Leaked: ATI Radeon HD 6990 specifications

[red1776]

Yup, like I said before, everything about it is just weird.

I await with keen interest the actual reviews....and which sites get to publish them at launch.

Care to take a shot at who gets the launch samples?

...or better yet, who doesn't?

 

[dividebyzero]

[H] (naturally), Guru3D, TPU, Bit-tech, Anand (wide circulation, vocal membership), Tweaktown, Tech Report, PC Perspective, Hardware.fr/Be.Hardware, Hexus (too influential not to include) - all nailed on certainties.

Techspot, Tom's, Bjorn 3D, OC3D, Legit Reviews -probable/possible- mainstream sites that don't court controvery

Hardware Secrets, Hardware Canucks, HT4U, Xbit, Benchmark Reviews - ???- All of these sites should have a review sample, but all have been either critical of AMD, or have published conclusions/editorials contrary to AMD's interests.

Hardware Heaven, Alienbabeltech - Both should have reviews, but like Hardware.fr and HT4U have large benchmarking suites and are traditionally a little later in posting their findings.

Just a rough guess off the top of my head.

EDIT: Maybe we should start a pool ! Too many sites to run a poll -and thats without taking into account some of the niche sites (OCIA, PureOC, VR Zone, HardwareOC, Ninjalane, Firing Squad etc.)

 

[red1776]

Thats a great idea Chef! maybe with the swing votes?

BTW I think Techpowerup gets one (because of their benchmarking suite). and I'm in for TS/Legion/Steve getting one, and Tom's ...because..... they're Tom's.

 

[dividebyzero]

2 x 8pin PCI-E power.
So long PCI SIG it's been nice knowing you!

[source] [and the ever reliable Kyle and Brent]

At least I had one Nostradamus moment even if my prediction of the GTX 590 being GF114 was off base.

Note the spacing around the 6pin connector in the picture. The shroud cutaway is clearly designed to allow the 6pin to be replaced by an 8pin connector
/Sherlock

And here's the full specs it seems.
350-375w @ 830/830/5000 and 415-450w @ 880/880/5000 (OC'ed setting BIOS setting #2)
So the 2 x 8pin power isn't just for show. This card better be considerably cheaper than buying two HD6970's because It's not going to match them in performance on these numbers.

Note the fine print in the second picture:

"AMD warranty does not cover damages caused by overclocking, even when overclocking is enabled via AMD software and/or the Dual-BIOS Function on the AMD Radeon™ HD 6990".

Not the most reassuring caveat I've ever seen

 

[red1776]

I read that in the we smalls at Hex i think it was. I think if you check, I had an even more clairvoyant Nostradamus moment about them running roughshod over the PCIE spec
Although...I did start o wonder if these guys (both Green and Red) found some real gems from the binning process. I have a sneaking suspicion that there may be a grand total of 20 cards available.

 

[dividebyzero]

That means all ten review samples will need to go back to AMD and re-boxed for sale :haha:

I do think you're definitely in the ballpark with those production figures (maybe with Asus DCII, MSI Lightning, Giga not-SOC we may get to triple figures).
It's going to be one very expensive board to produce- the extra power phases needed to handle ~450w are going to add some layers to the PCB, the extra power regulation circuits etc -for what should amount to slightly less than HD6970CFX performance thanks to the lower memory clock.

An engineering marvel to be sure- but I think this will be prohibitively priced to deter potential buyers and ensure that AMD/AIB's don't have produce many.

Should be interesting once the reviews hit the streets.
If I were a gambling man (and I am) I think I'd set a few lines:

Number of review sites getting two of these cards for CFX testing (over/under 2.5)
Number of sites topping 700w total system power (over/under 50% of reviews)
Number of reviews' benchmarks skewed from thermal throttling (over/under 75%)
Number of sites that play down the "warranty void if OC'ed" aspect (over/under 75%)
Number of review sites that admit to killing a card through overclocking (over/under 0.5)

EDIT: Just reading through some of the forums now. Amazing the sudden amount of acceptance for a power-hungry, hot and noisy card given the outrage and derision from the same people over the GTX 480 introduction. This is better than the sunday comics

 

[red1776]

EDIT: Just reading through some of the forums now. Amazing the sudden amount of acceptance for a power-hungry, hot and noisy card given the outrage and derision from the same people over the GTX 480 introduction. This is better than the sunday comics

I think they are taking the cue from the 1st round of Fermi releases. The (very few ) people who buy these cards are not going to care about its power consumption. actually IMO these cars are not going to be purchased by enthusiasts at all. A enthusiast more than likely has a board with a minimum of 2 PCIE slots ...probably 3 or four . and recognizes the Oc potential of a single card like this this is very limited and have a (to use a nascar term) restrictor plate on it. This a is an epeen, PR stunt. that can be outdone by a pair of the single GPU counterparts. Not to mention the history is one of driver problems and or slow remedy for it.
I would not invest in one of these foot long invitations to problems if you gave a 30% coupon.
But...it will be a fun launch day

Number of review sites getting two of these cards for CFX testing (over/under 2.5)
Number of sites topping 700w total system power (over/under 50% of reviews)
Number of reviews' benchmarks skewed from thermal throttling (over/under 75%)
Number of sites that play down the "warranty void if OC'ed" aspect (over/under 75%)
Number of review sites that admit to killing a card through overclocking (over/under 0.5)

Put me down for a hundred in the pick six exacta!

 

[dividebyzero]

My thoughts exactly. I would think that a HD 6990 is likely to sell for around the same price as two 6970's. Given that a good 6970 should be able to hit 1k core (a better cooled DirectCu card probably a lot higher) it doesn't bode well for the 6990 - I'm pretty sure a 1000MHz core on a 6990 would look something like this. Things I would suggest look even less rosier for the GTX 590 given the voltage/amps constriction will likely affect it more than is the case with the 6990.

Watching the fanboy wars will be fun though, as will be seeing what lengths both AMD and Nvidia will go to to claim a crown that none of their fans will probably want to wear.
The hypocricy levels have already reached Defcon 2 and we're still a few days away from launch. What would make it complete would be for the GTX590 to post lower performance but also use less power. It would be truly ROFL-worthy to see two fanboy bases switch arguments in the space of a week.

 

[dividebyzero]

Looks like my estimate of 20 cards was too conservative...who knew?!
Gibbo at overclockers.uk reckons he has 75% of the U.K.'s supply (~ 60 cards) in this picture, and total worldwide availability at less than a 1000 -so somewhat more of a collectors item than the Asus Ares.

And the review from Inpai- if anyone is interested.

 

[Guest]

NVIDIA can't never win GPU war,NVIDIA never managed to win hd5970

 

[Guest]

All they'll need to do now is engineer these GPUs out of Carbon Graphene nanotube material and crank the clockspeeds into the 300-500+ Giga hertz range with about 4-8 terabytes of 1024 bit 500-900+ gigahertz GDDR 10x memory. One GPU running at thoses speeds with that much memory available. Say hello to the ATI Event Horizon Series! Look up carbon graphene and its proposed applications to see what I'm talking about.

 

[dividebyzero]

Cool !
I look forward to playing the future console port and DirectX 9.0c API games on this new-super-duper-wicked-fast GPU.

 

[Guest]

isnt the teraflops is quadrillion calculation per seconds?? and the gigs is trillion?
you got it wrong dude

i thougt this card would cost about 550-600$