UniversalTruth
Veteran
edit: bad article anyways, delete it, not worth it to share.
Last edited by a moderator:
The AMD Execution Thread [2007 - 2017]
- Thread starter overclocked_enthusiasm
- Start date
-
- Tags
- amd
- Status
Bald Eagle is not mobile KVR, but rather the one aimed at the embedded market. One would guess that that one will come even later, since it might require more aggressive binning.
overclocked_enthusiasm
Regular
2014 WSA
Well, it is only 90 days late...
http://www.marketwatch.com/story/am...obalfoundries-2014-04-01?reflink=MW_news_stmp
"Under this amendment AMD expects to pay GLOBALFOUNDRIES approximately $1.2 billion in 2014. These purchases contemplate AMD's current PC market expectations and the manufacturing of certain Graphics Processor Units (GPUs) and semi-custom game console products at GLOBALFOUNDRIES in 2014. The 2014 amendment does not impact AMD's 2014 financial goals including gross margin."
So they purchased $960 million in 2013 for just CS products...ZERO GPU or cGPU products. This year they are committed to spend an additional $240 million up to $1.2 billion with "certain Graphics Processor Units (GPUs) and semi-custom game console products" being added to the mix. The fact that the $960 million from last year included neither GPU or cGPU, I think the CS group will be a train wreck this year. I just see nothing new in the tablet, mobile, or laptop space for AMD and wonder how committed they are to anything other than GPU products at this point.
As I suspected, the pound of flesh that GloFo got from AMD for being $160 million short of the 2013 WSA is moving GPU and cGPU production from TSMC to maintain volume while the CS production nosedives. So the question is how far off the cliff has CS sales dropped and can GVS sales compensate and drive profits?
Well, it is only 90 days late...
http://www.marketwatch.com/story/am...obalfoundries-2014-04-01?reflink=MW_news_stmp
"Under this amendment AMD expects to pay GLOBALFOUNDRIES approximately $1.2 billion in 2014. These purchases contemplate AMD's current PC market expectations and the manufacturing of certain Graphics Processor Units (GPUs) and semi-custom game console products at GLOBALFOUNDRIES in 2014. The 2014 amendment does not impact AMD's 2014 financial goals including gross margin."
So they purchased $960 million in 2013 for just CS products...ZERO GPU or cGPU products. This year they are committed to spend an additional $240 million up to $1.2 billion with "certain Graphics Processor Units (GPUs) and semi-custom game console products" being added to the mix. The fact that the $960 million from last year included neither GPU or cGPU, I think the CS group will be a train wreck this year. I just see nothing new in the tablet, mobile, or laptop space for AMD and wonder how committed they are to anything other than GPU products at this point.
As I suspected, the pound of flesh that GloFo got from AMD for being $160 million short of the 2013 WSA is moving GPU and cGPU production from TSMC to maintain volume while the CS production nosedives. So the question is how far off the cliff has CS sales dropped and can GVS sales compensate and drive profits?
overclocked_enthusiasm
Regular
BTW, the dGPU supply issue is fully resolved and pricing is waaaaay down
NewEgg
290x 10/10 IN STOCK ($569) down $80 in 3 weeks MSRP $549
290 10/10 IN STOCK ($459) down $90 in 3 weeks MSRP $399
280x 14/14 IN STOCK ($309) down $140 in 3 weeks MSRP $299
270x 19/19 IN STOCK ($199) down $50 in 3 weeks MSRP $199
"EDIT 3 weeks should be since Feb 27"
NewEgg
290x 10/10 IN STOCK ($569) down $80 in 3 weeks MSRP $549
290 10/10 IN STOCK ($459) down $90 in 3 weeks MSRP $399
280x 14/14 IN STOCK ($309) down $140 in 3 weeks MSRP $299
270x 19/19 IN STOCK ($199) down $50 in 3 weeks MSRP $199
"EDIT 3 weeks should be since Feb 27"
2014 will see the introduction of mobile Kaveri, Beema, Mullins, new server/embedded products, and possibly (hopefully) some new discrete GPUs. That's not so bad.
UniversalTruth
Veteran
http://www.techpowerup.com/199984/s...iver-multi-sourced-14-nm-finfet-offering.html
Samsung and GLOBALFOUNDRIES to Deliver Multi-Sourced 14 nm FinFET Offering
UniversalTruth
Veteran
http://www.zdnet.com/amd-says-it-wont-compete-against-intel-in-budget-tablet-market-7000028632/
AMD says it won't compete against Intel in budget tablet market
Summary: A company exec says it isn't interested in subsidizing cheap tablets like its chip rival does.
