The AMD Execution Thread [2007 - 2017]

780g (and its derivatives) is already pretty low power compared to the competing chipsets, but without a low power cpu to go along with that the extra 1-2W saved with an even lower power chipset doesn't matter much.

 

AMD IS becoming "asset lite"


http://biz.yahoo.com/bw/081006/2008....yahoo.com/bw/081006/20081006006639.html?.v=1
http://online.wsj.com/article/SB122334620497110417.html?mod=yahoo_hs&ru=yahoo
http://www.cnbc.com/id/27057968/site/14081545?__source=yahoo|headline|quote|text|&par=yahoo
http://www.bloomberg.com/apps/news?pid=conewsstory&refer=conews&tkr=AMD:US&sid=axKo70pBepvE

http://www.dailytech.com/article.aspx?newsid=13139

toldjaso

You can forget about AMD's execution now

1. AMD will continue to own 44% of The Foundry Co.
2. FoundryP&L are not part of AMD's books.
3. Probable Foundry Co customers should include AMD (both graphics and CPU), IBM, and TSMC overflow {*yes TSMC*}
4. One of AMD's most valuable IP products is APM, which they have ONLY licensed to The Foundry Co. This means that they are still free to license their IP.
5. Hector IS gone completely from AMD [thank goodness]

!!!

 

http://biz.yahoo.com/bw/081016/20081016006192.html?.v=1

 

Those are pretty reasonable results, all things considered. Very nice GPG results, up over 50% serially.

 

Bottom line is it is another loss just before the recession. The only way it is reasonable is that it is a lot less a loss than the last year or so

 

Looks like RV770 is literally keeping AMD afloat right now, because we all know their industry-low CPU ASPs aren't paying the bills.

Perhaps those that criticized or even ridiculed AMD's ATi purchase should re-think their positions now

 

Kind of weak accounting trick with the 193 million from the sale of old fab equipment being counted as processor group revenue, rather than some kind of one-time payment.

Most of that money has been payed and probably already spent since early in the year, when the money was actually payed.

 

LOL, gotta love accountants.

 

There aren't too many discontinuous trend lines in companies that big, are there?

 

They're eating nVidia's lunch for now, and deneb / shanghai looks relatively promising with better clocks, IPC, and much lower power draw. Hopefully one of the two, GPU or CPU, will pick up the slack at any given time.

 

They might have an opportunity in CPU's if Nehalem is as underwhelming in games as it looks.

 

Deneb would need to provide pretty signficant IPC enhancements in addition to their healthy clock increases to pull it off though. So far AMD has said 20-30% faster but when do statements like this ever turn out to be true?

I'm cautiously optimistic AMD will get close to Nehalem's single threaded IPC. Rather cautious.

 

They will be losing some of the high margin server business before bulldozer arrives almost for certain, though the much improved performance per watt is compelling.

AMD needs to bank on taking some of Intel's the consumer business with its other products before bulldozer and it looks like it's trying to push for just that with the launch of its Fusion brand. Too bad they didn't have the actual Fusion CPU/GPU on hand for the launch!

I think Fusion will have many design wins when it does finally show up though; inter CPU GPU communication would be much improved by this integration, so I think the performance ought to be pretty good. Maybe in the longer run the GPU portion will take over the CPU's floating point execution chores; and the CPU's schedulers could be used for GPU execution units. Unless nvidia and intel fuse their products onto a single core, the AMD part will be quite attractive.

Overall cost will be down too over fabbing two separate dies. It'd be a coup for sure if apple went for fusion instead of a pricey core2 + nvidia chipset! Apple certainly didn't shed many tears over dropping intel for their graphics or ditching PowerPC so here's hoping.

 

To recap recent events:

AMD has lowered Q4 guidance to 25% below Q3. This is a 25% drop for a quarter that is traditionally seasonaly stronger.

Its share price is struggling to maintain a plurality of dollars, revenues are now expected to drop below what analysts already expected as a poor showing.

