Where are people getting the 10.12's anyway?
He's saying the 10.12 INF identifies that card as a 6950.
That document says it's a 6970. Somebody is lying; either that document is fake, or ATI put a trick in the new Cats.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
ATI HD 6970 actual benchmarks
- Thread starter 96redformula
- Start date
That document says it's a 6970. Somebody is lying; either that document is fake, or ATI put a trick in the new Cats.
Well, if the Barts xt card really is 6738 and the pro part is 6739, then it makes sense that the Cayman xt part is 6718 and the Pro part is 6719. Thus I'm more inclined to think that the 10.12 INF, wherever it came from, is either wrong or fake rather than that document.
There's more GPUZ shots in the Guru3D thread confirming the same thing. http://forums.guru3d.com/showpost.php?p=3807087&postcount=240
Well, if the Barts xt card really is 6738 and the pro part is 6739, then it makes sense that the Cayman xt part is 6718 and the Pro part is 6719. Thus I'm more inclined to think that the 10.12 INF, wherever it came from, is either wrong or fake rather than that document.
Oh well, looks like I'm buying a GTX 570 for Christmas. Unless 10.12 is a magic driver.
Why not wait a few days when the real reviews come out before you tell your parents to buy you a GTX 570. You need to relax, smoke a pin, eat some din and chill you wont be dissapointed..
D
Deleted member 174368
Guest
Why? We don't have any solid performance numbers (besides 20% faster than 480 in F1 which puts it into 580 territory), a price point or anything else besides the number of shaders, yet you're already set on a 570?
ಠ_ಠ
Well, all the F1 benches I've seen show ATI winning by a large margin. It's just a favored game for them.
Price doesn't matter; Nvidia will slice their's.
The only thing that's left to see is power/heat/noise, assuming the GTX 570 vs 6970 stuff is true. And it looks to be true, given that it's coming from multiple sources at this point.
The only wild card left is the 10.12 driver, but even that's not plausible -- ATI wouldn't ship 6900's with gimped driver discs.
I won't buy a 6970 unless it's extremely aggressive in all other areas (other than performance, obviously). And by "extremely aggressive", I mean biblical proportions.
Even to reviewers to protect NDA?
Only distributors need the real drivers to produce them, but they've been crazy secretive.
The recent benches are from a store-bought 6970 in Germany, I think.Even to reviewers to protect NDA?
Distributors have the real drivers and have been crazy secretive.
That would be an actual vendor, not a reviewer.
If those are real then you're right. With only 3 days to go, I'll admit to covering my ears and closing my eyes
I believe these are it:
http://forums.guru3d.com/showthread.php?t=334165
However I just had a look myself and the INF doesn't specify the actual series number, which means there's no problem with the drivers, as far as card identification goes. Seems someone is just spreading FUD.
Well, one thing's for sure. The rumors went from overly optimistic to very pessimistic in just a few weeks. Let's just hope the card will land somewhere in between those rumors and at least be able to compete with the 580 in some games, even if losing out on average.
Yeah, maybe someone just assumed that since the lower card has 6719, which is higher than 6718, it would have to be the 6970.
96redformula
2[H]4U
- Joined
- Oct 29, 2005
- Messages
- 2,578
Well, that certainly doesn't bode well for the 6970.
Curious though, the F12010 benchmark with the 480 at stock and OC'ed. He actually managed to raise the fps from 56 to 78 or by almost 40%? Or was it done on another computer with faster CPU?
Based on the numbers that were just posted, the 6950 would probably be replacing the 5870 entirely (or slower than the 5870). The weird thing is, where exactly is that? The 5870 is being replaced by the 6870 already, or at least very close. Unless the 6970 is a $349 part, and the 6950 is something like $299.If that is true Amd just invented the R600 again... not much faster than the one year 5870, and that dosnt look well for the 6950.
D
Deleted member 174368
Guest
Yeah if those benches are true then the 6950 would actually be more or less in the 5870 performance range, which doesn't even make any sense considering the 6870 is nearly there already.
Oh well, we'll see in a few days.
e;fb
Oh well, we'll see in a few days.
e;fb
Yeah, makes you wonder. Could it be that the new shader arrangement actually messed everything up for ATI? When the 6870 came out so well, and the 6970 was said to feature even more improvements, I was really expecting a beast of a card. But this makes it seem like they'd be far better off just making a larger 6870.
96redformula
2[H]4U
- Joined
- Oct 29, 2005
- Messages
- 2,578
This probably mean that even with actual reviews, we will see it compete with GTX570.
Time for the Trombone : Wa Wa Waaaaaa
Last edited:
hmm, these benches are interesting.......sooo Germany has these cards 3 days before the NDA's expire? hell i need to move to germany i guess ........until a place like HardOCP does a review i think ill take most of these with a grain of salt, since we dont even know what the system they tested on(for all we know the rest of the hardware was favoring the nVidia card....)
note: did not read though the other threads, just this one, so the system specs for the testing may have been mentioned...
........until a place like HardOCP does a review i think ill take most of these with a grain of salt, since we dont even know what the system they tested on(for all we know the rest of the hardware was favoring the nVidia card....)note: did not read though the other threads, just this one, so the system specs for the testing may have been mentioned...
