AMD Radeon HD 7800 Series Specs. Table Leaked

[farquaid]

It's 40% more shaders not 29%, and 25% higher closks so there is a difference, but I dunno you could be right.

He said 29% less which is right, but when comparing he should have compared how many more shaders the one with most had compared to the one with lowest. Sinec that was done with clocks, the one with highest frequency compared to lowest

 

[Aquinus]

Looking nice. I'm really curious about the over-clocking headroom here as I've had amazing luck with my 6870 and the 7970 seems to over-clock like a champ as well.

Thanks for the information. Cheers.

 

[farquaid]

7850 should be a really good overclocker.

 

[Inceptor]

7870 vs 6950/6970

HD 6950 @ stock:
Pixel fillrate - 25.6 GPixel/s
Texture fillrate - 70.4 GTexel/s

7870 vs 6950:
Pixel fillrate - +25%
Texture fillrate - +13.6%

HD 6970 @ stock:
Pixel fillrate - 28.2 GPixel/s
Texture fillrate - 84.5 GTexel/s

7870 vs 6970:
Pixel fillrate - +13.5%
Texture fillrate - -5.6%


So, somewhere in-between the 6950 and 6970. Closer to the 6970.
I don't know, people. It's an improvement, but not a spectacular one, just a "meets expectations" kind of improvement. I guess it depends on how much overclock headroom there will be.

 

[zomg]

7850 should be a really good overclocker.

most likely hd7850 will be single 6-pin, hd7870 with double 6-pin will have more potential

also, there will be room for hd7850 oc editions with higher specs and double 6-pin for a little premium

 

[Super XP]

patiently waiting for 7800 reviews and such to pop up to see if its a worthwhile upgrade from my 6870

It should easily take out your HD 6870. The question is will it be faster than the HD 6970

 

[Hayder_Master]

7870 have same wattage usage of 7950, that's bad

 

[Hayder_Master]

Those results make no sense to me. Not because how they fare against the HD6000 cards, but because of how close they are to HD7950. You're telling me that 700+ SPs (and related texture units) make almost no difference? Fake results or broken architecture? IMO fake. We'll see soon I guess.

i like your avatar

 

[EpicShweetness]

Kinda thought it was gonna double the 7770. Well this gonna replace a (my) 5870. 400+ less shaders and more powerful amazing how far we've come.

 

[DarkOCean]

And Wow! Those slides are hilarious. HD7870 41% faster than GTX570. Not even the HD7970 is so much faster. Actual difference in W1zz's HD7970 review is 25-30%

They're comparing the 7870 2gb vram with gtx 570 1,25gb vram at 2560x1600 that's just silly.

 

[Benetanegia]

They're comparing the 7870 2gb vram with gtx 570 1,25gb vram at 2560x1600 that's just silly.

No but that's not an issue. The GTX 570 is a 1.25 GB card and it would be legit to demostrate it as a weakness, where it could be a weakness. Marketing trick, but legit.

The problem is that the results are simply exagerated beyond what's reasonable or have been really really skewed by something else other than resolution and AA. They do not represent reality in any way. If you go look W1zzard's review of HD7970, I don't think there's many games (if any) where the HD7970 as fast as those charts show in 2600x1600 4xAA, for HD7870 vs GTX570. The HD7970!! The HD7870 is not going to be anywhere close to that. So THAT is the issue, the fact that results are a blatant lie. Lying is usual in marketing, but to this proportions... AMD PR staff is high when they do these charts or something.

 

[hardcore_gamer]

They're comparing the 7870 2gb vram with gtx 570 1,25gb vram at 2560x1600 that's just silly.

Did you mean that gtx570 is not suitable for gaming at 2560x1600 ?

 

[Mr Verro]

the 7870 looks amazing ,i might get one if ican​

 

[Crap Daddy]

No but that's not an issue. The GTX 570 is a 1.25 GB card and it would be legit to demostrate it as a weakness, where it could be a weakness. Marketing trick, but legit.

The problem is that the results are simply exagerated beyond what's reasonable or have been really really skewed by something else other than resolution and AA. They do not represent reality in any way. If you go look W1zzard's review of HD7970, I don't think there's many games (if any) where the HD7970 as fast as those charts show in 2600x1600 4xAA, for HD7870 vs GTX570. The HD7970!! The HD7870 is not going to be anywhere close to that. So THAT is the issue, the fact that results are a blatant lie. Lying is usual in marketing, but to this proportions... AMD PR staff is high when they do these charts or something.

