The AMD Execution Thread [2007 - 2017]

[3dilettante]

There's only so long Valve are going to tie themselves to a perceived sinking ship. It's not like AMD have either the GPU or CPU superiority that led Valve to first take the side (and cash) of ATI.

If Nvidia ponied up the cash, has better tech, and Valve wanted to side with Intel (which presumably would be difficult with an AMD deal), it would be no surprise to me to see Valve switch to TWIMTBP in their ever-hungry quest for profits. Everyone else has done it.

If AMD ever manage to pick themselves up, then I'm sure any software company would look to again side with them as a market and technology leader, but right now (and for some time into the future) that leader doesn't look like it's going to be AMD.

I'm not convinced the performance argument is that large a part why such an event could happen (if confirmed), but I don't know Valve's reasoning.

Perhaps we should follow the money, or the possible inability to pay said money or a raised bid (thanks to a cash-flush competitor).

 

[Dave Baumann]

In record quantities no doubt

HP and Gateway have only recently announced their new Phenom systems in the past few days.

 

[INKster]

VALVE sending free gift to NVIDIA users, and joined TWIMTBP

(i not find any other thread to this news where its not fully off, maybe its deserve a own thread, this is a real shocker)

No need to go overboard about it. It's just a demo, not an indication of jumping over to TWIMTBP in detriment of ATI's GITG.


edit
Intel also has a sense of humor

http://legitreviews.com/article/645/1/

 

[MTd2]

AMD delays Phenom 9700 into Q2, preps low-power variant

Contrary to its earlier statements, AMD has now revealed that its 2.4GHz quad-core Phenom 9700 and 2.6GHz Phenom 9900 have both been delayed until the second quarter of 2008. AMD hasn't said how this will impact the introduction of the 2.8GHz and 3GHz parts that were originally projected for late in Q2, but these will almost certainly be delayed as well.

http://arstechnica.com/news.ars/pos...nom-9700-into-q2-preps-low-power-variant.html

 

[ShaidarHaran]

AMD doesn't even have the SKU headroom for >2.6GHz parts, so 2.6GHz appears to be the end of the line for Agena.

 

[vertex_shader]

No need to go overboard about it. It's just a demo, not an indication of jumping over to TWIMTBP in detriment of ATI's GITG.

I know its just a demo, but than what is this?

Here is the press release:

More Steam gamers, including myself, play on NVIDIA hardware than any other GPU," said Doug Lombardi, vice president of marketing at Valve. "That's a testament to the company's long history of innovation and quality. By working together with NVIDIA to expand our development, distribution, and marketing efforts, we're increasing our ability to serve the millions of NVIDIA customers logging onto Steam and help us prepare for our next generation of content.

Its sounds clear for me.

 

[Speccy]

Sounds like a mechanism for earning some more money off a title that has already been out and ensures further reach of the Steam client on to peoples PC's (which they already do with AMD via the bundling of the client install with Catalyst drivers).

Its called "marketing".

 

[OpenGL guy]

Consoles = consumer electronics.
Chipsets =? Breakdown not provided.

It's sort of a loop, is it not? I asked how much money ATI has made for AMD, you asked me to look at 10-Qs, I did and provide the numbers only to get "well, you are missing some data which is not included in 10-Qs" as the response. What was the point of that exercise?

You don't have all the data, so maybe you should reserve judgement. To put it simply, if GPG and consumer lost a total of $12 million, you don't think that could have been offset by chipset sales? Also, since GPG developed and supports the console projects, why are the revenues reported under consumer? I'll never understand that.

Yes, they are selling them, at the loss up until this point. That's hardly an adequate rebuttal to someone questioning whether a deal that has cost AMD over $6 billion dollars up until this point (in acquisition expenses, losses, chargers and interest) is a gamble that is going to pay off.

Another way of looking at the numbers is that the GPG and consumer results are the only bright spots. Look at the trend for the GPG results, look at the product launches, if you think that's not good, then I guess there's just no pleasing you.

 

[Geo]

I think most old school AMD people who dislike the ATI purchase do so because of the reduction of cash resources and increased debt incurred. The theory being that with AMD CPU financial results dropping like a rock that money would be nicer to have around right now than ATI.

It was meant to be a long-term strategy deal. Some people seem to agree with Lord Keynes about the long run.

 

[OpenGL guy]

I think most old school AMD people who dislike the ATI purchase do so because of the reduction of cash resources and increased debt incurred. The theory being that with AMD CPU financial results dropping like a rock that money would be nicer to have around right now than ATI.

You can't predict the future. If AMD and ATI had known about the problems that we had last year, things might have been different.

It was meant to be a long-term strategy deal. Some people seem to agree with Lord Keynes about the long run.

There are short-term benefits as well, unfortunately some product problems have reduced the impact of those benefits.

 

[AlexV]

You don't have all the data, so maybe you should reserve judgement. To put it simply, if GPG and consumer lost a total of $12 million, you don't think that could have been offset by chipset sales? Also, since GPG developed and supports the console projects, why are the revenues reported under consumer? I'll never understand that.

