• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[bit-tech] Radeon HD 7000-series rumoured for May production

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
AMD looks at fusion sales and giggles. Nvidia sees the writing on the wall. They are losing their market base, gaming is no longer able to float the company.
 
Last edited:
Arkadrel said:
@Silverforce11

You think AMD should aim for 380mm^2 for its topend?

And just double up on everything?

Is that even needed by anyone atm? imagine something thats as fast as 580 SLI on a single card.... what game needs that atm? Thats what a 380mm^2 28nm chip would be like.

I think AMD should aim at just makeing them smaller, aim for 150% of 580's performance, and make a chip thats like ~280mm^2 that they can sell for 150% 580's prices, and make huge profits on, until Nvidia has a 6xx series card out, and then AMD just lowers prices and competeings on performance/cost.

That would be the smarter bussines move to make. That would also mean theyd be sticking with 256bit bus for their cards again.
Click to expand...

We're talking about high end right? I'm not talking about mid-range stuff that can run most games at 1080p fine. High end cards have always been marketed at users who game at the max resolution of the day. That's 2560 or multi-res. A lot of games need a 6990 or 590 to run at these resolution.

For the previous generations, AMD has always focus on a single high end GPU that's within a certain TDP limit, somewhere around ~200W. So they can put 2 of these onto a card and claim the crown. With the 6970, they raised the bar to ~250W for a single card. AMD engineers design GPUs with a TDP limit in mind and also a die size. Both of these influence how much stuff they cram in, and they always cram in the max they can. Engineers don't think "Heck, we can double everything.. but why bother, no games need that power". 😀

There's very little chance they will back down for future generations when its going so well for them. "Best" or "Fastest" GPU reputation matters a lot for everything else in the franchise.

Edit: If they released a next gen high end GPU on 28nm that's ~50% faster than a 6970, they will fail when NV release Fermi 3, hard. NV have been improving their perf/mm and perf/watt and i bet they will continue focus on that for next gen.
 
Silverforce11 said:
We're talking about high end right? I'm not talking about mid-range stuff that can run most games at 1080p fine. High end cards have always been marketed at users who game at the max resolution of the day. That's 2560 or multi-res. A lot of games need a 6990 or 590 to run at these resolution.

For the previous generations, AMD has always focus on a single high end GPU that's within a certain TDP limit, somewhere around ~200W. So they can put 2 of these onto a card and claim the crown. With the 6970, they raised the bar to ~250W for a single card. AMD engineers design GPUs with a TDP limit in mind and also a die size. Both of these influence how much stuff they cram in, and they always cram in the max they can. Engineers don't think "Heck, we can double everything.. but why bother, no games need that power". 😀

There's very little chance they will back down for future generations when its going so well for them. "Best" or "Fastest" GPU reputation matters a lot for everything else in the franchise.
Click to expand...
Amd doesn't need to compete in bigger die sizes, Nvidia can have the compute market. It is cost prohibitive. This node is the last big one for amd the market isn't there.
 
Amd is thinking ahead. They have had some missteps but overall they have a future plan. They are competive and cautious. I believe they have a good plan. Nvidia has had theirs also. They are basically shut out of the decreasing market. They need innovation and they have positioned themselves in an unsustainable market.
 
Edit: If they released a next gen high end GPU on 28nm that's ~50% faster than a 6970, they will fail when NV release Fermi 3, hard. NV have been improving their perf/mm and perf/watt and i bet they will continue focus on that for next gen.[/QUOTE]

Yes this is true in a dying pc gaming market, The performance crown is minor compared to oem sales.
point is pc gaming is a moot point. crysis2, da2 and others have shown we are not a percentage that merits concern. This is a business we are a small part of that.
 
Last edited:
purefun1965 said:
Yes this is true in a dying pc gaming market, The performance crown is minor compared to oem sales.
point is pc gaming is a moot point. crysis2, da2 and others have shown we are not a percentage that merits concern. This is a business we are a small part of that.
Click to expand...

What else is new? PC gaming has supposedly been dying since Commodore 64 when Nintendo came out lol
 
load81 said:
What else is new? PC gaming has supposedly been dying since Commodore 64 when Nintendo came out lol
Click to expand...
Not only is it a tired statement, let's just put it to rest with one small thought: Every day during peak hours there are around 3 million gamers on Steam, and that's just Steam. Based on the dev costs of 95% of PC games, if you could make even $10 each off 3 million folks, you'd end up with a huge profit.
The revenue is there, you just need to know how to do it. Hint: At the moment, it's not by trying to compete with Call of Duty and Battlefield to make the next triple-A FPS game and spending $60m to net only $30m in sales. Is it any wonder that a dev studio that has a poor sales result has a bitter taste in their mouths regarding PC gaming that is tough to get rid of for years afterward? No. And their risky planning would be as much to blame as anything else.
Instead, we see the mobile appstore model success, where smaller studios (or even individuals) create catchy products that are simpler and cheaper and that get great reviews and sell millions of copies. And we have seen that evolve from/to the $10-20 PC/Mac indie game, which is more relevant to my point, as it is a more fully-featured product and where studios today can get their start and build up the capital to later attack the triple-A realm and its more expensive engine licenses and production costs.
 
