Official AMD Radeon 6000 Series Discussion Thread

[Benetanegia]

sense put into words id also like to say to bene that on a forum called tech power up nobody should EVER say permformance these days is enough cos how much is enough i aint seen enough of anything including high fidelity 3d games.

Actually I'm saying the opposite. I was being sarcastic if you didn't get that.

I was trying to demostrate that there's no point at which anyone can say "this is enough", because it will not satisfy everyone. Yet it is a fact that the majority of people play on the Wii, or in PC they play casual games, so if we have to draw a line as to where enough would be, shouldn't the line be there, low, and not where you need $300+ on hardware? For me "you need DX11, you need a lot of antialiasing, a lot of details, alot of bla and bla... and you need tesselation, but ey! beware, you only need enough tesselation and we draw the line and only we offer enough tesselation, the competitor is overkill." That makes no sense to me. Either I need it or not, and if I settle for "I need it and I'll pay for it" I want them to give me as much as is posible.

In a tech site called TechPowerUp it should be said the more powerful the cards the better, more tesselation the better (especially as it has been demostrated it can be done), , more more and more. Enough? Nah.

 

[f22a4bandit]

Sorry I meant that I was joining in with you on taking a pass this generation. I'm not currently unemployed but school bills are doing a good job of eating up just about all I have.

I'll keep you in my prayers.

To regards to 6850s.....
Amazon has the 6850 listed with specs.

Oh, no don't worry about it, no offense taken! I'll be right in that same boat with you when I get a job anyway, so it's almost a guarantee that I'll pass on this generation. Thanks for thoughtful prayers, btw, I'll do the same in return!

Until benchmarks come out, let's all just it up and relax...Hahaha, no I like eating my popcorn and reading the discussions, so banter on my good TPUers!

 

[EastCoasthandle]

6850 Unboxing. As much as you would rather see paint dry you know you want to watch this .

 

[SNICK]

Amd radeon hd 6850 gpuz shot (over clocked)
i have found this pic on some arab website so lets share it

 

[the54thvoid]

Can someone here that is informed and savyy say with a straight face that AMD is not trying to deceive customers, with the rebrand

Ok, with reference to the link below:

http://news.ati-forum.de/images/stories/Jekel/2010/amd-slide-leaks-02.jpg

The Juniper 57xx rebrand to 67xx is a great deceit if:
a) AMD dont tell anyone about it and/or,
b) They charge more for the cards.

The reason for the rebadge of the Barts core to 68xx is because it gives too high a performance to be badged as 67xx (low - mid) when the 6870 will apparantly be better than the 5850.

As the 67xx badge is vacant it makes sense to move the Juniper core to that new moniker.

However, i entirely agree with you that it is deceitful if AMD charge more for it. Let's face it, no one is being ripped off if the pricing stays logical given the performance of the new chips. However, there is speculation the 58 series will be discontinued as soon as Barts et al are out.

If they rename again for the 7xxx series, then it's truly just as bad as NV did. Although the 7xxx is a die shrink so it's tickety tock time. Old design, new tech type scenario.

 

[the54thvoid]

In a tech site called TechPowerUp it should be said the more powerful the cards the better, more tesselation the better (especially as it has been demostrated it can be done), , more more and more. Enough? Nah.

Fundamentally disagree. We as a scientific people are trying to create new green ways of saving energy while delivering performance. Tech does not equal power. Tech = advancement of the current state of knowledge, not the pursuit of pure power.

 

[cadaveca]

If they rename again for the 7xxx series, then it's truly just as bad as NV did. Although the 7xxx is a die shrink so it's tickety tock time. Old design, new tech type scenario.

If "AMD" charges more for 5770 rebadge, it's because the board the gpu sits on is more expensive. Clearly the current 68xx VRM design was a move to cut board costs, but at the same time, they added display connectors that can only increase overall cost. Seemingly they've still managed to reduce overall cost to the consumer though, and deliver similar performance as the last generation. That makes Barts a good move.

It's kind of a critical point...but in the end, it's much more likely that AMD is charging thier partners the same cost for the Juniper chip itself, and the partners are the ones charging more.

