NV: Everything under control. 512-Fermi may appear someday. Yields aren't under 20%

[Matrices]

Please, tell me where someone said that.

Look two posts above:

I don't think ATI ever imagined using their $260 5850 to compete with a $350 GTX470 because they thought GTX470 would compete with their 5870 and GTX480 would compete with their dual-gpu card.

 

[Matrices]

AMD has said themselves, they plan to make their highest end single GPU's for the 'performance' crowd. For ultra highend enthusiast area they plan on using multiple 'performance' GPU's on a single card. Nothing new here...

That amounts to saying nothing. Obviously, that is true of every generation and of either company ever since dual GPU cards were invented.

Again, read a few posts above. The point I am contending is this silly notion that ATI assumed the GTX 480 would compete with the ATI 5970, and that since the 480 didn't do that, it is a disappointment.

 

[yh125d]

That amounts to saying nothing. Obviously, that is true of every generation and of either company ever since dual GPU cards were invented.

Again, read a few posts above. The point I am contending is this silly notion that ATI assumed the GTX 480 would compete with the ATI 5970, and that since the 480 didn't do that, it is a disappointment.

It's well known that ATI doesn't intend to have their top single GPU card to be directly competitive with NV's top single GPU card. If you can't get that through your brain, it's your problem - not ours.

 

[Matrices]

It's well known that ATI doesn't intend to have their top single GPU card to be directly competitive with NV's top single GPU card. If you can't get that through your brain, it's your problem - not ours.

Who cares about this? Allow something to get through your own brain: for the third time, what I am contending is the notion that ATI expected the GTX 480 to compete with the 5970, which is what someone had said.

 

[Voo]

Who cares about this? Allow something to get through your own brain: for the third time, what I am contending is the notion that ATI expected the GTX 480 to compete with the 5970, which is what someone had said.

I think you take "compete" a little bit too precise. The 5970 surely is faster than a 480 but at the same time also more expensive, so they compete somehow.

Though I think it's not too far fetched to argue that they probably were quite surprised when the 60% larger chip didn't completely and uterly trashed their own...

 

[v8envy]

Who cares about this? Allow something to get through your own brain: for the third time, what I am contending is the notion that ATI expected the GTX 480 to compete with the 5970, which is what someone had said.

It's reasonable to expect ATI decision makers to position their flagship product against the competition's flagship product. It's less reasonable to expect something else. So, yes, the 5970 was targeted against the 380, 480, 580 or 5800 Ultra II -- whatever ATI strategists reasonably expected NV to have as their top SKU.

 

[Lonyo]

Who cares about this? Allow something to get through your own brain: for the third time, what I am contending is the notion that ATI expected the GTX 480 to compete with the 5970, which is what someone had said.

I think ATI could have expected that to be the case.
Given the scaling of the GTX480, if you make it a 512 shader part at 750MHz (what ATI expected from the GTX480), that's 14% more performance (since scaling seems pretty much 100% with core, memory doesn't have much impact).
That would make the GTX480 pretty competitive with the HD5970.

The HD5870 "competes" with the GTX480 despite being 10% or so slower on average. The HD5850 competes with the GTX470 despite being 10% or so slower.
To imagine that a 14% faster GTX480 would compete with an HD5970 doesn't seem far fetched.

I don't agree with that assertion, but it certainly is plausible.
If an HD5870 competes with a 10~15% faster GTX480 (the real one), then a 512 core/750MHz GTX480 which is around 15% slower than an HD5970 also competes with it.
Of course, you have to assume a GTX480 with 512 cores at 750MHz, which IIRC AMD at one point did... (it was on a slide somewhere I think).

 

[yh125d]

Who cares about this? Allow something to get through your own brain: for the third time, what I am contending is the notion that ATI expected the GTX 480 to compete with the 5970, which is what someone had said.

