So what did Rory Read do wrong?

[bononos]

Since he wasn't responsible for bulldozer, where exactly did he mess up to get himself fired?

Seamicro? Skybridge?
Not going into mobile?
Making too heavy cuts in engineering?

 

[Redentor]

Rory Read is a commercial. He was hired to hire the right people (Koduri, Keller, Papermaster, etc.), to fire the useless people and to take care of AMD balance sheet. After the work has been done, he has left AMD. Lisa Su, on the other hand, is an engineer.

 

[NTMBK]

He wasn't responsible for Bulldozer, but he was in charge during AMD's complete abdication of the high end. They surrendered the desktop market, while coming out with subpar laptop parts. The only bright spot in the CPU lineup was the Jaguar lineup, but that took a nosedive when GloFo screwed up their 20nm and left them with no replacement.

 

[BigDaveX]

Not going full-tilt on Steamroller and Excacator-based FXes was something that did make some degree of sense, since they'd have made next to no impact against Haswell or Skylake, but his big mistake was not having anything replace them, resulting in a four-year gap - which, for perspective, is about equivalent to the entire lifespan of the K7 family until the Athlon 64 arrived - between Piledriver and Zen. I mean, even when Intel cancelled Tejas they at least released Pentium D and Cedar Mill, even if they knew that neither would have a prayer against AMD's line-up.

Personally, I'd have skipped over Steamroller, but released a 4-module Excavator-based FX on Socket AM4 this year, just to get the infrastructure for Zen going and allow easy drop-in upgrades when it eventually does show up.

 

[Insert_Nickname]

Personally, I'd have skipped over Steamroller, but released a 4-module Excavator-based FX on Socket AM4 this year, just to get the infrastructure for Zen going and allow easy drop-in upgrades when it eventually does show up.

Just a 4 module FM2(+) option would have been nice. Or a 3 module with basic 128SP IGP.

 

[mahoshojo]

Rory Read was hired to fulfill the board's mobile dream, back in 2011.
Mobile meant arm and arm server became another plan.
Later they acquired seamicro, which only met <50% of their revenue target after the merge.
Skybridge and CMO and seattle were fine. Every AMDer knew those were joke and AMD only spent really limited resource on those topics.

To summarize:
1.believed arm server.
2.believed seamicro.
3.believed GF 20nm.

The execution on jaguar related products was good. And some cancelled gpu projects like bermuda, I don't know it is good or bad.

 

[Idontcare]

Since he wasn't responsible for bulldozer, where exactly did he mess up to get himself fired?

Seamicro? Skybridge?
Not going into mobile?
Making too heavy cuts in engineering?

To answer your question you must go back even further in time, before he was fired.

Why was he hired? Why was Dirk fired?

Recall the BoD fired Dirk for his lack of vision and leadership direction in the mobile arena.

Rory was selected because he would make up for Dirk's lackings.

Did he? Did he deliver for AMD's shareholders a more robust product portfolio, a more diverse product offering, any vision or leadership into the mobile space?

Not really. At least not to the extent that the Board of directors were looking for, and so Rory was dismissed as Dirk was dismissed, and Lisa was hired.

 

[mrmt]

Since he wasn't responsible for bulldozer, where exactly did he mess up to get himself fired?

Seamicro? Skybridge?
Not going into mobile?
Making too heavy cuts in engineering?

Just look at how Rory worsened AMD strategic drivers during his term. By the time he assumed AMD the Bulldozer disaster was still starting to unfold but Intel roadmap and direction was already clear: Efficiency and costs. Nvidia was also for efficiency.

As for AMD product portfolio in relation to their competitors, they were just a bit handicapped on GPUs and they knew (or should have known) that their CPU business was doomed, because Bulldozer would become sashimi after a few Core interations.

What did Rory do? He basically stopped GPU developed, with token improvements each generation but nothing close to what NVidia delivered and with development restricted to one high-performance, high-cost SKU. Nvidia went towards cheaper and power efficient products, and in the end it basically imploded AMD GPU business.AMD is milking GCN until today with a few improvements to talk about.

