Another AMD leaked slide shows upcoming BARTS is based on Radeon 5xxx architecture...

[Qianglong]

It seems our Chinese friends are rife with next generation AMD leaks and the latest one is here:
http://www.chiphell.com/thread-123528-1-1.html

It seems that dude has a friend who works at an OEM and got briefed by AMD on the upcoming products. From google translation, he is saying AMD had to scrap the plans for a brand new architecture because TSMC cannot get their 32 nm in order and they are forced to release an "upgraded" version of Radeon 5xxxx series just to compete with the GTX 460 which is giving AMD tons of headaches...

Anyone here thinks it makes sense? So isn't this a complete contradiction to what AMD has said that the 6000 series will be a completely new architecture - http://forums.anandtech.com/showthread.php?t=2103730

 

[Ancalagon44]

The next radeon has been in the works long before the GTX460 was released or even announced. So I dont think that can have affected it - by the time AMD say how good it was, it would be way too late to change anything about the upcoming release. At most it will make AMD make sure to release it soon and priced appropriately, but thats it. Wouldnt force a change of strategy.

The 32nm delay thing was the original rumour, looks like it was probably correct all this time.

 

[brybir]

It seems our Chinese friends are rife with next generation AMD leaks and the latest one is here:
http://www.chiphell.com/thread-123528-1-1.html

It seems that dude has a friend who works at an OEM and got briefed by AMD on the upcoming products. From google translation, he is saying AMD had to scrap the plans for a brand new architecture because TSMC cannot get their 32 nm in order and they are forced to release an "upgraded" version of Radeon 5xxxx series just to compete with the GTX 460 which is giving AMD tons of headaches...

Anyone here thinks it makes sense? So isn't this a complete contradiction to what AMD has said that the 6000 series will be a completely new architecture - http://forums.anandtech.com/showthread.php?t=2103730

Well, what should tip you off is that the 460 is not giving AMD "tons of headaches". AMD is or will be releasing an entire product line, most of which do NOT compete with the 460 at all. People in this forum sometimes get to thinking that the enthusiast market is the only one out there, and that when we perceive a card "doing well" we immediately begin to assume it must be the case and therefore the other companies competing against it must be worried. We dont know channel supply and demand, ASP's, production costs etc or even raw sales...best we can do is guess off of what little we know. That is why "feelings about a card doing really well" end up becoming a "fact" here in the forums, lack of information.

Further, it takes years to design a GPU, so whatever the 6XXX is going to be, it was planned a while ago. Even if it ends up being a mashed up hybrid of 5XXX tech and 6XXX tech, it was planned, designed and finalized quite some time ago. My point being is that the 6XXX cards were designed and finalized probably before AMD knew much of anything about the 460 at all, therefore saying the 6XXX series is a reaction to the 460 is probably not true.

 

[GaiaHunter]

That is why "feelings about a card doing really well" end up becoming a "fact" here in the forums, lack of information.

For example, the forum had a feeling AMD was doing good with the 4800 series even if the data available at the time wasn't showing that immediately but in the long run we have seen AMD position improved.

The GTX 460, on the other hand, even if it is a great card, especially from the price/performance perspective, might suffer from being to late to the market (I guess a similar situation to the X1800 series) and finding that market a bit saturated with 5850s.

After all the GTX 460 arrived almost 9 months after. Not only that, but the market had a gold age of price/performance for the 2 years before that with the 4800 vs GT200/9800GTX+ price wars.

If AMD gave GTX 460 time, the GTX 460 could indeed do good things for NVIDIA, but apparently AMD is not allowing that to happen (and I don't mean this 6000 series spawned out of thin air because of the GTX460, it was bound to happen sooner or later).

 

[thilanliyan]

Could this be some of the misinformation being spread by AMD to find leaks?

 

[GaiaHunter]

If you look closely you will see that powerpoint is in edit mode (couple of words have a red underline) so AMD could be changing specs on the fly to catch leaks.

 

[Keysplayr]

What Thilan said. Nothing can be trusted right now due to AMD's rumoured false info to weed out the leaks.

 

[Ben90]

brybir hit the nail on the head. GPU designs are planned out years and generations in advance. I find it comical yet agitating when people say X company made Y card in response to X2 companies Y2 card.

GF100 was in designing stages well before GT200 was even on the market.

 

[Genx87]

Well they can do redesigns rather quickly if needed. AMD redesigned the 5000 series due to process issues. They cut some functionality out of it.

