AMD HD7*** series info

[RussianSensation]

I might be remembering wrong, but aren't you overstating the 4890 -> 5870 performance increase?

Well 4890 was a refresh to 4870 so I think I should have stated 5870 vs. 4870 and 6970 vs. 4890. In that case 75-100% performance difference is still there, respectively. But ya, 5870 vs. 4890 was slightly less than 75% overall, closer to 60% faster.

HD6770 ~ HD5770 ~ HD4870.
HD5870 ~ HD6950

Looking at modern games:

http://www.computerbase.de/artikel/grafikkarten/2011/test-19-grafikkarten/4/

6950 1GB is 75% faster than HD6770 on average.
6970 2GB is 85% faster than HD6790 on average.

I don't see the point of linking "all resolutions" from TPU as relevant.

How about HD5870 vs. HD4870/4890 in games at 1920x1080 or higher:

1. Starcraft II => 2x faster than 4890
2. Mafia II => 2x faster than 4890
3. Call of Duty: Black Ops => 2x faster than 4890
4. Brink => 2x faster than HD6770 ~ HD5770 ~ 4870
5. Medal of Honor => 78% faster than HD5770 ~ HD4870
6. Avatar => 79% faster than 4870
7. BF:BC2 => 78% faster than 4870
8. AvP => 79% faster than 4870
9. Bioshock 2 => 78% faster than 4870
10. STALKER: COP => 82% faster than 4870
11. Shattered Horizon => 77% faster than 4870
12. Napoleon Total War => 78% faster than 4870
13. Just Cause 2 => 75% vs. 4870
14. GTA IV = 83% faster vs. 4870

There are at least 10 pages of benchmarks here.

On average HD5870 is at least 75% faster than HD4870 and HD6970 (refresh of 5870) is also at least 75% faster on average vs. HD4890. So HD7970 would need to be about 70-75% faster than HD5870 to be in line with what's expected imo.

But I am more upset that AMD isn't really increasing performance much in the mid-range. We had have nearly HD5850 level of performance for 2 years now (HD6870 was barely faster and it looks like HD7870 won't be faster than an overclocked 5850). That's pretty disappointing to me. While HD7950/70 cards will be a good deal faster, the mid-range continues to be lackluster, if these rumors are true.

HD6970 = 880mhz x 32 ROPs = 28.16 GPixels/sec pixel fill-rate
HD7970 = 1000mhz x 64 ROPs = 64.0 GPixels/sec (> 2x) That's a humongous increase in pixel fill-rate.

Still, if HD7970 is only 50% faster than HD6970, then how is AMD going to compete against Kepler with that? GTX580 is another 15% faster than HD6970 to begin with and Kepler is expected to be a massive performance increase over Fermi. Also, the entire HD7850/70 and lower end line-up seem to be re-badges of current generation chips, only at 28nm. I am not buying it. Seems way too conservative approach by AMD.

 

[WMD]

Well 4890 was a refresh to 4870 so I think I should have stated 5870 vs. 4870 and 6970 vs. 4890. In that case 75-100% performance difference is still there, respectively. But ya, 5870 vs. 4890 was slightly less than 75% overall, closer to 60% faster.

HD6770 ~ HD5770 ~ HD4870.
HD5870 ~ HD6950

Looking at modern games:

http://www.computerbase.de/artikel/grafikkarten/2011/test-19-grafikkarten/4/

6950 1GB is 75% faster than HD6770 on average.
6970 2GB is 85% faster than HD6790 on average.

I don't see the point of linking "all resolutions" from TPU as relevant.

