AMD: Southern Islands (7*** series) Speculation/ Rumour Thread

What both of you are saying is that AMD is likely setting their price so that demand doesn't outstrip supply. This may well be true.
However, this also means that AMD largely fails to capitalize on being months ahead of nVidia on the new process node. If your theory is right, AMD should be less than happy about the situation.

Then their failure to capitalize on their position (out of the gate first) is unrelated to product pricing.

loads of reasons for AMD to be unhappy. Pricing is NOT one of them.
 
Then their failure to capitalize on their position (out of the gate first) is unrelated to product pricing.

True. And yours and S_Bs position as far as I understand it is that the pricing is a consequence of being supply limited. (We can't tell if this is indeed the case, but it is an explanation which is consistent with what we observe).

What I pointed out was merely that if AMD is supply limited to an extent implied by the slow trickle of cards getting in stock and (even slower) getting sold, it is a lost opportunity to capitalize on being early out with working silicon at the 28nm node from a competitive point of view. With the current prices, people who are interested in the products are holding off, either because the prices simply are perceived as too high, or because they are unpalatable enough that consumers take a wait and see attitude. What they are waiting for is AMDs competitor, meaning AMD will loose market share in comparison to what they could have enjoyed. Neither AMD, nor their partners are likely to be happy about this.

If there are supply issues I wonder as to the cause. Have they booked too few wafers? (Would seem odd, but maybe TSMC are tightly constrained in terms of wafer starts). Do they have bad yields? (Possible of course, but it doesn't seem as if the current products are dangerously close to their operational limits).
So if we postulate supply constraints as the cause of current pricing, then there remains to explain just what causes the constraint.
 
I'm having a hard time understanding the complaints regarding pricing.

They're the fastest.

There is no competition.

Therefore, expensive.

Don't like to pay it? What are you going to do, buy something slower? Maybe. Wait for something else that might be faster, or might be cheaper? Maybe. If you want the fastest, and you want it now, your option is to pay the money that they're asking.

We can empirically prove they're selling at that price, so it isn't even a question. When sales get "slow enough" to warrant a price reduction, it will happen. Until then, I fail to see any other relevant information regardless of source availability or some perception of "customer goodwill."

As a publicly traded company, they owe "you the customer" nothing -- they owe shareholders a big pile of ROI. If you feel that high prices will anger customers to the point of never coming back, that's fine and dandy, but their marketing team obviously disagrees.
 
It seems that there is another factor that was not yet discussed here..

While AMD sold those HD 5770's very nicely, making boatloads of $$ off these cards, AMD is pricing their super-small HD 7770's (a lower-end chip given the size comparable against HD 5670 and 6670) at the same bracket as they did with HD 5770s.

One of the biggest reasons:

AMD is feeling a bit threatened by Nvidia's upcoming Kepler. AMD so desperately does not want to encourage Nvidia to launch their products at low, super-competitive prices. Nvidia was rather aggressive with GTX 460-768 cards, snatching back the marketshare that AMD worked so hard to gain with early DX11 cards. The 460-768 was even cheaper than HD 5770, and performed better. This was one of the best deals of the decade, having a card nearly as fast as the previous generation $500 flagship (GTX 280/285) for roughly $100, sometimes even $85 with a rebate. Those $100 deals did come around often enough to entice buyers to do an impulse buy.

So, AMD is trying to set the bar higher, to encourage NV to not feel pressurized to do aggressive prices (concerning the rumors of a $300 GK104). The Graphics department of AMD is the main profit source, IIRC.. so AMD desperately needs $$ from that dept. I'd just call it a subtle, indirect form of "price fixing", that AMD is trying to get NV to coordinate with. If NV does coordinate, then it gives AMD some room to make aggressive cuts in order to try to spoil NV's launches!

Right, the HD 5870 should've been $500 instead of $400, the price difference alone would've made a huge difference in the number of buyers. The price could've changed the overall mindset (Bah, it's only 30% faster than GTX 285, and the DX11 games suck.. just wait for Fermi, etc..) Plus AMD was actually increasing the prices anyways, while able to meet the demand by the time of Fermi's launch. The marketshare gain, and the large sales/profit were what AMD needed anyways, plus the spirit of AMD thrived in the gaming community ("AMD is king" once again after so long since the X1900XTX). If AMD didn't do it at a "shiny" price, like launching 9700 Pro at $500 instead of $400, then the spirit would've had a much harder time radiating in the gaming community (as is the case with HD 7970 today).
 
In my neck of the woods the 460-768 spawned at ~€200 and never really dipped below €160.
Supply also stopped a good 2-to-4 months after the launch (Depending on the brand) so this gives you just the idea why this product was put on the market.

As far as I can tell this is stock clearance (who wants to sell 460s when there are 560s on the market?) and not a nvidia strategy.
 
I'm having a hard time understanding the complaints regarding pricing.

They're the fastest.

There is no competition.

