AMD Q3 Results

[Ramses]

Nobody ever mentions that because it is just a small research project. Intel, IBM and Nvidia also have grants or in case of Intel even product projects aiming at exascale computing, something that AMD is not even close to get.

Seems if they are downscaling and have half the CPU engineers they did, it'd be a more significant project. That isn't small change, of the 60mil I read the gov doled out(recently), amd got half and NV and Intel and some other outfit got the rest. I'm a lot more interested in that than I am in a new CPU that runs the witcher really well myself.

More the point is to me, that we don't know what they are doing.
Trying to work backwards from the result to find the problem without any data seems.. Fruitless. I get that some people enjoy this sort of thing, but everyone is getting all uptight when nobody knows anything that isn't available via Google.

 

[Phynaz]

Will AMD stock ever go up again this year (and by "up" I mean to $4 levels)? What do you guys think? Does it even matter if AMD posts good results?

Not a chance. You're talking nearly doubling their market cap.

 

[Shehriazad]

I feel like Q2 2015 and Q22016 are what will either drive the last nails in the coffin or will save AMD.

Q2 15 because of the 300 series GPUs and Q2 16 because of Zen.

IF the high end models of the 300 series manage to actually release with HBM like was planned...BEFORE Nvidia releases Pascal...then that might be a huge income source...because with no competition in that regard they wouldn't be forced to sell ridiculously low just to sell their GPUs

As for Zen...no big hopes there...The Semicustom at least for consoles is just a little bandaid fix. They're probably doing it for a few cents (figuratively speaking) per sold console.



At this point my gut tells me (yes, absolutely no proof behind this) that AMD has about a 5-10% chance to ever become relevant again in the Desktop market. Half of that chance would've been APU...but with Carrizo out of the window...who knows.

 

[jpiniero]

I am convinced AMD is going bankrupt eventually; which of course wipes out shareholders. The question then will be whether they could emerge from bankruptcy and what they would have to give up to settle the $2B in debt and getting out of the WSA.

 

[mrmt]

Seems if they are downscaling and have half the CPU engineers they did, it'd be a more significant project.

It will be a more significant project for AMD, but not for the market itself. As a matter of fact the companies leading the pack are growing in size, while AMD is shrinking, and that means smaller projects. The "new" AMD won't be able to take projects of the same size and scope of its competitors, and not even projects that the "old" AMD had resources to do.

More the point is to me, that we don't know what they are doing.

Oh yes, we do. AMD isn't a skunk works division, they have public shareholders and are accountable to these shareholders, so they must say periodically how they are spending the shareholder's money. The fact that they aren't saying much about this project and a few others they should have on their R&D pipeline says more about the return perspective of these projects (small, if any) than any possibility of AMD working on the next big thing.

Given that Exascale will place a lot of burden on the interconnect and that AMD is completely outclassed in this area (QPI, NVLink), I wouldn't place too much faith on it regardless on how many exploratory projects they win.

 

[Abwx]

Given that Exascale will place a lot of burden on the interconnect and that AMD is completely outclassed in this area (QPI, NVLink)

Lol, so much outclassed that QPI, Quick Path to Innovation, is a carbon copy of Hypertransport, that is a point to point bidirectional bus....

 

[Arachnotronic]

Not a chance. You're talking nearly doubling their market cap.

AMD is quite volatile so big swings can happen (both to the up and down side), but I agree...doubling over the next year without a dramatic improvement in its fundamentals is unlikely.

 

[DrMrLordX]

Lol, so much outclassed that QPI, Quick Path to Innovation, is a carbon copy of Hypertransport, that is a point to point bidirectional bus....

Uh? Look, I'll be first to give AMD props for growing beyond the FSB (HT came first, by a long shot), but QPI is not a "carbon copy" of Hypertransport. It's a ring bus, for one thing . . .

 

[Ramses]

Not that I'm even remotely an authority on this, and am just taking it at face value, but last month..

"AMD will be developing an exascale node architecture using its own Heterogeneous System Architecture (HSA)-based APUs and a "new generation of memory interfaces," the company said."

and it's reported that this is the 3rd grant for this stuff from the same place in as many years. I'm not super confident in the gov or mil but I don't think they dump three years in a row if nothing was coming of it. Sure it's not desktop, isn't there a CRAY full of opterons that's pretty hot stuff still? Probably not a coincidence. I think it might be a hair naive in this day and age to think something related to the DOD/DOJ/DOE/ABCDE or whoever might not be kept less than public. I know that some large scale strategic planning is done on such huge computers and that stuff rapidly falls under a blanket of hush.

None of that helps you play video games directly or gets you in a Dell or whatever, but it's if anything more interesting, and provides some hope that AMD will stick around in some chip making capacity.

 

[Abwx]

Uh? Look, I'll be first to give AMD props for growing beyond the FSB (HT came first, by a long shot), but QPI is not a "carbon copy" of Hypertransport. It's a ring bus, for one thing . . .

HyperTransport (HT), formerly known as Lightning Data Transport (LDT), is a technology for interconnection of computer processors. It is a bidirectional serial/parallel high-bandwidth, low-latency point-to-point link that was introduced on April 2, 2001

Although sometimes called a "bus", QPI is a point-to-point interconnect. It was designed to compete with HyperTransport that had been used by Advanced Micro Devices (AMD) since around 2003.[5][6] Intel developed QPI at its Massachusetts Microprocessor Design Center (MMDC) by members of what had been the Alpha Development Group, which Intel had acquired from Compaq and HP and in turn originally came from Digital Equipment Corporation (DEC).[7] Its development had been reported as early as 2004

http://en.wikipedia.org/wiki/HyperTransport

http://en.wikipedia.org/wiki/Intel_QuickPath_Interconnect

 

[mrmt]

and it's reported that this is the 3rd grant for this stuff from the same place in as many years. I'm not super confident in the gov or mil but I don't think they dump three years in a row if nothing was coming of it. Sure it's not desktop, isn't there a CRAY full of opterons that's pretty hot stuff still? Probably not a coincidence. I think it might be a hair naive in this day and age to think something related to the DOD/DOJ/DOE/ABCDE or whoever might not be kept less than public. I know that some large scale strategic planning is done on such huge computers and that stuff rapidly falls under a blanket of hush.