And your contention is ridiculous, because most people realize that that was their intent precisely. They wanted a dual GPU card to be their flagship at $500+, to compete with NVs $500+ flagship single GPU

http://images.anandtech.com/reviews/video/ATI/4850/amd3.png

 

[happy medium]

Jeeeez it was just a thought or possibility, I didn't mean to get your guys blood boiling.

When all the refreshes are done and the leakages,heat,and power consumption problems are fixed, you guys honestly think a refreshed 58xx series will compete with the gtx series?

Sorry for sounding like a fanboy (which I'm not),but If you remember last generation, performance was dominated by nividia, if fact the generation before that too.

When the die shrink comes this winter , Nvidea will have a cooler ,less heat and more importantly faster cards accross the board.
I think ATI shot it's load already with the 5870 at 1000 core, the gtx will and does scale better with overclocking.

Wheres the gtx 595 960 sp's 22nm love?

ATI needs the 6xxx series to compete performance wise, simple as that.
And AMD will respond like it always does BY LOWING PRICES to compete.
The other option, like others have stated, is they will use dual gpu CROSSFIRE cards to compete, but that becomes exspensive, hot and loud. Sound familiar?

Edit: If you notice I have a ATI card in my rig, why because I got it for 85$ and its a great overclocker. I really wanted a 5830 (and later crossfired 5830's) but we all know why that idea got scrapped? . Now I'm looking for the gtx 460 for a upgrade. I'm a little to cash strapped for a 300+ upgrade. I have no real name preference when buying a card. So please keep the kid like fanboy comments to a minimum.

 

[RobertR1]

 

[Grooveriding]

What is surprising about this coming from NV's own mouth ? Their stock has been falling and they've been downrated, hence some damage control is in order.

They need to catch up somehow to ATI's six month development lead. As per ATI's stated scheduling the 6XXX series is coming end of this year, no iteration of Fermi, 512 shaders or not is going to compete. The dual gpu 6XXX series will be a massive performance contender until nv can get their next series out or put Thermi on 28nm.

 

[Genx87]

What is surprising about this coming from NV's own mouth ? Their stock has been falling and they've been downrated, hence some damage control is in order.

They need to catch up somehow to ATI's six month development lead. As per ATI's stated scheduling the 6XXX series is coming end of this year, no iteration of Fermi, 512 shaders or not is going to compete. The dual gpu 6XXX series will be a massive performance contender until nv can get their next series out or put Thermi on 28nm.

That is a lot of ifs. We really have no idea what the 6000 series will do and what process it will be on. The last I heard it was to be on the next process iteration. Who the hell knows when that will be ready. And what do you meant by development lead? These projects start years ago. Nvidia and ATI already habr successors in development for this and next generation. I have a feeling the 460 will help a lot in total units shipped for DX11 chips on the green side once it hits the streets in June.

People also seem to forget if OpenCL takes off AMD needs to play catchup in compute GPU in a big way. It is possible to do so will require they develop as complex a chip as Fermi with all the problems a huge monolithic chip brings from a manufacturing perspective.

 

[edplayer]

When all the refreshes are done and the leakages,heat,and power consumption problems are fixed, you guys honestly think a refreshed 58xx series will compete with the gtx series?

It competes with it right now.

Sure, it loses in performance which to many is the most important metric, but there are others that people consider also. Price, is probably 2nd most important and to some heat and power. Couple all those and you will see AMD's product are very competitive. Enough so that the most recommended card you see currently is the 5850. Only when someone asks for the high end/money no object recommendation do you hear people say to get the GTX480 in SLi. But thats over $1000.


How often do people recommend a single GTX 480 or 470? You can hear crickets chirping louder than people claiming that.


As for you being a fanboi, you sound like one to me. Not because you like Nvidia (I currently am using them also). But because you only bring up their positives and not the negatives and vice versa for ATi. You are not objective.

When the die shrink comes this winter , Nvidea will have a cooler ,less heat and more importantly faster cards accross the board.

