Nvidia uses GDDR5, Finally !!

Who pays for the cost of the cards if it isn't the consumers? The money doesn't appear out of thin air. Both NVIDIA and AMD are businesses with a bottomline, and having market share doesn't mean much if you're still hemorrhaging money. They create a product to sell for profit against the competition. Were you not here in June? AMD was able to take over NVIDIA because they could produce a competitive card more cheaply and sell it for less. Was this all due to just GDDR5? God no, but it's a part of it.

You realize you're agreeing with me right? The price to the consumer is a function of performance and features relative to the competition and the level of consumer demand that follows on from that. Nothing else matters. The cost of manufacturing is only relevant to the seller. It's a fundamental thing that you're missing. The seller's financial position is irrelevant to the price they can set. Don't you think AMD would love to sell Phenom for higher prices? Their manufacturing costs are higher than Intel's but that doesn't give them a license to charge more. Their performance is lower hence they have to sell it for less.

Let me remind you that your claim is "how you get the memory bandwidth makes a big difference in final cost to the consumer" That is false. Assuming performance parity of competing products (as is the case with 4870 vs GTX 260) using a more expensive approach does not translate to higher prices for the consumer. Nvidia has no pricing power there based on the higher relative manufacturing cost of GTX 260.

AMD could have easily and fairly priced the 4870 at $399 as well at launch. Their lower prices are more due to an attempt to reclaim market share than it is a reflection of lower manufacturing cost. No company sacrifices higher margins for kicks - in this case it was a grab for market share. Also, note that the actual manufacturing cost of a chip on a graphics card is only a part of the total cost to the company to bring that chip to market. We know that Nvidia's margins are lower than they would like but we don't know that AMD's margins are any higher.
 
i'm curious about something hypothetical. it doesn't seem like either nvidia or ATI is yet at a point where they need more memory bandwidth than GDDR5 on a 256-bit bus can offer. but if that point comes, is there's something to be said for nvidia's experience implementing wider memory buses? are wider memory buses more technically difficult to produce, or just simply more expensive?

They're probably not that difficult to design but PCB's and power circuitry will be more complex and expensive in order to handle the wider bus. But I think we'll see 512-bit buses on future Nvidia parts even after the move to GDDR5. How else will they keep supporting 4GB framebuffers on their professional cards a la Quadro FX 5800 ? It'll be interesting to see how they balance the need for such large framebuffers and the overkill 512-bit GDDR5 would bring to consumer parts.
 
You realize you're agreeing with me right? The price to the consumer is a function of performance and features relative to the competition and the level of consumer demand that follows on from that. Nothing else matters. The cost of manufacturing is only relevant to the seller. It's a fundamental thing that you're missing. The seller's financial position is irrelevant to the price they can set. Don't you think AMD would love to sell Phenom for higher prices? Their manufacturing costs are higher than Intel's but that doesn't give them a license to charge more. Their performance is lower hence they have to sell it for less.

Let me remind you that your claim is "how you get the memory bandwidth makes a big difference in final cost to the consumer" That is false. Assuming performance parity of competing products (as is the case with 4870 vs GTX 260) using a more expensive approach does not translate to higher prices for the consumer. Nvidia has no pricing power there based on the higher relative manufacturing cost of GTX 260.

AMD could have easily and fairly priced the 4870 at $399 as well at launch. Their lower prices are more due to an attempt to reclaim market share than it is a reflection of lower manufacturing cost. No company sacrifices higher margins for kicks - in this case it was a grab for market share. Also, note that the actual manufacturing cost of a chip on a graphics card is only a part of the total cost to the company to bring that chip to market. We know that Nvidia's margins are lower than they would like but we don't know that AMD's margins are any higher.
I'm agreeing that obviously supply and demand is still in effect. But, you have an extremely narrow view of economics and business. You're missing the point that the reason AMD was able to make such a bold move was because they were producing competitive cards more cheaply due to the 256-bit/GDDR5 design. Like I said, market share doesn't mean squat if you're still hemorrhaging money. If all you read is enthusiast sites then you might not realize that most graphics cards consumers could care about performance; cards are bought in OEM machines and the like and there the only thing that matters is bottomline dollar spent. The same is for the suppliers, AMD knows that and you can see it in the card's design.
 
