[donanimhaber] hd 7950 3dm11 results top 580, -300 points difference vs 7970!!!!!

IonusX

Senior member
Dec 25, 2011
392
0
0
today donanimhaber has leaked their result from their 1st hd 7950 test. its most promising.

in 3dmark 11 on extreme with a core i7 Nehealm extreme the 7950 managed to get X2385 while the 7970 was at x2635 and the 580 scored x2081 respectively
amdradeonhd7950gpuz2_dh_fx57.jpg

thats an impressive result to say the least. amd getting all fired up
source:
http://www.donanimhaber.com/ekran-karti/haberleri/DH-Ozel-AMD-Radeon-HD-7950-icin-ilk-test-sonucu-ve-GPUZ-goruntusu.htm
 
Last edited:

MisterMac

Senior member
Sep 16, 2011
777
0
0
Is that impressive?

Shouldn't the hd7970/7950 comletely SLAY gtx580 instead of best it by a..certain margin ?

Is the 6 series kepler also gonna be like this?
Marginal improvements?
 

Quantos

Senior member
Dec 23, 2011
386
0
76
Is that impressive?

Shouldn't the hd7970/7950 comletely SLAY gtx580 instead of best it by a..certain margin ?

Is the 6 series kepler also gonna be like this?
Marginal improvements?

1. Define impressive

2. Not necessarily. It's a new architecture, and it's a new fab process. Experiences past are not necessarily relevant.

3. We really cannot know how Kepler is going to be at all.

At 400 to ~$460, the 7950 currently bests the 580 with a lower price tag. Of course, the 580 is probably going to drop a bit. So far, so good I'd say.
 

ShadowOfMyself

Diamond Member
Jun 22, 2006
4,227
2
0
Huh... so now 30% is marginal?

It seems a lot of people on this forum expect Kepler to be the new 8800GTX... Good luck with that
 

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
It is illogical to ignore the difference in die size. I'm not positive, but from a cursory search it seems the 7970 would be roughly ~60% faster than the 6970 at this benchmark. TBH, if NVIDIA also gets this kind of gain for the die size then 28nm will be doing well even if we did have to put up with delays.

dkQkX.png


Is that impressive?

Shouldn't the hd7970/7950 comletely SLAY gtx580 instead of best it by a..certain margin ?

Is the 6 series kepler also gonna be like this?
Marginal improvements?
 
Last edited:

gramboh

Platinum Member
May 3, 2003
2,207
0
0
Hmm, cool if true. This could push down prices on the GTX560Ti as well eventually. Would like to pick up another one new for $160-175 for SLI in a few months, that should slay BF3 adequately for me.
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
Looking at the specs, I hold onto my previous statements. This card should be just slightly faster (5% faster) than the GTX 580 stock, and with overclocking that will obviously increase. That also puts it at 20% faster than an HD 6970 stock. The card I'm having some trouble with understanding is the HD 7890, though. That card should only be around 5% faster than the HD 6970, though with overclocking that will obviously increase. Given the landscape, it needs to be priced at under $350. It'd also be nice to see the 1.5GB HD 7950 at under $400; it's a perfect card for a single 1920x1200 display.
 

IonusX

Senior member
Dec 25, 2011
392
0
0
Looking at the specs, I hold onto my previous statements. This card should be just slightly faster (5% faster) than the GTX 580 stock, and with overclocking that will obviously increase. That also puts it at 20% faster than an HD 6970 stock. The card I'm having some trouble with understanding is the HD 7890, though. That card should only be around 5% faster than the HD 6970, though with overclocking that will obviously increase. Given the landscape, it needs to be priced at under $350. It'd also be nice to see the 1.5GB HD 7950 at under $400; it's a perfect card for a single 1920x1200 display.
rumors say pitcairn is basically just a clone of cayman xt/pro/le (hd 6970, 6950 and 6930 for the laymen)
so in short id expect the 7890 = 6970
kinda like how the hd 5830 - 5750 were just clones of exiting 4800 series cards
 

IonusX

Senior member
Dec 25, 2011
392
0
0
It is illogical to ignore the difference in die size. I'm not positive, but from a cursory search it seems the 7970 would be roughly ~60% faster than the 6970 at this benchmark. TBH, if NVIDIA also gets this kind of gain for the die size then 28nm will be doing well even if we did have to put up with delays.

dkQkX.png
a picture by vesku.. thats so factual :rolleyes:
a die shrink is what a card maker chooses to make of it. it can be greater or worse than expectations. me im hedging my bets on what most folks have said all along
gtx 600 = fermi 28nm
gtx 700 = kepler 28nm

and thats not anything new either folks at BSN, and many other places were saying that a while ago.. (picture doesnt work srry). anyway i say dont be suprised if your long awaited gtx 600 is little more than a ruse and as a result we see an fx5000 vs radeon 9000 all over again..
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
rumors say pitcairn is basically just a clone of cayman xt/pro/le (hd 6970, 6950 and 6930 for the laymen)
so in short id expect the 7890 = 6970
kinda like how the hd 5830 - 5750 were just clones of exiting 4800 series cards

Completely different arch. We're talking GCN compared to VLIW4. Not a clone at all, just same/similar specs.
 

