My take on the disaster that is the GFFX. (long post)

Think he's saying they should be fired for not MAKING it more expensive because of supply and demand. Not more expensive to manufacture, more expensive to buy.

 

Because why should Samsung as the seller do nvidia the buyer any favors. Samsung should realize that they have alomst complete pricing power in the short run. If Samsung is not charging nvidia (up fornt too, by the way) an arm, leg, & the other arm, then Samsung shareholders should bring suite & or heads should roll. It's all about maximizing profit. In other words, sell the least you can sell for the most profit.

I think that you think that the cost of manufacturing has a lot to do with the price that the manufactured good will sell for. To the extent that manufactuiring costs are fully recovered this is true. However, most prices are in fact market prices. These prices are typically many times the intrinsic value of the manufactured good & in fact are often times wholly contrived from the demand/supply curve. Right now Samsung has the plot point very far to the left & very high (with demand being a vertical vector & supply a horizontal one).


I would predict that the rollout cost that nvidia had projected for the NV30 have far surpassed what they expected, even at this late hour. Once again, I believe that nvidia is selling this initial batch of NV30's to their board vendors at cost or at best 3 to 5% markup. In other words, they are taking a financial bath. I do not think this is company killing, but it will hurt far longer than the 6 months that I see most posters postulate.


edit: disclosure statement
I do hold a very small amount of ATYT stock, & although I have tried to make my comments as unbiased as possible, full neutrality is, of course impossible.

 

a little knowlege is a dangerous thing.

 

I would agree with everything you siad apart from thie quoted bit.

There is no way at all ever the GFFX is selling at cost or 3% to 5% to its partners. Times are hard but when they get that hard you might as well quit.

Another thing.. NVIDIA dont sell DDR2 memory .. I think that killled your point dead.

 

NVIDIA do usually sell RAM, as well as other parts, along with the chips to ensure board vendors stay in line with the specifications -- they often relax this later in the product lifecycle though.

 

Samsung COULD be gouging NVIDIA, but I have severe reservations because:

-Gouged customers remember and would ditch Samsung immediately when another solution comes available (and it seems DDR2 is set to become a commodity).
-Samsung would also lose future business (which is likely more than whatever they'd earn by gouging)
-DDR2 is in competition with DDR1 and while is a unique product NOW, it is part of a commodity market (i.e. high speed memory) and there will certainly be more DDR2 vendors in the near future. Anybody designing in their part would work out the details with Samsung prior to putting their eggs in the basket, especially something where they expect to sell/buy over 1M pieces in a year. If samsung priced their part too high, it wouldn't have gotten the design win.

Speaking from experience, we've had companies tell us they won't deal with the competition because of allocation problems, or past price gouging when supply got short.

But, speaking of supply and demand, do you have any idea what the supply of DDR2 is that samsung is making and how many pieces are being desired by people?

I don't.

 

Where did I say that nvidia is selling memory? I was talking about the price that Samsung would charge.

 

Gouged, temporary monopoly power...these are just semantics...& yes this is exactly what every supplier does wtih new product when demand is very high & supply is very low

Only until the next big thing, which in this buisiness seems to be every 18 to 24 months.

My initial post advised near perferct pricing power. Samsung will make it hurt drill Seargeant, but not to the death.

What?!?!? I thought this discussion was of nvidia & the NV30 rollout. What other people are you referring to? The supply is Samsung, and the demand is nvidia. The fact that later Samsung's pricing power will be greatly eroded is irrelevant at this point. This is Capitalism in a nutshell

 

Perhaps you're missing the point(s):

1) You have no idea what the supply or demand is. (Though you seem certain that Samsung and NVIDIA are the only two people making and desiring the product)
2) NVIDIA would have determined the cost prior to going down the road (it didn't HAVE to use DDR2, now did it?)


And no, price erosion in the future is not irrelevant. You do not screw over your customers unless you like to have less customers.

 

The other aim is to average marginal costs & the possibility of arbitrage on the commodity side. OEMs don't like monopolists dictating bundled deals on commodity goods... It's unlikely that an IHV will use it's VDD to supply lower average costs to it's OEM/AIB partners - but it may make sense in the current context.