But, we are talknig a few dollars. With 6870 @ $269, 6850 is near $200, therefore 6770 has to be less than that, and is currently retailing @ $139 or so as 5770...

 

[Benetanegia]

Ok, with reference to the link below:

http://news.ati-forum.de/images/stories/Jekel/2010/amd-slide-leaks-02.jpg

The Juniper 57xx rebrand to 67xx is a great deceit if:
a) AMD dont tell anyone about it and/or,
b) They charge more for the cards.

The reason for the rebadge of the Barts core to 68xx is because it gives too high a performance to be badged as 67xx (low - mid) when the 6870 will apparantly be better than the 5850.

As the 67xx badge is vacant it makes sense to move the Juniper core to that new moniker.

However, i entirely agree with you that it is deceitful if AMD charge more for it. Let's face it, no one is being ripped off if the pricing stays logical given the performance of the new chips. However, there is speculation the 58 series will be discontinued as soon as Barts et al are out.

If they rename again for the 7xxx series, then it's truly just as bad as NV did. Although the 7xxx is a die shrink so it's tickety tock time. Old design, new tech type scenario.

It will never make more sense than with Nvidia. Nvidia at least did the first rename in order to catch up with Ati, because Ati jumped to new genenration for no reason (other than to differentiate themselves from the unpopular HD2900). No RV6xx should have ever been called HD3000 to begin with. It's there where the renaming game started. It was not faster and it was just a die shrink with some tweaks, if that was enough to warrant an new name, then 9800 was not a rebrand either as it was also a shrink and added some features, like compatibility with Optimus, newer video decode engine and such.

The second change to GT250 made less sense buta new naming model -> new name, and it's not sure that uninformed people will think that 250 > 9800, while 67xx > 5770 or 68xx > 58xx is almost a given, just like 38xx > 2900 or 9800 > 8800.

G2xx to G3xx makes no sense at all, just like HD57xx to HD6x77.

 

[the54thvoid]

It will never make more sense than with Nvidia

Never said it would.

Basically AMD are moving parts to fill a new (ish) architecture performance gap with old (ish) architecture cards. I'm not defending AMD. It would be swell if all the 67xx boxes had a strapline under the HD67xx name that said "formerly known as HD57xx".

But if that happens i'll eat your pee soaked pants. I'm really hopping that doesn't happen...

Can I add that at the time of the HD2900 I had a beautiful sli set up of 7950 GT's (silent heatpipe versions). They were my favourite cards ever (in terms of what they gave for no noise, bliss)

piccy here - http://www.overclock.net/appraisals/286913-appraisal-xfx-7950-gt-extreme.html

 

[War_God]

Wow the more PR slides I see more dissapointed that I am with AMD on this release. First the PR spin of claiming an improvement on perf/die area where it doesn't exist, using HD6870 and HD5850 to fake it out and now these

Sorry for being a little late to the game, but I read this post and I just have to ask.
How is it a fake?
I mean, from the slides, there is clearly a difference in die area.

5850»334mm
6870»255mm

Now, admitting that 6870 has a performance similar to the 5850, then, and having 6870 as a baseline:

255mm - 100%
334mm - X
Where X = 131%.
So, we have supposedly the same or better performance, minus 31% die area from the 6870 compared to the 5850.
Maybe I'm not looking at everything, but that, plus the fact that 6870 has considerably less stream processors than the 5850, due to architectural changes, really does seem to me like an improvement in die size/performance.

But please correct me if I'm wrong. I don't know all the details about this architecture.

 

[cadaveca]

Amd radeon hd 6850 gpuz shot (over clocked)
i have found this pic on some arab website so lets share it

Some food for thought:

 

[TheMailMan78]

Sorry for being a little late to the game, but I read this post and I just have to ask.
How is it a fake?
I mean, from the slides, there is clearly a difference in die area.

5850»334mm
6870»255mm

Now, admitting that 6870 has a performance similar to the 5850, then, and having 6870 as a baseline:

255mm - 100%
334mm - X
Where X = 131%.
So, we have supposedly the same or better performance, minus 31% die area from the 6870 compared to the 5850.
Maybe I'm not looking at everything, but that, plus the fact that 6870 has considerably less stream processors than the 5850, due to architectural changes, really does seem to me like an improvement in die size/performance.