How about HD5870 vs. HD4870/4890 in games at 1920x1080 or higher:

1. Starcraft II => 2x faster than 4890
2. Mafia II => 2x faster than 4890
3. Call of Duty: Black Ops => 2x faster than 4890
4. Brink => 2x faster than HD6770 ~ HD5770 ~ 4870
5. Medal of Honor => 78% faster than HD5770 ~ HD4870
6. Avatar => 79% faster than 4870
7. BF:BC2 => 78% faster than 4870
8. AvP => 79% faster than 4870
9. Bioshock 2 => 78% faster than 4870
10. STALKER: COP => 82% faster than 4870
11. Shattered Horizon => 77% faster than 4870
12. Napoleon Total War => 78% faster than 4870
13. Just Cause 2 => 75% vs. 4870
14. GTA IV = 83% faster vs. 4870

There are at least 10 pages of benchmarks here.

On average HD5870 is at least 75% faster than HD4870 and HD6970 (refresh of 5870) is also at least 75% faster on average vs. HD4890. So HD7970 would need to be about 70-75% faster than HD5870 to be in line with what's expected imo.

But I am more upset that AMD isn't really increasing performance much in the mid-range. We had have nearly HD5850 level of performance for 2 years now (HD6870 was barely faster and it looks like HD7870 won't be faster than an overclocked 5850). That's pretty disappointing to me. While HD7950/70 cards will be a good deal faster, the mid-range continues to be lackluster, if these rumors are true.

HD6970 = 880mhz x 32 ROPs = 28.16 GPixels/sec pixel fill-rate
HD7970 = 1000mhz x 64 ROPs = 64.0 GPixels/sec (> 2x) That's a humongous increase in pixel fill-rate.

Still, if HD7970 is only 50% faster than HD6970, then how is AMD going to compete against Kepler with that? GTX580 is another 15% faster than HD6970 to begin with and Kepler is expected to be a massive performance increase over Fermi. Also, the entire HD7850/70 and lower end line-up seem to be re-badges of current generation chips, only at 28nm. I am not buying it. Seems way too conservative approach by AMD.

Those rumoured specs may not be real but small incremental performance increase is the best way to milk the market. 40nm - 28nm is a big node jump. But they could use that shrink to primarily lower die size and improve margins instead of increasing performance. That will also bring about power improvements, another huge draw for the potential market. Look at how hugely successful the 6900 series are despite being only 10% - 15% faster than 5800. Don't expect them to give us another 3870 - 4870 repeat. I am sure both companies realize pushing the performance envelop too hard end hurting both sides. Why not play nice and take turns to milk the market like last year.

 

[arredondo]

Maybe they plan on impressing with more modest pricing for this lineup. For instance, if the 7970 launches for $350 at the rumored specs, it would be a hit with plenty of people despite "only" a 50% improvement over the 6970 (which launched at $370).

 

[happy medium]

I'll get the hype train started....

and suddenly the train crashes.....

Quote

"Radeon HD 7000 will like the current 6000 family launch in steps and with different base architectures. If we are to believe the latest information that is said to come from AMD's leaked roadmap the graphics card marker is aiming for a first launch in Q4 2011. Then it will be the latest VLIW4 architecture (Very Long Instruction Word) that is on the menu and it will most likely have the same base as the Radeon HD 6900 series"

"We have no concrete information on when Radeon HD 7900 series will launch, but most likely Q1 2012. AMD has chosen to use a more performance oriented node, TSMC's 28nm HP (High Performance) that will be ready later than HPL, but also have switched from GDDR5 to the pretty untested XDR2 technology."

"Either way it sounds likely that AMD will start its 28nm launch of Southern Islands with VLIW4 and later introduce Graphics Core Next."

End quote

http://www.nordichardware.com/news/...-new-architecture-and-xdr2-rambus-memory.html
Looks like mid range/ low end first guys.
Seems alot of us guessed right.

 

[Silverforce11]

Still, if HD7970 is only 50% faster than HD6970, then how is AMD going to compete against Kepler with that? GTX580 is another 15% faster than HD6970 to begin with and Kepler is expected to be a massive performance increase over Fermi. Also, the entire HD7850/70 and lower end line-up seem to be re-badges of current generation chips, only at 28nm. I am not buying it. Seems way too conservative approach by AMD.

Theres nothing even close to concrete on Kepler.

