Nvidia to blame for its 600 shortage?

[ShintaiDK]

NV dismissed the idea of sourcing from GloFo in 2009, but times change: http://forums.anandtech.com/showpost.php?p=33356836&postcount=19

In 2009 people hoped the arabs would pour endless oil billions into GF to catch up in processnode and technology progress. It didnt happend. Instead GF turned into what it is, a foundry business that needs to make profit to its arab owners.

I think some of us still remember how much hype there was about GF in the start. It was another rotten dream.

AMDs recent payment to get out of the GF contract shows the truth. For everyone else but Intel, there is essentially only TSMC. And thats where AMD hope to get its future CPUs made. GF couldnt even lift that task.

GF is about 4 years behind technology wise. And here we dont just talk processnode. TSMC is between 2½-3 years behind.

 

[SirPauly]

This seems highly unlikely. A high end card has never outsold a mid range card. Simple economics shows that there is a LOT more people willing to pay 200-300 for a video card than 500+. I would like to know just where you have this unofficial data that states a just released, limited quantity, top end card is outselling a cheaper, higher quantity, mid range card.

Real easy to find:

http://store.steampowered.com/hwsurvey/directx/?sort=pct

 

[ShintaiDK]

Real easy to find:

http://store.steampowered.com/hwsurvey/directx/?sort=pct

Arcording to those stats, GTX 680 outsells HD7970 9 to 1.

 

[3DVagabond]

Arcording to those stats, GTX 680 outsells HD7970 9 to 1.

Sorry, but no sales numbers were compiled in the calculation of those statistics.

 

[SirPauly]

Indeed, but certainly much more to go on then your slim and none view -- another rationalization I guess.

 

[3DVagabond]

If you want to believe that there aren't supply issues with 680's, that the shortage is because of massive sales. That TSMC can supply cards for 6 models of AMD cards, but sales of the 680 are just so huge that they are completely outstripping TSMC's ability to manufacture them, you go right ahead. I'm not that gullible.

 

[ocre]

Arcording to those stats, GTX 680 outsells HD7970 9 to 1.

Back at the start of DX11 it was commonly used and considered a reliable source that represented a real average. The steam survey was presented as proof that fermi was a failure and AMDs 5000 series a smash hit. By it, people claimed AMD had 95% of the dx11 market and that nvidia wasnt selling many cards.

While i dont think its 100% accurate, those number do represent real gamers. I think that there was information that could be gather back then, during the 5000series, and today. It does say a lot. The 680 must be selling well and there must be more than 1000s units shipped worldwide. to be so high on the survey already, there has to have been a lot of them sold. Its not been out very long and nearly 9 times more steam players use a gtx680 over the 7970.

Steam is no small sample. Most PC gamers other than MMO dedicate have a steam account. You cannot even buy some of the most popular games now without playing through steam. as a matter of fact in 2010 valve had a whopping 30 million accounts.

http://www.rockpapershotgun.com/2010/10/18/thirty-million-steam-accounts/

with a year-over-year new user growth of 178%. But if we only use that 30million (which is absolutely surpassed by now), you can with simple math come up with a number of gtx680s that are out there.

steam users in 2010 X GTX680s share as of APR = number of gtx680s in the wild
30,000,000 x .0022 = 66,000


This number is not accurate, its severely under estimated. Using 2010 figures you could calculate 66000 gtx 680s that are gaming. there is absolutely no way charlie has any truth to his claims. Its a complete fabrication. The 66000 figure is lower than whats actually out there. Steam has more users now, and there are some ppl who dont use steam. One could argue that the mass of 7970 users dont use steam and therefor not counted. But you cannot posibly say that only 1000 gtx 680s were shipped world wide. Its just flat out proven wrong.

I would estimate that steam has at least doubled its user base in 2 yrs. If not they are close. We are looking at a lot more than 66,000, since the 680 launched just a month and a half ago.

 

[ocre]

If you want to believe that there aren't supply issues with 680's, that the shortage is because of massive sales. That TSMC can supply cards for 6 models of AMD cards, but sales of the 680 are just so huge that they are completely outstripping TSMC's ability to manufacture them, you go right ahead. I'm not that gullible.

read my above post. And according to steam, the 680 is 9 times more popular. This would be done in 6 weeks with the 7970 being out since Jan. I mean there is data right there that clearly suggest the 680 demand is extremely higher than the 7970 and a lot more popular.

Being 9 times more popular on steam yet only selling for 6 weeks? is that not enough to account for its supply issues???

I mean there is clear data in front of us. The same data everyone had no problem believing proved fermi was a flop with.

 

[blastingcap]

The steam survey was presented as proof that fermi was a failure and AMDs 5000 series a smash hit. By it, people claimed AMD had 95% of the dx11 market and that nvidia wasnt selling many cards.