I remember the marketing graphs for the 7970, same BS. But as you can see the 7870 is positioned against the 570/6970 and the 7850 against the 560Ti/6950 so that's the performance they are aiming to achieve. The 7870 has a stock clock of 1GHz exactly for that reason.

Now the odd thing is what comes between the 7770 and the 7850 because something is missing there. The 7770 is close to 6850 but where's the replacement of the 6870, the very sweet spot of perf/price?

 

[Benetanegia]

I remember the marketing graphs for the 7970, same BS. But as you can see the 7870 is positioned against the 570/6970 and the 7850 against the 560Ti/6950 so that's the performance they are aiming to achieve. The 7870 has a stock clock of 1GHz exactly for that reason.

But HD7970 was a new falgship card in a new process. The claim could make sense. But once your flagship has been released and the actual performance is known, how can you release such a slide claiming it to be faster than your flagship card? AMD PR still incompetent or high all day long I tell you.

And you are right, probably the only meaningful info in the slide is which card they are comparing them with. HD7870 will be at least 10% slower than HD7950 and that puts it around GTX570 level. Probably slightly beating it thanks to clocking it as high as 1000 Mhz.

 

[DarkOCean]

Did you mean that gtx570 is not suitable for gaming at 2560x1600 ?

Something like that 1.25vram its not enough for gaming at that res, the 2.5vram variant its another story.

 

[Benetanegia]

Something like that 1.25vram its not enough for gaming at that res, the 2.5vram variant its another story.

Nah, it makes little difference in general*. Even at 2560x1600 only a few games see a substantial increase in performance, so the overall result would be what 5%? That's what I've seen in reviews at least. And in the slides they "tested" at 4xAA so less memory bound than 16xAA for example.

IMO the test conditions are OK in those slides. Results on the other hand are just hilarious.

* For average fps at least. Low fps or median low can be affected more (and thus, playability/smoothness), but almost nobody posts that kind of results.

 

[Horrux]

Here, some benchmarks:

http://www.3dcenter.org/

So, the 7870 si the new 6970 probably at the same price and the 7850 is the new 6950 proly at the same price.

It was expected that this new gen of cards, the 7800 series and under would bring the same performance for the same money as last gen.

The only reason to buy these cards is from someone coming from a 6850/6870/gtx560/460

That's exactly how AMD proceeds between generations most of the time anyway.

difference is only suspect in BF3, and if there using OLD data for the 7950 from before the BF3 patch that would explain alot, the patch for 7000 series gave 7900 owners an extra 15% and a driver update added another 5-10% so keep that in mind, if there using old data on hand that would explain it,

other titles look about right

theres also the law of diminshing returns to keep in mind as welel

BF3 is more GPU-limited than most games, so it stands to reason that its results would differ from most other games. At some point, some GPUs might get fast enough that some CPU bottlenecking might appear, making that GPU seem less powerful than it is...

Nah, it makes little difference in general*. Even at 2560x1600 only a few games see a substantial increase in performance, so the overall result would be what 5%? That's what I've seen in reviews at least. And in the slides they "tested" at 4xAA so less memory bound than 16xAA for example.

IMO the test conditions are OK in those slides. Results on the other hand are just hilarious.

* For average fps at least. Low fps or median low can be affected more (and thus, playability/smoothness), but almost nobody posts that kind of results.

Oh how totally wrong you are. I am saddled with two of those shitty GTX 570s and I can tell you about them. I have to lower the AA because of VRAM limitations. Or textures, or all kinds of things in order to maintain spike-free FPS. You look at AVERAGE FPS and you say "it makes little difference in general". That is completely untrue. Sure the AVERAGE FPS looks OK, but in real-life situations when you turn around in a game and the card has to load textures and shit from system RAM, you get a nice fraction of a second with no rendering as it waits for the textures. And I'M playing at 1680x1050, I cannot bear imagining what it would be like at 2560. Of course SLI'ing them makes the problem much worse, the GPUs being strong enough to render at 4xMSAA at high resolution, but the VRAM limiting everything.

