There's no way AMD can be blamed for the R600. After R600, ATi really became a well oiled machine. Under wings of AMD, that is. Whether or not there's causation there is a different question.
The AMD Execution Thread [2007 - 2017]
- Thread starter overclocked_enthusiasm
- Start date
-
- Tags
- amd
- Status
The R600 was pretty much done and dusted when the merger talks were taking place, AMD had no involvement there. They did show a lack of understanding the market by not waiting to see how the DX10 cycle goes...or Orton really REALLY sold them on the idea that the R600 was R300 reincarnated and ATI's value would quadruple or something (good for him if he did). When I spoke with ATI people around the RV770 launch it was quite clear to them that it's a make or break moment, so they were quite catalyzed by the whole R600 debacle, which might well have led to the well oiled machine that silent_guy mentions. Sadly neither themselves nor AMD proper had good staff handling the business side of things, so they never capitalised on anything, or took advantage of trends, or tried to climb above the knee of the curve.
3dilettante: from what I know (note, this might be wrong, so take it as such), the Apple deal story has substance to it, as do the reasons posited for its failure. It has to be noted that Apple is probably uncomfortable in dealing with a strong partner like Intel, and would have favoured AMD who would've done worse in terms of negotiating power. Moreover, Apple's CPU oomph requirements are rather spartan all things considered, and I'd expect the distortion field to have worked just fine in selling weaker Llanos instead of more powerful Sandys / Ivys, if needed.
Again, I'll note that my suggestion wasn't to have 32nm Phenom IIs hold the fort instead of BD, but rather continue shipping the 45nm SKUs, possibly with (minor) clock-bumps. In desktop that would not have harmed their position (see how their shipment mix looks for the past quarters), and it would have saved a bunch of cash. Of course, armchair expertise is a dime a dozen and we're all strategists in front of the keyboard, so take this as what it is, which is my (strongly worded, bien sur) speculation about hypothetical scenarios we'll never get to explore
.
3dilettante: from what I know (note, this might be wrong, so take it as such), the Apple deal story has substance to it, as do the reasons posited for its failure. It has to be noted that Apple is probably uncomfortable in dealing with a strong partner like Intel, and would have favoured AMD who would've done worse in terms of negotiating power. Moreover, Apple's CPU oomph requirements are rather spartan all things considered, and I'd expect the distortion field to have worked just fine in selling weaker Llanos instead of more powerful Sandys / Ivys, if needed.
Again, I'll note that my suggestion wasn't to have 32nm Phenom IIs hold the fort instead of BD, but rather continue shipping the 45nm SKUs, possibly with (minor) clock-bumps. In desktop that would not have harmed their position (see how their shipment mix looks for the past quarters), and it would have saved a bunch of cash. Of course, armchair expertise is a dime a dozen and we're all strategists in front of the keyboard, so take this as what it is, which is my (strongly worded, bien sur) speculation about hypothetical scenarios we'll never get to explore
.There is something I don't understand, actually 3dilletante tried to explained me once, but I still don't get get it.
Why AMD could not do an "in depth" rework of their architecture instead of moving to Bulldozer.
I mean starting from Nehalen, Intel has make quiet some significant changes in the followiung CPU and even sandy bridge, Ivy bridge, and the up coming Haswell are all different architecture, they share a lot of the same DNA (no matter what Intel calls tick and tock).
I don't understand why AMD step back from this incremental improvements approach. They were not lagging that far they quit improving their architecture. To me it almost look as if they entered paralysis, like hesitating between continue to grown on it or started new while in the mean time doing nothing. Ultimately they chose to go with BD buy the roadmap has stalled for a long while already without any significant modifications to their overal design.
I don't know but it really look like for a few years their engineers were in a lockstep for years till execs make up their mind on the matter.
For llano, I'm not sure about why it failed. In my opinion while there are no that great solution to solve the bandwidth constrain on APU they should have gone with way lesser IGP. At least the chip would have cost them less, they may had higher yields, etc.
In my opinion as they were to end at a disadvantage against Intel on pretty much every metric, they could have gone with a native tri core, and Caicos class of GPU, try to have a tiny chip and be in a situation where they can put a fight on pricing alone without digging them selves into the grave.
I don't really why it took them so long to get the fusion done. Again it look like they were in a lockstep actually waiting for the thing to get "really "green light" internally.
It is really a sad story I hope they survive. Overall I think the management has been dreadful and they all lacked "vision" to parrot their marketing term. Choosing an architecture ala Bulldozer to attack Intel on one of its strong points and on a segment where volume is not that high either (vs desktop and laptop) was more than brave,...
Why AMD could not do an "in depth" rework of their architecture instead of moving to Bulldozer.
I mean starting from Nehalen, Intel has make quiet some significant changes in the followiung CPU and even sandy bridge, Ivy bridge, and the up coming Haswell are all different architecture, they share a lot of the same DNA (no matter what Intel calls tick and tock).