And in terms of CPU? He also stopped the then-current CPU development, with Jaguar basically receiving a few tweakes and a planned die-shrink, which was canned because GLF was unable to deliver a production ready 20nm node. With the unmitigated failure family he killed the mainstream desktop products and went full steam ahead towards low power APU, the area which is a sore point for the Bulldozer family. As a result AMD CPU business also collapsed, not only they lost share in markets where they used to have a dominant position, they also failed to break into new markets they were trying to enter.

What of AMD cost structure, did Rory nail this one? No. Not only he lost the waiver that allowed him to manufacture CPUs at TSMC, he also misjudged the size of the cuts necessary to keep AMD in a sustainable path, so we watched more than one heavy personnel cut with deleterious effects on AMD R&D teams and supply chain.

But what about of the new ventures he sponsored: Semi-custom, SeaMicro, HSA, Zen, Mantle, Memory and SSDs, Skybridge, K12... and of those Semi-custom was a dud, because the only relevant contracts were the consoles, Skybridge was canned, HSA is vaporware, AMD is still a nobody on the memory and SSD market, SeaMicro was shut down, K12 was delayed and Mantle was shut down. Only Zen remains "on track" so far, and it seems that it is already being delayed;

With Rory AMD neglected the markets they were in and tried a shotgun approach to get new markets, and failed miserably. Today AMD is stuck where they are before Rory, x86 CPUs and GPUs but in a much more precarious position on every single market you can think about and even more tied to Globalfoundries, which quite frankly, is a tremendous liability for AMD.

Rory and his team (Lisa Su included) doesn't have much success to brag about, for someone who was boldly talking about resetting and reinventing the company, ended up smaller, cheaper and with a messed up pipeline is not acceptable.

 

[DrMrLordX]

I don't think any of you would have liked 2m/4c Excavator on the desktop assuming it was on 28nm + HDL. It has inferior voltage scaling compared to GV-A1 Kaveri past 2.6 GHz or so, at least according to The Stilt's tests on this TDP-unlocked Carrizo machine. The improvement in IPC is nice, but otherwise, meh. It's fine as a mobile processor, at least where it is not gimped by TDP limitations.

AMD had a shot at upsetting the apple cart (er, Intel cart?) with HSA, and they blew it. Read's greatest failure is in not executing on the Future of Fusion. So instead of seeing even a modicum of HSA-enabled software running quickly on Kaveri, instead we see a future unfolding in which OpenCL2.0 apps trickle out and run just as well on Intel Gen9 iGPUs as they do on GCN-based iGPUs. Or close enough that it doesn't really matter.

 

[Phynaz]

Rory Read was a salesman, not a CEO type. As IDC said, lack of vision was his shortcoming. Under his leadership AMD was scattershoting products all over the place. Sea Micro? Really?

Su rightfully took a hatchet to what Read built. She may be the best person AMD has had in the corner office in well over a decade.

 

[mrmt]

I don't think any of you would have liked 2m/4c Excavator on the desktop assuming it was on 28nm + HDL. It has inferior voltage scaling compared to GV-A1 Kaveri past 2.6 GHz or so, at least according to The Stilt's tests on this TDP-unlocked Carrizo machine. The improvement in IPC is nice, but otherwise, meh. It's fine as a mobile processor, at least where it is not gimped by TDP limitations.

This is the kind of strategic mistake I was talking about. Instead of pursuing desktop sales that could play into Bulldozer strengths, he tried to chase the mobile market with an inadequate product... and ended up losing the two markets.

AMD had a shot at upsetting the apple cart (er, Intel cart?) with HSA, and they blew it. Read's greatest failure is in not executing on the Future of Fusion. So instead of seeing even a modicum of HSA-enabled software running quickly on Kaveri, instead we see a future unfolding in which OpenCL2.0 apps trickle out and run just as well on Intel Gen9 iGPUs as they do on GCN-based iGPUs. Or close enough that it doesn't really matter.

Oh, I forgot AMD's favorite vapoware. Gotta edit the previous post.

 

[cbn]

Just a 4 module FM2(+) option would have been nice. Or a 3 module with basic 128SP IGP.

For mobile I got to thinking/wondering how the three modules with high performance libraries would have compared to two modules with high density libraries for performance per watt metric?