What die shrink? Apart from being laughed at now, you will be laughed at this winter when it doesn't show up.

I expect Nvidia to have a newer version of their cards (not a 512 shader, though that may come also) but I expect they would want to concentrate mostly on getting better yields. That may come with some better performance but I expect their efforts to concentrate more on getting yields higher than more MHz. Driver improvements will help also and Nvidia has a good track record there.

 

[edplayer]

Wheres the gtx 595 960 sp's 22nm love?

come on guy!

when do you think Nvidia will have a 22nm high end video card for sale? 2012? 2013?

 

[happy medium]

It competes with it right now.

Sure, it loses in performance which to many is the most important metric, but there are others that people consider also. Price, is probably 2nd most important and to some heat and power. Couple all those and you will see AMD's product are very competitive. Enough so that the most recommended card you see currently is the 5850. Only when someone asks for the high end/money no object recommendation do you hear people say to get the GTX480 in SLi. But thats over $1000.


How often do people recommend a single GTX 480 or 470? You can hear crickets chirping louder than people claiming that.


As for you being a fanboi, you sound like one to me. Not because you like Nvidia (I currently am using them also). But because you only bring up their positives and not the negatives and vice versa for ATi. You are not objective.







What die shrink? Apart from being laughed at now, you will be laughed at this winter when it doesn't show up.

I expect Nvidia to have a newer version of their cards (not a 512 shader, though that may come also) but I expect they would want to concentrate mostly on getting better yields. That may come with some better performance but I expect their efforts to concentrate more on getting yields higher than more MHz. Driver improvements will help also and Nvidia has a good track record there.

I did say heat, power cosumption problems? Other then that, I didn't think the gtx series had a problem. Did I miss something?

The ATI positives are ..............price per performance as usual?
But this time look at the percentage of perfromance you get for the price on the green side and it's really not to bad. Not like the gtx 280!

Its good for Nvidea that there skipping the 28nm process and going straight to 22nm. 40nm to 22 nm is a nice shrink , wouldn't you say? Say goodbye to the heat and power problems. Say hello to higher clocks and very good scaling.

Like I said ATI needs the 6xxx series to compete performance wise or some kinda dual card solution.

 

[happy medium]

come on guy!

when do you think Nvidia will have a 22nm high end video card for sale? 2012? 2013?

Last I read both chip manufacturers were going straight to the 22nm process. ATI and Nvidea.

So yea soon the next chips or refreshes.

edit: I was wrong there skipping 32mn and going right to the 28nm process not 22. My bad.

 

[edplayer]

price per performance as usual, yes

But that happens to be what MOST people consider to be the most important. Even you are using an $85 ATi card! The percentage of performance increase you get on Nvidia's side isn't bad like you say, but, to get it you have to spend $500.

Dropping down to 22nm should make a big difference but you are unrealistic in your expectations of when it will happen. I don't see how it will happen this winter. 2012 seems like it could be the earliest that we would see products for sale.

 

[SlowSpyder]

Happy Medium, why won't a refreshed 58xx card compete with the GTX series? They compete now afterall.

It's not like AMD is going to be stuck with the 40nm process and Nvidia is jumping to 22nm. If anything, it looks like AMD may have an advantage in process technology as they'll have both GF and TSMC to work with if they want. As of now Nvidia has said no to GF.

Even if both shrink their current parts to the exact same process, what would change? I believe power consumption will scale down for both AMD and Nvidia. This gives both camps the opportunity to go with the power savings or up clocks (generally speaking). What makes you think both AMD and Nvidia taking their current offerings to the same smaller process is going to give one the upper hand over the other? They should be at roughly the same proportion to each other as of now, at least that's how I understand it.

 

[SHAQ]

A 512 core Fermi at 800Mhz would compete nicely with the 5970. All they need is a good respin.