But please correct me if I'm wrong. I don't know all the details about this architecture.

Well first of all I didn't say anything you just said I said.

Actually I'm saying the opposite. I was being sarcastic if you didn't get that.

I was trying to demostrate that there's no point at which anyone can say "this is enough", because it will not satisfy everyone. Yet it is a fact that the majority of people play on the Wii, or in PC they play casual games, so if we have to draw a line as to where enough would be, shouldn't the line be there, low, and not where you need $300+ on hardware? For me "you need DX11, you need a lot of antialiasing, a lot of details, alot of bla and bla... and you need tesselation, but ey! beware, you only need enough tesselation and we draw the line and only we offer enough tesselation, the competitor is overkill." That makes no sense to me. Either I need it or not, and if I settle for "I need it and I'll pay for it" I want them to give me as much as is posible.

In a tech site called TechPowerUp it should be said the more powerful the cards the better, more tesselation the better (especially as it has been demostrated it can be done), , more more and more. Enough? Nah.

Ben I get where you are coming from on this and to an extent I agree. So before you call me a fanboy I want you to know chances are Ill be going green next time for a few reasons. But let me say this.

We have gone far enough with horsepower. Why because I would say 98% of gaming titles that are released today are in fact ports. They are ports from systems that have 5+ year old architecture in them. My 5850 or a 470 is pretty much overkill for gaming nowadays. It will be until the next NEXT generation of consoles are released. Thats just the sad fact. What we need is features. Things that our hardware can do to make the games better on a software level. Maybe its ATI and their "new" translacing. Maybe its Nvidias hardware Physx. Who knows? What I do know is more horsepower is a waste of money and resources IMO.

Oh and by the way the reason I may be going green next time is to support the underdog AND because Nvidia tends to offer more features then AMD. I don't care if no one uses them. Its the principle.

 

[Benetanegia]

Sorry for being a little late to the game, but I read this post and I just have to ask.
How is it a fake?
I mean, from the slides, there is clearly a difference in die area.

5850»334mm
6870»255mm

Now, admitting that 6870 has a performance similar to the 5850, then, and having 6870 as a baseline:

255mm - 100%
334mm - X
Where X = 131%.
So, we have supposedly the same or better performance, minus 31% die area from the 6870 compared to the 5850.
Maybe I'm not looking at everything, but that, plus the fact that 6870 has considerably less stream processors than the 5850, due to architectural changes, really does seem to me like an improvement in die size/performance.

But please correct me if I'm wrong. I don't know all the details about this architecture.

Because the HD5850 is not what the 334 mm^2 Cypress chip is, the HD5870 is the real 334 mm chip, which has 1600 SP and runs at 850 Mhz, but could had a 900 Mhz reference speed too, and at 900 Mhz is a good 30% faster than HD5850/HD6870 cancelling out the "advantage". The 5850 is a harvestd part, a crippled part and underclocked to 725 Mhz. In order to make a fair comparison they should have done HD5870 vs 6870 or 5850 vs 6850. In both cases HD5000 wins performance wise, and perf/die area is similar, so that wouldn't allow them the trick.

 

[Mr McC]

Because the HD5850 is not what the 334 mm^2 Cypress chip is, the HD5870 is the real 334 mm chip, which has 1600 SP and runs at 850 Mhz, but could had a 900 Mhz reference speed too, and at 900 Mhz is a good 30% faster than HD5850/HD6870 cancelling out the "advantage". The 5850 is a harvestd part, a crippled part and underclocked to 725 Mhz. In order to make a fair comparison they should have done HD5870 vs 6870 or 5850 vs 6850. In both cases HD5000 wins performance wise, and perf/die area is similar, so that wouldn't allow them the trick.

You're moving from improper implementation of tesselation to rebranding (incidentally I am totally against this as well) to now talk about "tricks", on practically the eve of the launch, shouldn't you also prepare the champagne just in case these turn out to be well priced, well made cards that offer excellent performance? You seem determined to be unhappy about cards that we as yet know nothing about.

