AMD Radeon HD 9970 Specifications Leaked – Twice as fast as GTX 780 (ChipLoco rumor)

[DarkKnightDude]

The chiphell poster appears to have stolen the pics from the Korean site that leaked the AMD ES card sample which now have benchmarks.

http://translate.google.com/transla...UTF-8&u=http://udteam.tistory.com/539&act=url

Now it all depends on pricing. If those benchmarks are true, it seems to be beating the Titan in some places by a nice gap, or trading places in others.

 

[twjr]

Interesting. I'm getting numbers around 450 mm^2 using this photo.

18.5 x 24.5 mm works out to 453.25 mm^2 which is what I measured. But I think there is at least +/- 0.5 mm of error based on the tools I had handy and the quality of that photo.

 

[Keysplayr]

450 is nothing to sneeze at that is for sure. Impressive if AMD managed that much.
We shall soon see. Maybe.
Unless there are two cores packed in there.

 

[HurleyBird]

450 is nothing to sneeze at that is for sure. Impressive if AMD managed that much.
We shall soon see. Maybe.
Unless there are two cores packed in there.

Just making a big die isn't really that impressive. What's impressive is (if the benchmarks on the previous page are correct, which they most likely are) that they are outperforming Titan with significantly less power and with a smaller die on an ES card.

 

[Saylick]

Just making a big die isn't really that impressive. What's impressive is (if the benchmarks on the previous page are correct, which they most likely are) that they are outperforming Titan with significantly less power and with a smaller die on an ES card.

Agreed. The performance is outstanding, if we are to consider the numbers as true, but the perf/W and perf/area is really the icing on the cake. 512-bit bus, too? Even better.

 

[Keysplayr]

Don't really see how that is going to happen, but hey, I'm with the you never can tell crowd. I'm open to anything. I do hope AMD offers up something good.

 

[AnandThenMan]

AMD has had two years to work on this GPU, so not really surprised at the reported performance. What IS surprising to me if the 512 bit memory bus, if true really taken aback by this. Bold move, wonder what other surprises are in store.

 

[Saylick]

Bold move, wonder what other surprises are in store.

New Never Settle bundle with BF4? :hmm:

 

[blackened23]

Information from chiphell is usually reliable, but with that said I'll wait and see what the final benchmarks are. While the (rumored) performance is good, the real eye opener for me was the fact that it was using less power at load than the Titan despite the 512 bit bus. Pretty good efficiency if that is the case, something I wasn't expecting.

Obviously, the information could be fabricated. I don't know. But if the benchmarks are true, kudos to AMD on having respectable efficiency given the performance. And the performance is obviously good as well (IF TRUE).

 

[HurleyBird]

Don't really see how that is going to happen, but hey, I'm with the you never can tell crowd. I'm open to anything. I do hope AMD offers up something good.

Did you miss the early benchmarks on the last page? They have pictures of the card, and the numbers look completely reasonable. Sure it could be an elaborate hoax, but chances are pretty low. Assuming the leak is legit though, there's a good chance benchmarks with launch hardware and drivers will paint a slightly better picture too.

 

[el etro]

maybe 512 can help power consumption once memory is clocked low. But i still believe in 290x superior efficiency.

 

[raghu78]

As I suspected the 512 bit memory is clocked very modestly at 1.25 Ghz to keep power consumption within 250W. This chip is going to shine at high resolutions and high AA like 1920 x 1080 8x MSAA, 2560 x 1600 4x MSAA and 2560 x 1600 2x SSAA. I think Titan is still the faster GPU on a clock for clock basis. though that will be known only when this chip falls in the hands of enthusiasts who clock this ah heck to 1.3 Ghz under water. this chip should scale like a beast due to massive bandwidth and reworked front end and back end.

Battlefield 3 results look impressive. AMD really looks all set to have a killer offering for BF4 with their expected game bundle.

http://www.chiphell.com/thread-862351-1-1.html
http://udteam.tistory.com/539

 

[HurleyBird]

I bet that card will do 1500MHz memory overclock easily.

 

[badb0y]

I want to believe.

 

[AnandThenMan]

If AMD was smart (they have not learned this lesson for how many generations now) they will have a higher clocked, cherry picked model to give to reviewers. You can bet that review sites will mysteriously have an Nvidia ringer card show up at their door at the last minute to spoil the AMD party.

 

[el etro]

A Titan Ultra can still win the 290X, but this power efficiency will allow the Dual Hawaii fight gtx 790(and surely win)!

And this time the 7xxx - R2 2xx will be more future proof than their Nvidia counterparts!

