New AMD Roadmap for 2013 (donanimhaber.com)

[jones377]

http://translate.google.es/translat...-ailesi-icin-2013te-guncelleme-gorunmuyor.htm

It looks legit, but the usual caveats and cautions apply. Some highlights:

* Steamroller is completely MIA.
* Piledriver for AM3+ continues throughout 2013
* New 3rd Gen APU "Richland" in 2013 based on Piledriver and Radeon Core 2.0 (?). No word on what process it's being built on.
* At least Kabini/Jaguar appears to be on track

 

[ShintaiDK]

No surprise really. Its pretty clear where AMD is going, screaming for a buyout. And x86 got no value in such a buyout since it will be cancelled.

 

[mrmt]

It looks legit, but the usual caveats and cautions apply. Some highlights:

* Steamroller is completely MIA.
* Piledriver for AM3+ continues throughout 2013
* New 3rd Gen APU "Richland" in 2013 based on Piledriver and Radeon Core 2.0 (?). No word on what process it's being built on.
* At least Kabini/Jaguar appears to be on track

IDC posted some slides where AMD pointed out complexity as one of the features that made AMD relinquish Bulldozer chips and move to Kabini and ARM, which fits the likely scenario that Kaveri is going to arrive in 2014 only. It seems that the gap between Intel and AMD is rising too fast. Until now AMD didn't have the resources to compete on servers and high end desktops/workstations, now they don't have resources to engineer proper mainstream chips. There is only the very bottom market for them.

The main implication of this roadmap, if it is correct, is that a band-aided Trinity will have to bear the brunt of the Haswell punch, and with that AMD will be further down in the price ladder. We can expect another bad Q313 and Q413 for AMD. Expect another round of personnel cuts until there.

For 2014, things cannot be any bleaker, Kaveri won't change things too much, as it be facing Broadwell for most of its life on the market, and not Haswell. As Broadwell is a die shrink, Intel will have more TDP and more die space to play with, which means more SKUs with on-die memory. That should be enough to significantly erode whatever advantage AMD has in iGPUs. There isn't much AMD can do about it, as they will be two nodes behind Intel by 2014.

2013 is the make or break year for AMD. They must pull out of traditional x86 markets otherwise they are toast. If they can get new revenues streams they are on business. If they don't, we probably won't see a chapter 11 filling but a chapter 7.

 

[Idontcare]

http://translate.google.es/translate?sl=auto&tl=en&js=n&prev=_t&hl=es&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fwww.donanimhaber.com%2Fislemci%2Fhaberleri%2FDH-Ozel-AMDnin-FX-islemci-ailesi-icin-2013te-guncelleme-gorunmuyor.htm

It looks legit, but the usual caveats and cautions apply. Some highlights:

* Steamroller is completely MIA.
* Piledriver for AM3+ continues throughout 2013
* New 3rd Gen APU "Richland" in 2013 based on Piledriver and Radeon Core 2.0 (?). No word on what process it's being built on.
* At least Kabini/Jaguar appears to be on track

Really not a surprise, even Intel abandoned Netburst in the middle of its roadmap, cancelling Tejas and planned successor Nehalem (the netburst project by the same codename as the now famous core 2 derivative, sharing hyperthreading in common and not much else).

AMD needs to focus on what was working, not focus on fixing what wasn't. Bulldozer wasn't working, bobcat was, gpu was, apu was.

 

[ShintaiDK]

Kinda fits what you said earlier IDC. Keep it on roadmaps till release. Then axe people/project when former product(pilediver) is released to avoid any problems before.

 

[nonameo]

AMD has secretly been working on revamping the stars core...

>_>

...

:-/

 

[Ajay]

AMD needs to focus on what was working, not focus on fixing what wasn't. Bulldozer wasn't working, bobcat was, gpu was, apu was.

So we get Trinity 2 (28nm PD + CGN, IIRC) instead of Kaveri. Bummer. So Jim (Keller) needs to shout down to his engineering team - 'we need warp factor 10 Scotty!'.