 

[Benetanegia]

You're moving from improper implementation of tesselation to rebranding (incidentally I am totally against this as well) to now talk about "tricks", on practically the eve of the launch, shouldn't you also prepare the champagne just in case these turn out to be well priced, well made cards that offer excellent performance? You seem determined to be unhappy about cards that we as yet know nothing about.

I don't care about the cards. I'm not bitching about the cards at all. I'm talking about the slides. I hate PR bullshit and AMD is raising the bar to levels I'd never thought were posible. And I didn't actually think it would be AMD who did that, maybe that's part of the problem, that it's not Intel or Nvidia who are doing it.

You know, I don't care being called Nvidia fanboy on these forums, it's been long siince I play the much needed devil's advocate role and I like it. But in TPU you cannot say anything that is remotedly bad for AMD or you get criticized constatly. Nvidia threads are full of "FAIL!""GTFU Nvidia" and similar posts, and nothing happens, but you say that AMD is lying when they claim and improvement in perf/die area in a well reasoning post and "oh boy, run while you can, because torches are lit and pitchforks are sharp".

 

[War_God]

@ TheMailMan78
Sorry. Quoted a quote, an honest mistake.

Because the HD5850 is not what the 334 mm^2 Cypress chip is, the HD5870 is the real 334 mm chip, which has 1600 SP and runs at 850 Mhz, but could had a 900 Mhz reference speed too, and at 900 Mhz is a good 30% faster than HD5850/HD6870 cancelling out the "advantage". The 5850 is a harvestd part, a crippled part and underclocked to 725 Mhz. In order to make a fair comparison they should have done HD5870 vs 6870 or 5850 vs 6850. In both cases HD5000 wins performance wise, and perf/die area is similar, so that wouldn't allow them the trick.

Thanks for clearing the issue.
But still, this is propaganda. It's Never absolutely honest. And basically, AMD did it right, because even though the hd5850 doesn't have it's chip in the full configuration, it still Does occupy 334mm^2 (it just doesn't use a part of it). It's a trick, indeed, but still, they Can get away with it.

Still on the topic of performance/size. The difference between the full-fledged and the trimmed Cypress in terms of usable die size is probably almost negligible (what, 10mm^2?) So, mathematics wise, the difference isn't that much. And still, they make do the same performance with much less stream processors. (1120 or so in the 6870 vs 5850's 1440) They may have put some tricks in the slides, but they Are based in facts.

And P.S.: I'm not criticizing. I just wanted to understand the reason why you said it was a fake, when in my admittedly limited knowledge, it didn't seem so. And if I don't agree with your assessment, well, that doesn't automatically mean I'm a "Pro AMD/Grave NV" torch holder.

 

[CDdude55]

@Bene
Of course it's all PR marketing, they're a company.

You know, I don't care being called Nvidia fanboy on these forums, it's been long siince I play the much needed devil's advocate role and I like it. But in TPU you cannot say anything that is remotedly bad for AMD or you get criticized constatly. Nvidia threads are full of "FAIL!""GTFU Nvidia" and similar posts, and nothing happens, but you say that AMD is lying when they claim and improvement in perf/die area in a well reasoning post and "oh boy, run while you can, because torches are lit and pitchforks are sharp".

I agree actually.

I am one of the may few who admit the forums obvious bias, but that's just how it is..

 

[Benetanegia]

@ TheMailMan78
Sorry. Quoted a quote, an honest mistake.



Thanks for clearing the issue.
But still, this is propaganda. It's Never absolutely honest. And basically, AMD did it right, because even though the hd5850 doesn't have it's chip in the full configuration, it still Does occupy 334mm^2 (it just doesn't use a part of it). It's a trick, indeed, but still, they Can get away with it.

Still on the topic of performance/size. The difference between the full-fledged and the trimmed Cypress in terms of usable die size is probably almost negligible (what, 10mm^2?) So, mathematics wise, the difference isn't that much. And still, they make do the same performance with much less stream processors. (1120 or so in the 6870 vs 5850's 1440) They may have put some tricks in the slides, but they Are based in facts.

Yeah for the $200 price bracket the HD6870 is a better card without a doubt. But that does not warrant the claim at all.

Also

1440 * 725 Mhz = 1044000
1120 * 900 Mhz = 1008000

As you can see it's very close, the HD6870 wins on real performance because its higher clock.