AMD should think about their margins first and foremost. Getting 6970 +10% performance from a ~200mm2 die at 120W is a winner for the market. Really, 2 of that on a card at ~200W would offer unbeatable perf/watt etc.

 

[Grooveriding]

Theres nothing even close to concrete on Kepler.

AMD should think about their margins first and foremost. Getting 6970 +10% performance from a ~200mm2 die at 120W is a winner for the market. Really, 2 of that on a card at ~200W would offer unbeatable perf/watt etc.

Pretty much. With Battlefield 3 coming out in October if AMD does it right, everyone will be buying a 7870 to play BF3 with.

 

[happy medium]

Theres nothing even close to concrete on Kepler.

AMD should think about their margins first and foremost. Getting 6970 +10% performance from a ~200mm2 die at 120W is a winner for the market. Really, 2 of that on a card at ~200W would offer unbeatable perf/watt etc.

If history repeats itself for the 4 time.

1.Nvidia will have the fastest card and the most expensive.
2. AMD cards will be at least 15% slower in every tier , but less expensive.
3. Nvidia cards will use about 30 more watts per performance tier vs AMD.

bla, bla, bla, ect, ect, ect,

we all know the rest..............

 

[Grooveriding]

If history repeats itself for the 4 time.

1.Nvidia will have the fastest card and the most expensive.
2. AMD cards will be at least 15% slower in every tier , but less expensive.
3. Nvidia cards will use about 30 more watts per performance tier vs AMD.

bla, bla, bla, ect, ect, ect,

we all know the rest..............

Nothing you listed in any of those three points is correct for the current gen or the prior.

 

[happy medium]

Pretty much. With Battlefield 3 coming out in October if AMD does it right, everyone will be buying a 7870 to play BF3 with.

Everyone that owns a overclocked 5850/gtx470 or above won't, unless the price of a 7870 is under 200$, thats not much of a upgrade.

 

[happy medium]

Nothing you listed in any of those three points is correct for the current gen or the prior.

nvm...

 

[AtenRa]

Well 4890 was a refresh to 4870 so I think I should have stated 5870 vs. 4870 and 6970 vs. 4890. In that case 75-100% performance difference is still there, respectively. But ya, 5870 vs. 4890 was slightly less than 75% overall, closer to 60% faster.

HD6770 ~ HD5770 ~ HD4870.
HD5870 ~ HD6950

Looking at modern games:

http://www.computerbase.de/artikel/grafikkarten/2011/test-19-grafikkarten/4/

6950 1GB is 75% faster than HD6770 on average.
6970 2GB is 85% faster than HD6790 on average.

I don't see the point of linking "all resolutions" from TPU as relevant.

How about HD5870 vs. HD4870/4890 in games at 1920x1080 or higher:

1. Starcraft II => 2x faster than 4890
2. Mafia II => 2x faster than 4890
3. Call of Duty: Black Ops => 2x faster than 4890
4. Brink => 2x faster than HD6770 ~ HD5770 ~ 4870
5. Medal of Honor => 78% faster than HD5770 ~ HD4870
6. Avatar => 79% faster than 4870
7. BF:BC2 => 78% faster than 4870
8. AvP => 79% faster than 4870
9. Bioshock 2 => 78% faster than 4870
10. STALKER: COP => 82% faster than 4870
11. Shattered Horizon => 77% faster than 4870
12. Napoleon Total War => 78% faster than 4870
13. Just Cause 2 => 75% vs. 4870
14. GTA IV = 83% faster vs. 4870

There are at least 10 pages of benchmarks here.

On average HD5870 is at least 75% faster than HD4870 and HD6970 (refresh of 5870) is also at least 75% faster on average vs. HD4890. So HD7970 would need to be about 70-75% faster than HD5870 to be in line with what's expected imo.

But I am more upset that AMD isn't really increasing performance much in the mid-range. We had have nearly HD5850 level of performance for 2 years now (HD6870 was barely faster and it looks like HD7870 won't be faster than an overclocked 5850). That's pretty disappointing to me. While HD7950/70 cards will be a good deal faster, the mid-range continues to be lackluster, if these rumors are true.