So I wonder where Keller is going to start. It probably takes ~five years to produce a CPU with a completely new uArch. They could cut it down, perhaps to three years, if there is something useful to use laying around on the bulldozer design team's cutting room floor. 14/20 nm 8-10 core Thurban w/HT + AVX2 + HSA, et cetera anyone?

 

[ShintaiDK]

I have a feeling Keller is parked in making IP products for ARM cores.

Anyway, 1 man is nothing on his own.

 

[podspi]

Makes you wonder how/if AMD is going to keep their promise of AM3+ compatibility for Steamroller...

 

[Ajay]

I have a feeling Keller is parked in making IP products for ARM cores.

Anyway, 1 man is nothing on his own.

I hope not.

1 man can make a big difference if he can *lead* and has a clear *vision*. I don't know if Keller is that kind of guy.

 

[pelov]

AMD needs to focus on what was working, not focus on fixing what wasn't. Bulldozer wasn't working, bobcat was, gpu was, apu was.

So what about the Opterons? Forget the desktop chips, they're not important. As is the case with Intel/AMD, if you want to find out what the enthusiast/workstation platform holds, look at the server lines.

 

[ShintaiDK]

So what about the Opterons? Forget the desktop chips, they're not important. As is the case with Intel/AMD, if you want to find out what the enthusiast/workstation platform holds, look at the server lines.

Opterons are actually their weakest place.

 

[Insert_Nickname]

* Steamroller is completely MIA.

It properly got steam-rolled...

Sorry, couldn't resist... ROFL :thumbsup:

 

[pelov]

Opterons are actually their weakest place.

Well BDs suck at perf-per-watt.

What I mean is, AMD has said they'll be offering Opterons in the future so wouldn't that subsequently mean desktop variants?

 

[ShintaiDK]

Well BDs suck at perf-per-watt.

What I mean is, AMD has said they'll be offering Opterons in the future so wouldn't that subsequently mean desktop variants?

ARM renamed Opterons? APUs sold as Opterons? Ye.

But there might not be a future.

 

[inf64]

Yes they will offer desktop variants as well. It's just that SR based CPUs are 2014 now,probably due to GloFo's process node troubles. And the only change to the roadmap is Kaveri since Vishera was always on the 2013 perf. desktop roadmap and there was never a mention of a "new" FX chip based on SR core in 2013. It's likely to come in 2014 on a new socket with DDR4 memory and 4+Modules.

 

[nemesismk2]

Makes you wonder how/if AMD is going to keep their promise of AM3+ compatibility for Steamroller...

I am thinking the same thing

 

[Idontcare]

I hope not.

1 man can make a big difference if he can *lead* and has a clear *vision*. I don't know if Keller is that kind of guy.

I think the point is "who" is left for Keller to lead? Their is the morale issue for those who remain, and then the question of what type of engineer is going to typically hang around versus moving on, etc, when so much writing is on the wall.

If the Dilberts, the Alices, and the Asoks have all fled or been laid off, leaving behind a bunch of Wally's, Teds and Toppers then it doesn't really matter if the pointy haired boss is replaced by Keller...the team itself isn't exactly going to be cut from the finest of cloths.

It is in this sense that I have my concerns that Keller may be too little too late. He may be a genius and he may have a nice budget to make a miracle CPU or two for AMD, but it is going to take some work just to get a team of skilled and motivated engineers to work in that environment.

Won't be anything like the environment that the engineers faced at Apple with Keller in charge of them designing the A6. Honestly you have to be pretty hard up on times to take on an assignment at AMD if you can get a job offer from pretty much anywhere else at the moment.

So what about the Opterons? Forget the desktop chips, they're not important. As is the case with Intel/AMD, if you want to find out what the enthusiast/workstation platform holds, look at the server lines.

I thought it was rather foreboding the fact that when the supercomputer Titan was still called Jaguar it had exactly the same number of CPUs (they were Magny-Cours though, not bulldozer based) but it pulled in ~2.3PFlops.