Nonsense - you describe only the HD7970, and only if compared to single chip solutions. It doesn't apply at all to the other members of the family. What's more, all new products aimed at upgraders have the competition of the owners existing solution.

Of course AMD is free to set their prices any way they damn well please - the point here was to try to figure out if we could pinpoint exactly why AMD isn't cutting nVidias legs out from under them. If AMD had been able to flood the market with cards that had an obviously attractive price compared to their larger process node competition, they would have maximized revenue from their lead on the new process not only to the benefit of themselves, but also to their business partners, and have cut their only competitor in the stand alone GPU market out of much of their upgrader revenue and market share - a double win.
But that is not what AMD is doing. I'm assuming there is a reason. Why?
 
Because hurting the competition is not always the best way to enrich yourself. Price wars rarely benefit anyone but the consumer.

What were AMD's profits like during the glorious RV770 days?
 
Of course AMD is free to set their prices any way they damn well please - the point here was to try to figure out if we could pinpoint exactly why AMD isn't cutting nVidias legs out from under them.
Occam's razor: Because they can.

What is the counter argument? Sure, you can construe things to the Nth degree -- maybe yields, maybe insufficient orders, maybe total cost of parts, maybe anticipated costing against NV's next and greatest, maybe a forced firesale of existing stock.

All of those make big, grandiose assumptions about things that we have no way to measure. The fact is, they can charge this and people will buy it. When the market decides that the cards are at an unsustainable price point, then the prices will come down -- or AMD will simply stop selling cards.

Since they're in the business of selling cards to deliver shareholder value, then I have to assume the "AMD will simply stop selling cards" option isn't truly an option.

I am personally a big fan of the 7950 / Tahiti Pro, and if the price made sense for me, I'd buy one. Sadly, the price is seriously at least $100 too high for me to even consider it, in spite of the nearly ~$3000 I have spent in the last 30 days on an i7-3930k, an 8-ram socket X79 board, a PCI-E 2.0 8x dedicated RAID card, six 240GB SATA3 SSD drives slaved into that RAID card, and 32Gb of ram to support it all. A $450 video card would be less than 15% of the entire cost of the finished rig, and yet I do not buy it.

I have my own concept of value, and right now AMD isn't hitting it. Does that forever make me hate AMD and vow to never buy their overpriced garbage for the entirety of my life on this planet? Nope. When NV or AMD provide me a video card option that meets or exceeds my own internal value bias, I'll buy it without hesitation -- just like all the other hardware I just listed.

Until then? Good on AMD for doing what they can to scrape up some money. It's not hurting me; it's obviously not hurting them (at least for now.) It's not 'sad" that my next video card might be a GK104, no more than it's "sad" that my prior three cards have been ATI/AMD. They provided the value proposition that met my own biases; NV had my business for about four more generations before ATI/AMD. I don't see it as a big deal.
 
Last edited by a moderator:
...Since they're in the business of selling cards to deliver shareholder value, then I have to assume the "AMD will simply stop selling cards" option isn't truly an option...

You mean money, not value. Money is NOT value. :LOL:
And then, "AMD will simply stop selling cards"- that's exactly what they do with last generation processes. Instead of using the well developed, much cheaper, with much more capacity 40 nm, they jump on the full of problems new unproven 28 nm. So, to me CV is a fail, simply because it appears way too early when the process and conditions are not stable enough.
 
Indeed, CV is going to destroy all those rebranded laptop fermis

Rebranded, or shrunk using 28 nm?
If they are only rebranded, then nvidia will have severe advantage both in quantities and price. AMD will not be able to meet orders.

And that's why (or one of the reasons) nvidia don't hurry so much with new processes. ;)
 
If they are only rebranded, then nvidia will have severe advantage both in quantities and price. AMD will not be able to meet orders.

This is not a fact: we don't know yield and pricing of the 28nm products, so it is possible that for a small 123 mm^2 chips supply will not be so tight and initial shipments could be allocated to the notebook market instead of flooding the channel. Eventually, higher pricing could be offset by lower system costs due to lower thermals and anyway CV will target price segments where the performance/mm^2 of Nvidia offers (rebranded) will be so amazingly lower that the comparison could be simply embarassing. I.e. the performance of CV is higher than a GF116 - 238 mm^2 (but it should be seen if the better thermals of CV will help in competing even with GF114 mobile chips -360 mm^2 chip - as a 70W CV will probably have the full clock of the desktop card while this is not the case for GTX580M). Now, it is clear that 28nm production has a price premium over the 40nm and that initial yields should be lower than those obtained on a mature process but I will bet that this premium is not even close to 100%.
In the lower segments Turks is already competitive with price (40nm) and performance and APUs/Ivy bridge are likely to render all lower ranked discrete GPUs useless.
 
Rebranded, or shrunk using 28 nm?
If they are only rebranded, then nvidia will have severe advantage both in quantities and price. AMD will not be able to meet orders.