AMD says it won't compete against Intel in budget tablet market
Summary: A company exec says it isn't interested in subsidizing cheap tablets like its chip rival does.
Meanwhile, back on Planet Earth...
I admit it's not much of a plan, but i cannot see them competing with Intel while Intel is subsidizing to gain market share.
Feel free to mention a better idea though. :smile:
I think the problem is bulldozer .
its performance is lack luster and just doesn't fit the mobile industry. Yes jaguar is nice but they are loosing out on deals like the surface because of the higher end chips. Jaguar just doesn't compete with an i5 .
If they actually spent time and fixed bulldozer they might get market share back
its performance is lack luster and just doesn't fit the mobile industry. Yes jaguar is nice but they are loosing out on deals like the surface because of the higher end chips. Jaguar just doesn't compete with an i5 .
If they actually spent time and fixed bulldozer they might get market share back
UniversalTruth
Veteran
Beema and Mullins (actually it is one and the same chip but with two codenames to distinguish different binning, wth?) seem okeyish and there could be no real-world excuses from OEMs for the potential lack of design-wins.
AMD could start manufacturing and selling their own tablets like that Discovery one without waiting for those monkeys to accept their products.
AMD could start manufacturing and selling their own tablets like that Discovery one without waiting for those monkeys to accept their products.
There are two pretty good excuses not to use Mullins, in spite of its apparently high technical merit:
1) AMD doesn't have Android support, and relatively few people seem interested in Windows tablets;
2) Intel subsidizes its chips so heavily they're basically free.
UniversalTruth
Veteran
What can you do on an Android tablet which cannot be done on Windows?
The other thing is another court case, I suppose. AMD needs to bring those guys to the court, again.
The other thing is another court case, I suppose. AMD needs to bring those guys to the court, again.
This is all true, but I would still pay a bit more for AMD chip inside my Windows tablet to get better overlay experience.
I'm aware my usage requirements are niche, but having had 3 days spent with Dell Venue 11 Pro based around Intel Z3770 chip I'm very happy with it's Windows/Internet experience, but not even close to be satisfied with gaming experience. For instance Trials Evolution is only semi playable at the lowest settings and 640x480 resolution but suffers from severe image quality issues like popping textures or missing whole objects. With AMD solution bringing 50%-100% more GPU power and properly working driver my son could play this game with better quality at higher resolution and looking as developer intended.
Another thing is, a lot of even older scene demos are not loading due to some OpenGL or DX functions not properly supported by Intel driver.
So as I said, niche case with son wanting to play some of my Steam library games on a tablet when other PC's are taken, but it exists.
My ideal tablet would be something with the power of mobile A10-4600 in the 5W TDP. Hopefully that's something they can do once on 14nm/16nm FinFET process.
That's probably a tall order on a CPU front, but I think it's very doable for the GPU. Yet I suspect that sort of solution may be more likely to come from the big core family, perhaps in a slightly higher TDP, but still <10W.
I think there's a decent argument drawing from the results of the revisions to the architecture and AMD's spreading out full design validation across multiple phases of the same basic chip (Trinity->Richland, Jaguar->Puma) that AMD can't really expend the necessary resources to "fix" Bulldozer, or that if it could that it wouldn't have the best ROI.
Bulldozer may have been a poor call at 32nm, but with the fab spinoff and the lack of high-performance MPU-focused processes by AMD's fab partners for the upcoming nodes makes me think Bulldozer is now fundamentally the wrong architecture.
32nm SOI is the high water mark for AMD's CPU circuit performance on a process that basically evolved with the CPU, something Bulldozer seriously needs to justify much of its design.
The future is cell-phone processes that give jack-all consideration to AMD's CPU needs, especially when the GPU is the only realy compelling element.
AMD isn't done degrading the starting point for its CPUs, not with the likely move to even more bog-standard bulk than the semi-specialized process Kaveri uses and a lowered TDP for Carizo.
The resources required for building the infrastructure for and validating a x86 server/desktop/laptop design are also incredibly high for a situation where AMD's fortunes in most of them are so poor and potentially sapped by the declining growth in some of them.
There's simply so much engineering that AMD has fallen behind on, and a number of partners it has burned severely over the years, meaning that even a return to competitive form would be hampered by slowed uptake.
If AMD is to fix Bulldozer to conform to the new realities going forward, I'm not sure it's a fix as much as a replacement or abandonment of its line. BD isn't good for any of the markets it was supposed to address, and I don't think AMD has the chops to produce a single line that can straddle this many segments.
- Status