As a result of all this, Abu Dhabi has either exercised some penalty clauses or decided to renegotiate the split.
AMD's voting interest is unchanged, but further stock shares are being granted to Mubadala and have reduced AMD's ownership of the fab company to about one third.
If stock ownership is the determining factor of the allocation of profits, the fab company is perilously close to ceasing to be a subsidiary under AMD's IP cross-licensing agreement, or at least as the agreement has been publically disclosed.
Capital calls have been indexed at 90% of AMD's net assets, which looks to restrict the fab company's ability to raise cash.

AMD's road maps, drawn up prior to this latest turn of the thumbscrews, already put Fusion into 2011 and at 32nm, which puts Fusion and Bulldozer as a delay of a delay of a delay of a delay.
Bobcat is dead.
We can expect to see Shanghai relatives for the next two years.


AMD's 45nm chips at least initially have shown modest per-clock increases and a few higher speed grades at the same TDP. The first products have socket compatibility with previous chips, which might serve to prevent further erosion in the more value-minded sectors.
At the desktop, AMD is trading blows with the low to high-middle range of Intel's products.
The highest Core2 quads appear to be slightly above, and i7 is above all else.

i7 is too high-end and requires pricey mainboards and memory, which will restrict its uptake.
AMD's pricing the highest PhenomII just below the lowest i7.

Atom is basically uncontested, and recent analysis has shown that while netbooks are cannibalizing some of the laptop low end, they are creating new sales where no units would have sold otherwise.
AMD's laptop share by units has declined, in part because of the cheap netbook sales diluting its numbers.
AMD has no specially-designed mobile processor, and its supposed answer to Atom will be stripped down desktop cores that cannot scale as far down as Atom.

At servers, Shanghai might have a little over a quarter's advantage at 4 socket.
One i7 in SpecInt is about as effective as a two-socket Shanghai.
At SpecFP, the single socket i7 is more than half as effective as doubled Shanghai, but not as catastrophically equivalent.

AMD's 4-socket, with the release of dual socket i7 boards next year, may well be gutted.

The higher core count boards, a niche within a niche, might persist for AMD for a while longer until being hit by Beckton at the same level, and low level IPF from above.

AMD has additional virtualization capabilities to deploy, but those will need new boards and sockets that will not be out for a while.

Graphics remains competitive, though the expectation is that discrete graphics will take a hit next year.

 

The funny thing is I asked how amd was doing financially here just a few days ago and was assured all was roses and sugar plums.

I hope something can be worked out though and quick as we need competition. I was actually thinking strongly about going amd around Febuary time frame, but honestly I find it somewhat difficult so far to justify since they are about to introduce a new socket (AM3) and unless there AM2+ chips are better than Intel's ddr2 lineup I don't see why I would bother.

Do you know if AM3 will take DDR2 as well as ddr3 or only DDR3?

I have extra DDR2 ram lying around so it is more likely I will get something that can use it than something requiring DDR3.

 

AM3 chips have memory controllers that support both types of memory, and can plug back into older AM2 boards.

I'm not sure if socket AM3 boards will support DDR2.

 

http://www.digitimes.com/news/a20081215PD214.html

 

You might want to buy a decent AM2+ mobo with a cheap X2 today coupled with some DDR2-1066 so that you can upgrade it to a 45nm Phenom when they will be available. The DDR2-1066 won't have much benefits over DDR2-667/800 now but might prove beneficial on a future quad-core.

 

AMD's taking another writedown on the ATI purchase, or rather, another another another writedown.

Either AMD was spewing crap the last time it had its final one-time charge related to the acquisition, or the outlook of the graphics side of the business now looks a whole lot worse with current economic projections.

Given the downward spiral of AMD's overall value, there appears to be precious little room for any value to be attributed to the former ATI, unless the x86 division is worth about ten cents. Perhaps this won't be the final last ultimate writedown, either.

 

I think the news said there will be more coming next year.