And that's why (or one of the reasons) nvidia don't hurry so much with new processes. ;)



Yeah, I wonder why nVidia didn't just keep using 130nm because of the "severe advantage both in quantities and price".
 
True. And yours and S_Bs position as far as I understand it is that the pricing is a consequence of being supply limited. (We can't tell if this is indeed the case, but it is an explanation which is consistent with what we observe).

What I pointed out was merely that if AMD is supply limited to an extent implied by the slow trickle of cards getting in stock and (even slower) getting sold, it is a lost opportunity to capitalize on being early out with working silicon at the 28nm node from a competitive point of view. With the current prices, people who are interested in the products are holding off, either because the prices simply are perceived as too high, or because they are unpalatable enough that consumers take a wait and see attitude. What they are waiting for is AMDs competitor, meaning AMD will loose market share in comparison to what they could have enjoyed. Neither AMD, nor their partners are likely to be happy about this.

You sort of got my point. Yes, pricing is dictated by predicted supply and predicted demand (which takes into account competing products).

But you completely missed another point. That we have no way of knowing whether sales are slow or not. Stock levels staying stable could certainly mean that there are say 5000 chips being produced a month (ridiculously low number) and only 2000 of them are selling. It could also mean there are 10,000,000 chips produced each month (ridiculously high number) and 9,750,000 of them are selling.

Both of those above extremes would result in the exact same stable stock situation that we see in retailers. Hence, we have no way to know whether it is selling well at retail or not. And that doesn't even take into account potential OEM orders for desktop and more importantly notebook computers.

Just because X number of people on forums say, "meh, I'm not going to buy one." Doesn't mean Y number of real world consumers aren't buying them. Forums don't always, and sometimes not very often, reflect what happens at retail.

At the end of the day, you could be right and perhaps the cards aren't selling well. On the other hand you're just as likely to be wrong as we don't have the necessary data to conclude anything except the following...

Demand is less than supply. That's it. Nothing else.

Heck, I'm not even seeing rumors of AIBs complaining about slow sales or inability to get stock. Something that invariably pops up anytime sales are unexpectedly low or supply is unexpectedly tight.

If demand is signficantly less than supply we'll potentially see a price drop in the next few weeks/months. But I'm not expecting that for a few reasons to follow...

So if we postulate supply constraints as the cause of current pricing, then there remains to explain just what causes the constraint.

There's many explanations but perhaps the easiest is just the fact that this is a very compelling chip for notebook computers. It has relatively high performance in a very small power envelope.

It's entirely possible that the majority of chip allocation has gone to OEM notebook designs. But again it's not something we'll be able to see for a few months yet.

There's also the possibility of yields being low, but that one isn't very high up on my list. Wafer allocation on a new process might be limited so that's another possibility. But really after looking at how the desktop variant performs, I can't help thinking just how attractive this must be to notebook manufacturers.

Because hurting the competition is not always the best way to enrich yourself. Price wars rarely benefit anyone but the consumer.

What were AMD's profits like during the glorious RV770 days?

Don't forget that same war also resulted in Nvidia having a rare quarter of posting an operating loss. It was ruinous for both companies. Hence why neither Nvidia nor AMD are particularly keen to start another one. AMD out to gain marketshare at all costs and Nvidia attempting to prevent AMD from getting marketshare at all costs. In the end both companies lost.

While I personally loved the 48xx and 58xx pricing (as it allowed me to fit enthusiast class cards in my yearly budget), I understand that it perhaps wasn't the healthiest thing for AMD.

Regards,
SB
 
And then, "AMD will simply stop selling cards"- that's exactly what they do with last generation processes.
Yes, that is exactly what everyone does with last generation processes. This isn't something new or different than every other time a process node shrink comes along, and NVIDIA has led just as many of those as ATI/AMD.

Instead of using the well developed, much cheaper, with much more capacity 40 nm, they jump on the full of problems new unproven 28 nm. So, to me CV is a fail, simply because it appears way too early when the process and conditions are not stable enough.
What kind of non-logic is this? If "well developed, much cheaper process" is your target, then why did anyone leave 65nm? 90nm? 130nm? 1µm?!? All were cheaper and "more mature" than their successors, but there's an obvious reason why we moved away from them -- just like the eventual move away from 28nm.

And as for stable? I'm sorry, I don't see any point you can make about instability. Please back this up with fact rather than your imagination...
 
Last edited by a moderator:
I don't remember reading 50 pages of price whining back during the NV40/R420 - G71/R580 days... Why on earth is it so controversial and wounding now?
 
Probably it has something to do with this:
LrwKI.jpg


Whenever you create a new marketing paradigm, chances are it'll backfire on you as soon as you don't adhere to it anymore.
 
I don't remember reading 50 pages of price whining back during the NV40/R420 - G71/R580 days... Why on earth is it so controversial and wounding now?

It isn't about price, it's price/perf over prevous generations. 9800 GT/GTS 250 rebrands probably offer better price/perf increase over their original 8800 gt/9800 GTX+ counterparts :D
 
Back
Top