You do realize that that number is mathematically true for at least one moment in time right? Because until Fermi showed up AMD had literally 100% of the DX11 GPU market. I also recall that even a quarter or so after the gtx480 debuted, AMD was still parading around a Mercury Research stat saying that AMD had 90% of the DX11 market. So it wasn't just Steam saying that.

It wasn't until the GTX 460 dropped that NV really sank its fangs into DX11 market share. But it did so in a big way.

 

[SirPauly]

If you want to believe that there aren't supply issues with 680's, that the shortage is because of massive sales. That TSMC can supply cards for 6 models of AMD cards, but sales of the 680 are just so huge that they are completely outstripping TSMC's ability to manufacture them, you go right ahead. I'm not that gullible.

I'm just offering some data to debate you on your slim and none point. There may be indeed issues from lower than expected yields and constraint issues -- I don't know.

 

[SirPauly]

The same data everyone had no problem believing proved fermi was a flop with.

I remember that but there needs to be caution with steam data and certainly not as accurate as Mercury Research or Jon Peddie but does indeed offer some insight for discussion.

 

[3DVagabond]

read my above post. And according to steam, the 680 is 9 times more popular. This would be done in 6 weeks with the 7970 being out since Jan. I mean there is data right there that clearly suggest the 680 demand is extremely higher than the 7970 and a lot more popular.

Being 9 times more popular on steam yet only selling for 6 weeks? is that not enough to account for its supply issues???

I mean there is clear data in front of us. The same data everyone had no problem believing proved fermi was a flop with.

No, not everyone accepted Steam data on Fermi as proving anything. I'm sure since you are referencing those discussions, you know that. We have no idea how or if Steam insures the accuracy of their numbers.

 

[Stuka87]

Arcording to those stats, GTX 680 outsells HD7970 9 to 1.

Just like we all told you in the last post where you paraded this around as fact, where are you getting this? There are ZERO numbers. Percentage points don't mean anything. Especially since AMD's chips have been on sale for 5 months. They cannot be compared on percentage points alone. Learn some basic math and stop spreading your false "Facts".

The only way we will learn how many are sold is when either IDC, or the companies themselves announce as much in their financial earnings for Q2.

As basing your "Facts" on an OPTIONAL Steam Survey is bogus. Most people I know (myself included) do NOT let Valve scan our machines to see what hardware and software we have. I consider it an invasion of privacy, as do many other people. This alone makes those percentage points worthless.

 

[SirPauly]

But one can see from the steam data there wasn't as much growth sequentially with April for the HD 7970 and this was the month both the HD 7970 and HD 7950 were kinda competing with the GTX 680 - AMD may of seen similar official data and decided upon price droppings and add some gaming titles to help spur sales.

 

[Arzachel]

Aww, not this crap again. How many times does it need to be repeated that the Steam HW survey isn't exactly what you'd call representative. No given sample size is quite an issue and if you believe that all 30,000,000 users have opted in and have been polled, I don't know what to tell you.

But one can see from the steam data there wasn't as much growth sequentially with April for the HD 7970 and this was the month both the HD 7970 and HD 7950 were kinda competing with the GTX 680 - AMD may of seen similar official data and decided upon price droppings and add some gaming titles to help spur sales.

Don't need no sales data to see that your products offer less for a higher price.

 

[ShintaiDK]

Just like we all told you in the last post where you paraded this around as fact, where are you getting this? There are ZERO numbers. Percentage points don't mean anything. Especially since AMD's chips have been on sale for 5 months. They cannot be compared on percentage points alone. Learn some basic math and stop spreading your false "Facts".

The only way we will learn how many are sold is when either IDC, or the companies themselves announce as much in their financial earnings for Q2.

As basing your "Facts" on an OPTIONAL Steam Survey is bogus. Most people I know (myself included) do NOT let Valve scan our machines to see what hardware and software we have. I consider it an invasion of privacy, as do many other people. This alone makes those percentage points worthless.

All I did was an observation. I didnt state it as fact.

Obviously it bothers you alot....wonder why. Oh I know why, its stated in your signature

So take a chillpill and relax.

 

[Arzachel]

All I did was an observation. I didnt state it as fact.

Obviously it bothers you alot....wonder why.

Observations don't get you out of having the burden of proof. Where are you seeing that 9 to 1 number? Because the survey shows a a third more 7970s than 680s.

 

[ShintaiDK]

Observations don't get you out of having the burden of proof. Where are you seeing that 9 to 1 number? Because the survey shows a a third more 7970s than 680s.

Look at the increase for the month.

HD7970 gained 0.02%. GTX 680 gained 0.18%.

 

[SirPauly]

Aww, not this crap again. How many times does it need to be repeated that the Steam HW survey isn't exactly what you'd call representative. No given sample size is quite an issue and if you believe that all 30,000,000 users have opted in and have been polled, I don't know what to tell you.



Don't need no sales data to see that your products offer less for a higher price.

Actually think both, at this time, offer incremental and evolutionary performance value for substantial and significant node and arch changes.

 

[ocre]

I remember that but there needs to be caution with steam data and certainly not as accurate as Mercury Research or Jon Peddie but does indeed offer some insight for discussion.

absolutely. but i think most ppl didnt even get my point. Clearly from the steam data you have got to come to the conclusion that the 1000 gtx680s worldwide cannot be true. In no way am I saying it is an accurate representation of total market share. What i am saying is it is clear that 1000 number is completely way way off.

The steam survey is gathered automatically, you dont take it. If you read the steam user agreement you agree to let them collect all sorts of information when you install it.

Steam collects/refreshes this data from the users that log on.

there are a lot more gtx680 users logging on than 1000. There are many more times that. There is enough data to conclude this much. The charlie D article is cannot possibly be the truth.

 

[Haserath]

I don't see what's wrong with steam's survey...

Who cares if some don't opt in? Isn't a GTX 680 owner just as much likely to opt out as an HD 7970 owner?

It has a huge sample size. What research firm has >30 Million gamers to choose from?

Who cares if it's just percentages? The percentage represents the number of gamers that have certain hardware out of the whole tested.

No research/calculation is ever 100% correct... An exact number would have to come from inside Nvidia/AMD to know exactly how many of their chips were sold. I'm sure sometimes in a massive operation like this, some inside numbers aren't quite right either(2 Billion transistors in Bulldozer?).

The only thing wrong with something like this is if Steam is fudging the numbers a little. It's not like their required to present factual data, but I don't know why they would do otherwise.

 

[SirPauly]

absolutely. but i think most ppl didnt even get my point. Clearly from the steam data you have got to come to the conclusion that the 1000 gtx680s worldwide cannot be true. In no way am I saying it is an accurate representation of total market share. What i am saying is it is clear that 1000 number is completely way way off.

The steam survey is gathered automatically, you dont take it. If you read the steam user agreement you agree to let them collect all sorts of information when you install it.

Steam collects/refreshes this data from the users that log on.

there are a lot more gtx680 users logging on than 1000. There are many more times that. There is enough data to conclude this much. The charlie D article is cannot possibly be the truth.

I would think so, considering the sheer amount of partners, default sku's, AIB differentiation sku's, and to try to feed demand for them based on its architectural balance with nVidia's feature differentiation through out the world. Not to mention, with a release of a GTX 670, with default and AIB differentiation sku's and the GTX 690. Sounds like very hard and very challenging work to me.

It may be a combination of lower than expected yields, constraints and a strong demand for the GK-104.

Would love a creation of an article called: The birth of the GPU -- from the planning stages; to actual silicon; to software; to finished product; to marketing and sales; to the work that actually goes into these products millions upon millions are so passionate about.

 

[Zebo]

In 2009 people hoped the arabs would pour endless oil billions into GF to catch up in processnode and technology progress. It didnt happend. Instead GF turned into what it is, a foundry business that needs to make profit to its arab owners.

I think some of us still remember how much hype there was about GF in the start. It was another rotten dream.

AMDs recent payment to get out of the GF contract shows the truth. For everyone else but Intel, there is essentially only TSMC. And thats where AMD hope to get its future CPUs made. GF couldnt even lift that task.

GF is about 4 years behind technology wise. And here we dont just talk processnode. TSMC is between 2½-3 years behind.

Nothing against arabs but ME isn't exactly what I'd call Silicon Valley save maybe Israel. GF was doomed under that direction. Anyway what about Samsung?

 

[DuddBudda]

I'd like to point out, for the chaps debating the steam stats, that the 7970 saw the exact same rate of adoption, in the first two months after launch, as the 680

now there's no raw data, but it's not unreasonable to assume that a sample of 30,000,000 produces sufficiently normalised results for our purposes


so we know that:

- after launch the 7970 had far less time out of stock than the 680
- the 7970 uses the same process
- the 7970 has a 35% larger die area
- something else I forgot while I was looking at data

as such, we must recognise that the 680 is experiencing manufacturing problems that the 7970 did not


PS Hi, I'm new here; I butt in on your shiz

 

[3DVagabond]

Just curious to those claiming Charlie said there were only 1000 worldwide. Can you point that out? I don't ever recall reading that myself.

He said 1000 in the initial order and less than 10,000 to date (link). That was on the 1st of May. Approx 1.5wks. ago. This is where someone states a fallacy as true and all of those who like the sound of it assume it's correct.

Saying that though, I still wouldn't put much into Charlie's figures. That still doesn't justify posting total BS just to attack his credibility. Which is what has happened here.