AMD Cayman info (or rumor)

[eric66]

me too

i doubt that that fugliness will fit to anybodies sig

[Heinz68]

Oh, very interesting, the giant hot ATI power hogging failure of a video card that is breaking apart at TSMC with an extremely expensive under 10% yield that drives prices through the roof and profit into extreme losses - and indicates a very, very flawed execution and core design that will cripple and cave in use as it cooks the case to melting.

EPIC FAILURE !

Can't you say BLOWBACK !!!!!!!! ????

roflmao I am SO HAPPY !

( thanks to all those hundreds from here that attacked NVidia's 480 production since that's how I know exactly the crap ATI has crooked up now and how pathetic it is )

God this is GREAT !

I checked Charlie's hate site and he's pretending to be dumb, blind, deaf, and stupid right now, not a word on the epic failure of cayman. ROFLMAO ! hahahahahhahahhaHhahaha

I am so HAPPY ! Oh man can you say " TOTAL FACEPALM !"
Hey ATI you finally did it ! >

First, I don't know how old are you but there is definitely about time for you to grow up.

Sure some people were critical and unhappy about the the Fermi delays but i don't remember people posting being happy about it as you keep repeating in your post.

Any product being delayed or having higher production cost is definitely not good news for the consumer, so what is there to be happy about?

Anyway lets wait till November 22 or something more official from AMD. Lets not forget this is Fuad news the same Fuad from Fudzila who was reporting Fermi coming out just about every months since November 2009 and it was about 6 months later on April 2010 when first Fermi cards arrived. We're still waiting for the dual Fermi cards which Fuad also reported to arrive in November 2009.

Lets also not forget the NVIDIA CEO Jen-Hsun Huang introducing the Fermi in October 2009 at GPU Technology Conference
"This Ladies and Gentlemen, this one here is Fermi." He forgot to mention it's fake with some wood screws.


So off-course people were disappointed about the many delays, I don't think anything close to it will happen to the HD 6900 cards.

[zerazax]

It doesn't even make logical sense for the card to be on single digit yields. I didn't even believe Fermi was in single digits, and Fermi is much larger and more complicated

So somehow AMD, which delivered the entire Evergreen family + Barts on time on 40nm, who is on their 3rd generation of 40nm products, is having issues with their chips much smaller than GF100/GF110... even Charlie couldn't write this story

Got a message a few days ago that the GTX 580's throne might not be as expected, performance of the 6970 is extremely impressive, at least on the same performance class as the 580.



6970 loses out on synthetics (Vantage is likely) vs 5970, but (way) better in games, bad drivers now, probably has ample room for improvement.


People have cards // NOT everyone has drivers yet

Sounds interesting. Way better than games would be awesome... single GPU > previous dual GPU is awesome when one considers the previous dual GPU was far and away the top dog

[ziggo0]

It doesn't even make logical sense for the card to be on single digit yields. I didn't even believe Fermi was in single digits, and Fermi is much larger and more complicated

So somehow AMD, which delivered the entire Evergreen family + Barts on time on 40nm, who is on their 3rd generation of 40nm products, is having issues with their chips much smaller than GF100/GF110... even Charlie couldn't write this story




Sounds interesting. Way better than games would be awesome... single GPU > previous dual GPU is awesome when one considers the previous dual GPU was far and away the top dog

Not to mention 2 weeks prior to launch... 'oh we are having issues all of a sudden' when around this time stuff should already be in transit to e-tailers and cards piling up left and right.

[Sn0wm@n]

Not to mention 2 weeks prior to launch... 'oh we are having issues all of a sudden' when around this time stuff should already be in transit to e-tailers and cards piling up left and right.

and up until the 22nd of nov we wont know if what you say is either true or not ...so let's wait ...

[zerazax]

and up until the 22nd of nov we wont know if what you say is either true or not ...so let's wait ...

Well, some folks do have pictures of the card already, so we know they're out there... the question is #'s

[Tomasis]

Oh, very interesting, the giant hot ATI power hogging failure of a video card that is breaking apart at TSMC with an extremely expensive under 10% yield that drives prices through the roof and profit into extreme losses - and indicates a very, very flawed execution and core design that will cripple and cave in use as it cooks the case to melting.

EPIC FAILURE !

Can't you say BLOWBACK !!!!!!!! ????

roflmao I am SO HAPPY !

( thanks to all those hundreds from here that attacked NVidia's 480 production since that's how I know exactly the crap ATI has crooked up now and how pathetic it is )

God this is GREAT !

I checked Charlie's hate site and he's pretending to be dumb, blind, deaf, and stupid right now, not a word on the epic failure of cayman. ROFLMAO ! hahahahahhahahhaHhahaha

I am so HAPPY ! Oh man can you say " TOTAL FACEPALM !"
Hey ATI you finally did it ! >

good example of this how well nvidia marketing works

[trinibwoy]

Not to mention 2 weeks prior to launch... 'oh we are having issues all of a sudden' when around this time stuff should already be in transit to e-tailers and cards piling up left and right.

Where did you get the impression that if there are yield issues that those issues appeared "all of a sudden"? I recall AMD making a statement about their next generation parts - was it something about coming out in 2010?