MrTeal

Diamond Member
Dec 7, 2003
3,919
2,708
136
rumors say pitcairn is basically just a clone of cayman xt/pro/le (hd 6970, 6950 and 6930 for the laymen)
so in short id expect the 7890 = 6970
kinda like how the hd 5830 - 5750 were just clones of exiting 4800 series cards

Rumours might says Pitcairn is a clone of Cayman, but most reputable sites like Anandtech are claiming Pitcairn will be a GCN part.
 

IonusX

Senior member
Dec 25, 2011
392
0
0
Rumours might says Pitcairn is a clone of Cayman, but most reputable sites like Anandtech are claiming Pitcairn will be a GCN part.
well im afraid anand also happens to include you. it lost all credibility just there..
 

LOL_Wut_Axel

Diamond Member
Mar 26, 2011
4,310
8
81
rumors say pitcairn is basically just a clone of cayman xt/pro/le (hd 6970, 6950 and 6930 for the laymen)
so in short id expect the 7890 = 6970
kinda like how the hd 5830 - 5750 were just clones of exiting 4800 series cards

The HD 7890 is based on a further cut down Tahiti die. Kind of like the very limited quantity HD 6930 and before the HD 6930, the HD 5830. This time they're calling these further cut down parts *890 instead of *930.
 
Last edited:

Vesku

Diamond Member
Aug 25, 2005
3,743
28
86
a picture by vesku.. thats so factual :rolleyes:

Can quibble over the performance average, but that picture shows die size comparison. If you have contra-factual information regarding die size and MSRP I'll be happy to look at it.
 

WMD

Senior member
Apr 13, 2011
476
0
0
today donanimhaber has leaked their result from their 1st hd 7950 test. its most promising.

in 3dmark 11 on extreme with a core i7 Nehealm extreme the 7950 managed to get X2385 while the 7970 was at x2635 and the 580 scored x2081 respectively
amdradeonhd7950gpuz2_dh_fx57.jpg

thats an impressive result to say the least. amd getting all fired up
source:
http://www.donanimhaber.com/ekran-karti/haberleri/DH-Ozel-AMD-Radeon-HD-7950-icin-ilk-test-sonucu-ve-GPUZ-goruntusu.htm

2 things to take note of. Notice the gpuz in the screenshot says 880mhz while the default frequencies of a stock 7950 should be 800mhz. Its also unlikely the 7970 is only a mere 10% faster than the 2nd tier card. Historically it was closer to 20% faster.

3dmark 11 results can favor AMD more than real world gaming. In games the 7970 is only 11% on average faster than gtx580 in DX11 mode and 2% faster in DX9 /DX10 mode.

http://www.tweakpc.de/hardware/test..._black_edition/benchmarks.php?benchmark=pfd11

http://www.tweakpc.de/hardware/test...70_black_edition/benchmarks.php?benchmark=pfd

This puts the 7950 on par with the gtx580 in real world performance at best.
 

thilanliyan

Lifer
Jun 21, 2005
12,072
2,280
126
3dmark 11 results can favor AMD more than real world gaming. In games the 7970 is only 11% on average faster than gtx580 in DX11 mode and 2% faster in DX9 /DX10 mode.

:hmm:

Serious? A lot of reviews say you're wrong.

I can pick out a review that shows a 6970 is just as fast as a GTX 580...would you take that as gossip?
 

MrTeal

Diamond Member
Dec 7, 2003
3,919
2,708
136
well im afraid anand also happens to include you. it lost all credibility just there..

I'm not sure exactly what you meant by that, but since it wasn't clear I'll take it as a compliment. Thanks.

I find Anand's articles incredibly informative, and while the site generally doesn't break stories like some of the other sites, when information is posted it's usually accurate. If Anand claims in his article that AMD is telling him Tahiti, Pitcairn and Cape Verde are all GCN, I'm inclined to believe him over a site that changes it's story every couple weeks.
 

WMD

Senior member
Apr 13, 2011
476
0
0
:hmm:

Serious? A lot of reviews say you're wrong.

I can pick out a review that shows a 6970 is just as fast as a GTX 580...would you take that as gossip?

Did you read the links? Tweakpc.de tests the most comprehensive number of titles. Look at DX11 performance index. I am sure BF3 performance matters alot to most people and that is where AMD failed to shine.
 

thilanliyan

Lifer
Jun 21, 2005
12,072
2,280
126
Did you read the links? Tweakpc.de tests the most comprehensive number of titles. Look at DX11 performance index. I am sure BF3 performance matters alot to most people and that is where AMD failed to shine.

Like I said, I can find a review where a 6970 is as good as a GTX580...would you take only those results into consideration?

You can't quote one review and make a generalization...on average (ie. MORE than ONE review) a 7970 is more than 2% and 11% faster in DX9/10 and DX11 titles, respectively.
 

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
I like that review site, expect there is no testing methodology part that I can see so I can't see their test setup.

Aside from that great layout and use of click setting/res graphs with a very nice number of games tested.
 

WMD

Senior member
Apr 13, 2011
476
0
0
Like I said, I can find a review where a 6970 is as good as a GTX580...would you take only those results into consideration?

You can't quote one review and make a generalization...on average (ie. MORE than ONE review) a 7970 is more than 2% and 11% faster in DX9/10 and DX11 titles, respectively.

The problem is Anandtech and Pcperspective also had mediocre results from the 7970 we already discarded them in favor of results from [H]. Isn't that some form of cherry picking as well?