Thanks nVidia for making a powerful, but VERY SHITTY videocard with no future whatsoever. At least AMD has been smart enough to pack 2Gb of VRAM in their cards for a while now... STANDARD, no "super special edition $150 more than the normal, order today, we'll have it in stock within 20 days!" type shitty deal the way nVidia makes them.

 

[Crap Daddy]

Oh how totally wrong you are. I am saddled with two of those shitty GTX 570s and I can tell you about them. I have to lower the AA because of VRAM limitations. Or textures, or all kinds of things in order to maintain spike-free FPS. You look at AVERAGE FPS and you say "it makes little difference in general". That is completely untrue. Sure the AVERAGE FPS looks OK, but in real-life situations when you turn around in a game and the card has to load textures and shit from system RAM, you get a nice fraction of a second with no rendering as it waits for the textures. And I'M playing at 1680x1050, I cannot bear imagining what it would be like at 2560. Of course SLI'ing them makes the problem much worse, the GPUs being strong enough to render at 4xMSAA at high resolution, but the VRAM limiting everything.

I can tell you that I'm playing at the same res and had no such problems whatsoever with just one 570 on Ultra preset. I am talking about BF3 which uses as much VRAM as it's available.

 

[Benetanegia]

Oh how totally wrong you are. I am saddled with two of those shitty GTX 570s and I can tell you about them. I have to lower the AA because of VRAM limitations. Or textures, or all kinds of things in order to maintain spike-free FPS. You look at AVERAGE FPS and you say "it makes little difference in general". That is completely untrue. Sure the AVERAGE FPS looks OK, but in real-life situations when you turn around in a game and the card has to load textures and shit from system RAM, you get a nice fraction of a second with no rendering as it waits for the textures. And I'M playing at 1680x1050, I cannot bear imagining what it would be like at 2560. Of course SLI'ing them makes the problem much worse, the GPUs being strong enough to render at 4xMSAA at high resolution, but the VRAM limiting everything.

I can tell you that I'm playing at the same res and had no such problems whatsoever with just one 570 on Ultra preset. I am talking about BF3 which uses as much VRAM as it's available.

Yeah whatever the problem is not memory. I use a GTX460 1GB, so a lot weaker than GTX570 and don't have such problems at such a low resolution and 4xAA. Don't blame the memory if you can't find what the problem is.

And also what you described in the second part is exactly what I said was the exception (low frames and low median frames), so I don't know what the hell are you talking about me being wrong: I was talking about the slides, they show average fps so no amount of memory is going to make them look much faster there, whatever the real experience would be in real life.

 

[Horrux]

I can tell you that I'm playing at the same res and had no such problems whatsoever with just one 570 on Ultra preset. I am talking about BF3 which uses as much VRAM as it's available.

How much VRAM is it using? I'm betting you are having this problem but are too blind to see it. But then again, Ultra on a single 570 is almost unplayable, so it isn't surprising.

Yeah whatever the problem is not memory. I use a GTX460 1GB, so a lot weaker than GTX570 and don't have such problems at such a low resolution and 4xAA. Don't blame the memory if you can't find what the problem is.

And also what you described in the second part is exactly what I said was the exception (low frames and low median frames), so I don't know what the hell are you talking about me being wrong: I was talking about the slides, they show average fps so no amount of memory is going to make them look much faster there, whatever the real experience would be in real life.

OK about the second paragraph but the first is wrong. I'm playing at high and 4xAA and my cards are loaded to nearly 1200Mb of VRAM on the larger maps in BF3. And if I crank anything up, I can tell when my video card loads stuff from memory. Not my fault if you can't see it. You of the "the difference between 30 and 34 fps is MASSIVE" conviction. Sigh.

And for the record, I play at 60 to 100 fps. And yes, I can readily tell when my PC skips a frame or two due to lack of VRAM. It's not hard, there's a hiccup, and Afterburner reports VRAM being full... You figure out the rest.

 

[Crap Daddy]

How much VRAM is it using? I'm betting you are having this problem but are too blind to see it. But then again, Ultra on a single 570 is almost unplayable, so it isn't surprising.

Well mister, I don't know what you're talking about. At 1680x1050 although VRAM is at max I have very smooth gameplay in multiplayer around 45-50 FPS in the most demanding maps 4MSAA ULTRA all, FOV max, vsync on. In metro, bazaar, seine crossing the frames are almost all the time closer to 60. And I'm not blind, I like to see all the eyecandy and don't need 100FPS.