[bill_d]

come on guys quote the troll nevermore

[marten_larsson]

Where did you get the impression that if there are yield issues that those issues appeared "all of a sudden"? I recall AMD making a statement about their next generation parts - was it something about coming out in 2010?

When Barts was released there was some info about Cayman as well which clearly indicated November launch. That was a month ago, surely if they had yield issues they'd have known by then..

[UNTAMED]

"Nonetheless, building a 3 billion transistor chip was never going to be easy, a fact clearly demonstrated Nvidia's GF100."

-so is he telling us this chip will be the apox. same size as the gtx 480 ?

[tajoh111]

I wonder how true the delay is. There would be alot less credibility if fud was the only source, but he wasn't the original source. If there are yield problems, what could it be a result of? Size would be one I could imagine but how about the memory controller? Isn't 1.5 or 6ghz ghz crazy fast for a memory controller?

[tbone8ty]

amd/ati drivers team was caught with there pants down, AIB's have the card there jsut waiting on bios and drivers. apparently the drivers r bad.

[Iconyu]

"Nonetheless, building a 3 billion transistor chip was never going to be easy, a fact clearly demonstrated Nvidia's GF100."

-so is he telling us this chip will be the apox. same size as the gtx 480 ?

AMD have greater transistor density then nvidia on TSMC's 40nm process, they can hit 3 billion transistors and still be below 500mm^2. However, I haven't a clue what kind of complications this could incur. If the 580GTX is based on the GF104 then it's probably below 3 billion transistors.

Edit: GF110 is 3 billion, GF100 is 3.2 billion

[japamd]

amd/ati drivers team was caught with there pants down, AIB's have the card there jsut waiting on bios and drivers. apparently the drivers r bad.

Or AMD it's waiting to see what kind of performance GTX 580 gives to finally adjust clocks/voltages and so on. For what can be seen on techPowerUp! it's going to have low clocks and voltages

[Sn0wm@n]

"Nonetheless, building a 3 billion transistor chip was never going to be easy, a fact clearly demonstrated Nvidia's GF100."

-so is he telling us this chip will be the apox. same size as the gtx 480 ?

so cayman is a 3 billion transistor chip ???? since when ??

[Nintendork]

Sadly the tpu "review" is a mess of drivers xD (compare 580 preview with HD6870 review numbers).

[Dimitriman]

the tpu review is kinda messy but things seem the point out than gtx 580 is around 5970 perf beating it in some games and losing in others while mostly around the same perf +-5%

[tbone8ty]

Or AMD it's waiting to see what kind of performance GTX 580 gives to finally adjust clocks/voltages and so on. For what can be seen on techPowerUp! it's going to have low clocks and voltages

true, nice review :p

[maoklein]

i've been reading this thread since page 1 -- actually, others cayman related as well -- and i believe i haven't heard of crossfire scalability on caymans yet.

surely, that would be something that only reviewers and costumers with the real cards at hand could attest to, but since this is a rumor and speculations thread ...

in topics (cause i'm not very articulated ):
- what got me thinking is that the gtx580 will be tri-sli capable (not sure if this is a cap* or just common sense )
- but since we all know that despite poor scalability quad xfire is a reality, regarding the cypress series, specially on the hd5970 (x2) -- + amd will be releasing another dual chip version (antilles)
- and i`ve read on hardwarecanucks that the barts cards offer an "awesome" scalability in crossfire (x2)

from that i`ve got 2 inquiries:
have anyone heard of this sli limitation on the 580 series?
would that awesomeness on xfire scalability be exclusive to the barts series (and so, more hardware bound, so to speak) or have amd already tweaked the drivers anticipating this ...and so extending its effects throughout the 6k series ...?

what i mean is this: suppose the 580`s are capped to x3 sli, suppose catalyst new installments are "awesome" and they got rid of the bottlenecks that affected the x4 gpu`s (even yielding inferior performance against x3 (5870`s) on some games benchies) ...couldn`t we assume that the ultimate e-peen rig would be composed of x4 caymans (and thus negating the possible 5-10% superior performance of a 580... x3)?

sorry, i`m all over the place and not sure if made my point across. as you can see, lots of doubts ^^ (maybe i could summarize it by saying: any word/expectation on xfire for caymans yet?)

*my doubts are substantiated by this 'rock solid' argument: i saw that on a gtx580 galaxy box, written on a sticker, that was posted on the 580 thread (which would add further to the vagueness of this limit -- and i also know for sure that quad sli`s are possible on the 480`s)

ps it got me 2 years to get a login, another year to post this. so, my next post will be in 6 months back to lurking

edit to add: but then again we`ve got this new fud of driver issues
response to (@)nintendork -- cause i do not wish to surpass my 1 post count : "ultimate e-peen rig" + IF that tri-sli cap is true (for 580`s) then antilles (in xfire, virtually x2, effectively x4) would definately smoke them ...depending on the drivers, i guess. maybe =p

[Nintendork]

What would you run on 4 caymans? Just two of them with 2gigs each must deliver enough for an eyefinity setup specially with MLAA.

