Rumour: AMD cancels Steamroller/Excavator

[Game_boy]

http://semiaccurate.com/2012/11/19/amd- ... excavator/

"It looks like the big AMD core line that went Bulldozer, Piledriver, Steamroller, Excavator, has been revised to Bulldozer, Piledriver, the end. Pretty much the entirety of the AMD big core line was a casualty of the last round of layoffs. It looks like it is game over in that whole segment for the green team."

Cancellation of these amounts to withdrawal from the desktop and server markets, at least until ARM can compete there. There is an official AMD denial, but Charlie doesn't make up stuff on this scale.

[sschaem]

did you look at the update ? AMD denied it officially.

I actually wonder if this was "AMD bait" ... AMD is very secretive, and I guess this is the only way to get them to talk.

[Game_boy]

did you look at the update ? AMD denied it officially.

I actually wonder if this was "AMD bait" ... AMD is very secretive, and I guess this is the only way to get them to talk.

I already put that in my post. Denials don't mean much from a company whose stock is down 75% since March and has had three rounds of layoffs. They also shed no light on its veracity because AMD would not allow a rumour site to break that news, they will wait until the next earnings call or Analyst Day.

Given the layoffs, do you think AMD is capable of carrying on developing Bulldozer, Bobcat AND ARM products? That's three full CPU cores.

[chuckula]

I already put that in my post. Denials don't mean much from a company whose stock is down 75% since March and has had three rounds of layoffs. They also shed no light on its veracity because AMD would not allow a rumour site to break that news, they will wait until the next earnings call or Analyst Day.

The official denial actually added to the weight of Charlie's story in my mind. He regularly posts similar hyperbolistic crap about other companies who completely ignore him while what little is left of AMD's PR department bends over backwards to deny this one particular story....

I actually only semi-believe semi-accurate on this one (and that's more than I believe a lot of their stories). When it comes to pure-CPU systems that you as a consumer want to buy, then AMD is rapidly leaving that business. Don't expect anything post-Vishera for the AM3+ platform and don't expect there to be a successor to the AM3+ platform.

While Kaveri is definitely not a 2013 product, I think Charlie is jumping the gun when he claims that it has been cancelled and that there is no future for AMD in APUs above where the low-end Jaguar systems are aimed. AMD recently released statistics showing that 75% of its sales are from APUs and the majority of those are Trinity... you don't kill the product that is doing the best.

[I.S.T.]

There is an official AMD denial, but Charlie doesn't make up stuff on this scale.

I'm pretty sure there are at least ten examples someone could find...

Funniest one to me, and one first coming to mind is Fermi runs tessellation in software...

[Game_boy]

There is an official AMD denial, but Charlie doesn't make up stuff on this scale.

I'm pretty sure there are at least ten examples someone could find...

Funniest one to me, and one first coming to mind is Fermi runs tessellation in software...

http://semiaccurate.com/2010/02/20/semi ... 80-scores/

"There is one bright spot, and it is a very bright spot indeed. The tessellation performance in Heaven. On that synthetic benchmark, the numbers were more than twice as fast as the Cypress HD5870, and will likely beat a dual chip Hemlock HD5970."

He wasn't claiming it was slow, he was claiming Fermi didn't have a dedicated unit for it. And he was partly right: it scaled with shader count, AMD's was fixed. I think that's more him not understanding the tech.

I don't care to defend his record, undoubtedly he has been wrong, but I see no case that he's deliberately lied or made stuff up. His greatest weakness is the trashy style he writes the articles in which is basically trolling for hits.

[Captain Ned]

Charlie don't surf.

[just brew it!]

The whole site is very tongue-in-cheek, starting with the name; and if you still have any doubt about the intended double meaning of the name, look at the logo. I don't know why people get so bent out of shape that a site named SemiAccurate isn't always... erm... accurate.

He does get things right (or at least close to right) from time to time though. I hope this is one of the times he's wrong!

[ronch]

SemiAccurate is worse than Fudzilla.

However, I do acknowledge that AMD is spreading themselves too thin these days, with SR/EX, APU, Jaguar, GPU and ARM projects going on and with just peanuts in the bank. I won't be surprised if one or more of their projects get canned.

[MadManOriginal]

I'd bet that it got pushed back because GloFo wouldn't have had their hybrid 14nm/20nm process ready in time. If I'm remembering correctly, they are going straight from 32nm to that process and it is due out in 2014. Ouila, no Steamroller in 2013.

[Meadows]

Ouila, no Steamroller in 2013.

Perhaps you mean "voila".

Like I said on TR's FX-8350 review, it's the first AMD product I've considered buying in a long time, and at this rate, I might do it. (Originally I held a tentative "wait and see" approach for Steamroller. I might still wait for the new Piledriver stepping in 2013. Partly depends on how soon it comes.)
And if the rumour turns out true, it may become my last AMD processor.

