The end is near. When products cost more in the U.S. than in Canada, you know shit's about to hit the fan.
So, do we know anything about RV670 yet?
- Thread starter nicolasb
- Start date
The end is near. When products cost more in the U.S. than in Canada, you know shit's about to hit the fan.
Wake up and face reality. Nvidia has higher margins than ATI. Stop spreading FUD.
From vr-zone, RV670 is 192 mm^2. G92 is over 300 mm^2. If you think that area doesn't affect margins, then I'd like some of what you are smoking. The difference in area is over 50%...
And I am talking RV670 vs. G92 only not nvidia vs. ATI (what is ATI, btw? Maybe you meant AMD?).
-FUDie
So what ? There is a bigger die.
But there's also (likely) a higher yield at 65nm than at 55nm, and TSMC probably charges more for the use of the 55nm half node anyway.
The memory used by the HD3870 is obviously more expensive than 8800 GT's (2.4GHz GDDR4 -running at 2.25GHz- vs 2.0GHz GDDR3 -running at 1.8GHz-).
Finally, the 8800 GT is able to command a price premium because it's faster, so the net profits also increase accordingly.
The profits come from the whole card, not from the GPU core alone.
Last edited by a moderator:
They've had higher margins for the last generations and nodes, sure, but how on earth does that necessarily translate to higher margins for G92 vs. RV570?
And how on earth does speculation to the effect that a chip ~50% smaller than the other has better margins turn into FUD?
Now, we obviously don't know the margins for AMD at 55nm yet, but it doesn't seem too far fetched to assume that (given the die size advantage) they'd be at least as good as nVidia's. Certainly your roll eyes emoticon and extrapolation based on older (and irrelevant) data is out of line.
This thread was doing so well...
nexus_alpha
Newcomer
Yields between node and corresponding 1/2 nodes are usually closely comparable, so this argument does not really hold water. Even if tsmc charged more the difference in die size more then compensate.
So since 65nm is more mature, manufacturing costs will likely decrease little, unlike 55nm where manufacturing costs will likely decrease more rapidly. And do we really know how the yields are on G92 vs. RV670? For all we know, RV670 is yielding very well.
If it's being priced at the same price point as RV670, where is the price premium? Hello? This was my whole point! nvidia won't want to sell G92 at the same price point as RV670.
-FUDie
See my updated post on the cost of onboard memory...
Fudie,
Many have tried to explain this to you already. Two critical aspects of margins is yield percentage and cost of production; the other aspect is the sales price. The yield percentage is higher and cost of production is lower for 65nm. The yield percentage is lower and cost of production is higher for 55nm. Nvidia is able to command a higher sales price for their chip while mainting higher yields and lower cost of production. That leads to higher margins.
So you think the Nvidia CEO is telling lies to the StockHolders? You do know doing so opens up the entire company to lawsuits and investigations by the SEC which will have a very large negative effect on the company.
Wake up and step into reality.
Many have tried to explain this to you already. Two critical aspects of margins is yield percentage and cost of production; the other aspect is the sales price. The yield percentage is higher and cost of production is lower for 65nm. The yield percentage is lower and cost of production is higher for 55nm. Nvidia is able to command a higher sales price for their chip while mainting higher yields and lower cost of production. That leads to higher margins.
So you think the Nvidia CEO is telling lies to the StockHolders? You do know doing so opens up the entire company to lawsuits and investigations by the SEC which will have a very large negative effect on the company.
Wake up and step into reality.
Read each companies quarterly report on what their product margins are. We'll know for certain after the next quarter.
Really? Nothing to do with layout or transistor count/die size? Woohoo! Let's go back to 90nm node!
Guilding the lilly... See Arun's comments.
Oh please, it's a CC not a regulatory filing... He's not actually giving enough detail to nail him to the cross over it, either. These are projections based on projected ASPs/yields/margins, etc, too.
Yes, and? That's what I said. We know nothing concrete about RV570 vs. G92 margins, which is contrary to your statement.
Your extrapolation from past data is like someone calling as statement before this season that McLaren would be in contention this year FUD pointing to the 'fact' that Renault obviously have the faster car based on the last couple of cycles worth of data.
Throwing around rolleyes and accusations of FUD based on a flawed comparison stated as 'fact' doesn't help discussion.
Bigger die = less/300mm wafer -> higher chance of > defect %. We'd need hard numbers to make any assertions.
We'd expect so, but the noise is that yields for NV on the 65nm node with 750m trans could be better. Weak evidence is release of 7C part first. We'll see more with volume of 8C availability & any 6C ASIC. AMD is using the 55nm node with 90m less trans & obviously different layout/libs, etc.
Volume bundling kills these direct comparisons. It's a case of price to market. I guess the argument proffered is that a higher ASP for 8800GT will only allow it to tread water against a lower priced 3870.
Yeah, but the margin on texolite for 512bit boards isn't going to differ much. Indeed the PCB appears cleaner for AMD series. Don't know what PWM power stages cost relative to analog, though. The big AIBs have been unhappy for some time with the bundling of GPU/RAM/HSF. Even less so when the entire board was outsourced to Flextronics, et al. Even the re-marketing outfits would like more price competition. Looks like AMD is allowing more flexibility this generation. Out of necessity, probably. Interesting aside.
If Nvidia has higher margins on the G92 than their corporate average, how does it stand that they are losing money on it, as stated by FUDie? Sure smells like FUD to me...
nice comparison of die size between HD2900XT and HD3870 at Hexus.
http://www.hexus.net/content/item.php?item=10415&page=5
http://www.hexus.net/content/item.php?item=10415&page=5
I've been told by a reliable source that AMD is pushing much lower cost than NV are in all areas. This allows them to compete on price without cutting into their margins like NV is doing already with G92. RV670 is 73% smaller than G92 and NV need twice as much wafers to get the same volume to the market. Partners also realise this.
Woah, that is impressive. Hopefully those kinds of prices will translate nicely to other countries too!
Yes.