 

[extra]

What Thilan said. Nothing can be trusted right now due to AMD's rumoured false info to weed out the leaks.

Yep and with AMD, Nvidia, Intel, etc, you always hafta wonder where leaks end and marketing begins.

 

[RussianSensation]

Does anyone know what "Full Cypress AA" mean in the slide?

 

[Qianglong]

So you guys think this is some kind of hidden witch hunt to sort out the leaks where different OEMS are given different numbers?

 

[Kenmitch]

brybir hit the nail on the head. GPU designs are planned out years and generations in advance. I find it comical yet agitating when people say X company made Y card in response to X2 companies Y2 card.

GF100 was in designing stages well before GT200 was even on the market.

This is a true statement but on the other hand is not so true when you step back and look at the big picture.

The part that makes it true is it does take a certain amount of time to plan out and design a gpu. It takes time to make pre-production samples for testing. Most of the time I'd think some faults are detected and you kinda start over in the design area to fix them. Once you get it up to snuff is the end of the true statement.

The part that makes it wrong is as follows " I find it comical yet agitating when people say X company made Y card in response to X companies card. "

You might ask why and I'll tell you why. When you are in competetive mode and your opponent releases a product you have something to shoot for a goal a person could call it. Your future released products are not set in stone yet. A company may choose to push performance down the line to one up the opponent or to be more competetive in a certain price point.

 

[Elixer]

Why bother to speculate ?
These kinds of tactics have been used in the past, and they will be continued to be used in the future, since it creates doubt. Then the fence sitters tend to go with what is out now, instead of waiting.

 

[Idontcare]

Well they can do redesigns rather quickly if needed. AMD redesigned the 5000 series due to process issues. They cut some functionality out of it.

No two redesigns mean the same thing.

Technically you could classify a mere stepping increment involving back-end wiring changes as a "redesign", not that many would but you could and still argue that you are right.

But usually the moniker "redesign" is reserved for when things literally go back to the drawing board.

Bulldozer is a redesign, they started with 45nm objectives and tossed it to go back to the drawing board for a 32nm intersection point. (not my speculation, they have stated this)

The work they did on Evergreen following the reworking of RV740 is not nearly as drastic. Look how long it took Nvidia to iterate through four steppings on Fermi, and they were all "A" steppings (meaning BEOL changes, not FEOL which is what a redesign would entail)...no way we could argue that AMD did a redesign on Cypress in under a year.

 

[Seero]

Well, what should tip you off is that the 460 is not giving AMD "tons of headaches".

I bet someone can bring in the sales chart from July to show now the sales of 460 impact other video cards.

AMD is or will be releasing an entire product line, most of which do NOT compete with the 460 at all.

460 penetrated the every discrete video card markets. Unless the entire 6xxx series are IGPs, it is going to compete with 460.

People in this forum sometimes get to thinking that the enthusiast market is the only one out there, and that when we perceive a card "doing well" we immediately begin to assume it must be the case and therefore the other companies competing against it must be worried.

The irony is many in this forum believed that 6xxx series is going to flip Nvidia's boat because Nvidia doesn't seem to have a counter product. Now you are saying that it is perfectly fine for AMD not to have a counter product against 460.

We dont know channel supply and demand, ASP's, production costs etc or even raw sales...best we can do is guess off of what little we know. That is why "feelings about a card doing really well" end up becoming a "fact" here in the forums, lack of information.

You are trying to turn "facts" into "fiction". You are saying that it isn't true that 460 is selling well, much better than any other GPUs in the market atm.

Further, it takes years to design a GPU, so whatever the 6XXX is going to be, it was planned a while ago.

The plan takes years, the execution does not. Nvidia planned Fermi for a very long time, which is not the problem. The problem was the 6 month delay with the execution. In fact, Nvidia planned several more 2xx cards that never made it to the market. Because of this, Nvidia is said to be dropping the balls. To consumers, we never knew its existence. To other companies that had planned to use those chips, are big hits.

Even if it ends up being a mashed up hybrid of 5XXX tech and 6XXX tech, it was planned, designed and finalized quite some time ago.

This particular rumor is about TMSC not able to deliver, which is not something within the original plan of AMD.

My point being is that the 6XXX cards were designed and finalized probably before AMD knew much of anything about the 460 at all, therefore saying the 6XXX series is a reaction to the 460 is probably not true.