 

[EarthwormJim]

Um, did you miss the memo about ATIs business plan? It's that way by design. They don't want a single dominant GPU.
http://www.anandtech.com/show/2556/2

That AMD managed to have the single GPU crown for 6 months wasn't a design decision or a goal they had, it was the result of NV being late.

That was AMD's strategy for the 4xxx series. They really didn't follow that strategy for Cypress. They built a huge GPU, not as big as Nvidia because they don't care about computing quite as much with a GPU, but huge for a gaming/graphics card.

http://www.anandtech.com/show/2937/3

 

[happy medium]

Happy Medium, why won't a refreshed 58xx card compete with the GTX series? They compete now afterall.

It's not like AMD is going to be stuck with the 40nm process and Nvidia is jumping to 22nm. If anything, it looks like AMD may have an advantage in process technology as they'll have both GF and TSMC to work with if they want. As of now Nvidia has said no to GF.

Even if both shrink their current parts to the exact same process, what would change? I believe power consumption will scale down for both AMD and Nvidia. This gives both camps the opportunity to go with the power savings or up clocks (generally speaking). What makes you think both AMD and Nvidia taking their current offerings to the same smaller process is going to give one the upper hand over the other? They should be at roughly the same proportion to each other as of now, at least that's how I understand it.

Perforrmance aside..............I look at the track record , It seems ATI is happy in second place.
It's almost like ATI and Nvidea planned it this way.

The usual cards in performance order and usually in price too.................>>

1. Nvidia} ultra HIGHEND 500$ but faster dual card
2. ATI } 450$ but 10% slower dual card

3.nvidea } high end 350$ but faster
4.ATI } 300$ 10% slower

5. nvidea and ATi about even in performance but Nvidiea 20$ more. (mid range)
6. lower end........do we really care.

This is the basic breakdown. Sometimes I realize there were exceptions.

 

[SlowSpyder]

That was AMD's strategy for the 4xxx series. They really didn't follow that strategy for Cypress. They built a huge GPU, not as big as Nvidia because they don't care about computing quite as much with a GPU, but huge for a gaming/graphics card.

http://www.anandtech.com/show/2937/3

Read the page your linked. They planned on building a huge GPU, they didn't. They built a more middle of the road GPU (going by higher end GPU sizes). Larger than the previous generation, but still much smaller than the competition.

 

[Martimus]

Happy Medium, why won't a refreshed 58xx card compete with the GTX series? They compete now afterall.

It's not like AMD is going to be stuck with the 40nm process and Nvidia is jumping to 22nm. If anything, it looks like AMD may have an advantage in process technology as they'll have both GF and TSMC to work with if they want. As of now Nvidia has said no to GF.

Even if both shrink their current parts to the exact same process, what would change? I believe power consumption will scale down for both AMD and Nvidia. This gives both camps the opportunity to go with the power savings or up clocks (generally speaking). What makes you think both AMD and Nvidia taking their current offerings to the same smaller process is going to give one the upper hand over the other? They should be at roughly the same proportion to each other as of now, at least that's how I understand it.

From what I have read Northern Islands is a new architecture, so I am not sure we can derive improvements based on the current 5870 architecture. nVidia will likely continue to refine their current architecture, and we will see next year how they compare. I do wonder if NI will have similar problems to Fermi/R600 when they were first introduced, and won't really impress until a few iterations, like the R600 architecture (and what I expect from Fermi as well)? It is entirely possible that it will start out well too.

 

[thilanliyan]

That's exactly what I was thinking lol. What else would anyone expect nVidia to say other than everything's fine and dandy? I'm not saying it isn't fine and dandy but they would probably say it even if it wasn't.

 

[happy medium]

A 512 core Fermi at 800Mhz would compete nicely with the 5970. All they need is a good respin.

In a nutshell thats what I was saying, but I just dont see a refreshed 58xx series scaling as good as a gtx series by looking at the overclocked benchmarks.

I see Nvidia exstending there performance lead and ATI once again giving up and lowering prices and settling for second.:hmm: