Radeon 7970 Specifications & Release Date slide leak?

Status
Not open for further replies.

Firestorm007

Senior member
Dec 9, 2010
396
1
0
It kind of resembles the stock batman style coolers on the 5 series.... I like it...Now just take my $$$ AMD! :p
 

tviceman

Diamond Member
Mar 25, 2008
6,734
514
126
www.facebook.com
32 ROPs?!?!

That is only half of what was touted in the rumored thread. Weird. The <3 watts idle is very interesting. I wonder if it is somehow shutting down the GPU almost entirely when it's not in use - a la Nvidia's optimus. But in the same breath, I don't like the <300 watt power usage sign. I was hoping AMD would tiptoe back toward the power usage scale of the 5800 series cards. But it only confirms what I've speculated many times - as GPU cores become more and more complex, they're going to use more power.
 
Last edited:

toyota

Lifer
Apr 15, 2001
12,957
1
0
http://4.bp.blogspot.com/-lmM_C5I4lQk/Tuj2x5x2ryI/AAAAAAAABh4/h0P2cr-FyU0/s1600/7970.png

Shaders: 2048
Compute Power: 3.5TFLOPs (SP)
Clock speed: 925MHz
ROPs: 32
Frame Buffer: 3GB GDDR3
Memory Bus Width: 384 bit
Memory Speed: 5.5Gbps

Absolute beastly. I think I'll take two, please (once my AMD stock recovers from the shitter, that is...)
how does just 33&#37; more shaders for a lower process node next gen gpu sound beastly? AMD usually doubles the amount of shaders and even did that one time before while staying on the same process.

and as others have said 32 ROPs makes no sense with a 384 bit bus.
 
Last edited:

Anarchist420

Diamond Member
Feb 13, 2010
8,645
0
76
www.facebook.com
This doesn't mean jack dempsey s*** if their filtering isn't fixed and if they still don't do their z calculations the same as nvidia does.

However, if they do fix both, I'll get a new power supply and add one of these to my PC (although I'm not going to get a reference design) and alternate between drivers. nvidia's drivers have been lacking for a long ass time and AMD's are better in some of those regards.

FYI, the 32 ROPs will do a lot more if they do 8 zixels per clock like nvidia does.

Unfortunately, their shader cache apparently still isn't going to be clocked very high compared to the competition.
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,605
6
81
So many errors (3W idle is unrealistic, wrong Teraflop number, the missing R)...not buying it.

Now he wrote it is "outdated".
But I still want one :D
 
Last edited:

nitromullet

Diamond Member
Jan 7, 2004
9,031
36
91
I don't care if the slide is legit or not, I still kinda want one.

Although, for $500 the 7970 needs to lay some smack down on the GTX 580, and at $400 the 7950 needs to at least meet/beat it. Considering that's what NV has been offering for over a year at that price point, I'm hoping for some big numbers from AMD. I'd really like to run 2560x1440 with a single card, and my GTX 580 just barely squeaked by for most of 2011. Once BF3 came out it was definitely time for an upgrade.

Whatever happens it's nice to actually have something to talk about on AT video again. :)
 

Lonyo

Lifer
Aug 10, 2002
21,938
6
81
One would think that Embago=Embargo.

It says "embago" on the bottom of the pic.

One would think that someone who really worked for AMD would spell check their slides, especially when the text appears on every slide.
 

toyota

Lifer
Apr 15, 2001
12,957
1
0
One would think that someone who really worked for AMD would spell check their slides, especially when the text appears on every slide.
they have made mistakes on their slides before. this one seems pretty fake though with the misspelling and listing of 32 rops for a 384 bit bus.
 
Feb 19, 2009
10,457
10
76
VLIW4/5 typically got 75&#37; of its SP utilized since their design is difficult to code for. The move to flexible independent shaders means it should be identical to NV, every shader will be used. The change to a hardware scheduler rather than via drivers also ensures this happens.

2048 1D shaders should be compared to Cayman as if it had ~1150 (factoring inefficiencies). Thus, the perf advantage should be huge.

It can't have 32 ROPs, that is way too low for the perf of the rest of the card, they will be bottlenecked hard. Also, if it still has 2 prim/clk, that would suck hardcore. I don't believe these specs, it makes no sense.
 
Last edited:

ocre

Golden Member
Dec 26, 2008
1,594
7
81
VLIW4/5 typically got 75% of its SP utilized since their design is difficult to code for. The move to flexible independent shaders means it should be identical to NV, every shader will be used. The change to a hardware scheduler rather than via drivers also ensures this happens.

2048 1D shaders should be compared to Cayman as if it had ~1150 (factoring inefficiencies). Thus, the perf advantage should be huge.

It can't have 32 ROPs, that is way too low for the perf of the rest of the card, they will be bottlenecked hard. Also, if it still has 2 prim/clk, that would suck hardcore. I don't believe these specs, it makes no sense.

LOL!!!!! identical to nvidias? that would mean its like 2048 nvidia shaders so it would be like 4x more powerful than the 580! Yea, right!!!

But for real you make all kinda predictions on specifications that you know dont add up. You know its wrong. The slide dont add up. Its a fake for hits. Clearly AMD knows how to spell. And clearly they dont label the power draw as >300w to 3w. A mean wth does that mean? 299, 240, 210w? BTW surely they know how to spell. There is very very little information on the slide and most of it doesnt add up.
 
Feb 19, 2009
10,457
10
76
When i say identical to NV, i mean as in utilization efficiency, which VLIW4/5 is clearly inefficient, ~3.4 per 5 sp cluster gets used. GCN SP from the fusion summit presentation points towards complete utilization as each SP can process every instruction and designed for thread rather than code parallelism (very similar to NVs architecture), unlike their previous design.
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,605
6
81
Why is the ROP count important? You don't even know what a new ROP is capable of. And they could be decoupled from the memory bus, this is possible.
 

tviceman

Diamond Member
Mar 25, 2008
6,734
514
126
www.facebook.com
When i say identical to NV, i mean as in utilization efficiency, which VLIW4/5 is clearly inefficient, ~3.4 per 5 sp cluster gets used. GCN SP from the fusion summit presentation points towards complete utilization as each SP can process every instruction and designed for thread rather than code parallelism (very similar to NVs architecture), unlike their previous design.

It's not anything like Nvidia's cores though. Nvidia's gtx580 512 cores outperform AMD's 1536 Cayman cores. Simply comparing them 1 to 1 is like comparing a watermelon to a strawberry.
 
Status
Not open for further replies.