I don't understand why AMD step back from this incremental improvements approach. They were not lagging that far they quit improving their architecture. To me it almost look as if they entered paralysis, like hesitating between continue to grown on it or started new while in the mean time doing nothing. Ultimately they chose to go with BD buy the roadmap has stalled for a long while already without any significant modifications to their overal design.
I don't know but it really look like for a few years their engineers were in a lockstep for years till execs make up their mind on the matter.
For llano, I'm not sure about why it failed. In my opinion while there are no that great solution to solve the bandwidth constrain on APU they should have gone with way lesser IGP. At least the chip would have cost them less, they may had higher yields, etc.
In my opinion as they were to end at a disadvantage against Intel on pretty much every metric, they could have gone with a native tri core, and Caicos class of GPU, try to have a tiny chip and be in a situation where they can put a fight on pricing alone without digging them selves into the grave.
I don't really why it took them so long to get the fusion done. Again it look like they were in a lockstep actually waiting for the thing to get "really "green light" internally.
It is really a sad story I hope they survive. Overall I think the management has been dreadful and they all lacked "vision" to parrot their marketing term. Choosing an architecture ala Bulldozer to attack Intel on one of its strong points and on a segment where volume is not that high either (vs desktop and laptop) was more than brave,...
In regards to an earlier comment... buying XGI wasn't an option for AMD as ATI has already effectively bought them.
http://www.xbitlabs.com/news/graphics/display/20060306144301.html
http://www.xbitlabs.com/news/graphics/display/20060306144301.html
I think that ATI bought that chunk of XGI that used to be Trident (confusingly enough some sort of Trident still seems to be around
), leaving the old SiS graphics division in place...which SiS re-absorbed at some point. I think XGI kept on rolling some 2D cores for servers and still retained a fair chunk of its IP. But we can substitute XGI with something else that was cheap and enough for AMD's needs, I think.
), leaving the old SiS graphics division in place...which SiS re-absorbed at some point. I think XGI kept on rolling some 2D cores for servers and still retained a fair chunk of its IP. But we can substitute XGI with something else that was cheap and enough for AMD's needs, I think.
Not sure how buying a "smaller guy" with no track record (or failed track record) would steer the competitive situation away from the CPU/process paradigm to one of genuine differentiation. Adding crappy graphics (that will likely not be as area efficient and certainly wouldn't have been as compatible from a software/ecosystem perspective) would probably just things on par with Intel graphically and not change much at all.
Where do AMD go from here?
Do they continue trying to fix Bulldozer? Streamroller was supposed to be megawesome but then it fell off the right hand side of the roadmap. Beefed up Jaguar, making up for performance per thread with a ton of cores? Something completely new? Nothing?
Do they continue trying to fix Bulldozer? Streamroller was supposed to be megawesome but then it fell off the right hand side of the roadmap. Beefed up Jaguar, making up for performance per thread with a ton of cores? Something completely new? Nothing?
At this point, I am not sure that anyone knows.function said:
If I were AMD, I might start by trying to figure out what not to do.
If we look at their competition (Intel), its clear that their manufacturing advantage is significant and will likely be maintained. If that is the case, then trying to compete with them in absolute (x86) performance is simply not feasible. So that leaves two other important metrics by which they might compete: performance/watt and performance/dollar.
Honestly, AMD can not compete with Intel on performance/dollar if Intel doesn't want them to. However, they don't really have to win that metric as long as they can offer a lower absolute cost and "good enough" performance.
Performance/watt is another matter. Here is where they absolutely must be competitive with Intel in both single-threaded and multi-threaded cases. This is where Bulldozer & Piledriver have really failed. I don't think significant power savings were to come until Excavator, so that may be why Steamroller has been delayed or cancelled. At this point, they should probably just skip to Excavator & 20nm. If at that point they still can't get competitive performance/watt, it's time to call a spade a spade and ditch the Bulldozer line.
Alternatively, they could just scrap the Bulldozer line now and evolve the Jaguar line. This might make more sense as the Desktop PC market at this point appears terminal. By the end of 2014, we could be seeing 4 (or even 8) core Armv8 tablets & phones with hybrid memory cube RAM, and 128 GB+ of storage. At that point, the casual/office use desktop is dead. An interesting consequence of this is that as the consumer PC becomes something you carry in your pocket, it will also likely cease to be x86 and Windows based.
That leaves ultrabooks, servers, & HPC. AMD's x86 offerings will then be competing against their ARM offerings... At this point, it should be clear whether or not their is a reason to continue x86 development at all.
If they can weather 2013 and the fact they have no real answer to Haswell, I think their long term prospects are decent.
To sum up:
Performance/Watt > *
Performance/Watt/Transistor (important for profitability)
Be on board with HMC [I don't know if this product will live up to all of its promises, but if it does, it isn't something you want to lag behind on]
Focus on what you are good at [Right now that is GCN and presumably GCN2.]
Figure out a way to survive 2013 & Haswell.
caveman-jim
Regular
Selling off the austin campus and leasing it back is a short term win, long term loss, expect all the distributed campuses to follow with either closures and consolidation or lease back if they can find someone willing to invest.
If I had connections I'd be trying to find out if MS and Sony are still thinking about APU or AMD GPU technology or not, or are asking/being offered to buy the IP of their design outright, and if not where they're looking for alternatives.
If they can't find someone to bankroll them, then it's either merger time or they're done. If by end of Q1 '13 we haven't heard about a cash injection/further new restructuring, new partnership, or spectacular raison d'etre, AMD will be out of the game and in bankrupt 'ops mode' by Q3 '13 fulfilling existing contracts. They maybe could scale back into being an embedded supplier, just focus on the mini apu business and not try to make GPU's, CPU's or big APU's. Stuff the small APU's into Alternative NUC marketing.
Bummer, really.
If I had connections I'd be trying to find out if MS and Sony are still thinking about APU or AMD GPU technology or not, or are asking/being offered to buy the IP of their design outright, and if not where they're looking for alternatives.
If they can't find someone to bankroll them, then it's either merger time or they're done. If by end of Q1 '13 we haven't heard about a cash injection/further new restructuring, new partnership, or spectacular raison d'etre, AMD will be out of the game and in bankrupt 'ops mode' by Q3 '13 fulfilling existing contracts. They maybe could scale back into being an embedded supplier, just focus on the mini apu business and not try to make GPU's, CPU's or big APU's. Stuff the small APU's into Alternative NUC marketing.
Bummer, really.
Their GPU market share means they can and actually should make GPUs.
Their APU are, for the desktop market, still a way better solution than any Intel one.
Their FX line sucks. Well, their BD architecture was flawe dand they got the fix for Steamroller, which may (or may not) come.
They paid alot of money to get rid of GF. You, as a company, dont pay 500M$+ to get rid of a contract UNLESS you know you will get them back somehow.
Well, you can always consider managers as utterly incompetents - but usually not this much (but nokia one, ofc).
If you want to get rid of GF, it means you have to produce many CPU for someone, and you already have that someone. Or you wouldnt do that crazy thing.
Sunnyvale and Markham were sold years ago, from what I understand.
1CV (Markham ATI "newer" office / HQ) was sold back to the developer the day ATI moved in.
caveman-jim
Regular
This ignores the economic realities of the company, despite being true.
See above, but add significant investment in ISV's and marketing to consumers and OEM's.
Best of bad lot of choices, there was a lot of hope about Steamroller (well, some, people still remember hype before BD) but that fizzled with the dead cores reporting.
Unless your business is being dictated to by someone other than persons interested in the business, i.e. looking to pull a part of the company out and make it their own.
This is what I understand, too.
Didn't know that, but figures.I wonder about Orlando, Boston etc.
It might be a standard practice, but everything I'm hearing indicates to my admittedly untrained and naive eye that without a big 'something' in Q1 to drive revenue and sales, by Q3 the coffers will be empty (except for golden parachute money) and it'll be Operations Mode time. The Fellows moving out left right and center kinda indicates it's done tom turkey time. I hope I'm wrong, I really do, but I'm not getting information about hope on the horizon.
UniversalTruth
Veteran
Something similar to this from vr-zone:
ARM - Possible Beginning of the End for AMD? - Part 1
ARM - Possible Beginning of the End for AMD? - Part 2
with some interesting (at least to me) bits
Last edited by a moderator:
To briefly elaborate on BRiT's post:
Costs:
- development of two CPU core families (Bulldozer & Bobcat), a GPU architecture, numerous chips using various combinations of CPU cores and GPU IP with I/Os, including ARM cores now;
- chipsets;
- drivers and software;
- sales and marketing.
Revenue:
- low-margin APUs, from Ontario/Zacate/Hondo to Trinity;
- relatively low-margin desktop CPUs (Bulldozer);
- somewhat high-margin server CPUs, but in low volume (small market share in a small market);
- high-margin professional graphics cards in very low volume (very small market share in a small market);
- a bunch of IP in consoles, which is pure revenue but probably quite small.
That said, AMD's situation is also a product of the contracting PC market. All in all, they're not doing all that poorly.
UniversalTruth
Veteran
The problem is not only the low-margin, but as one could imagine- the performance of those low-margin products. AMD needs something revolutionary, something innovative with which it can beat Intel and Nvidia, so even with low-margin but extra high quantity and market share, they could maintain a good shape. As hinted by VR-zone, AMD needs to change the game, to enter something completely new. If they play Intel's or Nvidia's game, they will always lose.
Well, thanks for warning people to not bother clicking the link. Or was this supposed to be a joke by the author? Not clear from limited context.
UniversalTruth
Veteran
Well, I don't know whether it's a joke. But if you think logically- less people-> smaller market-> lower amount of sales-> less profit.
Should be something like this, I presume.
Oh, and it's your choice to click it or not. I have just put here the thing which impressed me the most.
- Status
Similar threads
