The AMD Execution Thread [2007 - 2017]
- Thread starter overclocked_enthusiasm
- Start date
-
- Tags
- amd
- Status
With Seamicro and now this, AMD is taking at face value around billion dollars in charges, with more than half of it being cash.
I'm not sure from the description of the GF deal how much of this goes into the total AMD expects to pay GF this year, or how much of this charge was money AMD was going to pay anyway per the prior wafer agreement.
The money going to the fab is expected to go from 900 million to 1.5 billion.
There's an interesting dynamic with these two events.
Seamicro has a number of former AMD staff that AMD is paying big money to make exclusive use of, while GF has a number of former AMD staff that AMD is paying big money for the priviledge of not making use of.
I'm not sure from the description of the GF deal how much of this goes into the total AMD expects to pay GF this year, or how much of this charge was money AMD was going to pay anyway per the prior wafer agreement.
The money going to the fab is expected to go from 900 million to 1.5 billion.
There's an interesting dynamic with these two events.
Seamicro has a number of former AMD staff that AMD is paying big money to make exclusive use of, while GF has a number of former AMD staff that AMD is paying big money for the priviledge of not making use of.
It looks like they're paying to adjust the wafer agreement in order to change pricing and to allow AMD to not use GF exclusively for some range of APU products at 28nm. Breaking the earlier links apparently came with that price.
There seems to be a lack of clarity as to what each change is for, and this may be partially due to a lack of visibility on what AMD would have been doing if not for this new agreement.
The amount of cash being paid for the gift of not using GF doesn't sound like something to be proud of if you're the fab. Whatever benefit AMD gets from being freed up to use other foundries for its APUs has to overcome a hefty penalty.
Perhaps all of AMD's plans cannot be satisfied by GF's volumes, pricing, yields, or some combination thereof.
It's not clear if the release from certain exclusivity clauses at 28nm means things become close again at 20nm, or the final break was meant to happen then.
At least some irony is in play because AMD is paying to ditch a process whose features were decided upon prior to the spinoff.
edit:
The take-or-pay change in the agreement could also be a sign of a lack of faith in AMD's volumes or ability to pay.
GF is going to offer its services according to some schedule, and AMD can either accept or pay some amount of money.
It would seem better if AMD took up on the offer so it could sell some of those chips, as opposed to not taking them and paying money anyway.
I'm not sure what kind of replacement volumes at 28nm AMD is expecting to get from elsewhere, such that it can turn its nose up at GF's product.
That might mean that it doesn't expect to sell as much, can't afford to pay for full volume, or the price/yield is so askew that it's better to pay a fee and not get something from GF.
There seems to be a lack of clarity as to what each change is for, and this may be partially due to a lack of visibility on what AMD would have been doing if not for this new agreement.
The amount of cash being paid for the gift of not using GF doesn't sound like something to be proud of if you're the fab. Whatever benefit AMD gets from being freed up to use other foundries for its APUs has to overcome a hefty penalty.
Perhaps all of AMD's plans cannot be satisfied by GF's volumes, pricing, yields, or some combination thereof.
It's not clear if the release from certain exclusivity clauses at 28nm means things become close again at 20nm, or the final break was meant to happen then.
At least some irony is in play because AMD is paying to ditch a process whose features were decided upon prior to the spinoff.
edit:
The take-or-pay change in the agreement could also be a sign of a lack of faith in AMD's volumes or ability to pay.
GF is going to offer its services according to some schedule, and AMD can either accept or pay some amount of money.
It would seem better if AMD took up on the offer so it could sell some of those chips, as opposed to not taking them and paying money anyway.
I'm not sure what kind of replacement volumes at 28nm AMD is expecting to get from elsewhere, such that it can turn its nose up at GF's product.
That might mean that it doesn't expect to sell as much, can't afford to pay for full volume, or the price/yield is so askew that it's better to pay a fee and not get something from GF.
Last edited by a moderator:
Instead of editing my edits further, one additional thought is that this agreement is partly the fallout from AMD's cancellation of Krishna and Witchita, whose volumes could have been locked in prior to their cancellation.
Another round of uncertainty is whether the changed wafer agreement is also in reaction to uncertainty in volume for 32nm product, or even the 2013 28nm products like Kaveri, which if they planned at GF would have begun ramping in the latter part of 2012.
As buddy-buddy as PR goes from AMD and GF, it doesn't seem that warm when it comes to actions.
Another round of uncertainty is whether the changed wafer agreement is also in reaction to uncertainty in volume for 32nm product, or even the 2013 28nm products like Kaveri, which if they planned at GF would have begun ramping in the latter part of 2012.
As buddy-buddy as PR goes from AMD and GF, it doesn't seem that warm when it comes to actions.
Bug found in amd cpu's
http://permalink.gmane.org/gmane.os.dragonfly-bsd.kernel/14518
http://permalink.gmane.org/gmane.os.dragonfly-bsd.kernel/14518
All CPUs have bugs, it's no big deal.
Most CPU bugs are relatively minor and in most case can be handled by software. Also, many CPU have "micro-code update" which can correct many types of CPU bugs.
The most serious bugs are those occurred rather frequently and can't be patched via micro-code update or software (at least without serious performance problem). The infamous FDIV bug is one, and another example is, early AMD Phenom CPU have a bug in its TLB which can cause hard lock (though very rarely), so they have to disable TLB in BIOS and OS, which degrades performance.
The most serious bugs are those occurred rather frequently and can't be patched via micro-code update or software (at least without serious performance problem). The infamous FDIV bug is one, and another example is, early AMD Phenom CPU have a bug in its TLB which can cause hard lock (though very rarely), so they have to disable TLB in BIOS and OS, which degrades performance.
Except that in this particular case, the bug has no workarounds (other than modifying the compiler to avoid the offending code sequence) and impacts normal execution (the bug is trigger by normal popping register from the stack prior to returning from a subroutine), so it's actually quite nasty, as you can't possibly know if you're going to hit the problem or not unless you recompile all your applications (which isn't possible for most people).
well you learn something new every day
Want to bore yourself to death? Check out the epic list of errata for my i7-3930k
With all these AMD-ATI folks leaving I get the feeling, right or wrong, that the AMD-CPU folks are running the show in a way that is making them jump ship. Which, again as an outsider, seems odd as ATI is the division doing the best, has the best technology, and is the best asset to leverage across multiple markets and could really be a trend setting for the future. Just doesn't seem... right.
Godfrey Cheng is now no longer with AMD.
Godfrey Cheng is now no longer with AMD.
Godfrey was working in the CPU division for the past few years.
There seems to be a migration
http://www.forbes.com/sites/brianca...md-veteran-positioning-itself-as-intel-rival/
http://www.forbes.com/sites/brianca...md-veteran-positioning-itself-as-intel-rival/
Q1/2012 number are out:
http://phx.corporate-ir.net/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=1685255&highlight=
AMD has IMO one of the best product stacks in their history plus a competitive advantage in GPU business, yet they made a loss. How?
http://phx.corporate-ir.net/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=1685255&highlight=
AMD has IMO one of the best product stacks in their history plus a competitive advantage in GPU business, yet they made a loss. How?
AMD said:
AMD said:
- Status