Most of the sales come from the mid-range anyway.
To be honest, having the fastest GPU hasn't been ATI's goal for years. To some degree, they probably have no interest in beating the 580.
Given the amount of time, and where the 5870 is sitting, it seems to me that it would be logical for the 6970 to at least tie the GTX 580. That's not expecting too much, I think.
Thats just a silly assumption. They have no interest in beating the 580? They are in a heads up competition with another company, do they really want to let the "fastest single card" title slip past them?
You make it sound like AMD has a loser's attitude, "oh its okay if we arent the fastest..."
They want the fastest cards period and they want the best card for the price point as well, its all about market share, they aren't going to let nVidia seize the high end market without putting out their own serious offering.
Yeah, they do it every time, typically.
They always just strive for the best price/performance. Look what happened with the 4870 and GTX 280. 4870 was significantly slower, but caused a HUGE price cut on the 280's.
No, they want to make money.
If that means having a slower card with better perf/$, then they'll do it.
Well, you also have the eVGA Hydro Copper 2. Palit/Gainward have a GTX 570 with an added Display Port connector and custom cooler so there seems to be a good chance of a GTX 580 version of that, but nothing yet.
octoberasian
2[H]4U
- Joined
- Oct 13, 2007
- Messages
- 4,082
I'm having a hard time thinking these benchmarks are true because if you think about it:
Is it worth getting the 6970 if it's just a slight improvement over the 5870 but is barely at the same level or above a GTX 480?
If it isn't priced right for it to sell, it wouldn't be worth to buy it if the GTX 570 performs just about the same as the 6970. And, if going by these benchmarks the 6970 is no where near a GTX 580.
That then comes down to the 6950. If the 6870 performed between a 5850 and a 5870 of the previous generation, and came very close to a 5870 in performance, where does the 6950 slot in? At the same level as the 5870? But, if true, then that wouldn't make sense to get either if 6970 is just a few percentage points higher than that.
In other words, the performance gap between a 5870-6970 and a 6870-5870/6950-6970 isn't going to make this worth buying if one is expecting a noticeable performance difference between the products. You may as well stick to the Radeon 5870 and overclock it, or move to the GTX 570 or the GTX 580 even.
That is if these benchmarks are true.
Is it worth getting the 6970 if it's just a slight improvement over the 5870 but is barely at the same level or above a GTX 480?
If it isn't priced right for it to sell, it wouldn't be worth to buy it if the GTX 570 performs just about the same as the 6970. And, if going by these benchmarks the 6970 is no where near a GTX 580.
That then comes down to the 6950. If the 6870 performed between a 5850 and a 5870 of the previous generation, and came very close to a 5870 in performance, where does the 6950 slot in? At the same level as the 5870? But, if true, then that wouldn't make sense to get either if 6970 is just a few percentage points higher than that.
In other words, the performance gap between a 5870-6970 and a 6870-5870/6950-6970 isn't going to make this worth buying if one is expecting a noticeable performance difference between the products. You may as well stick to the Radeon 5870 and overclock it, or move to the GTX 570 or the GTX 580 even.
That is if these benchmarks are true.
Last edited:
http://techpowerup.com/reviews/Powercolor/HD_6870_PCS_Plus/27.html
I'm thinking those slides are fake though.
Also, keep in mind the settings on which these cards were benched.
Something looks wrong.
The dotted line in the 6970 picture is the GTX 580 (15% faster than 480). 6970 = 15% Average, 580 = 15% Average, so the cards trade blows.
The dotted line in the 6950 picture is the GTX 480 (20% faster than 470). 6950 = 20% Average, 480 = 20% Average, so the cards trade blows.
6970 = 580
6950 = 480 = 570
In other words, the flat 480 normalized line in the 6970 picture is the equivalent of the dotted line in the 6950 picture. The cards performance is quite... close.
In fact, look at Metro 2033. It looks like the 6950 is tieing the 6970 on that game.
The dotted line in the 6970 picture is the GTX 580 (15% faster than 480). 6970 = 15% Average, 580 = 15% Average, so the cards trade blows.
The dotted line in the 6950 picture is the GTX 480 (20% faster than 470). 6950 = 20% Average, 480 = 20% Average, so the cards trade blows.
6970 = 580
6950 = 480 = 570
In other words, the flat 480 normalized line in the 6970 picture is the equivalent of the dotted line in the 6950 picture. The cards performance is quite... close.
In fact, look at Metro 2033. It looks like the 6950 is tieing the 6970 on that game.
Last edited:
I agree, fake or extremely exaggerated.
They're marketing slides from ATI, after all. Assuming they're real.
Well, if the Barts xt card really is 6738 and the pro part is 6739, then it makes sense that the Cayman xt part is 6718 and the Pro part is 6719. Thus I'm more inclined to think that the 10.12 INF, wherever it came from, is either wrong or fake rather than that document.
Let's see if Ducari is the trick
Michaelius
Supreme [H]ardness
- Joined
- Sep 8, 2003
- Messages
- 4,684
I think you have short memory.
ATI was very happy to compete for performance crown in radeon 9700/ 800XT/ 1800XTX/1900xtx times.
They started PR bullshit about how they don't care when they couldn't compete anymore due to lack of ability to produce high performance core.