still wasn't much software-wise from K.Y. HO. It's funny that nearly immediately after NVidia bought SGi, ATI launched the Radeon branding, already back then a bit behind NVidia.

It's not an easy road for anyone at the helm for AMD, to be honest. Intel's killing it with tablet silicon, NVidia's got GPGPU tightly leashed, and next-gen consoles are a ways off yet. I see the next few years as stationary for AMD; they can't do a whole lot of R&D, and their products do sell within their capacity, a big change from the past when they couldn't meet demand. They need to have a win in some segment, and for some time I thought that maybe holding onto Qualcomm might have saved them...

 

[HumanSmoke]

I don't think NVidia would have been so successful with CUDA if that hadn't taken over @ Stanford when AMD bought ATI. UP till that point, GPGPU was primarily ATI's forte, and even Gabe Newell complained about NVidia's lack of math capabilities.

Wasn't that more in relation to floating point calc and gaming? I remember SGI's Visual Workstations being pretty much the gold standard at the time - and while you have to take the whole system into account, their VPro VR cards were pretty much standard fit out (except on the 540 which I think used the Cobalt graphics chip). The VPro VR7 is just a rebadged Nvidia Quadro2 Pro

Yet that stands as the perfect example of how years of investment on both hardware, and software paid off for NVidia, while when it was ATI that had the strong footing, there still wasn't much software-wise from K.Y. HO. It's funny that nearly immediately after NVidia bought SGi, ATI launched the Radeon branding, already back then a bit behind NVidia.

That whole chain of events that led to SGI eviscerating themselves is berserk . Handing over a profitable business to Nvidia for basically nothing was the least of it - Spinning off Cray, ditching MIPS for Itanium, deciding that selling Windows based workstations was a bad idea !?!?! These guys could teach kamikaze pilots a thing or two about ritual suicide.

It's not an easy road for anyone at the helm for AMD, to be honest. Intel's killing it with tablet silicon, NVidia's got GPGPU tightly leashed, and next-gen consoles are a ways off yet. I see the next few years as stationary for AMD; they can't do a whole lot of R&D, and their products do sell within their capacity, a big change from the past when they couldn't meet demand. They need to have a win in some segment, and for some time I thought that maybe holding onto Qualcomm might have saved them...

And that win needs to be in a high profile reasonably high margin sector. From what I've read, AMD are putting a LOT of faith in GloFo executing on time and in quantity. Hopefully for them (and us as consumers) it all works out. If AMD can bridge the process node gap, they'll have a much better shot at increasing their presence.

 

[cadaveca]

SGi was bought for the IP, not the staff. We both know that, as should most level-headed people.


And yeah you're right in your last comment. But let me tell you this... I'm in school now. my class has 33-34 people (depending on the class). 3-4 know what Intel is. The others think information about a subject, clandestinely achieved. That's 10%.

Intel isn't a household name, and neither is AMD. Yet how many houses have AMD-based APUs in their consoles?

Processes and performance won't give AMD anything, but perhaps some respect from PC enthusiasts. Nobody else really seems to care.

 

[HumanSmoke]

SGi was bought for the IP, not the staff. We both know that, as should most level-headed people.

You mean Nvidia getting SGI's pro graphics division ? That's a given. The company learned fairly early that collecting IP was the best way to both protect itself and generate revenue. Nvidia bought 3Dfx's property for the same reason, and it had less to do with SLI than the other IP that 3Dfx had developed and accumulated (It's probably also a given that the tiled resources patent that Nvidia just hit Samsung and Qualcomm with originates from GigaPixel IP that 3Dfx bought just before they went belly-up when 3Dfx was still smarting from Sega ditching their Blackbelt program in favour of PowerVR's own tiled solution for the Dreamcast)

And yeah you're right in your last comment. But let me tell you this... I'm in school now. my class has 33-34 people (depending on the class). 3-4 know what Intel is. The others think information about a subject, clandestinely achieved. That's 10%.

