AMD: R9xx Speculation

If this talk turns out true I can see how that can be compared to Nvidia intensive renaming.
It's a new product, with different arch coming at a different price...
it's about positioning... :LOL:
 
That's a fair point, starting with 38xx. But it hasn't been long enough to establish a trend. The 9800, x800 and x1800 before them were most definitely positioned higher than the 3870 was, the latter being more a consequence of poor performance than any strategic positioning. I don't think anybody has a problem with 6800's returning to the price points of old though, there's only an issue if it's slower than the 5800 series....



:rolleyes: Wasn't RV670 a change in strategy in order to fight NVidia with more but cheaper chips?
Here 6800 will be the biggest anomaly, simply because it will not be the highest performing chip...
 
Come now, we wouldn't want B3D to get a reputation for suppressing dissent against AMD for the same things Nvidia was mutilated over. Fair and equal nerd rage please :)

If and when it happens, sure. But at this point, when we have no idea how fast Barts is, let alone what the rest of the lineup is going to look like, it's just noise.
 
:rolleyes: Wasn't RV670 a change in strategy in order to fight NVidia with more but cheaper chips?
Here 6800 will be the biggest anomaly, simply because it will not be the highest performing chip...

It's the sweet spot, stupid!*


(it's the economy, stupid!)
 
Sometimes commercial realities have to manifest themselves.

They do not necessarily have to manifest in deceptive marketing though.
While naming mobile solutions exactly the same as higher performance desktop parts may seem like a good marketing idea short term, I'm pretty sure that those who buy what they believe is higher performance only to find that they didn't get what they thought aren't going to be happy.

Arguing that deceptive marketing is par for the course in the industry and thus OK says quite a lot about the industry and how it regards its customers. We observe and take note.

The naming of mobile parts always struck me as particularly ill advised, because it can sway consumers both ways. Just as it can make a consumer stretch to a higher end part, it can also make them go lower since they don't feel they need the performance or cost of the highest end. So it can act to downsell just as well as upsell, and neither of these two categories are going to be happy to learn that they didn't get what they thought. I find it difficult to believe that the market gains really outweigh the longer term losses.
 
@ neliz:
I don't know what you are smoking but definitely... I don't want it.

RV790 is 282 mm2 and I think it lies in that "sweet spot". What is the size of Barts? Do we expect so big chip, or will it underperform so badly? I think the only people that care about that freaky "sweet spot" are the guys at AMD themselves and no one else. I am looking forward to the upcoming PR slides and how AMD will manage to convince people that "sweet spot" is responsible for their so slow 6800 series. :LOL:

I think there will be a lot of cases like this:

There is a big shiny box in the store and it is written with big letters 6800 series. The average Joe buys that and returns back home, opens that box and starts his PC with the new card (of course with high expectations). But his friend next door has a 5800. And they compare their results. The average Joe will be terrified that his new card shows lower FPS. :LOL::LOL::LOL:
 
:rolleyes: Wasn't RV670 a change in strategy in order to fight NVidia with more but cheaper chips?

Change in strategy? HD 3870 was priced at $200 because that's the performance level it provided compared to Nvidia's parts. Unless of course you believe that R600's poor performance was a strategic move on ATi's part.

If and when it happens, sure. But at this point, when we have no idea how fast Barts is, let alone what the rest of the lineup is going to look like, it's just noise.

Yes, agreed. That's why everyone is saying that it will be misleading IF performance isn't up to par. There's no need for pre-emptive defense until we have the details.
 

Hmm, anyone notice the GT which I take will be the 30?

41596155.jpg
 
Well x800 has meant a lot of different things these past few years. It has been a cheap part, it has been a middle range part and it has been an expensive part. It does make sense moving forward to make it consistant generation to generation. So I understand now why they would move Barts to that level even though it may be weaker than a previous generation.

The price in itself is a positioning signal. I don't think people will be that blind as to think that it will be say nearly half the price of a 5870 and offer +50% additional performance. Beyond this the non enthusiast people don't upgrade their chips every generation. I don't think there are many people who would be misled by this and the end justifies the means if the entire lineup can be more consistant from here on out generation to generation.
 
I hope it can help to stop the pointless discussion...

With only 2 exceptions, names of the ATi's products launched in last 7 years, always corresponded to die size:

101 - 150mm² => x6xx
  • RV830 (Radeon HD5500): 104 mm²
  • RV635 (Radeon HD3650): 120 mm²
  • RV530 (Radeon X1600): 150 mm²
  • RV630 (Radeon HD2600): 150 mm²
  • RV730 (Radeon HD4600): 150 mm²

151 - 190mm² => x7xx
  • RV410 (Radeon X700): 156 mm²
  • RV840 (Radeon HD5700): 170 mm²

191 - 340mm² => x8xx
  • RV670 (Radeon HD38x0): 192 mm²
  • R430 (Radeon X800XL): 240 mm²
  • RV770 (Radeon HD4800): 256 mm²
  • R420 (Radeon X800XT): 281 mm²
  • RV790 (Radeon HD4890): 282 mm²
  • R520 (Radeon X1800): 288 mm²
  • RV870 (Radeon HD5800): 331 mm²

341 - 500mm² => x9xx
  • R580 (Radeon X1900XT): 342 mm²
  • R580+ (Radeon X1950XT): 34x mm²
  • R600 (Radeon HD2900): 420 mm²

The only two exceptions in this system were two "pipe-cleaner" parts, which were used for testing of new manufacturing process and because of that, they were significantly smaller, than other GPUs of that family: RV570 (80nm), RV740 (40nm).

We expect Cayman to be sized around 400 mm² and Barts under 300 mm². What name should these two parts use? I think AMD isn't going to change anything. Maybe the users misinterpretted naming convetions.
 
Then why not name it 6700?
Oh, wait, then people won't be fooled because 8>7 ...

Anything that AMD does is good :rolleyes:.
In case 6800<=5800 this will be cheating and taking advantage of simple uneducated people.
Good for business? Maybe.
Tells much about some people's bias.

We will see at launch what the real reason is, but there are more reasons than yours for why it would be labelled an 8.

For example, Barts is going to be a bigger chip than Juniper. It is going to yield a lot worse probably. Common sense says it's going to be more expensive than Juniper and for the same reasons it will be cheaper than Cypress.

So AMD is left with the choice of "overpricing" a 7 card or "underpricing" an 8 card. There is no real winning method here as "overpriced" cards would just make people would say AMD was price gouging because of their lead.

That's not to say renaming it an 8 is the lesser of the two evils, however it's easily the lesser in AMD's case. I'm thinking this wasn't a lightly taken decision.
 
With only 2 exceptions, names of the ATi's products launched in last 7 years, always corresponded to die size

All you did there was come up with arbitrary die size ranges to fit reality. If you did it based on "fastest chip" or some other more concrete criteria your argument falls apart. Besides, aren't we talking about product positioning (and the branding that communicates it). Obscure details like die sizes aren't a factor.

There's really no way to spin it as not being misleading if 6870 < 5870 so it's a waste of time trying. The good thing is that once it's done it should be a one time deal and things will again be consistent going forward.
 
Whats funny I think is that if they had simply named the 5970 the 5870X2, all this wouldn't be an issue... people would just think... "oh sweet, they're making a 6970 single GPU? That must mean they're returning to the high end larger GPU designs!" and everything would be easier to swallow
 
All you did there was come up with arbitrary die size ranges to fit reality.
Try to check it once again. These are not hand-picked exaples, but all >100mm² GPUs launched in past 6-7 years. Die size definitely is the factor. Or do you consider to be more likely, that I posted a list of ~20 coincidence examples? :smile:

This list also explains, why was R600 called HD2900 despite it was barely waster than previous generation and name like HD2700-2800 would reflect it's performance more adequately. Performance level seems to be irrelevant...
 
AnarchX: It would happened, because it would be pushed by another 32nm cheaper-to-made product. But 32nm process was cancelled and 40nm process is still expensive. It was the most expensive TSMC's process at launch and its price even raised since the launch. Price of memory modules raised, too. It is no wonder, that the HD5800 isn't cheap.
Memory is the same like on $129 HD 5750. And the die is smaller than a GF100/GF104, which are sold at lower prices (except GTX 480 SKU). So there should be very high margin or the contract with TSMC is much worser than NVs one.

Cypress (aka RV870 in 2008 drivers) is a good RV770-replacement performance- and feature-wise, but on the price-side it looks very worse. But its not only AMD failing in this way, NV has the same problem, although they are now keeping up to a normal market expectation with GF104/GTX460. You have to remember, that in early 2009 (~18 months ago) you could buy a HD 4870/90/GTX 260/275 card for <$199.
So if we are now going to a HD 6870 (with summarized performance below HD 5870) I get an uncomfortable feeling.:???:

So the hope is, that 28nm at GF/TSMC runs much better, that we can go back to a yearly >50% performance increase in the important market segment of ~$199, which was there in the last 10 years, except 09 to 10.


.
 
Back
Top