I'm agreeing that obviously supply and demand is still in effect. But, you have an extremely narrow view of economics and business. You're missing the point that the reason AMD was able to make such a bold move was because they were producing competitive cards more cheaply due to the 256-bit/GDDR5 design. Like I said, market share doesn't mean squat if you're still hemorrhaging money. If all you read is enthusiast sites then you might not realize that most graphics cards consumers could care about performance; cards are bought in OEM machines and the like and there the only thing that matters is bottomline dollar spent. The same is for the suppliers, AMD knows that and you can see it in the card's design.

And I believe you don't understand economics and business at all in this context but our opinions are hardly relevant now aren't they? You have no idea what my academic or professional background is and I don't know yours so the empty insults are really unnecessary.

AMD were being battered into the ground and had no choice but to fight back for market share through lower pricing. This was not a "bold" move; it was a necessary one regardless of how cheap or expensive it was to manufacture the chips. They also made a "bold" move and dropped the price of the 2900XT and guess what - that chip was not cheap to manufacture. They had to drop the price because performance sucked. Did Nvidia drop its prices on G71 although it was vastly smaller than R580? No and you know why? Because they didn't have to since performance was there. These two recent examples totally contradict what you're attempting to say. Your claim that market share doesn't mean squat if you're hemorrhaging money is also counter to your central point. If this was the case why didn't AMD sell the 4870 for $399 and take advantage of the higher margins to stem the losses they're still currently enduring?

I'm not debating whether lower manufacturing cost gives a seller more flexibility with pricing. That's obvious. What I'm saying is that AMD's price cuts were not a result of lower cost, they were necessary to fight back for their place in the market. They would not have fared as well were they to tackle the GTX 260 on price and performance parity since Nvidia had the lion's share of mindshare at the time.

Again, let me remind you of your previous claim: "how you get the memory bandwidth makes a big difference in final cost to the consumer". That claim is easily debunked by the mere existence of the similiarly priced 256-bit GDDR5 4870 and 448-bit GDDR3 GTX 260 both with a bandwidth of ~115GB/s.
 
They're probably not that difficult to design but PCB's and power circuitry will be more complex and expensive in order to handle the wider bus. But I think we'll see 512-bit buses on future Nvidia parts even after the move to GDDR5. How else will they keep supporting 4GB framebuffers on their professional cards a la Quadro FX 5800 ? It'll be interesting to see how they balance the need for such large framebuffers and the overkill 512-bit GDDR5 would bring to consumer parts.
Most likely they'll use a modular design. Just make quadros with a double set of memory controllers on the die vs the geforce parts.
 
And I believe you don't understand economics and business at all in this context but our opinions are hardly relevant now aren't they? You have no idea what my academic or professional background is and I don't know yours so the empty insults are really unnecessary.