Now that it has been "upgraded" to use 16-core bulldozer-based operton CPUs, the compute power that comes from the CPUs themselves has actually fallen from ~2.3PFlops to only ~2PFlops...the Nvidia GPU's account for 90% of the spec'ed 20PFlops capability of Titan.

Titan consists of 200 cabinet-sized Cray XK7 supercomputers. All told, it includes 18,688 nodes, with each node comprised of a 16-core AMD Opteron processor and an Nvidia Tesla K20 GPU. Nvidia tells us Titan can achieve over 20 petaflops of peak performance, and over 90% of those flops come from its GK110 GPUs.

source

That is quite the upgrade from AMD's side of the equation, 33% more cores and only 10% lower performance :thumbsup:

Is there any wonder why so few opterons are sold in the server segment anymore? Who wants them when that is what they have to offer versus the competition (or just stick with your existing stars-based opterons you bought 2 yrs ago).

 

[krumme]

+No mentioning of Kaveri on last earnings call from AMD.

Shit, it probably means i have to pay double for my next high performance cpu

Lets hope for a miracle or we will get a return to the pricing structure 20 years ago

This is worse than a stoneage market where you at least can buy sheeps from 3 different farmers. Hell even Apple have more competition.

 

[pelov]

That is quite the upgrade from AMD's side of the equation, 33% more cores and only 10% lower performance

Is there any wonder why so few opterons are sold in the server segment anymore? Who wants them when that is what they have to offer versus the competition (or just stick with your existing stars-based opterons you bought 2 yrs ago).

Well, when it comes to dumb FP operations your best bet is a GPU anyway. I guess this is just going to spark another CMT debate

 

[podspi]

Now that it has been "upgraded" to use 16-core bulldozer-based operton CPUs, the compute power that comes from the CPUs themselves has actually fallen from ~2.3PFlops to only ~2PFlops...the Nvidia GPU's account for 90% of the spec'ed 20PFlops capability of Titan.

Really? They don't gain anything from AVX, XOP, or FMA4? This is HPC after all, they ARE more than happy to recompile (or tweak assembly) for performance...

If that is true, I hope everyone involved got a refund...

 

[meloz]

That is quite the upgrade from AMD's side of the equation, 33% more cores and only 10% lower performance :thumbsup:

Oh, burn.

Anyway, I see this move as preparation for eventual sale to Apple. Rather blatant, too. AMD are going to marry ARM designed CPUs with their own proven GPUs, and hope (pray?) Apple bite before bankruptcy strikes.

AMD64.....they couldn't care less about that, now could they. With their restricted resources AMD simply cannot develop two different product lines.

Congratulations to Intel on their x86-64 monopoly. What tumultuous times we live in...

 

[ShintaiDK]

+No mentioning of Kaveri on last earnings call from AMD.

Shit, it probably means i have to pay double for my next high performance cpu

Lets hope for a miracle or we will get a return to the pricing structure 20 years ago

This is worse than a stoneage market where you at least can buy sheeps from 3 different farmers. Hell even Apple have more competition.

Its an illusion that price will change. Simply because higher price=lower profit for Intel due to lower volume.

 

[Idontcare]

Anyway, I see this move as preparation for eventual sale to Apple. Rather blatant, too. AMD are going to marry ARM designed CPUs with their own proven GPUs, and hope (pray?) Apple bite before bankruptcy strikes.

That is something I had not considered before. It makes a whole lot of sense!

AMD could also become a design foundry for Apple, turning out designs for them in exchange for royalties on a per-chip basis like the console gaming market has instituted.

I think the bottom line is that without diversifying away from x86, AMD's options are limited going forward barring some serious stumble by Intel. By diversifying AMD creates opportunities to make revenue and survive to fight another day or at a minimum create IP that makes it more attractive for a buyout.

 

[mrmt]

AMD could also become a design foundry for Apple, turning out designs for them in exchange for royalties on a per-chip basis like the console gaming market has instituted

What AMD brings to the table that would be unique? Because they won the console battle but this does not mean anything, as we know that console wins are a race to the bottom.