Last edited:
RussianSensation said:
Actually, I also don't understand what point you are trying to get across.

Based on your assessment of HD4870 --> HD6870, what does this comparison yield in terms of the expected performance increase from HD6970 to HD7970?

(1) You seem to acknowledge the facts that HD6870 does not have dual graphics engines, does not have HD69xx's ROPs, does not have Cayman's memory controller, does not support double precision, etc. Then, what meaningful information can you derive from looking at the HD6870 that will allow us to make a reasonable prediction of HD7970's performance, itself based on the Cayman and not Barts architecture? All of those features resulted in a much larger die space on the 6970; and they will make their way into the 7970.

Take a step back and look at HD6870 vs. HD5870. (3) After removing 480 SPs, 24 TMUs and reducing memory bandwidth by 12.5% compared to the 5870, the 6870 is only about 10% slower. We may perhaps conclude that the original Cypress VLIW-5 architecture was not a very balanced one. (2) However, this doesn't tell us anything about HD7970's architecture or die space.
Click to expand...

(1) Are you kidding? You need to take a step back. You're way too caught up in the details.

--Theory of the hypothetical thought process that can be used to derive educatedly guessed hypothesis.--
R770/790 to Barts:
-Same die size
-Different chips separated by one node shrink

Analysis: What kind of "improvement" do you get from shrinking a node but keeping a similar die size? Improvement includes all attributes, not just performance.

Barts to Cayman:
-Different die size
-Different chips on the same node

Analysis: What kind of improvement do you see by increasing the die size?

Cayman to Southern Islands high end:
-Die size?
-Different chips separated by a node shrink

Analysis: Using the previous analysis, we can hypothesize what AMD could improve with this chip. There is a correlation between HOW MUCH of an improvement there is and the die size.
---------------------------------------------------------

(2) What a silly statement. Your conclusion about Cypress Balance... of course has no relation to actual, real Southern Islands information nor does it have any relation to whatever argument you're making here. Silly straw man. Stop doing that. The only way to get conclusive information about the 7970's architecture and die size is from AMD themselves. But what you did, what everyone is doing here, what I'm doing, and what Lonyo was doing, was use past information to come up with a trajectory of what COULD happen. COULD HAPPEN. Not what WILL HAPPEN. COULD. Could, because there are variables in the equation and their values are unknown at this time. So we're looking at the past, to see a trend, to come up with one of many answers to the equation. And we're doing that by simply plugging in different variables. No one is saying X variable absolutely equals "this" value. Well I'm definitely not. I'm suggesting X variable could (COULD) be this value, and this is what the equation will look like with that value for X being used.

How hard is that to understand?

(3) Well it's good that are utilizing the process I've been talking about. But you do know the details you just illustrated can (CAN) point to another trend? The first iteration of these high end chips aren't the most efficient per mm. The 4890 increased performance per mm. The 6870 increased performance per mm. GF110 increased performance per mm. These changes are on the same node. It's highly likely Southern Islands won't be at peak efficiency, either.

So I really have no idea why you even bothered listing those differences, because they don't back you up. The way you said it was basically a straw man diversion. "Oh look at this fact BECAUSE IT HAS NOTHING TO DO WITH WHAT I'M GOING TO SAY NEXT BUT WHAT I'M GOING TO SAY NEXT PROVES WHAT YOU'RE SAYING HAS NOTHING TO DO WITH WHAT I'M SAYING."


And to bring it all back, I first engaged in this debate because you said it was MORE appropriate to compare the 4870 to the 5870 and that you couldn't compare it to the 6870 at all. Just about everything I've been trying to say since is illustrate both methods are equally valid.
 
Last edited:
Rezist said:
Everyone assumes they want to double everything,
Click to expand...

When i say AMD could double the 6970 at 28nm, im only talking about Transistor count.

So at 28nm we could have double the transistor count of Cayman (HD6970) 2x 2640M = 5280M transistors and keep almost the same die size (389mm2). I will have to say that they will try and make it smaller than that.

Even if AMD would literally double the Cayman, the chip would be bellow the 400mm2 at 28nm and would have an exceptional performance but i dont think they will need a 512-bit memory nor will they double the ROPs.

What i really want to know is if they will double the Front End, will we have 4 Tessellation units or will they try to save die space and keep them at two ?? That part will tell us the philosophy of the design and what root they would like to pursuit.

One more think i would like to know is if they will keep the same SIMD design with 64 SPs (4-VLIW) and 4 Texture units, because if they will double the SP count then 79xx could come up with 192 Texture units.
 
cusideabelincoln said:
--Theory of the hypothetical thought process that can be used to derive educatedly guessed hypothesis.--
R770/790 to Barts:
-Same die size
-Different chips separated by one node shrink

Analysis: What kind of "improvement" do you get from shrinking a node but keeping a similar die size? Improvement includes all attributes, not just performance.

Barts to Cayman:
-Different die size
-Different chips on the same node

Analysis: What kind of improvement do you see by increasing the die size?

Cayman to Southern Islands high end:
-Die size?
-Different chips separated by a node shrink

Analysis: Using the previous analysis, we can hypothesize what AMD could improve with this chip. There is a correlation between HOW MUCH of an improvement there is and the die size.
Click to expand...

Since the architectures you comparing are so much different and 55nm to 40nm didn’t have 100% transistor density as 40nm to 28nm will have, the comparison to 4870 vs Barts (die size, node process) and Barts vs 6970 (die size, same process) will not get you any ware in relation to 6970 vs 7970.

If you want to make an analysis from 4870 onwards for High End chips you really need to take out of the equation the Barts architecture. So in order to come up with a valid conclusion, we have to compare 4870/90 to 5870 and 6970 to 7970.

cusideabelincoln said:
So we're looking at the past, to see a trend, to come up with one of many answers to the equation. And we're doing that by simply plugging in different variables. No one is saying X variable absolutely equals "this" value. Well I'm definitely not. I'm suggesting X variable could (COULD) be this value, and this is what the equation will look like with that value for X being used.
Click to expand...

I will agree with what you saying but we have to know what to take from the past and make that trend.

cusideabelincoln said:
The 4890 increased performance per mm. The 6870 increased performance per mm. GF110 increased performance per mm. These changes are on the same node. It's highly likely Southern Islands won't be at peak efficiency, either.
Click to expand...

And here comes that part,

4890 increased the performance by increasing the core frequency from 750MHz (4870) to 850MHz (4890) but they kept the same die size. So the increase came from the frequency and not by the design.

They took off silicon and features (no FP64, different memory controller) from 5870 and they created the 6870. So in order to raise the performance per mm they lost features and that don’t make a valid comparison. If 5870 didn’t support FP64 and had the same Front End with 6870 they could have the same performance per mm.

GF110 increased performance with a frequency boost from 700MHz(480) to 772MHz(580) plus 32 more Cuda cores. If GTX480 would have the same Cuda cores and Frequancy I will bet they would have the same performance per mm.
 
AtenRa said:
GF110 increased performance with a frequency boost from 700MHz(480) to 772MHz(580) plus 32 more Cuda cores. If GTX480 would have the same Cuda cores and Frequancy I will bet they would have the same performance per mm.
Click to expand...

Almost the same, GF110 has some cosmetic improvements to z-culling engine and it filters some texture formats two times faster then the predecessor.
 
purefun1965 said:
Yes this is true in a dying pc gaming market, The performance crown is minor compared to oem sales.
point is pc gaming is a moot point. crysis2, da2 and others have shown we are not a percentage that merits concern. This is a business we are a small part of that.
Click to expand...

Not this crap again. You are incorrect. First of all Crysis 1 never sold well to begin with. Crysis 2 may not be the game you want it to be and yes we are a fraction of the console kiddies but we are here to stay.

You are like one of those guys holding a sign saying judgement day is almost here. There have been many of you before for years and years and there will be many more to come after you. Holding the sign does not make it true.

There are plenty of PC games that are selling well. The biggest issue for big game developers is that many of the games are social and indie games.

Do not think that PC AAA titles are not turning a profit though, they have many benefits from the PC that they miss out on in a console. They have no cost associated with hardware license rights like xbox and ps do, they have very little cost for a digital distribution model - which brings me to another point that often in provided numbers digital distributions like steam and d2drive are not counted and PC has a much larger base of this than xbox and ps. It only takes a fraction of the sales to make profit on pc.

There are still games that cater to the PC and do very well. Many blizzard titles: sc2, wow, etc; Battlefield 3, Diablo 3, The new star wars MMO, warhammer 40k series..

Is it true that Bioware has started creating games for consoles as well as the PC? Yes, however it is assinine to say that the changes made to those games were only based on consoles. You see - DA sold quite well on the consoles and was essentially the same as the PC version. Therefor why blame changes to DA2 on consoles? Crysis did not sell well to begin with, no one bought the game. Of course they are going to change up their strategy. They found that when you make a game that people cannot play at high settings with high end graphics cards - no one buys your game. So they turned their focus away from graphics. Are there carryover graphical downgrades because of the consoles? Yes there are BUT they are releasing the SDK and dx11 - as DA2 did too so to say they don't care at all about the PC is silly.

The only game series I can say has shown they do not care about the PC community is MW2. Which was overrun by hackers and cheaters and never did anything about it.

PC gaming is here to stay. The markets will shift back and forth. Delevopers will come and go. We are here to stay.
 
digitaldurandal said:
Not this crap again. You are incorrect. First of all Crysis 1 never sold well to begin with.
Click to expand...
Not this crap again. You are incorrect. First of all, Crysis 1 is one of the best selling PC games of all time.

http://en.wikipedia.org/wiki/Crysis

By May 2010 the game has sold over 3 million units (and its standalone expansion about 1.5 million units)
Click to expand...
That said, Crysis 2 is doing spectacular sales
http://www.destructoid.com/crysis-2-huge-success-xbox-360-dominates-sales-197396.phtml
 
digitaldurandal said:
There are plenty of PC games that are selling well. The biggest issue for big game developers is that many of the games are social and indie games.

Do not think that PC AAA titles are not turning a profit though, they have many benefits from the PC that they miss out on in a console. They have no cost associated with hardware license rights like xbox and ps do, they have very little cost for a digital distribution model - which brings me to another point that often in provided numbers digital distributions like steam and d2drive are not counted and PC has a much larger base of this than xbox and ps. It only takes a fraction of the sales to make profit on pc.

...

PC gaming is here to stay. The markets will shift back and forth. Developers will come and go. We are here to stay.
Click to expand...

Well put.
 
^^ I don't actually see a sales figure there only some link to videogamer that also doesn't have a sales figure.
 
If the 7 series is really going to be good. then ill ditch the idea of getting a 6950 and wait for the 7950 or 7970. So a May Production means they will release sometime in the summer? like June or July??
 
Last edited:
So there is 1, perhaps 2 "sources" that said something about a may production. Now thats spreading around to the FUD sides and then to this forum?


28nm "whatever" isnt even here in force, not sure if its here at all. Maybe that should be focused on some more.
 
nvidiaintelftw said:
If the 7 series is really going to be good. then ill ditch the idea of getting a 6950 and wait for the 7950 or 7970. So a May Production means they will release sometime in the summer? like June or July??
Click to expand...

I think this is too optimistic. I expect we'll see the 7970 somewhere from mid August through to mid September.
 
AtenRa said:
Since the architectures you comparing are so much different and 55nm to 40nm didn’t have 100% transistor density as 40nm to 28nm will have, the comparison to 4870 vs Barts (die size, node process) and Barts vs 6970 (die size, same process) will not get you any ware in relation to 6970 vs 7970.

If you want to make an analysis from 4870 onwards for High End chips you really need to take out of the equation the Barts architecture. So in order to come up with a valid conclusion, we have to compare 4870/90 to 5870 and 6970 to 7970.
Click to expand...

Relative transistor density is simply another variable to add to the equation I've presented. It does not make the comparison invalid, like you keep trying to say.

And... have you even been reading what we've been saying? We are looking at 4870 to 5870 to see what 6970 to 7970 can be. But we're ALSO looking at R770 to Barts and Barts to Cypress to give us a more complete picture. Jesus Christ.

AtenRa said:
4890 increased the performance by increasing the core frequency from 750MHz (4870) to 850MHz (4890) but they kept the same die size. So the increase came from the frequency and not by the design.

They took off silicon and features (no FP64, different memory controller) from 5870 and they created the 6870. So in order to raise the performance per mm they lost features and that don’t make a valid comparison. If 5870 didn’t support FP64 and had the same Front End with 6870 they could have the same performance per mm.

GF110 increased performance with a frequency boost from 700MHz(480) to 772MHz(580) plus 32 more Cuda cores. If GTX480 would have the same Cuda cores and Frequancy I will bet they would have the same performance per mm.
Click to expand...

AMD tweaked the silicon for R790 in order to get increase the clockspeed. Nvidia did the same to GF100. And keep power consumption manageable. So those two examples work to show efficiency.

Why does losing the features in 5870 to Barts make the comparison invalid? IT DOESN'T. Jesus. The simple fact they lost features is YET ANOTHER VARIABLE VALUE TO CONSIDER. Seriously... how do you not see it?

Man I'm done with this.
 
nvidiaintelftw said:
If the 7 series is really going to be good. then ill ditch the idea of getting a 6950 and wait for the 7950 or 7970. So a May Production means they will release sometime in the summer? like June or July??
Click to expand...

It's looking like AMD will make an announcement at E3... for both HD7970 and FX 8000 Bulldozer.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top