You see AMD as some benevolent company that graciously passed its cost savings onto the consumer. That is not what's happening. They were being battered into the ground and had no choice but to fight back for market share through lower pricing. This was not a "bold" move; it was a necessary one regardless of how cheap or expensive it was to manufacture the chips. They also made a "bold" move and dropped the price of the 2900XT and guess what - that chip was not cheap to manufacture. They had to drop the price because performance sucked. Did Nvidia drop its prices on G71 although it was vastly smaller than R580? No and you know why? Because they didn't have to since performance was there. These two recent examples totally contradict what you're attempting to say.
Are we speaking the same language, or are you reading every other word I write? Where did I say that AMD was benevolent? I said it's a business and will operate like one (obvious to most, but evidently not to you). I also said they produced a competitive product more cheaply (third time at least I've said this now) and used their advantage to aggressively take a market share. The example you gave of the 2900XT has no relevance to the situation in June of the 4800 series vs. GTX series. Let's start at the beginning. NVIDIA makes the GTX 260 and GTX 280. These cards are very expensive to produce due to the extreme size of the die, and consequently low yields, as well as the 512-bit memory bus PCB design, etc.. They're released first, at a price point of $600 and $400, a point where NVIDIA feels it can maximize its revenue. AMD goes a different route, producing the much cheaper, 55nm-based 4870 and 4850, using a cheaper 256-bit bus, thinner PCB, and GDDR5 to make up the bandwidth difference. Because the cards are cheaper to produce but perform competitively, AMD releases them at $300 and $200, and they become an instance hit. The fact that the cards were cheaper to produce allowed them to make this bold move of releasing $100 below the MSRP of the competition. If the 4800 series was as expensive to make as the GTX's, AMD wouldn't have released $100 below for shits and giggles and marketshare, that's just stupidity.
Your claim that market share doesn't mean squat if you're hemorrhaging money is also counter to your central point. If this was the case why didn't AMD sell the 4870 for $399 and take advantage of the higher margins to stem the losses they're still currently enduring?
But AMD wasn't hemorrhaging money due to the 4800 series being so much cheaper to build, so what is the relevance in that point? Like I said, you need to study more economics. AMD released the 4870 at a price to maximize revenue, due to profit from each card x cards sold. Study of supply and demand (this is economics 101 here) will reveal a "sweet spot" whereby a company can price its product to maximize revenue due to units sold. AMD saw that at $300 dollars this was it. If they released at say $200, they might have sold more cards, but they also might not have made enough profit on each to make up for the reduced (if any) revenue from each card. If they had priced it at $400, they might have made more per card, but they would have sold fewer cards and might not make up the difference.
I'm not debating whether lower manufacturing cost gives a seller more flexibility with pricing. That's obvious. What I'm saying is that AMD's price cuts were not a result of lower cost, they were necessary to fight back for their place in the market. They would not have fared as well were they to tackle the GTX 260 on price and performance parity since Nvidia had the lion's share of mindshare at the time.
That is absolutely ridiculous. The amount of money they would have lost by introducing a a video card at $100 below competition (PER CARD) would be absolutely insane if they were only trying to capture marketshare. The only way they could do this is because they had a product with a superior price/performance ratio when viewing manufacturing costs (i.e. cheaper to produce), so they were still making money at the lower price point (a point where NVIDIA probably wouldn't be making as much, so an excellent move by AMD really).
Again, let me remind you of your previous claim: "how you get the memory bandwidth makes a big difference in final cost to the consumer". That claim is easily debunked by the mere existence of the similiarly priced 256-bit GDDR5 4870 and 448-bit GDDR3 GTX 260 both with a bandwidth of ~115GB/s.
What was the original MSRP of the GTX 260? Oh yah, $400. And the 4870? Oh yah, $300. Thank you for proving my point, and I hope you now realize it. Due to supply and demand (that funny little principle I just talked about), the cheaper 4870 dragged the price of the GTX 260 down. Now in NVIDIAs case, they HAD to cut prices to KEEP some marketshare. AMD produced similar bandwidth with a much cheaper design. That's my damn point, I hope you finally understand. And honestly, if you ignore or don't get the points that I've made in this post, and exhaustively tried to explain, then I give up. I can't dumb it down any more.
 
AMD produced similar bandwidth with a much cheaper design. That's my damn point

No that was not your point. This was your point: "how you get the memory bandwidth makes a big difference in final cost to the consumer". The fact that Nvidia was able to lower its price to the same level proves that your claim is false and has gotten you upset.
 
No that was not your point. This was your point: "how you get the memory bandwidth makes a big difference in final cost to the consumer". The fact that Nvidia was able to lower its price to the same level proves that your claim is false and has gotten you upset.
You seriously cannot be this dumb, are you kidding at this point? You can't understand that by going a different route through using a 55nm core, thinner PCB, and a 256-bit bus and GDDR5, AMD created a performance competitive card more cheaply? Nor can you understand that cheaper manufacturing costs then allowed them to sell this performance at a cheaper price, $300, which benefited us consumers? Furthermore, you can't understand that this forced NVIDIA to then lower its prices to stay competitive, which brought down video card prices all around, and benefited consumers even more? Wow, just wow.
 
You seriously cannot be this dumb, are you kidding at this point? You can't understand that by going a different route through using a 55nm core, thinner PCB, and a 256-bit bus and GDDR5, AMD created a performance competitive card more cheaply? Nor can you understand that cheaper manufacturing costs then allowed them to sell this performance at a cheaper price, $300, which benefited us consumers? Furthermore, you can't understand that this forced NVIDIA to then lower its prices to stay competitive, which brought down video card prices all around, and benefited consumers even more? Wow, just wow.

The petty insults can't mask your fundamental problem my friend. You made a claim that was easily debunked and now you want to change your story. It would have been a lot less painful for you to just admit that. All of these rants and points against which I have not argued are really a waste of your time.

If you want to have a separate conversation about the motivation behind AMD's lower introductory price on the 4870s we can do that. But your original claim about manufacturing cost being a significant driver of price to the consumer is still going to be false. If it was correct G71 and R600 would have followed very different pricing structures than they did.

Oh, and in terms of how much AMD is benefitting financially from these lower costs and lower prices you might want to compare Q3 results between the firms and check some of those margin numbers. I'm sure someone of your great economic talent will have no problem analyzing the data. Once you do you might be surprised.
 