[clone]

Cancellation of these amounts to withdrawal from the desktop and server markets, at least until ARM can compete there. There is an official AMD denial, but Charlie doesn't make up stuff on this scale.

while some hate Charlie I don't believe he ever outright lies, might get it wrong, may have a favorite, may misinterpret but no lies.

regarding AMD's X86, it's no surprise, Bulldozer wasn't really meant for desktop anyway and did more to signal AMD's departure than it did an effort to regain prominence.

[RobotHamster]

Personally I doubt that they are leaving, besides all the stuff about trading standards and all probably won't let them leave the laptop/desktop sector as then Intel would have a monopoly.
The fact that they nearly have that already is beside the point but they will probably kick up a fuss about them dropping their CPU lines.

Take Seagate for example (at least I think it was seagate) who bought HP's drive business was forced to sell all the manufacturing capabilities to and share all HP design Patents with Toshiba so that there wouldn't be a Duopoly of WD and Seagate.

Since there is already a Duopoly in the CPU space they would be hard pressed to allow AMD to leave the desktop sector.

Maybe they should try releasing a processor with 2 FP units and 1 int unit instead of 2 int and 1 FP unit, perhaps it will perform much better in games that way.

[just brew it!]

Maybe they should try releasing a processor with 2 FP units and 1 int unit instead of 2 int and 1 FP unit, perhaps it will perform much better in games that way.

That would probably be wasteful, I'm guessing most instruction mixes would not be able to keep both FP units fed. Keep in mind that while the FP units do the "heavy lifting" computationally, the integer units perform all of the logic. If they aren't kept balanced, one or the other is going to be waiting much of the time; all you're doing is wasting die space and power.

[Game_boy]

I'd bet that it got pushed back because GloFo wouldn't have had their hybrid 14nm/20nm process ready in time. If I'm remembering correctly, they are going straight from 32nm to that process and it is due out in 2014. Ouila, no Steamroller in 2013.

Steamroller products were meant to be GF 28nm, not 20nm or 14nm.

[clone]

Personally I doubt that they are leaving, besides all the stuff about trading standards and all probably won't let them leave the laptop/desktop sector as then Intel would have a monopoly.

while a valid moral position I'm not sure a business case is available for AMD to make given what they've got to offer.

[MadManOriginal]

I'd bet that it got pushed back because GloFo wouldn't have had their hybrid 14nm/20nm process ready in time. If I'm remembering correctly, they are going straight from 32nm to that process and it is due out in 2014. Ouila, no Steamroller in 2013.

Steamroller products were meant to be GF 28nm, not 20nm or 14nm.

Well, same reason then - GloFo is skipping 28nm. (Part of the same story - skipping 28nm for 14nm/20nm process which won't be ready in 2013 so even if AMD reworked Steamroller for that 14nm/20nm process it wouldn't be out in 2013.)

[RobotHamster]

Personally I doubt that they are leaving, besides all the stuff about trading standards and all probably won't let them leave the laptop/desktop sector as then Intel would have a monopoly.

while a valid moral position I'm not sure a business case is available for AMD to make given what they've got to offer.

Well what I was refering to was more of the agencies involved in trading standards won't let AMD quit the desktop CPU game as there is already an effective monopoly run by Intel but there is some competition from AMD so it can still be classed as a duopoly and for the past 8-10 years this has been the case, in the past year WD was in a situation to buy the Storage division of Hitachi (yes I got the companies totally wrong originally) and for the deal to be approved by the regulating agencies they had to give an asset package to Toshiba, read the rest here http://techreport.com/news/22553/toshib ... hi-storage

Since these agencies don't like to allow a duopoly in the hard drive business I think they would have a fit if AMD dropped the higher end desktop CPUs from their lineup leaving a monopoly of Intel at every performance point above AMD's A10 trinity APU's so they might either make intel allow more license of its products or provide financial aid to AMD

[clone]

I don't disagree but that would have to be done via an incentive not an imposed position, it's not like their is an agency that has the power to compel AMD to continue losing cash in a segment they can't compete in just to maintain a presence.

I believe the way the market has handled it in the past is by partners stepping in and sharing the risks, IBM worked with AMD to get SOI working in their fab process and AMD is announcing a partnership with ARM with 64..... HP & IBM "have been there" for AMD in the past, IBM by building supercomps with AMD processing when AMD had no platform division to support it and HP has scooped up a lot more AMD cpu's than any other OEM over the years, the U.S. government has made bulk orders with AMD as well using their product in some defense contracts and I wonder if that has had anything to do with it.... worth consideration.

in doing that they keep the development costs/risks at AMD and don't have to answer to shareholders as to why they took the risk in building cpu's and competing directly with Intel.... just throw a little coin every so often to keep em up and running.

AMD is still a far more interesting company than ARM by any comparison, their IP is 100 fold more expansive, I've found it disappointing that some ppl keep spouting off how much they want AMD gone so that a new player can magically appear so that they can continue to root for Intel..... it doesn't work that way and they are as stupid as they are naive for wishing it.