The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
Seems like Zacate's smooth roll-out is some evidence in favor of ease of execution with this synthesized approach.

Zacate is hardly a complex core, and whether or not its rollout was smooth is, shall we say, unknown?
 
And the story was just updated on Semiaccurate to note that Trinity is having yield problems, though allegedly not as bad as Llano's.
 
Bulldozer was to be the acid test of the new and improved AMD. Fusion was a wobbly start, with signs of decent demand AMD could not meet, and it's not a case of selling out of everything AMD wanted to build, it produced far fewer than it wanted.

Now we get news that even AMD's 45nm capabilities were compromised due to 32nm tool sharing.


Whatever the dwindling number of customers may say about AMD fighting the Intel monopoly bogeyman, it rapidly loses this utility if it cannot be trusted to meet its obligations or projections. What's worse for a company trying to build product, an Intel that overcharges for its chips, assuming it can do so without throttling the upgrade cycle, or an AMD that will leave you twisting in the wind when it screws up yet another product transition?
Better yet, how do they feel about an AMD whose product transitions screw up their current supply capabilities?
 
GF's process is probably at fault for a lot of this. The power numbers for slightly different Llano SKUs look like they are due to variability, and that has been predicted years ago as a likely problem with gate-first.

However, it is AMD's only option for high-performance CPU silicon, so as far as the CPU market is concerned, it's all the same.
 
Zacate was by no means simple in an absolute sense, simpler than Llano sure but we're still talking about an extremely densely packed ~400 million transistor design w/ a brand new OoOE CPU micro-architecture w/ all the bells and whistles and a new bus architecture. Getting it to work is no mean feat, esp. when the only experience you've had w/ TSMC was fabbing GPUs which have higher tolerances than CPU/GPU combos. Anyway, there certainly wasn't as much noise about delays and full revisions as with BD / Llano.

Then again, they were both announced at the same analyst's day in '05 and are all appearing in the same year so there probably were some teething issues w/ Zacate (or maybe this is Zacate 2 we're looking at). It is also possible they pulled engineers off of Zacate to work on Llano which is why it's only appearing this year. We'll see what moves AMD makes in the future re: layouts.
 
Last edited by a moderator:
Better yet, how do they feel about an AMD whose product transitions screw up their current supply capabilities?

That's the part that's truly inexcusable; they doubled down and lost their house. Atleast they had the sense to outsource some work to TSMC, a definite benefit of the new fab-lite AMD.
 
It is weird I was really annoyed AMD bought ATI back in the day b/c I thought it would kill off Nvidia and graphics competition. Now it looks like AMD is just trying to kill themselves and ATI may be all that is holding them above water. The best case for me would be AMD has to sell off ATI to stay afloat and then manages to make a decent CPU as well so we have competition all around again :)
 
I think ATI will spin off their money guzzling CPU division into a separate entity. Then HP will buy it, run it for 10 months, then shut down the division, sell all remaining chips for 29.99 and then proceed to fire it's CEO.
 
A significant role in intels succes is also the ability to work with own fabs, the RND they spent there and patents they own.
Bulldozers blame is probably not just AMD-s design but also GloFo-s 32nm.
 
Status
Not open for further replies.
Back
Top