In case you don't know, 480/470 is the first generation of Fermi, which is planned to counter the Cypress series. Nvidia stated that the delay of the first generation of Fermi will not affect the original plan time of the 2nd generation of Fermi, and 460 is the 2nd generation of Fermi. On the other hand, the 6xxx series from AMD is planned to counter the 2nd generation of Fermi.

I am not saying this rumor has any truth to it. However, it will be easier to believe that TSMC still hasn't recover from the precious problem then to believe that everything is fixed.

If AMD planned to release 5xxx as 40nm chip and 6xxx as 32nm chip, then it is not unbelievable that 6xxx will have problems as TSMC stated that they are going to skip 32nm and go directly to 28nm. If TSMC is telling the truth, then AMD may have a problem following the plan, which made this rumor plausible.

 

[RussianSensation]

Nvidia stated that the delay of the first generation of Fermi will not affect the original plan time of the 2nd generation of Fermi, and 460 is the 2nd generation of Fermi.

No it isn't. GF104 is 1st generation of Fermi, hence GF1xx where GF stands for Graphics based on Fermi architecture.

I see this mixed up once in a while. GF100/104 and 106 are all the same architecture (i.e., same generation) with slight tweaks in 104/106. Just like G80 and G92 is the same architecture with tweaks in the latter. In which case the 2nd generation of G80 architecture is actually GT200.

Fermi 2 is what we expect NV to counter HD6000 series with in 2011.

 

[Seero]

No it isn't. GF104 is 1st generation of Fermi, hence GF1xx where GF stands for Graphics based on Fermi architecture.

I see this mixed up once in a while. GF100/104 and 106 are all the same architecture (i.e., same generation) with slight tweaks in 104/106. Just like G80 and G92 is the same architecture with tweaks in the latter. In which case the 2nd generation of G80 architecture is actually GT200.

Fermi 2 is what we expect NV to counter HD6000 series with in 2011.

I am not making this up. The code name Fermi refers to an architecture. If GT100 is the first generation(chip) based on this architecture, then gt104 is the second generation(chip) based on this architecture. While the word "Generation" is arguable, but the fact stays. GF100 was planned to be released before the end of 2009 and GF104 at the 2nd quarter of 2010. GF100 was 6 month off the mark but not GF104.

 

[edplayer]

using your logic the gf106 is now 3rd generation and the gf108 is the 4th. A couple more generations and Nvidia will be at NES level graphics


I think their 2nd generation products will include a replacement for the GTX 480 so it has not come out yet and will not this year. Maybe next year...

 

[Caribbean Geek]

You are trying to turn "facts" into "fiction". You are saying that it isn't true that 460 is selling well, much better than any other GPUs in the market atm.

I doubt it, if its selling well, why is dropping in prices in a weekly base? You don't drop prices of a sucessful product if its selling well, it won't do good to your profits and the company image. Don't state your opinion as granted facts, do you have proof of the GTX 460 is selling great?

Back on topic, if its true that Bart XT is based on Cypress, something to improve performance were done, because the HD 5770 shares almost the same specs as the Barts PRO and its quite behind of the GTX 460 768MB in terms of performance, faster IPC?

 

[cfedu]

using your logic the gf106 is now 3rd generation and the gf108 is the 4th. A couple more generations and Nvidia will be at NES level graphics

lol!!! I almost pooped my pants

 

[brybir]

I bet someone can bring in the sales chart from July to show now the sales of 460 impact other video cards.

Yes, sales of one product affect the sales of other products. That is not in dispute. The OP said that the 460 specifically was causing problems, to which I said we dont really know that because we don't have that information. Yes, we could bring in sales numbers, but until someone does, we do not have that info, hence, my point. If you have those numbers, please share, otherwise, you have no factual idea either.

460 penetrated the every discrete video card markets. Unless the entire 6xxx series are IGPs, it is going to compete with 460.

The 460 competes in one product segment. It does not matter to people that want the best of the best and will spend $500 on it, as they are not in the market for a 460. As with all things in an economy, your completion is for people that would buy your product, not those that would NOT buy your product.

The irony is many in this forum believed that 6xxx series is going to flip Nvidia's boat because Nvidia doesn't seem to have a counter product. Now you are saying that it is perfectly fine for AMD not to have a counter product against 460.

I have no idea where this came from. My only point, and the only thing I said, was that the forums "feelings about the success of cards" has very little correlation to what is actually happening.

You are trying to turn "facts" into "fiction". You are saying that it isn't true that 460 is selling well, much better than any other GPUs in the market atm.

That is one heck of a read on what I am saying. Again, we DO NOT have any factual information about sales, demand, supply etc and have no way to know how well any card, including the 460, is selling.