The petty insults can't mask your fundamental problem my friend. You made a claim that was easily debunked and now you want to change your story. It would have been a lot less painful for you to just admit that. All of these rants and points against which I have not argued are really a waste of your time.

If you want to have a separate conversation about the motivation behind AMD's lower introductory price on the 4870s we can do that. But your original claim about manufacturing cost being a significant driver of price to the consumer is still going to be false. If it was correct G71 and R600 would have followed very different pricing structures than they did.

Oh, and in terms of how much AMD is benefitting financially from these lower costs and lower prices you might want to compare Q3 results between the firms and check some of those margin numbers. I'm sure someone of your great economic talent will have no problem analyzing the data. Once you do you might be surprised.
My attempts at leading you by the hand through logical progression have been a completely unproductive. I have proven my point over and over again, and yet it still isn't sinking in. However, you haven't debunked anything, you haven't made a single point, you haven't shown anything besides not agreeing with me for the last few posts. Like I said, at this point you either have gotten my point and are ignoring it or you're just not getting it and I'm wasting my time anyway. Either way, I could care.
 
Heh, no I hear exactly what you're saying and agree with most of it as most of it is obvious and widely accepted. But you refuse to defend your original point and I choose to ignore all the other points that you've raised that I never contested in the first place. So we're not going anywhere with this.....
 
Heh, no I hear exactly what you're saying and agree with most of it as most of it is obvious and widely accepted. But you refuse to defend your original point and I choose to ignore all the other points that you've raised that I never contested in the first place. So we're not going anywhere with this.....
Let me post this again:
You can't understand that by going a different route through using a 55nm core, thinner PCB, and a 256-bit bus and GDDR5, AMD created a performance competitive card more cheaply? Nor can you understand that cheaper manufacturing costs then allowed them to sell this performance at a cheaper price, $300, which benefited us consumers? Furthermore, you can't understand that this forced NVIDIA to then lower its prices to stay competitive, which brought down video card prices all around, and benefited consumers even more?
What don't you get? AMD went a different route to get the same bandwidth/performance. Because of this cheaper design used in the 4800 series, they produced a competitively performing card more cheaply. They then could sell this more cheaply, which benefited us by not only giving us this performance more cheaply, but also brought down prices in the video card market in general. All you've said in the past posts are "no," but you haven't stated what you specifically can't connect. I've tried to show this as many ways as I can, but I'll try again if you can pinpoint where you don't see it.
 
It's obvious what AMD did with RV770...those are facts and the history of it is there for everybody to see so I'm not sure why you keep repeating it as that's not what I'm debating at all hence I'm not responding to it. My original "No" was to your claim that lower manufacturing costs necessarily translate to lower prices for the consumer. I gave you three past cases where this was not the case - G71/R580; R600/G80 and Phenom/Core 2 and you chose to dismiss them.

Since you're an econ hotshot you should be able to follow a simple scenario. I'm sure we can both agree that a consumer's value proposition is not in any way impacted by the cost to the supplier. The value the consumer places on the product is a function of performance and features, power consumption, noise, form factor, brand loyalty etc etc and those things determine the price they're willing to pay.

Given that this is the case, when a manufacturer finds itself at a cost advantage it has two primary options - maintain the price level and improve margins/profits or cut prices in an attempt to steal market share away from the competition by increasing the value to the consumer.

We differ on a very subtle point. What you're saying is that the manufacturer will always lower the price. That would imply that manufacturers focus on cost cutting solely to lower prices which of course is totally in contrast with their reason for being - to turn a profit. You made the argument earlier about maximizing returns by finding the balance between price and quantity sold. That's obvious but it's irrelevant when your competitor can price match you at any time. Which Nvidia can do because they are in a much stronger financial position right now and sitting on a significant amount of cash. That sets up a scenario where no market share is transferred in the long run but the price level falls for all manufacturers. So in the end the lower cost manufacturer gains nothing except making its competition hurt a little more. But it also hurts itself in the process as the new equilibrium price level is lower than the original one. And this is why companies don't cut prices simply when they have a cost advantage unless that advantage is significant enough to drive the competition out of business or if they're desperate. If AMD was in a stronger position in terms of mind and marketshare I can assure you that any cost advantage would have gone to profits and not to us as consumers. As was the case with G71 and the ridiculous margins Nvidia pulled out of that chip.
 