Another way of looking at the numbers is that the GPG and consumer results are the only bright spots. Look at the trend for the GPG results, look at the product launches, if you think that's not good, then I guess there's just no pleasing you.

Whilst I agree with you that you're the only ones doing quite well within AMD, I don't think that the GPG can keep the company running by itself. Either the core business of making CPUs makes a comeback and starts bringing some cash by itself, either things aren't going to get better, and if anything you(the GPG)'re going to end up suffering due do reduced R&D expenditures, reduced marketing expenditures and so on. IMHO.

 

[3dilettante]

I agree that few would have predicted the extent of AMD's product troubles, but something tells me there were those in both companies who knew there would be problems in their respective spheres.

It seems clear the CPU division would have been aware of issues after failing to replace K8 several times and settling on putting together Barcelona on a tighter timeline than AMD had used for K7 and K8.

I'm a little more hazy on the timeline for ATI, but I think there were a few whispers in the rumor mill that would have been louder for those within the company.

 

[Geo]

Y'know, I'm not so sure about that re ATI. Given the deal was announced in July 2006.

 

[3dilettante]

There may not have been any real indicators of trouble, but I don't have as many signposts as the CPU business has.
I recall vaguely some statements about the delay and evolution of R400 into what was finally released much later for Xenos and whatnot, but I can't make any qualitative judgements from my memory of those events.

If we believe R600 was about 6 months late, then there could have been some indication of some form of delay, given some unknown amount of lead time needed for validation and manufacturing to hit that initial date.

AMD's latest CPU troubles have been a glaring public lesson in the timing of chip steppings, design efforts, and fab turnaround.

 

[OpenGL guy]

There may not have been any real indicators of trouble, but I don't have as many signposts as the CPU business has.
I recall vaguely some statements about the delay and evolution of R400 into what was finally released much later for Xenos and whatnot, but I can't make any qualitative judgements from my memory of those events.

If we believe R600 was about 6 months late, then there could have been some indication of some form of delay, given some unknown amount of lead time needed for validation and manufacturing to hit that initial date.

Sounds like a bunch of speculation to me.

AMD's latest CPU troubles have been a glaring public lesson in the timing of chip steppings, design efforts, and fab turnaround.

What, exactly, does this mean?

 

[3dilettante]

Sounds like a bunch of speculation to me.

I've admitted as much by stating I'm only going by rumors for the ATI side of the business.
I did not and do not have a direct line of communication with the internals of ATI/AMD.

What, exactly, does this mean?

The CPU division and AMD's own marketing, in their zeal to reassure customers that the next miracle stepping is just around the corner, have made a number of public comments from which we can infer a lot about AMD's design, process transitions, and fab times.

The entire BA/B1/B2/B3 stepping debacle and the succession of unmet release and re-release dates is a pretty good collection of data points concerning how long it takes to finalize a revision and how long it takes AMD to get finished silicon.

A whole slew patents, comments by industry insiders, a number of departures of former CPU design leads, a lot of PR statements and convenient PR statement omissions have been very helpful.

Hector Ruiz himself pegged Barcelona as 6 months late in an interview.
(edit: The patents and turnover of deisgn leads indicate K8's replacement was restarted at least twice).

I don't need to speculate as much for AMD's CPU side, because the situation they're in has basically beat any margin for spin out of them.

People have already predicted successfully the delay of Bulldozer and the likelihood that Shanghai is a late 4Q 08 or 09 product based on the existence of Montreal (AMD's official slides) and the admission that Shanghai hasn't even taped out yet (a PR statement that they expect first silicon for a product that usually takes at least a year, and the fact that the design center responsible for Shanghai wasn't even open as of December last year), respectively.

AMD has helpfully confirmed all of that, unless Phil Hester, Dirk Meyer, and Hector Ruiz are lying to investors in violation of how many laws.
They can't provide mysteries needed for speculation because their guidance has been so wrong recently that they're pushing the line for shareholder lawsuits.

 

[Speccy]

I'm also running across rumors that the B3 stepping has failed to fix the TLB bug

Or perhaps not, eh?

http://www.theinquirer.net/gb/inquirer/news/2008/01/11/barcelona-b3s-fine

 

[MTd2]

How many GHz should Phenon attain to get competitive in the hight tier? 5GHz?

 

[3dilettante]

It's nice to see some less than catastrophic news for AMD. I'm willing to lend credence because someone went on the record, though I don't follow what Charlie meant by verifying the fix on B2 silicon for B3.

Perhaps AMD has a few machines like Intel showed in a tour that allows them to rewire a chip in the test lab.

Going by this, B3 works, but it's going to be later by a number of weeks. It's sort of like half the delay of a new stepping.


On a side note, where's the news on Griffin?

edit:
Going back over Charlie's story, he's indicating B3 is the last stepping. The 65nm Barcelona core has already passed middle age before being fully available.

 

[OpenGL guy]

It's nice to see some less than catastrophic news for AMD. I'm willing to lend credence because someone went on the record, though I don't follow what Charlie meant by verifying the fix on B2 silicon for B3.

I don't know what was done, but you can sometimes use FIB to test a fix if the change is small.