Brand recognition and awareness are sometimes much more subtle in a general context. A great many people might not know the company if asked, but I'd guarantee that a lot more people recognise the "Intel Inside" and "Core" case stickers and have some remembrance of the five-note Intel jingle. The company stands consistently high in brand awareness valuations.
Let me turn your question around a little....How many people come to tech forums for buying advice and include something along the lines of " I heard that Intel [or Nvidia] are better..." They have to had heard that somewhere, and many of these people are new forum account holders. I don't think that many are lurkers. People I build systems for (or upgrade) actually specify Intel without knowing what part(s) of the system Intel actually makes - as far as they are concerned they want a box that allows them to do a (pretty narrow) range of tasks and aren't tech savvy in the slightest.

Intel isn't a household name, and neither is AMD. Yet how many houses have AMD-based APUs in their consoles?

The difference there is that when you come to buy a computing product, the logo and the advertising where you buy it are front and centre. Past associations tend to kick in at that point. As for consoles, most people that own consoles only associate the vendor (Sony, MS, Nintendo etc) with the hardware. Even fewer people are aware of Kingston and Micron as brands, but their products ship in a vast number of systems.

Processes and performance won't give AMD anything, but perhaps some respect from PC enthusiasts. Nobody else really seems to care.

Unfortunately true. All it could do is get an APU sufficiently higher in the perf/watt stakes to be considered for a wider range of products. That being said, AMD's best bet is also to cultivate better OEM relations. Selling to the DIY'ers is inconsequential next to the volumes that Lenovo, HP, Dell etc. order. That could possibly be part of the brand awareness issue as well. Intel has sufficient power to have its brand elevated alongside that of the OEM. AMD being a smaller company doesn't have the same "position power" to ensure that.

 

[cadaveca]

The difference there is that when you come to buy a computing product, the logo and the advertising where you buy it are front and centre. Past associations tend to kick in at that point. As for consoles, most people that own consoles only associate the vendor (Sony, MS, Nintendo etc) with the hardware. Even fewer people are aware of Kingston and Micron as brands, but their products ship in a vast number of systems.

Kingston and Micron sell in high number because of OEM/store relationships only, IMHO. You barely see these brands on forums.

The last thing is that I have a Haswell-based tablet. AMD just cannot compete here. APUs, while neat, are too late. Battery life is just too important.

 

[HumanSmoke]

Kingston and Micron sell in high number because of OEM/store relationships only, IMHO. You barely see these brands on forums.

The same could be said for AMD's APU's inside the consoles. Does AMD have any marketing presence for the average XBone or PS4 user? Anything on the machine or packaging to indicate that the machine is AMD powered?
I'd go out on a limb and say a sizable number of forum members (who should be more tech savvy than average) couldn't quantify the APU hardware without a quick Google.

The last thing is that I have a Haswell-based tablet. AMD just cannot compete here. APUs, while neat, are too late. Battery life is just too important.

You're more than likely right. Carrizo is starting to look obsolete before it even debuts, and I doubt AMD will get much of a window of opportunity in any case. Worst case scenario for Intel, its employs economies of scale to drop prices and flood the market, The company has a lot of excess fab capacity.

 

[GhostRyder]

I'm in school now. my class has 33-34 people (depending on the class). 3-4 know what Intel is. The others think information about a subject, clandestinely achieved. That's 10%.

Same back in school for me, heck even in the data center 90% of the staff who literally work on the machines do not even know what's inside. A logo is about all people may know and sometimes its pretty interesting to see what they come up with.

Intel isn't a household name, and neither is AMD. Yet how many houses have AMD-based APUs in their consoles?
Processes and performance won't give AMD anything, but perhaps some respect from PC enthusiasts. Nobody else really seems to care.

As long as the machine works people do not care in most cases. OEM's really are the deciding factor because no amount of commercials, guys stomping on video cards, or whatever is going to be as helpful as walking into Best Buy and picking up a machine that has said processor and graphics card inside. Normal people are just not going to pay enough attention to that or care enough to spend the time and make a decision. Even with AMD APU's inside the consoles how many people who have no idea are still going to associate based on just what a big group keeps saying that Intel-Nvidia is the way to go without even realizing that there are alternatives. Most people also associate slow machines with the basic processor being the root cause of all the issues. Heck most people think tablets are superior to laptops/desktops without even realizing its the SSD inside making them faster at least that I run into on a daily basis.