-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD 69XX delayed officially ?
- Thread starter Martini1991
- Start date
Associate
- Joined
- 27 May 2007
- Posts
- 1,346
- Location
- Kent
) would seem a wast of money throwing away.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,226
Since when have amd been worried about having the fastest gpu crown? I think the last time they worried about that was hd2900 which turned out pretty bad. They still have the worlds fastest card and will continue to have that by a big margin when antilles comes around. This delay could be down to any number of reasons just wish amd would confirm something.
Have you got proof of this big margin, Links etc ?(What is a big margin in your eyes)
Maybe they need the time to get all of the box's reprinted with the words 'Even Faster than the worlds fastest dx11 GPU'
Or (more likely) theyve delayed the release to have a larger stockpile of cards for sale in a few weeks time.
Depends, if its a new component that does a unique, specific job, then yes it is possible that this has delayed things.....
Quite possibly. The only people that know the real reason are AMD and they won't tell so I'm guessing we will never find out.
It's too late in the day for AMD to still be tinkering with the card though so I doubt it's any last minute nvidia orrientated performance tweaking as some people have rumoured (just my own opinion so need to flame me if I turn out to be completely wrong).
Out of interest did AMD ever actually set a firm release date for this card or was it merely 'expected/rumoured' to arrive one month after the 6800 series cards?
Common sense? Even if a 6990 was just two 6870 GPUs, it'd have quite a performance lead against a GTX580. Now considering the fact that Cayman is supposed to be a nice chunk faster than a 6870 (minimum rumoured specs should put a 6970 at 580 performance) a 6990 is going to have a ridiculous lead in performance over a 580. If a 6970 has GTX580 performance, then a 6990 should give GTX580 SLI performance, or there about.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,226
Well considering 2x6870's in crossfire are quicker than a gtx580 by a decent margin in some games i would assume a cayman based antilles will be somewhere in the region of 50% faster which for me is a big margin. Tbh i don't think i need proof that a dual gpu is gonna be a good deal faster than a single chip. Most people think a 6970 is going to be as fast as the gtx580 never mind 2 chips based on either 6970/6950 in crossfire.
As Dutch guy said, it's about disabling part of the chip so that you can increase the percentage of chips that are viable...
Your GPU is divided into a number of semi-independent units. It's possible to disable one or more of these and the GPU still works, although by doing this you obviously reduce its performance slightly.
When manufacturing the chip, there is a good probability of a fault developing - this is what people mean when they refer to the "yield" of a chip. The yield is the percentage of chips that come through with zero faults. Now, if the fault lies in some vital area (like a global data share, memory controller or cache) then the chip is a dud. If it lies in one of the modular processing areas (...the most likely outcome) then you can simply disable that module and the chip can still function. Most of the chip is modular in one way or another, so most manufacturing faults don't lead to an outright "dud" chip.
Now, obviously all of your 6970s need to have the same spec - you can't have some of them with 1920 shaders and others with 1856 or 1792 (by disabling one or two blocks of 64 shaders). So, by choosing to disable a few extra clusters on the 6970 spec, you increase the percentage of chips that are viable (i.e. you improve the yield).
This is commonly done for the derivative parts - i.e. a 5850 is a full 5870 chip that has some of the clusters disabled. This way AMD can still make money on the chips that have developed a fault, rather than just throwing them away. However, if you are having difficulty producing a good number of full chips, you might disable some of the clusters even on the 'full' chip.
Nvidia did this with the GTX480, for example... On the early 40nm process at TSMC they could not get a good enough yield of the full 512 shader part, so they disabled a bank of 32 shaders on all GTX480s to increase the percentage of viable chips (i.e. to "improve the yield"). The assumption here is that AMD could be doing the same with Caymann - after all it's an extremely complex chip based on an updated architecture, so it's to be expected that initial yields won't be great... As part and parcel of this, they may be increasing the clockspeed and/or voltage of the chip in order to guarantee that it still beats the GTX580 (for a "PR victory" or to allow them to price the 6970 to compete directly with the GTX580).
Last edited:
Soldato
- Joined
- 16 Jan 2003
- Posts
- 10,975
- Location
- Nottingham
Sourcing components that nobody has the specs for and doesn't even show up on TI's website? I very much doubt Nvidia is using the exact same component that is mentioned in the article.
Ahh ok mate, i did not know if you had seen something somewhere!
Soldato
- Joined
- 28 May 2007
- Posts
- 10,226
Lol i wish i had all this rumour business is doing my head in.
aaarrgghhhhh I want them soon
Same here
First thing first!
This is a BIG delay. 3 weeks? Not big right?
Missing the crucial Xmas window is BIG!
Whoever says that window doesn't matter is either a complete retard or has never been in retail. Let's no also forget 6990 had Been delayed too.
So let's see the excuses -
1) Yeald issues, it's possible... depend just how big the 69 series is. But unlikely! the 40nm has matured in the last 2years and amd ppm is second to none.
2) tweaking 69## so it can beat this and that. Complete and utter BS. What delayed 6990 then? Tweaking it to beat what? Utter bs, that a company wants to tweak cards so it can have 70fps instead 69fps and miss the Xmas window..
3) Texas instruments and other parts missing...
Very probable theory. Something that's outside of Amd's control.
And I tell u what. Had they yspent more time on 69## series instead of ****ing about with 68## maybe they would have had a decent product out already by now.
As it stand Nvidia just pulled carped under amd and stole the show with the "inferior" unmanufacurable fermi. Hence 580 gtx on order.
Last edited:
Yep - raising clockspeeds and voltages will increase power draw (and so also heat output). Nothing comes for free...
But, it's worth remembering that disabling some components within the chip will reduce power draw... So, assuming that AMD are aiming for a particular max thermal output, if they disable some components of the 'full' chip then they will have a little extra headroom to increase clockspeeds.
Associate
- Joined
- 6 Sep 2006
- Posts
- 728
Yes to both, but since we don't know how hot/energy hungry they are now, it may be that there is ample headroom for such tweaks.
All guesswork, but my money's on ^Andy^'s suggestions of a combination of building up stock levels and tweaking drivers. Of all of the issues noted, these are two where ATI's track record has been less than perfect and if I was already sitting on a clear performance win, I could see the benefit of fixing these to avoid the references to a paper launch / woeful buggy drivers.
I don't believe that stockpiling products is a valid reason to delay, particularly when a number of potential sales are going to your biggest rival for every day that you delay... It would be better from a sales point of view to have the product "out there" with limited availability. You don't increase your overall sales by waiting until you have a large surplus of products...
Tweaking and optimising drivers could be a very valid reason for a delay though. If the architecture is significantly different from the 5-series then there could be a lot of room for improvement with driver tweaks (just like there was with Fermi). This could tie in with a desire to beat the GTX580. It's entirely possible that AMD feel a few specific driver tweaks could "put them over the edge" to match or beat the performance of the GTX580.