Nvidia misses Revenue estimate - Huang says AMD to blame

[DimasRMDO]

*Wonders how their revenue estimation misses*

*Blames competitor*

 

AMD:

[Lathlaer]

1 hour ago, S w a t s o n said:

Steam doesn't show "gamers dont like the rx580" what editorialized bullshit is that?

I believe that when @Tech Enthusiast said that the survey shows "gamers don't like the RX580" he meant that the card isn't selling that good in the gaming department. Well-received - as in "well sold" and not actually "badly reviewed".

 

Same as you can say that an excellent movie with good reviews was poorly received because it actually lost money because people didn't want to go and see it.

 

14 minutes ago, DimasRMDO said:

*Wonders how their revenue estimation misses*

*Blames competitor*

If you read the article then you know that he actually doesn't wonder and doesn't blame but explains how the inventory mechanism works which was what analysts based their predictions on. 

 

It literally says in the article:

 

I talked to 

Jensen Huang, CEO of Nvidia, who was puzzled at the reaction to the earnings miss, which included a $57 million charge related to the “post-crypto falloff.

 

That means he knew exactly that analysts overshot their estimations but it was an interview so he had to explain it anyway. 

[mr moose]

3 hours ago, S w a t s o n said:

He can pull claims out of his ass but without proof it's just a pipe dream

Only becasue you don't want it to be true.   Seriously,  do you expect everyone who makes a claim to back it up otherwise you call them a liar?  Did you call Huddy a liar when he spoke about Gsync?  I assume because AMD had no evidence when they called out GPP that you also accused them of pulling it out of their ass?

[leadeater]

5 hours ago, Tech Enthusiast said:

And that is exactly why i exclude workstations. They don't use a low-midrange GPU build for gamers.

Depends who's buying, RX 580 are options for HP, Dell and Lenovo workstations. Not everyone can afford a Quadro or Radeon Pro and when you just need a GPU, literally any GPU, for application acceleration the cheapest will do. There's a metric ton of business computers with what we would call completely illogical GPU choices in them, like GT 730's.

 

But you missed the actual point, none of these will ever show up in a Steam hardware survey and Nvidia doesn't know any more about AMD GPUs sales than anyone else with the public sales data and distributor industry contacts, which market analyst have.

 

You can tell just how far off the Steam survey is simply based on discrete GPU shipment market share.

 

We can theorize all day long how much of that is mining sales but need I remind you of the countless images online of large scale GPU mining using GTX 1060 and GTX 1070 GPUs as well, the giant warehouse kind who buy out everything they can for any vendor, any supplier, any GPU in that price and performance range. I can say with decent confidence Nvidia is either understating the mining sales of their GPUs or are more unaware than they realize.

 

Steam survey is about the most clear cut example of biased data, I don't mean that in a bad way just in the mathematical/statistical sense. There is no sample size given or standard deviation. There is no way for anyone analyzing the data to know how accurate and representative it is, limited in scope of participants and the way the human mindset works the more likely you will submit your results if you think you have a good system. Brag factor is very much a thing, RGB lighting exists for a reason .

 

If this data gives you any kind of perception that RX 580 are essentially irrelevant to gamers and the gaming market then so are the GTX 1070 Ti, GTX 1080 and GTX 1080 Ti. How you perceive the data is up to you but for me it's nothing more than interesting to look at and see how it changes but it's statistically irrelevant due to how the data is presented.

 

"Gaming" GPUs don't exist, there's only professional and non professional (with a recent side order of mining GPUs). What makes everything even less clear is that non professional cards are purchased and used for professional use cases. All the GPUs the researchers use here are gaming GPUs because it's financially irresponsible to buy the professional cards when they offer nothing to them and they work on very tight budgets, you're better off relying on failure rates (which are very low) and covering failures yourself and you will always be in a massive net financial win. I hate that this is the way it actually is, being in IT the idea of something equivalent as a desktop being used as a server for multi million dollar purposes is just wrong but running CUDA code on a GPU is a very far cry from something like that even when it involves the same kinds of money.