 

[blackened23]

If AMD was smart (they have not learned this lesson for how many generations now) they will have a higher clocked, cherry picked model to give to reviewers. You can bet that review sites will mysteriously have an Nvidia ringer card show up at their door at the last minute to spoil the AMD party.

Most review sites with integrity would do reference to reference comparisons - that is the most fair way to review. Throwing factory overclocked models in there, when they won't be available initially for the R9X (i'm assuming) wouldn't be a fair comparison. I'm sure R9X will have factory OC'ed models in time, but it would just be ridiculous for a website to compare a reference R9X to a factory OC'ed 780. It needs to be either aftermarket vs aftermarket or reference vs reference, period. That is the most fair basis of comparison, really.

 

[HurleyBird]

A Titan Ultra can still win the 290X

Assuming there is no performance uplift at launch, which is a pretty poor assumption.

Looking over the leak again, it's possible that boost wasn't even enabled in the benchmarks. Every card that has boost has a "+" listed after the core clock, while the 290X does not. On the other hand, it's possible they simply weren't able to detect when the card was in a turbo state.

 

[Saylick]

So I grabbed this image from csbin's thread and used it to approximate the die area again since this image was of better quality:

To make sure the image isn't a composite, I checked the size of the RAM chips in both cards and they more-or-less matched up.

To the best of my judgement, I measured out the die area of Tahiti and compared it to the 290X die and noticed that the width of Tahiti's die was slightly smaller than the 290X's die. I figured this was within the margin of error.

I looked at two cases:
A) The 290X die is the same width as Tahiti's (in RED) => 290X = ~454 mm^2
B) The 290X die is slightly wider than Tahiti's (in CYAN) => 290X = ~476 mm^2

I was unable to produce a case where the 290X die was smaller than ~450 mm^2, but again, this could just be me.

 

[skipsneeky2]

A Titan Ultra can still win the 290X, but this power efficiency will allow the Dual Hawaii fight gtx 790(and surely win)!

And this time the 7xxx - R2 2xx will be more future proof than their Nvidia counterparts!

I love nvidia but lets face it,the titan ultra could very well be a $1,000 card while the 290x will most likely be faster by a mile and either the same price or cheaper....without some major refresh from nvidia who is gonna give a rats ass about the ultra at this point?

Nvidia has milked the living hell out of their kepler cards,think most of the folks here are about tired of it.

 

[Keysplayr]

Did you miss the early benchmarks on the last page? They have pictures of the card, and the numbers look completely reasonable. Sure it could be an elaborate hoax, but chances are pretty low. Assuming the leak is legit though, there's a good chance benchmarks with launch hardware and drivers will paint a slightly better picture too.

Heh, so, "if" that is indeed a picture of the upcoming AMD GPU, how exactly to you derive the performance as reasonable from a picture? And like people have been saying, this could be fabricated. It would be great if true, but I'm cautioning you not to get your hopes up. Let it be a pleasant surprise for you if it is exactly what it "appears" to be. I have zero problems with it being faster than Titan so I don't really care. Unlike others here who REALLY want it to be faster. It would be kewl though.

 

[AnandThenMan]

It's not that simple. With boost/turbo now the norm, it's easy to have cherry picked GPU's perform better even with the same board design/card model. Some of the review sites got really aggressive boost speeds on the Kepler cards, not hard to be suspicious as to why.

All I'm saying is AMD has been really poor at playing the PR game, we'll see if they get burned by it again this round.

Most review sites with integrity would do reference to reference comparisons - that is the most fair way to review. Throwing factory overclocked models in there, when they won't be available initially for the R9X (i'm assuming) wouldn't be a fair comparison. I'm sure R9X will have factory OC'ed models in time, but it would just be ridiculous for a website to compare a reference R9X to a factory OC'ed 780. It needs to be either aftermarket vs aftermarket or reference vs reference, period. That is the most fair basis of comparison, really.

 

[Keysplayr]

I love nvidia but lets face it,the titan ultra could very well be a $1,000 card while the 290x will most likely be faster by a mile and either the same price or cheaper....without some major refresh from nvidia who is gonna give a rats ass about the ultra at this point?

Nvidia has milked the living hell out of their kepler cards,think most of the folks here are about tired of it.

I agree. 1000.00 vid cards are a niche. I'd personally never pay that much for a GPU. I would pay 550 tops for a top of the line GPU in any given generation.

 

[Tempered81]

R9 290X at 1020/1250 Vs. GTX Titan at 993/1500

http://udteam.tistory.com/539

 

[Carfax83]

Daaaaang!

Looks like NVidia is going to be doing a price cut soon :awe: