Why do you think they underproduced the GTX 1070 and 1080 GPUS?

[JacobusRex]

I mean, you think the folks at Nvidia would have made sure they had enough to go around for a year, let alone the first 2 weeks... What the heck happened? Got so excited to finally buy my build and all the reasonably priced 1070 and 1080 graphics cards are sold out. Sad Panda for sure.

[Deleted]

Cartmanland marketing strategy

[mrgreenthump]

I mean, you think the folks at Nvidia would have made sure they had enough to go around for a year, let alone the first 2 weeks... What the heck happened? Got so excited to finally buy my build and all the reasonably priced 1070 and 1080 graphics cards are sold out. Sad Panda for sure.

They probably felt that AMD had a too good of a deal with the RX 480 and needed to rush their cards out so they could sell them as the first 14/16 nm FinFET cards. It's not the first time this has happened and it's certainly won't be the last.

Also they already have the majority of the high-end market, so outside of making more money, they really have no incentive to make more than retailers can sell.

[Lordbrutus]

Every generation is the same thing.

[MatthewOU2015]

They wanted to rush and be the first to introduce a new line of graphics cards.

It's been like 6 weeks? Hell even the iPhone is in steady supply after a month. Nvidia messed up big time.

[JacobusRex]

I can't imagine why Nvidia is even the least bit intimidated by AMD. Cheaper price doesn't get buyers, selling superior products for a reasonable prices does. There was a time, long time ago, when I would have gladly purchased AMD products... until I experienced the hassle of their buggy drivers, command center non-sense and overheating issues. They're cheaper for a reason, that's what I always say about anything in technology, food and automotive products.

I just find it a bit... shady... that they said they "didn't anticipate the demand for the 1080 and 1070 graphics cards". Please forgive my noobiness as I am unfortunately not an "enthusiast" level PC consumer. I didn't realize they have these issues but then again it gives credence to the idea that they do it on purpose to make more money off of it. I'm a cynic already and this just makes my cynical senses vibrate like Hitachi...

- - - Updated - - -

They wanted to rush and be the first to introduce a new line of graphics cards.

It's been like 6 weeks? Hell even the iPhone is in steady supply after a month. Nvidia messed up big time.

That makes sense. Has it been 6 weeks? They should have fixed the problem by NOW then... yeeeeesh... Shady, shady, shady.

[mrgreenthump]

I can't imagine why Nvidia is even the least bit intimidated by AMD.

Because they can only loose market share, which from an investor's PoV is like poison.. AMD is currently far better option.

Also AMD is doing far better than they were a few months ago. When you are the leader in any market, the worst thing is to underestimate the capabilities of your competition.

[Fascinate]

I dunno but i wish i could buy an evga SC or asus strix for msrp, gogo. TBH ill prob wait on 1060 but if i saw one for MSRP id jump on it.

[Vash The Stampede]

I don't think the 16nm process has high yields. That and the 16nm process isn't ready.

[Deleted]

seems very much nvidia decided to push their releases ahead 1 month, despite not being totally ready.

initially all the rumors were nvidia and amd were both announcing at computex, then suddenly on short notice nivdia announces about a month earlier.
and now there are rumors going around that the 1060 launch is also being pushed ahead 1 month cause of the 480.

[chazus]

I don't think the 16nm process has high yields. That and the 16nm process isn't ready.

Pretty much this, with anything similar (Like CPUs)

The smaller the process two things happen.

1) They become more expensive. Theoretically, smaller units = more units = money saving... But that's not true. The new tech costs more, and as they get smaller, they lead to...

2) They have lower yields. Again, theoretically, smaller units = more units... However the smaller they get, and more advanced they are, the higher failure rate.

Lets say you have 10,000 units on a wafer, and 40% fail. You then shrink them by 40% (As they did from Maxwell to Pascal). You now have 14,000 units, at a 60% failure rate. You're literally producing less sellable units, at a higher cost, which take longer to produce (2.5 months vs 3 months)

Not only that, but foundries are having problems with 16/14nm process in general.

[Karon]

Pretty much this, with anything similar (Like CPUs)

The smaller the process two things happen.

1) They become more expensive. Theoretically, smaller units = more units = money saving... But that's not true. The new tech costs more, and as they get smaller, they lead to...

2) They have lower yields. Again, theoretically, smaller units = more units... However the smaller they get, and more advanced they are, the higher failure rate.

Lets say you have 10,000 units on a wafer, and 40% fail. You then shrink them by 40% (As they did from Maxwell to Pascal). You now have 14,000 units, at a 60% failure rate. You're literally producing less sellable units, at a higher cost, which take longer to produce (2.5 months vs 3 months)

Not only that, but foundries are having problems with 16/14nm process in general.

More Cheese = more holes

More holes= less Cheese

More Cheese = less Cheese

[chazus]

More Cheese = more holes

More holes= less Cheese

More Cheese = less Cheese

Thats... not accurate or even remotely applicable.

Unless you were to say.. cheese slices... more thinner cheese slices.. and every time you have holes in the cheese slice that slice gets tossed out.

If they could maker the wafer larger (and they do, in time), it would help to some degree. But they can only get so big. Every time the wafer size changes, the entire production plant needs to be rebuilt, costing hundreds of millions.

[JacobusRex]

So for us who don't know what you're referring to by "16nm process", "units", and "wafer", can you us simpler words? :P

Are you saying that basically the smaller the card and therefor the connections and things on the card are, the more meticulous it is to make it, thus slowing down the production? And are you also saying they launched a month early so they didn't have time to create a stockpile?

[justflayin]

So for us who don't know what you're referring to by "16nm process", "units", and "wafer", can you us simpler words? :P

Are you saying that basically the smaller the card and therefor the connections and things on the card are, the more meticulous it is to make it, thus slowing down the production? And are you also saying they launched a month early so they didn't have time to create a stockpile?

They're talking about the actual size of the chip (GPU). Chips sizes reduce with each new architecture.

[TyrianFC]

So for us who don't know what you're referring to by "16nm process", "units", and "wafer", can you us simpler words? :P

Think of the whole situation like this: If your name was "AMD", then Nvidia would be pissing on you every chance it gets.

[Biggles Worth]

All I know is in a rush to eat dinner I accidentally chose the default 4-7 business shipping method on a Friday and holiday weekend. BibleThump

[Remilia]

So for us who don't know what you're referring to by "16nm process", "units", and "wafer", can you us simpler words? :P

Are you saying that basically the smaller the card and therefor the connections and things on the card are, the more meticulous it is to make it, thus slowing down the production? And are you also saying they launched a month early so they didn't have time to create a stockpile?

Well, when it comes to silicon manufacturing there's a process node which is what's being referred for 14/16nm, which is the distance between each transistor essentially. So theoretically going from a 28nm node (previous generation) to 14nm node would reduce size by 1/2.

So when it comes to manufacturing silicon chips, they're produced into a gigantic circular board, which is called a wafer. Each wafer is fixed in size however, so a chip that's 314mm^2 (GP104 aka GTX1070/1080) vs a 232^mm2 (Polaris 10/RX 480), there'd be less chips per wafer on the GP104 than there is for a Polaris 10 one.

The other issue is yield, like every process there are potential defects that occur. Since this 14/16nm are relatively new to the playing field compared to 28nm, they're still working out the minor kinks in the process to increase yield, node maturity essentially.

Another issue is they also share factory (TSMC) with Apple, MediaTek and many other fabless companies, so it means they won't be getting as many wafers. Combine that with less chips per wafer and low yield, you get less chips and therefore less cards.

[Artorius]

Well, when it comes to silicon manufacturing there's a process node which is what's being referred for 14/16nm, which is the distance between each transistor essentially.

It's the distance between the drain and the source of a single FET actually. Which translates into the length of the gate.

@Missionary Jacobus
Since using words might be a little hard to imagine something from it:



It's essentially the same thing for GPUs.

[Remilia]

It's the distance between the drain and the source of a single FET actually. Which translates into the length of the gate.

Oh, my bad.