Sapphire Readies Radeon HD 7970 Cards with 1335 MHz Core Clock, 6 GB Memory

[ACEkombatkiwi]

1700/8000 is good enough?

http://news.mydrivers.com/1/213/213602.htm

these scores are completely sick

 

[TheoneandonlyMrK]

wow look genuine too though could be otherwise , has anyone seen anything similar with waterblocks on not Ln2

 

[ensabrenoir]

I .....may..... not be going nvdia on my new build. Gonna go back to school to learn how to design something that will take advantage of two of these in xfire...more likely. When it comes to video cards amd rocks. Its all that ati dna.

 

[Psychoholic]

Impressive!

I think they should bump it up 2 more mhz, for 1337 mhz

 

[alexsubri]

6GB is a little ahead of its time, however higher resolutions require higher vram

 

[THANATOS]

crazyeyesreaper

with 1335mhz core that is a 31% overclock taking W1zzards overclock results and extrapolating the data that would be a 29% performance boost

essentially 15% oc w1zz saw a 14% performance bump double those numbers = 31% 925 / 1335 = .69 x 100 = 31% so if 15% oc is 14% performance 30% would be 28 and 31 = 29%

this should result in around a 5-6% performance boost OVER the HD 6990 / GTX 590

It also results in the HD 7970 being a comfortable 40% faster then a GTX 580

1335/925=44%
what you got(31%) means how much slower would be default clocked 7970 compared to this superclocked one. 100/144=69% -> 100%-69%=31%
So Here is a new calculation.
44% lets say it will give 40% boost
so Relative performance (1920*1200)
GTX580 84%
7970 100%
HD6990 117%
GTX590 122%
OC 7970 140%
This card would be 140/122=15% faster than the second fastest card.
Comparing to GTX580: this card would be 67% faster.

 

[the54thvoid]

crazyeyesreaper

It seems someone doesn't know math.
1335/925=44%
what you got(31%) means how much slower would be default clocked 7970 compared to this superclocked one. 100/144=69% -> 100%-69%=31%
So Here is a new calculation.
44% lets say it will give 40% boost
so Relative performance (1920*1200)
7970 100%
HD6990 117%
GTX590 122%
OC 7970 140%
This card would be 140/122=15% faster than the second fastest card.

lol, was doing a post to correct you then realised you are right. The 410MHz diff between clocks should be divided by the stock clock of 925 to gauge the speed faster than the stock (which is 44%).

 

[OneCool]

6GB is a little ahead of its time, however higher resolutions require higher vram

How about the 6gb version and one of these
http://www.wired.com/gadgetlab/2011/12/lg-84-inch-4k-display/

 

[Trigger911]

Isnt it rule of thumb to have sys ram to cache the video ram ... guess people will need to jump up to an 8gig setup if they buy one of these beast

 

[Delta6326]

I Want you!

 

[KooKKiK]

From the slide, so Tahiti is partially disabled after all ???

 

[EpicShweetness]

*reads post*
*opens side panel*
*Yells obscenities at old 5870*
*closes side panel*
*cries in a corner*

 

[Millennium]

Trigger: For DX9 you need to cache the entire GPU memory, so yes! Madness, 6gb GPUs 4tw.

In general, I really like this new architecture and performance from AMD. And the amazing overclock for a retail card shows they really nailed this new process this time.

Now (JF-AMD if you are reading) please please make the 7950 unlockable, even partially unlockable, and AMD will have a huge hit on their hands. Please please make this happen! OMG!111

 

[bear jesus]

How about the 6gb version and one of these
http://www.wired.com/gadgetlab/2011/12/lg-84-inch-4k-display/

Just one? more like 3 in eyefinity, would never get playable frame rates with a single 7970 but it may eat in to that 6 GB.

 

[crazyeyesreaper]

crazyeyesreaper

It seems someone doesn't know math.
1335/925=44%
what you got(31%) means how much slower would be default clocked 7970 compared to this superclocked one. 100/144=69% -> 100%-69%=31%
So Here is a new calculation.
44% lets say it will give 40% boost
so Relative performance (1920*1200)
GTX580 84%
7970 100%
HD6990 117%
GTX590 122%
OC 7970 140%
This card would be 140/122=15% faster than the second fastest card.
Comparing to GTX580: this card would be 67% faster.

lol, was doing a post to correct you then realised you are right. The 410MHz diff between clocks should be divided by the stock clock of 925 to gauge the speed faster than the stock (which is 44%).

yea hadnt slept yet so you can save the insult about math, i did it in google search lolz after not sleeping for 2 days. still 30% is a more realistic performance bump as i would expect at 1335 the ROPs will become a limiting factor much like 5850 vs 5870 with intense overclocks

 

[the54thvoid]

yea hadnt slept yet so you can save the insult about math, i did it in google search lolz after not sleeping for 2 days. still 30% is a more realistic performance bump as i would expect at 1335 the ROPs will become a limiting factor much like 5850 vs 5870 with intense overclocks

Yeah , was jumping in to defend your math and then realised myself. Quickly deleted my post and then reposted. But no need for the math insult, i agree.

 

[badtaylorx]

does anyone else here get a chubby when thinking about the possibilities of an asus ARES II???

 

[raptori]

Ok time to sell my 580.

 

[Valdez]

In general, I really like this new architecture and performance from AMD. And the amazing overclock for a retail card shows they really nailed this new process this time.

And this is only the first generation of the 28nm process. The 55 and also the 40nm process saw a rebirth.

I also agree, this was a strong start, maybe we are going to see a full 2304 sp tahiti with higher clock speed than 925 mhz later.

 

[v12dock]

crazyeyesreaper

It seems someone doesn't know math.
1335/925=44%
what you got(31%) means how much slower would be default clocked 7970 compared to this superclocked one. 100/144=69% -> 100%-69%=31%
So Here is a new calculation.
44% lets say it will give 40% boost
so Relative performance (1920*1200)
GTX580 84%
7970 100%
HD6990 117%
GTX590 122%
OC 7970 140%
This card would be 140/122=15% faster than the second fastest card.
Comparing to GTX580: this card would be 67% faster.

PLUS driver optimizations >70%

 

[dj-electric]

My reaction when i woke up and saw the article title:

This picture sais everything, i dont need to add a single word

 

[v12dock]

 

[alwayssts]

I love the guys explanation, as its totally plausible.

Essentially, Release 7870 as the fastest single card, even if conservative and only by 10% at stock. Exploit for high dollars (and overclock versions) while building yields. As the price drops down to $300 over time, release the stock full tahiti at a price right above the overclocked versions of 7870. Call it the 8000 series. Make more money than releasing the full version right away.

They must figure thats essentially what nvidia did last gen, and it worked out well. Without figuring in the fact gf100 and 104 were essentially broken, gf110 and 114 were just fully-enabled designs released at a later time to maximize yields and revenue.

When nvidia released 460, it had no direct native competition. When Barts came along and 460 prices were in the toilet, so did the 560ti. 480...cayman...580.

I wonder if 8970 will use that shiny new .28ns gddr5 samsung is sampling to...somebody

 

[Hayder_Master]

So 6Gb, for eyefinity setup.

 

[mediasorcerer]

I was reading that oc,d they can get up to almost 6990 perf in some games.

Pretty nice unit.