Your statement that the 460 is selling better than any other GPU must be based on something, although I have not seen it publicly released how many IC's AMD or Nvidia are selling. Care to share your source?

The plan takes years, the execution does not. Nvidia planned Fermi for a very long time, which is not the problem. The problem was the 6 month delay with the execution. In fact, Nvidia planned several more 2xx cards that never made it to the market. Because of this, Nvidia is said to be dropping the balls. To consumers, we never knew its existence. To other companies that had planned to use those chips, are big hits.

I think you are agreeing with me. The design for the 6XXX series has been in place before the 460 came to market and were probably designed around the same time. Check out IDC's post about respins and redesigns, as I think you may have the two concepts a bit backward.

[

This particular rumor is about TMSC not able to deliver, which is not something within the original plan of AMD.

How do you know this? Again, you are asserting this as fact, but as far as any of us can tell, it is just your opinion. You do not know what AMD's plans were with this design. In any event, they certainly had backup plans and would have known about node problems before the general public (i.e. us).

[

In case you don't know, 480/470 is the first generation of Fermi, which is planned to counter the Cypress series. Nvidia stated that the delay of the first generation of Fermi will not affect the original plan time of the 2nd generation of Fermi, and 460 is the 2nd generation of Fermi. On the other hand, the 6xxx series from AMD is planned to counter the 2nd generation of Fermi.

I am not saying this rumor has any truth to it. However, it will be easier to believe that TSMC still hasn't recover from the precious problem then to believe that everything is fixed.

If AMD planned to release 5xxx as 40nm chip and 6xxx as 32nm chip, then it is not unbelievable that 6xxx will have problems as TSMC stated that they are going to skip 32nm and go directly to 28nm. If TSMC is telling the truth, then AMD may have a problem following the plan, which made this rumor plausible.

I think other posters have explained that your statement is false. By that logic, the 5770 cores are a different generation than the chips that power the 5830, 5850 and 5870. They are somewhat different *designs* but they share a common architecture (building blocks) as their larger brothers. Same goes for the Fermi family.

 

[OCGuy]

Wow the (mis)information on 6XXX is flying everywhere!

I think that bogus vantage score started it, and everyone else is just piling on, trying to get some page hits.

 

[Seero]

I doubt it, if its selling well, why is dropping in prices in a weekly base?

This is your doubt, your guess, but not the reason of the price drop.

Price drop can be the result of the reduction of the cost of BoM. It isn't a myth Nvidia wants to regain the number of people using Nvidia product over the immediate revenue. In fact, since the drop of price increases demand. If supply is not a constraint, a price adjustment can lead to better profit.

You don't drop prices of a sucessful product if its selling well, it won't do good to your profits and the company image.

This is your logic reasoning, not facts. See above.

Don't state your opinion as granted facts, do you have proof of the GTX 460 is selling great?

Steam Hardware
Look that the number of people in percentage who uses 460, and the change between July and August.
NVIDIA GeForce GTX 460 +2.86%, from 0.57% to 3.43% in a single month. Compare this with any other cards and you will see the growth of 460 surpasses all other cards.

I understand this isn't a proof, but more like a fact which can be used to support my statement. Can you now get a piece of fact or evidence indicating 460 isn't selling well?

Back on topic, if its true that Bart XT is based on Cypress, something to improve performance were done, because the HD 5770 shares almost the same specs as the Barts PRO and its quite behind of the GTX 460 768MB in terms of performance, faster IPC?

I don't recall I stated anything about speed.

 

[Caribbean Geek]

This is your doubt, your guess, but not the reason of the price drop.

And your reasoning is right?

Price drop can be the result of the reduction of the cost of BoM. It isn't a myth Nvidia wants to regain the number of people using Nvidia product over the immediate revenue. In fact, since the drop of price increases demand. If supply is not a constraint, a price adjustment can lead to better profit.

That's your logic reasoning, not facts. Do you have proof of such claims? I somehow doubt it.

Steam Hardware
Look that the number of people in percentage who uses 460, and the change between July and August.
NVIDIA GeForce GTX 460 +2.86%, from 0.57% to 3.43% in a single month. Compare this with any other cards and you will see the growth of 460 surpasses all other cards.

You downplayed the steam survey when I used as a reference of AMD's marketshare and now suddenly you use it as a fact to support your statement, pot meet kettle?

I don't recall I stated anything about speed.

You don't, that was just my apportation to the thread.