HD6970 = 880mhz x 32 ROPs = 28.16 GPixels/sec pixel fill-rate
HD7970 = 1000mhz x 64 ROPs = 64.0 GPixels/sec (> 2x) That's a humongous increase in pixel fill-rate.

Still, if HD7970 is only 50% faster than HD6970, then how is AMD going to compete against Kepler with that? GTX580 is another 15% faster than HD6970 to begin with and Kepler is expected to be a massive performance increase over Fermi. Also, the entire HD7850/70 and lower end line-up seem to be re-badges of current generation chips, only at 28nm. I am not buying it. Seems way too conservative approach by AMD.

It is invalid data to show benchmarks one or two years after the initial release of the Graphic Card's and also, showing benchmarks at 2560x1600 and high filters only shows the difference of Higher/more ROPs and higher Memory Bandwidth
and it is not representative of the Architecture performance difference between the Cards.

When AMD's 58xx series was released, and that's the time we want to have the performance data, at 1920x1200 it was 63% faster than HD4870 1GB and 43% faster than HD4890.

If one year later drivers were better and gave another 10-20% that is another story, but at release date HD5870 was only 43% faster than HD4890 and that was the card it replaced.

http://www.techpowerup.com/reviews/ATI/Radeon_HD_5870/30.html

 

[blackened23]

edit: duplicate

 

[blackened23]

So how about some real benchmarks instead of the "relative performance" nonsense? Looks like pulling numbers out of thin air to support an argument. Would be nice if people on both sides didn't pull 3 benchmarks out of 20 to support their argument....anyway, i'll approach the 79xx like I did the 69xx: patiently await for a real product and real benchmarks

So GTX 580 wins some, 6970 wins some, conclusion? 6970 is cheaper and uses less power.. Is that important to you? Maybe not, maybe so. Crossfire scales better than SLI. Need more kittens.

 

[Mallibu]

So how about some real benchmarks instead of the "relative performance" nonsense? Looks like pulling numbers out of thin air to support an argument. Would be nice if people on both sides didn't pull 3 benchmarks out of 20 to support their argument....anyway, i'll approach the 79xx like I did the 69xx: patiently await for a real product and real benchmarks

So GTX 580 wins some, 6970 wins some, conclusion? 6970 is cheaper and uses less power.. Is that important to you? Maybe not, maybe so. Crossfire scales better than SLI. Need more kittens.

Infact 580 crashes 6970 in every game except batman & f1 2010,i wouldn't call that "even"

 

[blackened23]

Infact 580 crashes 6970 in every game except batman & f1 2010,i wouldn't call that "even"

Hmm, i'm trying to find where I claimed they were even. Oh wait, I didn't !

GTX 580: 490$

6970: 330$

6970x2 crossfire: 660$

GTX 580 vs 6970x2 question mark ??

 

[Mallibu]

Hmm, i'm trying to find where I claimed they were even. Oh wait, I didn't !

GTX 580: 490$

6970: 330$

Well you said "Gtx 580 wins some,6970 wins some" which is giving the impression the cards have slight differences,when the truth is 580 wins 40 of those 44-45 tests and most times with a large margin.

 

[blackened23]

Well you said "Gtx 580 wins some,6970 wins some" which is giving the impression the cards have slight differences,when the truth is 580 wins 40 of those 44-45 tests and most times with a large margin.

The 6970 xfire is also similarly priced to GTX 580 and trounces it in every test.

Anyway, rumor has it that tahiti XT (7970) is coming out Q1 2012. Depressing indeed.