 

[dj-electric]

Amd radeon hd 6850 gpuz shot (over clocked)
i have found this pic on some arab website so lets share it

If this card will overclock like this ill sh*t bricks

 

[War_God]

1440 * 725 Mhz = 1044000
1120 * 900 Mhz = 1008000

As you can see it's very close, the HD6870 wins on real performance because its higher clock.

I may be hard-headed, but aren't you dismissing the architectural improvements? This will all be corrected when we have the real reviews in front of our eyes, but consider overclocking the 5850 to the same 900Mhz. Even if it's performance is +5%, even +10% than the 6870, that will be a negligible difference comparing the absolutely huge die size difference between chips.

 

[Benetanegia]

I may be hard-headed, but aren't you dismissing the architectural improvements? This will all be corrected when we have the real reviews in front of our eyes, but consider overclocking the 5850 to the same 900Mhz. Even if it's performance is +5%, even +10% than the 6870, that will be a negligible difference comparing the absolutely huge die size difference between chips.

My friend, but you keep comparing a full chip to a crippled part. And if you OC a HD5850 to 900 Mhz it is faster than a HD5870, not 5% faster than HD6870.

The new architecture itself might be better, because of the dual engine it posibly is more efficient, but Barts (chip =! architecture) is not showing that even on AMD's slides, which of course have been made to favor the HD6870. But like I said, because of the dual engine, Cayman might be more efficient than Cypress, because Cypress has never demostrated good scaling up to 1600 SPs, i.e a HD5850 clocked to exact same clocks is just as fast, BUT these inneficiency showed up between 1440 and 1600 SP not lower. AMD's slides are not showing any real improvement on Barts and they are internal benchmarks! We should wait for real benches? Of course, but don't expect real benches to show better performance than AMD ones.

 

[motasim]

I don't care about the cards. I'm not bitching about the cards at all. I'm talking about the slides. I hate PR bullshit and AMD is raising the bar to levels I'd never thought were posible. And I didn't actually think it would be AMD who did that, maybe that's part of the problem, that it's not Intel or Nvidia who are doing it.

You know, I don't care being called Nvidia fanboy on these forums, it's been long siince I play the much needed devil's advocate role and I like it. But in TPU you cannot say anything that is remotedly bad for AMD or you get criticized constatly. Nvidia threads are full of "FAIL!""GTFU Nvidia" and similar posts, and nothing happens, but you say that AMD is lying when they claim and improvement in perf/die area in a well reasoning post and "oh boy, run while you can, because torches are lit and pitchforks are sharp".

Bene, I'm sorry for hitting hard at you mate, and I'm really sorry for calling you an nvidia fanboy

I can confirm to you without doubt that I am not biased towards any company in any capacity. I am building a gaming rig next December/January and I'll buy the best GPU that is available at the time (regardless if it was made by ATI or nVidia) noting that I have a maximum budget of $425 set for the GPU. My decision will be only based on benchmarks in GAMES and nothing else! I really hope that by that time I'll be able to choose between AMD's 6950/6970 and nVidia's GTX 475/490/495/580 or whatever it is called

 

[the54thvoid]

Theres gonna be great brand new cards from AMD out then, the 67xx series, get two and crossfire them.

<sarcasm>

Any serious reply to this post gets shot

Just thought, as it's well known NV scales better with sli, we should start calling crossfire 'misfire' until it reaches parity with NV.

 

[motasim]

... Just thought, as it's well known NV scales better with sli, we should start calling crossfire 'misfire' until it reaches parity with NV.

+1 ... Pure Gold ...

 

[Benetanegia]

Just thought, as it's well known NV scales better with sli, we should start calling crossfire 'misfire' until it reaches parity with NV.

You know what's rare about that:

Nvidia = better at multi-GPU scaling, also when it comes to dual_gpu cards historically Nvidia ones consumed less comparatively than Ati ones, which usually consume horribly in comparison to single ones. For example, HD4870 was around 200w, X2 380w. GTX275 230w, GTX295 300w.

AMD = better on manufacturing and yields and single card power consumption

You'd really think that Nvidia should be following AMD's strategy (smaller chips, multi-GPU) and AMD Nvidia's (big and well packed chips).