Key Quote: The Taiwan Semiconductor Manufacturing Company Ltd. (TSMC) is struggling to get up to speed manufacturing AMD's 5800 series, 40-nm GPUs (graphics processing units), according to Jim McGregor, an analyst at In-Stat. He added that the foundry is in full production but so far yields are below expectation.
Matt Davis, a spokesman for AMD, confirmed to Computerworld that TSMC is having issues in ramping up production of the chips. He added that it's not clear how far behind the foundry is on production expectations.
"The design is sound. It's just a matter of trying to get TSMC to a point where they can yield. They're feeling the manufacturing crunch," said Davis. "We're a little bitter under yield but we're working back into a manufacturing schedule we want for these parts. TSMC can only kick them out so fast at this point."
Davis also said AMD is working with TSMC on the issue and hopes to have production up to speed by year's end.
*******************************
I wonder if this is the same problem that Nvidia is/are having with Fermi samples. If 40nm yields are low on AMD's (relatively) smaller 5xxx part, and depending on what the specific issues are, imagine the impact it may be having on Nvidia with their much larger fermi samples.
I wonder if maybe Nvidia was aware of these issues and when planning Fermi production decided that instead of cramming a large chip on a weak yield process, to take the time to do some extra spins and work out the bugs while waiting for profitable yields. I suspect Nvidia needs yeilds to be very solid with the size of their chip to make a decent return.
Last edited by brybir; 11-06-2009 at 10:09 PM.
Reason: I cannot copy and paste...
NVidia disabled ATI Manufacturing in their Process Drivers.
Damn you NNNNNNNVIIIIIIIDDDIIIIIIIIIIIIIAAAAAAAAA!!!!!! Damn you to HELL!
__________________
In the event that this post proves useless, it also doubles as a bump! Sandorski's Wager: Why waste time preparing for an afterlife, when the only Life you absolutely know you have may be the only chance you get to affect the Universe positively? Wowza2
nvidia disabled ati manufacturing in their process drivers.
Damn you nnnnnnnviiiiiiidddiiiiiiiiiiiiiaaaaaaaaa!!!!!! Damn you to hell!
lol
__________________
"My right flank has been turned, my left flank has disintegrated and my rearguard is in chaos. Situation excellent. I shall attack."
System: E8400 (4050mhz, 450x9 @ stock volts), Abit IP-35 Pro XE, 4gig RAM, HD4890, Vista Ultimate x64
Here, every vendor has the 5xxx products listed, but they're not in stock, they're all "upon availability from supplier". That typically means 2-3 weeks here for hard to source products. And I thought I'd be getting one for christmas as the supply would have stabilized (as I mistakenly assumed).
Perhaps next time they (nvidia aswell) should be looking to Global Foundries and not TSMC...?
You can bet all options are routinely assessed and reassessed, but who is to say GF would do any better of a job than TSMC?
We can all bitch and moan about the lack of 40nm production capacity at TSMC but the simple fact of the matter is that currently the 40nm production capacity at IBM, GF, UMC, Chartered, Samsung, etc is zero.
So had AMD selected any other foundry for their 40nm GPU lineup guess how many chips would be available right now?
Any non-zero number of 5xxx supply is better than zero supply. Sure a larger non-zero supply would be better than the currently small non-zero supply, but TSMC is the only one right now who can deliver a non-zero supply of 40nm parts.
You can bet all options are routinely assessed and reassessed, but who is to say GF would do any better of a job than TSMC?
We can all bitch and moan about the lack of 40nm production capacity at TSMC but the simple fact of the matter is that currently the 40nm production capacity at IBM, GF, UMC, Chartered, Samsung, etc is zero.
So had AMD selected any other foundry for their 40nm GPU lineup guess how many chips would be available right now?
Any non-zero number of 5xxx supply is better than zero supply. Sure a larger non-zero supply would be better than the currently small non-zero supply, but TSMC is the only one right now who can deliver a non-zero supply of 40nm parts.
Yup, GF is a no-go at the moment, BUT, I tihnk this will make others realise that at the cutting edge they need fab choices, and so it may cause more people to look at GF in the future than may have otherwise.
If GF can get their 32nm going well they would surely be able to compete with TSMC, which can only be good for almost everyone.
This whole issue just highlights the problem with relying on one company to produce something.
This is playing out favorably for Nvidia. Had AMD been able to supply regular amounts of their new cards, they would have had a significant sales jump on the high end and mid range DX11 parts. Instead, it appears that by the time TSMC solves the yield issues, Nvidia will be on the market along side AMD.
I'm anxious to see all dx11 parts out, and what kinds of price fluctuations they'll experience in the first 8-12 months.
@tviceman: Agreed. This is probably a nice turn of events for Nvidia, given their current problems (I know they aren't exactly strapped for cash right now, but it's not exactly all peachy for Nvidia at the moment and they could use this lucky break).
BTW: AMD, ATI... is Nvidia also "NVIDIA" (all caps like AMD/ATI) or just "Nvidia"? In articles, I find little consistency in the "internetz" whether it's NVIDIA or nvidia or Nvidia.
You can bet all options are routinely assessed and reassessed, but who is to say GF would do any better of a job than TSMC?
We can all bitch and moan about the lack of 40nm production capacity at TSMC but the simple fact of the matter is that currently the 40nm production capacity at IBM, GF, UMC, Chartered, Samsung, etc is zero.
So had AMD selected any other foundry for their 40nm GPU lineup guess how many chips would be available right now?
Any non-zero number of 5xxx supply is better than zero supply. Sure a larger non-zero supply would be better than the currently small non-zero supply, but TSMC is the only one right now who can deliver a non-zero supply of 40nm parts.
One thing I wondered about, with Global Foundries ties to AMD, would Nvidia really want to use them? I think AMD owns somewhere around 1/3 of Global Foundries, and I remember (maybe incorrectly?) some number of the board of directors have ties to AMD as well, half maybe? I can't remember exactly.
Anyway, Nvidia would have to share a lot of sensitive information with people that have a lot of ties to AMD if they were to ask Global Foundries to build chips. Would it be possible that say with Fermi2 Nvidia would give Global Foundries the schematics, or however they put the order in for the wafer, and then it's possible that AMD would know before launch what they're going to be competing against? Sure it'd be too late for AMD to change their competing design, but maybe they'd know a clock speed goal they'd like to acheive, or have a better idea of what Nvidia's chip will be capable of, and where to price their part. I'm sure Nvidia would demand in the contract that AMD can't know what their specs are, but I would think it wouldn't be too hard for someone at Global Foundries/AMD that is in a lofty position to find out and possibly use that information to better their (AMD's) competitive position.
Care to shed any light on the possibility of that happening?
__________________
Steve
PhenomII 940 currently at 3.68GHz/2070MHz uncore under a Scythe Ninja Copper
MSI K9A2 Platinum
4x2GB A-Data Vitesta DDR800 4-4-4-12
Sapphire Toxic Edition Radeon 4870 512MB @ 820/4100
Westinghouse L2610NW (25.5" 1920x1200)
ABS Tagan BZ 900 watt power supply
Logitech Z-5500's
One thing I wondered about, with Global Foundries ties to AMD, would Nvidia really want to use them? I think AMD owns somewhere around 1/3 of Global Foundries, and I remember (maybe incorrectly?) some number of the board of directors have ties to AMD as well, half maybe? I can't remember exactly.
Anyway, Nvidia would have to share a lot of sensitive information with people that have a lot of ties to AMD if they were to ask Global Foundries to build chips. Would it be possible that say with Fermi2 Nvidia would give Global Foundries the schematics, or however they put the order in for the wafer, and then it's possible that AMD would know before launch what they're going to be competing against? Sure it'd be too late for AMD to change their competing design, but maybe they'd know a clock speed goal they'd like to acheive, or have a better idea of what Nvidia's chip will be capable of, and where to price their part. I'm sure Nvidia would demand in the contract that AMD can't know what their specs are, but I would think it wouldn't be too hard for someone at Global Foundries/AMD that is in a lofty position to find out and possibly use that information to better their (AMD's) competitive position.
Care to shed any light on the possibility of that happening?
What you say is absolutely plausible, but understand that there are no secrets at any level in this industry because there are so many people involved and certain percentage of those people have ego's that can only be satiated by them constantly talking about how great their company is (and by extension, themselves in their eyes) with the 3rd party vendors (tool suppliers, software suppliers, cafeteria food suppliers, there is no end to the list) who also just so happen to supply the competition or are married to them.
Plus you have constant headcount flux from one company to the other and back again between regular job-hopping and layoffs.
In other words the people at AMD who stand to benefit from knowing info about NV's plans likely already know that info regardless whether NV is using TSMC or GF as their foundry.
Have you by any chance been following the ongoing public saga of the TSMC vs. SMIC court-case debacle?
Quote:
Jury Finds in TSMC's Favor in TSMC vs. SMIC Trade Secrets Trial
The jury in the two-month-long TSMC vs. SMIC trade secrets trial ruled in TSMC's favor this morning after several days of deliberation. The jury found that SMIC has been misappropriating TSMC's trade secrets since a 2005 legal settlement was reached between the two foundries.