Obviously high density library is better at low voltage, but then with three modules AMD could have spread power consumption over more cores (which helps performance per watt in a different way).

In any event, I do wish there was a better CPU upgrade for FM2+ owners. Also would be great to see Vishera prices drop, although I do not know how far that could go due to the large 315mm2 die created by having the large L3 cache on board?

 

[bononos]

To answer your question you must go back even further in time, before he was fired.
Why was he hired? Why was Dirk fired?
Recall the BoD fired Dirk for his lack of vision and leadership direction in the mobile arena.
.........

I have a vague idea about why Read was fired but I remember someone posting about Read's 2 big blunders and I was curious if there was really 2 really big reasons why he was canned.

..........
..........
And in terms of CPU? He also stopped the then-current CPU development, with Jaguar basically receiving a few tweakes and a planned die-shrink, which was canned because GLF was unable to deliver a production ready 20nm node. With the unmitigated failure family he killed the mainstream desktop products and went full steam ahead towards low power APU, the area which is a sore point for the Bulldozer family. As a result AMD CPU business also collapsed, not only they lost share in markets where they used to have a dominant position, they also failed to break into new markets they were trying to enter.

What of AMD cost structure, did Rory nail this one? No. Not only he lost the waiver that allowed him to manufacture CPUs at TSMC, he also misjudged the size of the cuts necessary to keep AMD in a sustainable path, so we watched more than one heavy personnel cut with deleterious effects on AMD R&D teams and supply chain.

But what about of the new ventures he sponsored: Semi-custom, SeaMicro, HSA, Zen, Mantle, Memory and SSDs, Skybridge, K12...........

Should Read be blamed for Jaguar's 20nm failure which was Glofo's fault?

I agree that he cut alot of engineering staff which hurt AMD in the long run and I'm curious whether the BoD were in the loop that K12/Steamroller big core were never going to be developed, but Read continued to imply that they were back then, maybe to lessen the negative impact on morale/stock prices.

And wasn't Jim Keller the point man on K12/Seamicro/Skybridge? Was he also fired for supporting Rory? Just not as soon because AMD needed him on Zen?

 

[NTMBK]

But what about of the new ventures he sponsored: Semi-custom, SeaMicro, HSA, Zen, Mantle, Memory and SSDs, Skybridge, K12... and of those Semi-custom was a dud, because the only relevant contracts were the consoles

I wouldn't call it a dud; the consoles are still good business, even if they're nowhere near mobile levels of revenue. It's a stable revenue stream, and AMD needs all of those that it can find.

Skybridge was canned,

The GloFo 20nm screw ups didn't help with this either.

HSA is vaporware,

...and will remain so for the forseeable future. AMD don't have market share to push radical changes to software model.

AMD is still a nobody on the memory and SSD market,

Do they even still make them?

SeaMicro was shut down,

Such a waste of money.

K12 was delayed

...at the minimum...

and Mantle was shut down.

Mantle formed the basis of Vulkan, and helped inform DX12- if it weren't for Mantle, I doubt we would be seeing features like async compute (which are a big help for AMD), and perhaps not the "thin" API style which limits the damage from AMD's driver teams.

 

[Idontcare]

I have a vague idea about why Read was fired but I remember someone posting about Read's 2 big blunders and I was curious if there was really 2 really big reasons why he was canned.

AMD's biggest problems has always been their board of directors. A commonality they share with HP, unfortunately.

Rory may have been hired on a whim, and fired on a whim, for all we really can be certain of.

When they hired him I thought it was so he could leverage his business connections and assist AMD in developing a marketing and sales team that could secure more design wins at OEMs. That was just my own personal view/hope, not anything based on rumors or whispers.

In conclusion, I wish I could be fired from a job and retain as much financial compensation as Rory. Hell I'd happily be fired for 1/10 the cash!

 

[Soulkeeper]

I'm with Idontcare on this one.
Every since sanders left the company has essentially been run by the bod, the CEOs have been puppets. Some might dissagree with me on this, but I actually think dirk was the best of the puppets and got canned cause he didn't share the bod's vision. Although I doubt anything would be much better had he remained. Also canning a CEO gives a false sense of hope to shareholders.

 

[mrmt]

Should Read be blamed for Jaguar's 20nm failure which was Glofo's fault?