 

[Benetanegia]

OK about the second paragraph but the first is wrong. I'm playing at high and 4xAA and my cards are loaded to nearly 1200Mb of VRAM on the larger maps in BF3. And if I crank anything up, I can tell when my video card loads stuff from memory. Not my fault if you can't see it. You of the "the difference between 30 and 34 fps is MASSIVE" conviction. Sigh.

And for the record, I play at 60 to 100 fps. And yes, I can readily tell when my PC skips a frame or two due to lack of VRAM. It's not hard, there's a hiccup, and Afterburner reports VRAM being full... You figure out the rest.

I don't play BF3. Hiccups occur no matter how much ram you have and in fact SLI results in that kind of hiccups far more often than vram does. TechReport did an article on that. When it comes to hiccups and similar "artifacts" 1 GPU >>> 2 GPU, so maybe there lies your problem.

I can see a difference between 30 and 34 fps, so I play at setting that ensure higher fps and of course I can see hiccups sometimes, and it happens even when Afterburner reports only half of my vram is in use and in older games that should not push the GPU so much, so I can definitely say that Vram is not the (only) problem when it happens.

So like I said if you have too many hiccups and the likes, you should look for a way to fix them, vram is not provoking them.

And if I crank anything up, I can tell when my video card loads stuff from memory.

Besides this is a HDD or CPU problem not vram problem. The GPU is constantly loading stuff to vram seamlessly, constantly or do you really think that a complete map's geometry and textures only take 2 GB?? 1 GB or 2 GB while it makes a little difference it's not as big as you are making it to be, and nearly no one has this problem. Even when a particular game only uses 50% of your ram it's still loading stuff constantly, that's why the memory bandwidth is so freaking high on GPUs, if it uses only half the memory is not because it contains everything it needs, it's because the game was designed for that footprint. Conversely if a game takes as much vram as you have available it does not mean that it needs all of what's there at that time, the process of loading things never stops and in cases where bandwidth is sufficient (most cases if you chose the correct settings to ensure good fps) it does not matter if 1GB or 2GB have been filled. It's of little help but not so much. Think of it like a warehouse between a fab and transportation, you can make it bigger but won't help if packets get out as fast as they get in, and that's a GPU's everyday (everysecond). 1-2 GB is such a small number compared to the 160 GB/s that are moved in just one second... think about it, really.

 

[Horrux]

I don't play BF3. Hiccups occur no matter how much ram you have and in fact SLI results in that kind of hiccups far more often than vram does. TechReport did an article on that. When it comes to hiccups and similar "artifacts" 1 GPU >>> 2 GPU, so maybe there lies your problem.

I can see a difference between 30 and 34 fps, so I play at setting that ensure higher fps and of course I can see hiccups sometimes, and it happens even when Afterburner reports only half of my vram is in use and in older games that should not push the GPU so much, so I can definitely say that Vram is not the (only) problem when it happens.

So like I said if you have too many hiccups and the likes, you should look for a way to fix them, vram is not provoking them.



Besides this is a HDD or CPU problem not vram problem. The GPU is constantly loading stuff to vram seamlessly, constantly or do you really think that a complete map's geometry and textures only take 2 GB?? 1 GB or 2 GB while it makes a little difference it's not as big as you are making it to be, and nearly no one has this problem. Even when a particular game only uses 50% of your ram it's still loading stuff constantly, that's why the memory bandwidth is so freaking high on GPUs, if it uses only half the memory is not because it contains everything it needs, it's because the game was designed for that footprint. Conversely if a game takes as much vram as you have available it does not mean that it needs all of what's there at that time, the process of loading things never stops and in cases where bandwidth is sufficient (most cases if you chose the correct settings to ensure good fps) it does not matter if 1GB or 2GB have been filled. It's of little help but not so much. Think of it like a warehouse between a fab and transportation, you can make it bigger but won't help if packets get out as fast as they get in, and that's a GPU's everyday (everysecond). 1-2 GB is such a small number compared to the 160 GB/s that are moved in just one second... think about it, really.

Sigh, you must be one of those who complain about "lag" when their videocards are overloaded.

I'm not going to argue with you, it's a waste of time and energy. You are filled with your own blindness and think you are right. So be it then.

 

[DarkOCean]

don't buy today's 7970 save for tomorow 8970