Most likely they'll use a modular design. Just make quadros with a double set of memory controllers on the die vs the geforce parts.

Yeah but since they have to equip the chips with a 512-bit bus anyway for the professional line they aren't realizing any savings by disabling half of it on the desktop parts. Might as well keep it enabled and use cheaper memory. They could always use the cheapest GDDR5 at 3.2Gbps which would take them over 200GB/s on a 512-bit bus. Whereas AMD would have to shell out for the most expensive GDDR5 at 6Gbps+ if they stick to 256-bit.
 
It's obvious what AMD did with RV770...those are facts and the history of it is there for everybody to see so I'm not sure why you keep repeating it as that's not what I'm debating at all hence I'm not responding to it. My original "No" was to your claim that lower manufacturing costs necessarily translate to lower prices for the consumer. I gave you three past cases where this was not the case - G71/R580; R600/G80 and Phenom/Core 2 and you chose to dismiss them.

Since you're an econ hotshot you should be able to follow a simple scenario. I'm sure we can both agree that a consumer's value proposition is not in any way impacted by the cost to the supplier. The value the consumer places on the product is a function of performance and features, power consumption, noise, form factor, brand loyalty etc etc and those things determine the price they're willing to pay.

Given that this is the case, when a manufacturer finds itself at a cost advantage it has two primary options - maintain the price level and improve margins/profits or cut prices in an attempt to steal market share away from the competition by increasing the value to the consumer.

We differ on a very subtle point. What you're saying is that the manufacturer will always lower the price. That would imply that manufacturers focus on cost cutting solely to lower prices which of course is totally in contrast with their reason for being - to turn a profit. You made the argument earlier about maximizing returns by finding the balance between price and quantity sold. That's obvious but it's irrelevant when your competitor can price match you at any time. Which Nvidia can do because they are in a much stronger financial position right now and sitting on a significant amount of cash. That sets up a scenario where no market share is transferred in the long run but the price level falls for all manufacturers. So in the end the lower cost manufacturer gains nothing except making its competition hurt a little more. But it also hurts itself in the process as the new equilibrium price level is lower than the original one. And this is why companies don't cut prices simply when they have a cost advantage unless that advantage is significant enough to drive the competition out of business or if they're desperate. If AMD was in a stronger position in terms of mind and marketshare I can assure you that any cost advantage would have gone to profits and not to us as consumers. As was the case with G71 and the ridiculous margins Nvidia pulled out of that chip.
Ah, I see the problem, you're reading words that aren't there. I never said manufacturing costs always lower prices, I just said they did in this specific case. Hence all your examples were pointless because they never addressed this case. Honestly, we beat around the bush for a good 5+ posts that could have been avoided if you just made that clear.
 
I never said manufacturing costs always lower prices, I just said they did in this specific case.

No actually you made a broad statement about memory configuration and price. I would quote you again but I'm sure you can pull up your original post. If you can show me where you indicated that you were speaking specifically about RV770 in that comment I'll accept your clarification.

In any case, we still differ on the specific case of RV770 if you want to narrow the discussion to this particular chip. Again, it's a subtle difference. You're saying AMD lowered prices because RV770 was cheap. Lets establish something - are you or are you not implying a causality there of lower manufacturing costs leading to lower prices?
 
No actually you made a broad statement about memory configuration and price. I would quote you again but I'm sure you can pull up your original post. If you can show me where you indicated that you were speaking specifically about RV770 in that comment I'll accept your clarification.

In any case, we still differ on the specific case of RV770 if you want to narrow the discussion to this particular chip. Again, it's a subtle difference. You're saying AMD lowered prices because RV770 was cheap. Lets establish something - are you or are you not implying a causality there of lower manufacturing costs leading to lower prices?
Actually I think I'm finished with this discussion. If you aren't trying to put words in my mouth you are generalizing the hell out of everything I say in some lame attempt to be "right." There's no point in my trying to carry on a discussion with someone who has little to no communication skills. I can see why Kyle dismisses Beyond3D forums if you're all this obnoxious over there.
 
I can see why Kyle dismisses Beyond3D forums if you're all this obnoxious over there.

Heh, yes we are and love it :) It's not about being right or wrong though. But yeah I think we've thoroughly exhausted this particular argument.

Why can't I have 512bit AND GDDR5. I wont my bandwidth with extra sauce please.

You just might get it on the first DX11 parts. At least from Nvidia.
 
Back
Top