ATI Availability at Anand's

[Geo]

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2317&p=16

So much for an "Availability refresh". And that crack about it coming from the partners instead of ATI makes it more likely to be true has got to sting a bit. . .

Apparently NV isn't doing so hot on PCie availability either tho.

 

[Bouncing Zabaglione Bros.]

Apparently Dell has bought them all again, and OEMs are getting priority over the retail market.

 

[AlNom]

Why is there such shortage in the first place versus last gen? Is it the 130nm process (that everyone is using it?)

 

[Galduta]

Despite what ATI promised, none of the Radeon X850 line are available, even as we start 2005. The Radeon X850 XT will be available at the end of January with the X850 XT PE available a week later in February.

There is no word on the Radeon X850 Pro, which confirms our suspicions that ATI hadn't thought their positioning through on that part before announcing it.

Even after the X850 line is officially launched, we are told that availability will be extremely limited.

In Spain ....

http://www.izarmicro.net/

the store is very serious.

-

 

[Geo]

Seems to be ample and growing evidence that ATI is slowly eating thru their store of goodwill with a pretty important group of folks --graphics/gaming site webmasters. Not at crisis level yet, but they better get it turned around sooner rather than later.

Those folks typically got into the biz because they love the subject, love being the recipient of authentic "inside info", and have a natural 'splainer predisposition in their personality that makes them highly value their own credibility, even when passing on second-hand info. ATI has been chipping away at least two of those, and another. . .making them bust their butts to do reviews on cards that then either don't show up on the market or show up much later than the webbie tells his readers they are going to.

In the grand scheme this is small beer to the larger community. . .but rumor has it that most webmasters are people too, and these little indignities build up in their craw.

 

[Apple740]

Well, at least there is one thing that Ati delivers on time each time; their Cat drivers.

 

[ANova]

Let's not forget nvidia isn't doing much better. It's almost impossible to find PCIE 6800s, as the article mentioned. The 6800 Ultra is fairly hard to find at times as well. All this seems to suggest there is something else causing the delays. My guess would be complications among DDR3 quantities and/or the fab process of either the chips themselves, the memory modules or both.

 

[Bjorn]

It's almost impossible to find PCIE 6800s, as the article mentioned. The 6800 Ultra is fairly hard to find at times as well.

Is very easy to find 6800 PCIE's. The 6800 GT/Ultra PCIE versions are another thing though

All this seems to suggest there is something else causing the delays. My guess would be complications among DDR3 quantities and/or the fab process of either the chips themselves, the memory modules or both.

I seriously doubt that RAM modules are causing the problems. Why would the AGP versions of the 6800's have such good availability then ? (There are lot's of 6800 GT's and even Ultras in stock at komplett.se f.e)

 

[phenix]

I am guessing that the reason for the availibilty crisis that plagued the whole industry is due to mass transition of the industry to 90nm process production. So my prediction is, all will be solved when the transition is complete due to second half of the year when taiwanese foundaries start full scale production of the new process.

 

[Rys]

I seriously doubt that RAM modules are causing the problems. Why would the AGP versions of the 6800's have such good availability then ? (There are lot's of 6800 GT's and even Ultras in stock at komplett.se f.e)

Indeed. It seems that the number of now basic mid-range boards apparently being shipped with 1200MHz GDDR3 puts paid to the rumours that Samsung can't make enough to go round.

 

[Dave Baumann]

I am guessing that the reason for the availibilty crisis that plagued the whole industry is due to mass transition of the industry to 90nm process production.

I would be more included to suggest that there are multiple factors that are combining - we have the issues that modern processes take 3 months to turn around, such that if you have a popular part and want to order more wafer for that part it'll be three months before the extra supply is seen, we have complxexity problems at the high end, we have a transition to a new interface and to top it all off we have a massive OEM demand because OEM's are more frequently looking to differentiate by graphics rather than CPU - this demand does also put pressure on the supply of all components as well (memory, flipchip substrates, etc.)

 

[Fodder]

modern processes take 3 months to turn around, such that if you have a popular part and want to order more wafer for that part it'll be three months before the extra supply is seen

Can you expand (in layman's terms ) on why it takes quite so long? Or do you just need to book (a long time) in advance?

 

[phenix]

How much time does it take to switch a semiconductor foundry to a more advanced process (installing the new hardware, testing, optimisations and ramping up the production)? Does the foundry produces any wafer with the old tech during the transition process? Because if it doesn't, this means the foundry will sit idle during this period causing a decrease in production.

I dont have a cow myself but I am imagining that huge technology transitions in the overall industry, such as this 90nm transition effects the overall production potential of semiconductor companies. I may be way off with this idea of course, still too early in the morning for me.

 

[PatrickL]

I remember reading that ATI was securing in past august foundry space for January. If it is true it is awfully long.

 

[ANova]

I seriously doubt that RAM modules are causing the problems. Why would the AGP versions of the 6800's have such good availability then ? (There are lot's of 6800 GT's and even Ultras in stock at komplett.se f.e)

Indeed. It seems that the number of now basic mid-range boards apparently being shipped with 1200MHz GDDR3 puts paid to the rumours that Samsung can't make enough to go round.

Yeah you're right. I forgot about the 6600 GT.

 

[Nappe1]

it always can be argued, what has really happened to availability, but somehow I feel like saw this coming, though I wasn't quite sure how it would be visible to consumers.

almost exactly year ago, Bloodbob asked me via PM about how I see the situation, and I replied that fab processes can't keep up with the development of the chips and that must be visible somehow or other. (unfortunately I have lost my reply, and I only have the original PM from bloodbob left.) In this case, it's more like visible as problems to deliver enough working high end chips than that lines wouldn't be able to make them at all. I am not sure if the yields are good enough or if the manufacturing capacity is just lacking the need (or perhaps both partially.), but in any case, the situation to consumer is pretty much same.


Also, anyone ever thought that both big ones are trusting single semiconductor manufacturer? if anything troubles arises with TSMC (strikes, natural catastrophes, etc...) what happens to availability of chips? How big storages companies have to keep up if something happens?

 

[101998]

Personally the shortages are no surprise to me. If you look at the age of atari/nes/snes generation they are just coming into the early/mid/late 20's and are getting into jobs that pay well and can afford the high end systems, wither they know shit about computers or not they still want to have the best. That is puting a huge strain on high end product that was not there in the previous generations. Plus look at all the high profile games that came out this year; people who want to see the "pretty graphics" think they need the highest of the high end to truly enjoy those types of games. Then you get the people who are tried of the influx of bullshit TV and turn to MMOs for entertainment (WoW has about 350k active subscribers and EQ2 has 310k active) they as well want to see 'teh pretty' or be the l33t d00ds in and out of game so they go into Fry's/Comp/Best Buy/Dell and say "I want the best you can give me". Like others said, oems get first dibbs and if they are selling more we are seeing less.

 

[Geo]

Also, anyone ever thought that both big ones are trusting single semiconductor manufacturer? if anything troubles arises with TSMC (strikes, natural catastrophes, etc...) what happens to availability of chips? How big storages companies have to keep up if something happens?

Hopefully Taiwan's army won't pick TSMC's plants to make their last stand against the Red Army, when the time comes.

 

[overclocked_enthusiasm]

In my opinion it is simply an inventory issue. Neither company has shown the ability to forcast or manage the correct inventory levels at the right time. ATI first grabbed all the Tier 1 OEM wins with PCIe because they had the first native solution. NVDA responded by entrenching themselves in retail with something ATI did not have...next gen AGP parts. ATI continued the OEM dominance with barely enough .13 low k capacity to satiate the OEM PCIe demand. There was no extra capacity for AGP so they left $60 million in unrealized revenue from Q4 on the table and ceded the next gen retail market to NVDA. ATI simply countered with older legacy AGP parts at retail to try and buy time before they could switch some capacity to .11 micron with the refresh and hopefully make more cores. As the goodwill and mindshare started to erode, they rushed some AGP parts into the channel and now you can usually find some online. Neither AGP nor PCIe next gen parts have made it to brick and mortar retail. They overguessed PCIe demand and underguessed AGP demand and mismanaged both inventories.

NVDA on the other hand was late to the PCIe party because the HSI chip is for whatever reason less effective AGP to PCIe in gaining OEM share. The only bridged PCIe product that was in OEM was the 6800 GTO at Dell. My guess is they were too big, too hot and too expensive. As a result, NVDA was shut out of OEM until they came up with a native PCIe solution like 6600 or 6200. They therefore concentrated on the retail AGP market and made a killing from ATI's leaving that market completely abandoned. Now that they have a successful retail product in either native AGP, native PCIe or bridged PCIe to AGP they can start to penetrate the OEM market once again with native PCIe solutions. The HSI bridged PCIe to AGP work great in retail.

In sum, ATI missed the retail opportunity and NVDA missed the OEM opportunity. Neither could have pulled off both OEM and retail from what I have seen due to yields, capacity and overall demand. ATI has been forced to use the OEMs to try and penetrate retail and NVDA has been forced to use retail to trty and penetrate OEM. My how things change...NVDA the retail darling and ATI the OEM darling. Hard core NVDA fans are pissed cuz they can't buy top PCIe NVDA systems and ATI fans are pissed because retail availability for add in cards suck. Take your pick as both companies have plenty of blame to go around. The proof will be in the financial pudding in coming quarters.

 

[ben6]

The AGP to PCI Express transition is a strange one. At least when moving to AGP you could still use your PCI video cards. Now only specialized motherboards like ECS's 915P-A take both "Hacked AGP"