 

[SirPauly]

What can you do?

http://www.newegg.com/Product/Produc...HD6970&x=0&y=0

339 dollars - with rebate

http://www.newegg.com/Product/Produc...TX+580&x=0&y=0'

429 dollars -- with rebate

 

[exar333]

I'd generally agree, but performance per watt is quickly becoming a more important factor than flat out performance and, especially with Nvidia, adding more and more compute features is definitely high up on their list. Another obstacle AMD and Nvidia are facing with releasing hardware that is simply faster is that with so many current users running hardware that is "fast enough" for everything they do, new functional features becomes a more crucial way to distinguish products from their own previous lineup and competitor's current lineup as well. Hence why AMD is touting triple display with 1 card, and Nvidia is more aggressively touting tessellation, 3D vision, and physx.

Did I miss this? Power usage has gone through the roof on GPUs over the past 2-3 years. It makes sense if AMD/NV want to re-use these for mobile purposes, but I sure have not seen either of them focus on power efficiency much at all. That's why we have a dozen plugs go into each GPU now.

 

[exar333]

I have to agree with the comments regarding GTX460 made by other posters. It was superior to HD6850 and HD6870 for many many months. Most of those GTX460 768/1GB cards were available for $120-180 and lower with rebates. Add far superior overclocking headroom and HD6850/6870 weren't a slam dunk. HD6870 became superior only in the last 3 months or so when its price started to dip to $150-155. Not to mention HD6870 initially was a disappointing card since it launched at around $240 at the time when GTX460 FTW was about $200 and GTX470 was $240.

You could also get the 5870's on their way out for about $200 when this was launched.

 

[LOL_Wut_Axel]

I might be remembering wrong, but aren't you overstating the 4890 -> 5870 performance increase?

Yep. Almost everyone does, though, so I can see why. At 1920x1200, overall the HD 5870 is 43% faster than the HD 4890.

100/70: 43%

Given this, I'm not expecting anything more than a 60% increase in performance in comparison to the HD 6970. 75% is out of the question.

 

[exar333]

Makes me glad I picked-up the 5870 shortly after launch, it has been a great card for a while now. Kind of sad there isn't much of an upgrade until Q1 2012.

 

[Spikesoldier]

The 6970 xfire is also similarly priced to GTX 580 and trounces it in every test.

Anyway, rumor has it that tahiti XT (7970) is coming out Q1 2012. Depressing indeed.

sad because you would guess they want to cash in on the holiday season and bf3's release.

guess were getting the pipe cleaning cards first to test 28nm.

 

[RussianSensation]

It is invalid data to show benchmarks one or two years after the initial release of the Graphic Card's and also, showing benchmarks at 2560x1600 and high filters only shows the difference of Higher/more ROPs and higher Memory Bandwidth
and it is not representative of the Architecture performance difference between the Cards.

If you clicked my links, most of the benches I linked were actually 1920x1080, not 2560x1600. Either way, to show true representation of performance difference between a modern generation and an older one, you should be testing them in high resolution, not in "all resolutions" averaged as you keep linking from TPU. Both 4870 and 5870 were intended for 1920x1080 at launch since they were high end cards. Also, of course you aren't going to get 75-100% performance difference if you keep averaging 7 4-6 year old games in there!

If one year later drivers were better and gave another 10-20% that is another story, but at release date HD5870 was only 43% faster than HD4890 and that was the card it replaced.

That's because TPU loves to include older games in their reviews - and we didn't have a ton of modern games when they did that initial 5870 review. As more modern games came out, HD5870 was really able to show its prowess over the slower 4870. And that's the point: AMD designed HD5870 to perform well in more modern games, not in older games.

If you look at a lot of the games they tested, an HD5870 would not be able to show its true performance since the games they tested were not intensive / ancient that reduced the true performance delta:

- UT3
- Dawn of War 2
- Call of Juarez 2
- ET:QW
- Far Cry 1
- Quake 4
- Prey
- Stalker 1
- TF2

^ Who upgraded from an HD4870 $130 to a $350 HD5870 to get more performance in these games? TPU's decision to include these old games is diminishing the performance advantage that HD5870 has.

Not to mention they include 3dMark as part of their averages and those aren't games.....