Price gouging is not at issue. Pricing to market is.

 

Price for a similar configuration at the same clockspeeds (eg. 350Mhz) is approximately 35% more for DDR-2 over DDR-1. This delta goes up by a fair bit as the rating of the memory goes up. I don't have an exact number, but my guess is that 500Mhz DDR-2 is roughly 50% more than 350Mhz DDR-1.

The memory represents a very significant portion of the BOM costs for any given board, particularly with DDR-2. After that comes the cost of the ASIC and in the case of NV30-500, the cooling mechanism.

 

Right, my bad

However I still believe sumdumyunguy is way off when he said that NVIDIA were to make from 0% to 5% profit even at this stage on the GFFX. Does anyone know who manufactures the cooling system by the way?

 

Russ are you being obtuse on purpose? Both your points seem to make the assumption that there are many sellers & buyers for the NV30 memory subsystem. Also at this late date, alternative (even if better) products are completly out of the question. AFAIK we are talking about the (delayed) NV30. I would bet my next years mortrgage payments that Samsung is quite aware of the need for nvidia to release that product without further delay. The only customer at this point for memory on the NV30, in light of the fact that nvidia announced that they (nvidia) would handle the initial rollout means they (nvidia) are the only customer for DDR2 memory on the NV30 board att. Thus Samsung has some positive pricing potential for the short term. Samsung would be entirely remiss not to maximize this advantage. Also as you yourself stated, memory is a commotatized product. There is no such thing as "long term" relationships as each transaction is based upon market conditions at the time of the transaction[/] Oil, pork bellies, & memory chips are bought & sold on ontime contracts that bear no relatioships to past history. A common idiom of investing is that, past performance is no indicator of future profits. A real world example is this. Here in Atlanta unleaded gasoline prices have increased $0.10 in 3 days in my local area. BP is not worried about "gouging" me now & that I won't come back this september. Neither is Samsung.

A more intresting aspect of our little debate is that you believe nvidia will not accept 3 to 5% margins at rollout. Are you not aware of the 'loss leader" markteting concept? Whereby the seller will take little or no margins or sometimes even a slight loss in order to (mostly) entice you purchase said discounted product now in the hopes of higher margin purchaes & also (what I think in this case) to gain marketshare. AMD has been successful enough at this to be a valid competitor against a behomenth such as Intel. I see no reason (at this point in time) why nvidia would be immune to such a strategy vis a vis other graphics board distributors.

Furthermore, companies screw over their customers all the time. We have anit-trust & anti monopoy laws because of this very aspect of human nature.

 

We dont know how the FX is going to overclock and what about driver revisions... they might have another 20% lurking under the hood.
Who knows they might even supply some ear muffs at that price. (lol)

 

So the only people interested in DDR2 is NVIDIA?

Not router manufacturers like Cisco and Lucent. Not high end server manufacturers. Nobody?

You keep calling it "NV30 memory subsystem". DDR2 is simply a high speed memory, not designed specifically for the NV30.

(And no, I didn't say anything about your 0-5% margin thing, you've got me confused with somebody else)

 

ATI doesn't seem to support it..

http://www.ati.com/companyinfo/press/2002/4544.html

 

I thouht that our conservation was about the NV30 & its memory subsytem with components manufactured by Samsung & in particular how that relates to the initial rollout of the FX!

I am sorry if I attributed that to you. I was under the impression that you & I were the only ones really carrying on this debate. As you can tell by my low post count, this is really the first time at this board that I have had such an interaction. I thank you for that. However, that being said, we seem to have strayed afar from "3D Technology & Hardware" & into the world of just general business practices. Lets say we agree to disagree & call it a goodnight.

 

Did you guys miss CMKRNL's post or something?

He pretty much settled this issue.

 

As far as the reference board with "standard" cooling goes, yes, we do. [H]OCP hit 530/535 stable; Anand had somewhat similar results (although he states it in terms of 540/550 being unstable).

So it seems that GFfx Ultra is not much of an overclocker in its current form. We'll see if actual shipping cards can do better, but it doesn't seem promising, especially given how, um...robust the cooling system is already.

 

sorry , i have some questions ,what is "BOM costs" ? what does it mean? need some comment in advance