Barts HD6800 scale better than fermi in multigpu, cayman should not be different.

[Dimitriman]

i've been reading this thread since page 1 -- actually, others cayman related as well -- and i believe i haven't heard of crossfire scalability on caymans yet.

surely, that would be something that only reviewers and costumers with the real cards at hand could attest to, but since this is a rumor and speculations thread ...

from that i`ve got 2 inquiries:
have anyone heard of this sli limitation on the 580 series?
would that awesomeness on xfire scalability be exclusive to the barts series (and so, more hardware bound, so to speak) or have amd already tweaked the drivers anticipating this ...and so extending its effects throughout the 6k series ...?

what i mean is this: suppose the 580`s are capped to x3 sli, suppose catalyst new installments are "awesome" and they got rid of the bottlenecks that affected the x4 gpu`s (even yielding inferior performance against x3 (5870`s) on some games benchies) ...couldn`t we assume that the ultimate e-peen rig would be composed of x4 caymans (and thus negating the possible 5-10% superior performance of a 580... x3)?

ps it got me 2 years to get a login, another year to post this. so, my next post will be in 6 months back to lurking

edit to add: but then again we`ve got this new fud of driver issues

From my understanding nothing will stop you from using 4x gtx 580 in a setup although you would need a separate 1200w power supply for that. The barrier comes also from drivers not being too efficient in distributing the load among 4 cards and its the same deal with 2x 5970 although less obvious in the latter. All gtx480 owners I spoke with say the best possible setup is tri-sli and quad is a waste of money.

We dont know how Cayman will come out but we can definately expect that it will at least close the gap vs the gtx580 better than did the 5870 vs the 480. So having tri 580 or tri 6970 will be a comparible setup and possibly the same level of 2x Antilles due to reasons given above.

btw if you wanna see a real epen system it doesnt get much better than this:

http://www.youtube.com/watch?v=-s9LM..._order&list=UL

[Lanek]

i've been reading this thread since page 1 -- actually, others cayman related as well -- and i believe i haven't heard of crossfire scalability on caymans yet.

surely, that would be something that only reviewers and costumers with the real cards at hand could attest to, but since this is a rumor and speculations thread ...

in topics (cause i'm not very articulated ):
- what got me thinking is that the gtx580 will be tri-sli capable (not sure if this is a cap* or just common sense )
- but since we all know that despite poor scalability quad xfire is a reality, regarding the cypress series, specially on the hd5970 (x2) -- + amd will be releasing another dual chip version (antilles)
- and i`ve read on hardwarecanucks that the barts cards offer an "awesome" scalability in crossfire (x2)

from that i`ve got 2 inquiries:
have anyone heard of this sli limitation on the 580 series?
would that awesomeness on xfire scalability be exclusive to the barts series (and so, more hardware bound, so to speak) or have amd already tweaked the drivers anticipating this ...and so extending its effects throughout the 6k series ...?

what i mean is this: suppose the 580`s are capped to x3 sli, suppose catalyst new installments are "awesome" and they got rid of the bottlenecks that affected the x4 gpu`s (even yielding inferior performance against x3 (5870`s) on some games benchies) ...couldn`t we assume that the ultimate e-peen rig would be composed of x4 caymans (and thus negating the possible 5-10% superior performance of a 580... x3)?

sorry, i`m all over the place and not sure if made my point across. as you can see, lots of doubts ^^ (maybe i could summarize it by saying: any word/expectation on xfire for caymans yet?)

*my doubts are substantiated by this 'rock solid' argument: i saw that on a gtx580 galaxy box, written on a sticker, that was posted on the 580 thread (which would add further to the vagueness of this limit -- and i also know for sure that quad sli`s are possible on the 480`s)

ps it got me 2 years to get a login, another year to post this. so, my next post will be in 6 months back to lurking

edit to add: but then again we`ve got this new fud of driver issues
response to (@)nintendork -- cause i do not wish to surpass my 1 post count : "ultimate e-peen rig" + IF that tri-sli cap is true (for 580`s) then antilles (in xfire, virtually x2, effectively x4) would definately smoke them ...depending on the drivers, i guess. maybe =p

We don't get the performance of the 6970 or the Antilles 6990... or even a real idea on how scale the GTX580 in Tri-SLI ... it's just really impossible to respond now ....

[zerazax]

I wonder how true the delay is. There would be alot less credibility if fud was the only source, but he wasn't the original source. If there are yield problems, what could it be a result of? Size would be one I could imagine but how about the memory controller? Isn't 1.5 or 6ghz ghz crazy fast for a memory controller?

Techeye is the only one who has reported this. Fud just picked it up from them. Not to discredit techeye, but I've never even heard of them... if this were Digitimes or Toms or somewhere I'd be more concerned about this rumor

[spursindonesia]

Even on a pessimistic perpective, i don't buy the low yield rumor, the chip has been in development for quite a long time, already taped out sometime ago, and if there was a low yielding situation it would already came out, not just leaked right when the product is gonna hit the street. But the difficult driver development, those i can somewhat weigh on, because the chip will be a new mArch that won't be easy to extract performance upon early on.