If you tested a bunch of old games with minimal AA at lower resolutions today with GTX580 vs. GTX460, you also will see much lower performance delta than if you were testing modern games for which GTX580 was intended in the first place.

Yep. Almost everyone does, though, so I can see why. At 1920x1200, overall the HD 5870 is 43% faster than the HD 4890.

Like I said TPU prefers to include a variety of older + newer games in their reviews. When I upgraded my HD4890 I didn't do so to go from 100-200 FPS in TF2. If you look at performance delta of HD5870 vs. 4870/4890 in more modern games for which people purchased the HD5870, the performance is far more than 43%. My HD6970 is nearly 2x faster than 4890 in Crysis 1 for example.

When I linked 14 or so modern games, the HD5870 is 75-100% faster. And there are 8 more pages at GameGPU.ru that test <2 year old games that all show this. If you test 5-6 year old games, of course you are only going to have 40% performance difference because you aren't even pushing the HD5870 (i.e., or not exposing the bottlenecks in the 4870/4890).

In a review of modern games: BF3, Witcher 2, Crysis 2, Skyrim, Shogun 2, Dragon Age 2, I would expect HD7970 to be at least 50% faster than HD6970.

 

[LOL_Wut_Axel]

If you clicked my links, most of the benches I linked were actually 1920x1080, not 2560x1600. Either way, to show true representation of performance difference between a modern generation and an older one, you should be testing them in high resolution, not in "all resolutions" averaged as you keep linking from TPU. Both 4870 and 5870 were intended for 1920x1080 at launch since they were high end cards. Also, of course you aren't going to get 75-100&#37; performance difference if you keep averaging 7 4-6 year old games in there!



That's because TPU loves to include older games in their reviews - and we didn't have a ton of modern games when they did that initial 5870 review. As more modern games came out, HD5870 was really able to show its prowess over the slower 4870. And that's the point: AMD designed HD5870 to perform well in more modern games, not in older games.

If you look at a lot of the games they tested, an HD5870 would not be able to show its true performance since the games they tested were not intensive / ancient that reduced the true performance delta:

- UT3
- Dawn of War 2
- Call of Juarez 2
- ET:QW
- Far Cry 1
- Quake 4
- Prey
- Stalker 1
- TF2

^ Who upgraded from an HD4870 $130 to a $350 HD5870 to get more performance in these games? TPU's decision to include these old games is diminishing the performance advantage that HD5870 has.

Not to mention they include 3dMark as part of their averages and those aren't games.....

If you tested a bunch of old games with minimal AA at lower resolutions today with GTX580 vs. GTX460, you also will see much lower performance delta than if you were testing modern games for which GTX580 was intended in the first place.



Like I said TPU prefers to include a variety of older + newer games in their reviews. When I upgraded my HD4890 I didn't do so to go from 100-200 FPS in TF2. If you look at performance delta of HD5870 vs. 4870/4890 in more modern games for which people purchased the HD5870, the performance is far more than 43%. My HD6970 is nearly 2x faster than 4890 in Crysis 1 for example.

When I linked 14 or so modern games, the HD5870 is 75-100% faster. And there are 8 more pages at GameGPU.ru that test <2 year old games that all show this. If you test 5-6 year old games, of course you are only going to have 40% performance difference because you aren't even pushing the HD5870 (i.e., or not exposing the bottlenecks in the 4870/4890).

In a review of modern games: BF3, Witcher 2, Crysis 2, Skyrim, Shogun 2, Dragon Age 2, I would expect HD7970 to be at least 50% faster than HD6970.

No, it wasn't. And performance going from 43% higher to 55-60% higher in newer games doesn't make that much of a difference given the argument we're trying to make. It's still a ways off from 75-100%, and like I said, AMD didn't milk everything they could out of 55nm with the 4890. That's not the case for 40nm, as the 6970 was the biggest they could make without having a huge die. Again, thinking you'll get 75% more performance out of a single GPU in comparison to the HD 6970 with 28nm is simply nonsensical. 50% is more in line with reality.