Yes, he should. When he assumed help AMD had a waiver which would allow AMD to manufacture Skybridge at TSMC, today they don't.

I agree that he cut alot of engineering staff which hurt AMD in the long run and I'm curious whether the BoD were in the loop that K12/Steamroller big core were never going to be developed, but Read continued to imply that they were back then, maybe to lessen the negative impact on morale/stock prices.

I think Rory misjudged the amount of cuts he would have to make in order to make AMD sustainable because Rory screwed up big time in predicting the amount of sales they could get with their current product portfolio, especially on three points: He seems to be caught off guard by the bay trail ramp up, which wiped out the Jaguar business in a couple of quarters and he didn't seem to think Kaveri would fare that bad against Haswell, and he thought he could charge premium prices on GPUs like Nvidia and lost a lot of share to learn that he couldn't.

When those three events impacted the business, eating gross profits that should have funded R&D, AMD had to make further rounds of cuts and thus screwing up the R&D pipeline. I don't think Rory couldn't hide this fact from the BoD, if anything because this reality would be plain simple to explain: If you ask for a resource and don't have it anymore either you won't be able to accomplish the job or you didn't need the resource in the first place. So I don't think he could have rid this situation from the BoD.

In the end, Rory was unable to give direction to the company and neither improve the commercial chain he knows so well, what's the point of him being there?

 

[AtenRa]

To answer your question you must go back even further in time, before he was fired.

Why was he hired? Why was Dirk fired?

Recall the BoD fired Dirk for his lack of vision and leadership direction in the mobile arena.

Rory was selected because he would make up for Dirk's lackings.

Did he? Did he deliver for AMD's shareholders a more robust product portfolio, a more diverse product offering, any vision or leadership into the mobile space?

Not really. At least not to the extent that the Board of directors were looking for, and so Rory was dismissed as Dirk was dismissed, and Lisa was hired.

I really do not think Dirk and Rory were fired for those reasons.

Dirk created the small core Cat family, the AMD was the first with x86 Windows Tablets in 2011-12 when Intel had nothing even close to it. They clearly had the best product in 2014 with Beema and Mullins against any Intel ATOM based APU but because of the famous Intel contra-revenue policy, Intel completely destroyed the AMD Tablet competition.

Rory was not fired because he didnt deliver new hardware, you cannot create new hardware in just 2-3 years. Rory changed how AMD is operating from now on. AMD is smaller and more maneuverable than it was in 2011.
I believe hes job was to make AMD survive for 3-4 years until the new designs and products will hit the market in 2016 and later. Also to make Su ready for the big chair.
They managed to win the consoles under Rory and that is what will keep the company afloat until ZEN and GCN 2.0 hit the road.

And just to remind everyone, we really dont know if both Rory and Keller were fired or left on their own. We only guessing at this time.

 

[AtenRa]

When they hired him I thought it was so he could leverage his business connections and assist AMD in developing a marketing and sales team that could secure more design wins at OEMs. That was just my own personal view/hope, not anything based on rumors or whispers.

That was my thoughts as well back then. It turned out to be something else i believe.

 

[mrmt]

I wouldn't call it a dud; the consoles are still good business, even if they're nowhere near mobile levels of revenue. It's a stable revenue stream, and AMD needs all of those that it can find.

Depends how one look at it. Is semi-custom bringing much needed money to AMD? Yes, it is. But is semi-custom the transformational business Rory was talking about circa 2012, the business that Rory was keen to slave every single other AMD project (Everything must be reusable, everything must be foreign-IP friendly), the business that would be half of AMD *before* the Jaguar and Kaveri crash? I don't think it is.

The GloFo 20nm screw ups didn't help with this either.

They had a waiver, they lost the waiver under Rory.

Do they even still make them?

they're on sale at newegg.

Mantle formed the basis of Vulkan, and helped inform DX12- if it weren't for Mantle, I doubt we would be seeing features like async compute (which are a big help for AMD), and perhaps not the "thin" API style which limits the damage from AMD's driver teams.