Cray dumped AMD circa 2012 not only they sell only Xeon chips, their interconnect business was sold to Intel. Today most supercomputers are either Nvidia Tesla or Xeon Phi, the processors are either POWER, SPARC or Xeon, Opterons are totally out of favor in the HPC market. I think AMD and Globalfoundries should be the last places where Opteron is seriously considered for server duties.

About the DOE grant, we are talking about a 32 million project, that's peanuts for Nvidia, let alone Intel (although it might not be peanuts for AMD). If Intel wanted to spend 100 times as much as this grant for a project they felt it was necessary for them they would. Don't read too much into it.

 

[mrmt]

Uh? Look, I'll be first to give AMD props for growing beyond the FSB (HT came first, by a long shot), but QPI is not a "carbon copy" of Hypertransport. It's a ring bus, for one thing . . .

Didd you not know that? QPI is a shoddy implementation of AMD HT, and hyper threading is a shoddy implementation of AMD CMT.

 

[NTMBK]

Lol, so much outclassed that QPI, Quick Path to Innovation, is a carbon copy of Hypertransport, that is a point to point bidirectional bus....

How the hell else do you expect to connect a pair of NUMA nodes? :\

Seriously, AMD is way outclassed in servers. They just don't have competitive perf/W, and in servers cooling and power bills are big expenses.

 

[Ramses]

Cray dumped AMD circa 2012 not only they sell only Xeon chips, their interconnect business was sold to Intel. Today most supercomputers are either Nvidia Tesla or Xeon Phi, the processors are either POWER, SPARC or Xeon, Opterons are totally out of favor in the HPC market. I think AMD and Globalfoundries should be the last places where Opteron is seriously considered for server duties.

About the DOE grant, we are talking about a 32 million project, that's peanuts for Nvidia, let alone Intel (although it might not be peanuts for AMD). If Intel wanted to spend 100 times as much as this grant for a project they felt it was necessary for them they would. Don't read too much into it.

I think I was refering to #2 there http://www.top500.org/lists/2014/11/
Again, not an authority just observing online info.

 

[kagui]

im just gonna leave this here
http://www.green500.org/news/green500-list-november-2014?q=lists/green201411
and this
http://www.top500.org/

ramses beat me to post the link

 

[mrmt]

I think I was refering to #2 there http://www.top500.org/lists/2014/11/
Again, not an authority just observing online info.

Have a look at the rest of the list.

 

[Abwx]

How the hell else do you expect to connect a pair of NUMA nodes? :\

Seriously, AMD is way outclassed in servers. They just don't have competitive perf/W, and in servers cooling and power bills are big expenses.

Dont know, my point was that QPI wasnt an innovation and that it s much inspired from Hypertransport, wether it s possible to do otherwise or not.

The perf and perf/watt side of thing is another debate, i dont know their server CPUs and chipsets caracteristics, so i cant say the extent of their competetivness, or lack of, in this department.

 

[Enigmoid]

I think I was refering to #2 there http://www.top500.org/lists/2014/11/
Again, not an authority just observing online info.

Yep, and the CPUs were upgraded in 2011.

CPUs had been in place for over a year

 

[Ramses]

I see that, still seem to be doing well.



like my FX chip

 

[mrmt]

I see that, still seem to be doing well.



like my FX chip

Yes, a relic of the past. Not a processor you would look if you were to build a new system today. But even if you *wanted* to build such supercomputer with AMD processors, you would have to look for another company other than Cray, as they no longer build system with AMD processors.

 

[Ramses]

Is it still in #2 because they used a crap ton of CPU's or something? I saw on that other link it was #1 in like 2012.

Man we are a flighty fickle bunch of "what have you done for me today" SOB's about hardware. It's almost the epitome of throw-away culture. I don't know if it's california or my wife's influence or old age but it's getting less attractive the older I get. It used to be fun and if anything hardware moved faster in years past. Weird..

 

[ElFenix]

http://en.wikipedia.org/wiki/HyperTransport

http://en.wikipedia.org/wiki/Intel_QuickPath_Interconnect

everything is just old DEC designs!

 

[CHADBOGA]

Dont know, my point was that QPI wasnt an innovation and that it s much inspired from Hypertransport, wether it s possible to do otherwise or not.

Much like how a Lexus was inspired by a Lada. :hmm:

 

[boxleitnerb]

Interesting, thanks for your opinions. How do you think the AMD stock will develop this year? Will it go further down, stay at $2.00 to $2.20 or will it go up at least a little?

 

[5150Joker]

If you take a look at this: https://www.google.com/finance?q=NYSE:AMD&fstype=ii how long does anyone think AMD really has left? They got wiped out on the CPU side of things and with discrete GPUs (notebooks + desktops) they only have at best 30% share and it's shrinking. Their next bold move is a 300W desktop (oven) GPU that needs watercooling. D: Kinda reeks of desperation like another company I knew that made this: http://www.3dfx.ch/gallery/d/16283-2/3dfx+Voodoo+5+6000+AGP+128MB+Rev_A1+1500+Octa+fan+card+a.JPG I'd give them max 2 years before they go under.