That article doesn't do the history justice though, if you do go to the link be sure and scroll to the bottom of the webpage and peruse thru the "Related Contents" links, some of the specific examples of blatant corporate espionage are funny.
Back to the topic of Nvidia and Globalfoundries versus TSMC...TSMC offers one distinct experience advantage that GF's simply lacks the experience to offer and that is yielding and packaging ridiculously large die. There are scaling issues in packaging that are not trivial to solve, nor are they readily foreseeable by simple extrapolation of lessons learned on solving packaging problems posed by smaller die, which globalfoundries simply can't be expected to be able to master at time zero should Nvidia ever migrate to GF with their 500mm^2 monsters.
That said, IMO GPU's on SOI stands to make a compelling performance impact when we consider that the limitations on today's top-end GPU's are almost entirely TDP related. Anything that can be done to reduce power-consumption per xtor and per GHz will be directly fed back into releasing a higher clocked, higher xtor count GPU.
So IF AMD were to migrate to GF's SOI for GPU production at 32nm and beyond then I think Nvidia would be at a distinct disadvantage to AMD in the process technology department in a way that would force their hand to migrate as well simply to even-up the field again.
BTW: AMD, ATI... is Nvidia also "NVIDIA" (all caps like AMD/ATI) or just "Nvidia"? In articles, I find little consistency in the "internetz" whether it's NVIDIA or nvidia or Nvidia.
Part of the problem is the 5770. Lots for sale and if they were a little quicker they would have taken the pressure off the 5850. That said I will likely get one.
I would imagine Nvidia is having the same problems.
__________________
"Communism can be defined as the longest route from capitalism to capitalism."
"Capitalism is the unequal distribution of wealth. Socialism is the equal distribution of poverty"
"The Democrats seems to be the kind of people who switch to Geico and lose money." -Jon Stewert
NVidia disabled ATI Manufacturing in their Process Drivers.
Damn you NNNNNNNVIIIIIIIDDDIIIIIIIIIIIIIAAAAAAAAA!!!!!! Damn you to HELL!
Hahaha, I lol'd Nice one. I needed a laugh. Lol. And yeah poor ati, they get a good card going on and nooooo their supplier can't make enough of them lol.
__________________
more active culture than a carton of yoplait.
Hahaha, I lol'd Nice one. I needed a laugh. Lol. And yeah poor ati, they get a good card going on and nooooo their supplier can't make enough of them lol.
The silver lining is (thought it's not much of a lining, admittedly) is that they'll capture all the "market surplus". So all those who can't wait and are willing to pay the higher price (due to the recent price increase thanks to high demand and low supply) will do so. Normally, they'd have lost the high-end market surplus (I mean those willing to pay the higher price).
The "right range" (normal price) and the lower-priced market surplus (those willing to pay only a lower price than what AMD set the regular price of the 5xxx cards) are already in the bag, c/o rebates. It's the "willing-to-pay-higher" market surplus that normally doesn't get milked (oh, and how capitalists hate that), which is what will happen now (I mean they will get to milk all the impatient, rich people).
Of course, given that most of the profit in video cards don't come from high-end, expensive unit sales, it's not like this "silver lining" will mean buckets of money for AMD now. But at least, there's something nice (for them, not for us, I'm not one of those impatient rich people) for them anyway however minor, to console themselves while they work with TSMC to remedy the yield problems.
I think we'll be seeing a light at the end of the tunnel soon according to this HardOCP article. At least as far as 5800 series card availability.
Looking at the bullet points, it's surprising how little cards have shipped. The good thing is TSMC seems to have gotten their crap together and ATI expects "tens of thousands" of cards to ship per week in late November. I think we'll see most of the pent up demand alleviated in early December. But for now it seems you guys looking for cards will still have to do some scrounging.
Another interesting bullet point is how they say TSMC has, once again, solved their 40nm manufacturing problems and production has ramped up. They also say that TSMC has new machines since September and that production is now starting to come online. The new machines are probably a major reason for the production of 5800's to go from a few thousand a week to tens of thousands per week.
So this bodes well for anyone looking for 5800 cards next month but you still seem to be SOL if you're looking for one in the next 2-3 weeks.
For ATI, this is not the best situation to be in, being supply constrained, when you have your competitor in a "Johnny come lately" status but at least they will still have some time before nVidia comes out with Fermi.
AMD/ATI Confirms TSMC Supply Problems for 5XXX
Ouch