I think Mantle was more a PR move than anything. AMD is basically selling 2011 technology today with a few tweaks here and there, it's the same basic architecture and in some cases the same actual chip of that time, so how to make these old chips "sexy" again? A software piece that would give AMD cards an edge would be a nice idea in theory, and how to do it quickly? Publishing a DX12 lite, since the real DX12 specs would be already frozen and AMD would have to be testing it anyway in order to build the drivers withing WHQL levels of quality.

It's no surprise that Maxwell and Skylake Gen9, two newer architectures, support DX12.1 while AMD GCN does not. Intel and Nvidia wouldn't be able to support it if Microsoft didn't have finished the DX12 specs when those two architectures were still on the design phase.

 

[NTMBK]

I think Mantle was more a PR move. AMD is basically selling 2011 technology today with a few tweaks here and there, it's the same basic architecture and in some cases the same actual chip of that time, so how to make these old chips "sexy" again? A software piece that would give AMD cards an edge would be a nice idea in theory, and how to do it quickly? Publishing a DX12 lite, since the real DX12 specs would be already frozen and AMD would have to be testing it anyway in order to build the drivers.

Oh come on, if DX12 was ready that early, they would have shipped it on the XBox One from day 1. Instead they went with Direct3D 11.X. Battlefield 4 was a day 1 launch XBox One game, and yet DICE helped build Mantle at the same time. If you look at it from DICE's perspective, they pretty obviously weren't happy with the state of DirectX (both 11.2 and 11.X), and wanted to influence the direction of DX12.

 

[mrmt]

Oh come on, if DX12 was ready that early, they would have shipped it on the XBox One from day 1. Instead they went with Direct3D 11.X. Battlefield 4 was a day 1 launch XBox One game, and yet DICE helped build Mantle at the same time. If you look at it from DICE's perspective, they pretty obviously weren't happy with the state of DirectX (both 11.2 and 11.X), and wanted to influence the direction of DX12.

With a 18+ months gap between the XBO and the W10 launch I think Microsoft had to do something if they were to keep the XBO launch date, and that thing was Direct3D 11.X. It's either this, or thinking that Microsoft would scrap whatever they had in development and bringing DirectX 12 in its current form in just about 18 months.

 

[NTMBK]

With a 18+ months gap between the XBO and the W10 launch I think Microsoft had to do something if they were to keep the XBO launch date, and that thing was Direct3D 11.X. It's either this, or thinking that Microsoft would scrap whatever they had in development and bringing DirectX 12 in its current form in just about 18 months.

Given that AMD were the GPU vendor, they will have worked closely with Microsoft on developing an API (and more importantly, toolchain) to exploit their hardware. DICE worked on D3D 11.X games, AMD at the very least worked closely with Microsoft on implementing D3D 11.X, so Microsoft's latest thinking (11.X) most likely helped inform Mantle's design, which then in turn helped inform the design of DX12. It's not like it's a one way street.

 

[Dribble]

Mantle development was probably never ahead of DX12, it's just that they released it early. It's much easier to release something that's only required to work for a few specific chips (a few GCN chips) and only in a few specific ways (a few frostbite games) and then fix the bugs for those use cases. MS had to make DX12 fully functional across many different architectures and for many different purposes. That takes a lot more effort to get it right.

 

[flopper]

I think Mantle was more a PR move than anything. AMD is basically selling 2011 technology today with a few tweaks here and there, it's the same basic architecture and in some cases the same actual chip of that time, so how to make these old chips "sexy" again? A software piece that would give AMD cards an edge would be a nice idea in theory, and how to do it quickly? Publishing a DX12 lite, since the real DX12 specs would be already frozen and AMD would have to be testing it anyway in order to build the drivers withing WHQL levels of quality.

It's no surprise that Maxwell and Skylake Gen9, two newer architectures, support DX12.1 while AMD GCN does not. Intel and Nvidia wouldn't be able to support it if Microsoft didn't have finished the DX12 specs when those two architectures were still on the design phase.

What?

Mantle = Dx12/Vulkan a world dominATIng software from AMD.
PR?
Dx12 suits GCN better than Maxwell from nvidia, how can you say it does not?
AMD designed Mantle so it could run on their hardware they designed to be run for.
Flexible compute dx12 goodness.

AMD have been mismanaged for the last 10 years.
not even sure if Lisa Su can fix it.