 

Why do you think Nvidia tried to stop people using gaming GPUs in servers and datacenters? Because people with good education, care about cost and know there is no difference at all will always opt for the cheaper choice. They can not like it all they want but the only way it'll stop is to lower the price of professional cards and that won't happen.

 

It really is not that easy to analyze the market like people want to see or try to do, just go in to assuming that you'll be wrong or what you'll see is wrong to some degree. Sales and shipment data is a step up more reliable than Steam survey is though.

[leadeater]

4 hours ago, Tech Enthusiast said:

The numbers are right there. The context does not make this go away. Without the mining, AMD would probably be dead on the ground. And that mining craze is kinda gone now. So shutting our eyes and ears is not gonna help much.

 

But how can you say the RX 580 is well-received by gamer if the freaking 1080ti outsells it by a factor of 3. The 1060 (same segment) by a factor of 30. You can't just mix and match your arguments to make it sound better. You insist on not bringing up steam, but you also insist on gamers loving the GPU. Steam shows that gamers don't like the RX 580. In fact, you could argue a market share of 1% (480 and 580 combined!) means they avoid it.

Seems to me Nvidia themselves are saying that with the mining demand gone sales are matching even between them. With that kind of statement, from a competitor, it would be more reasonable to say that without the mining craze far more gamers would have RX 580 than they do and that in no way would they be dead.

 

I would certainly never suggest that it would actually be a 50/50 split of sales for GTX 1060s and RX 580s though, more people want Nvidia so I would expect the more traditional 70/30 or 60/40 type split.

[Tech Enthusiast]

5 hours ago, S w a t s o n said:

You what? Are you just an Nvidia shill? You think there is no market for the RX580? Do you think everyone lives in tech tuber land with triple 4k monitors? Steam doesn't show "gamers dont like the rx580" what editorialized bullshit is that? It's highly recommended by people within it's price bracket. Nvidia enjoys a mindshare advantage but the card was well received. Maybe put on your rational thinking cap next time

You don't seem to want to understand.

I even said that we, the techy people, need to help AMD out by letting others know the RX 580 is a great product.

 

But just because we know it is a great product, does not mean it sells well.

How you can still insist on it being widely used by gamers, even after seeing its share of less than 0,5% is just mind boggling. The meme with a burning house and a dog comes to mind to be honest.

 

To make it 100% clear: I know the RX 580 is great. The reviews know it is great. The people buying gaming GPUs don't care and buy 1060 anyways. Facts are right there in the survey. Claiming Steam is not a valid source for gaming market share is just delusional.

[Tech Enthusiast]

41 minutes ago, leadeater said:

Steam survey is about the most clear cut example of biased data, I don't mean that in a bad way just in the mathematical/statistical sense.

That is exactly the point why it is so great. It shows gamer data only.

I don't claim the RX 580 is selling badly at all. I even think it is the most sold GPU for mining ever. But we really can't tell.

 

What we indeed can tell is that Steam is, by far, the biggest gaming Platform for PCs and if they show a market share for a GPU, that means it is representative for gamers.

If we are talking about popularity across gamers, there really is no debate to be had. The little possibilities you mentioned like bragging rights would a very valid reason for a discussion if the numbers would be close enough that such things could be a determining factor. But we are looking at a very, very one sides set of numbers here. 

 

I am all up for letting possibilities and reasoning be part of a discussion why numbers may be off by a few percent, but most definitely not by THAT much. We are talking a factor of 1.05x or even 1.5x here, but a factor of 30x. No way in hell can you argue a factor of 30 is actually just a 70/30 share.

 

Again: This is for gamers only. I don't care about workstations or anything else, because we don't have any numbers backing up any claims we may try to make. We do have those for gamers only.

[leadeater]

34 minutes ago, Tech Enthusiast said:

What we indeed can tell is that Steam is, by far, the biggest gaming Platform for PCs and if they show a market share for a GPU, that means it is representative for gamers.

They don't list sample size or SD though, without these for all we know any data points below 2% fall outside of confidence and it's no better than a random guess.

 

But separate from this you used the data to posit that RX 580 are irrelevant to gamers when the source information for this story more closely supports that gamers have just been unable to buy the GPU and current sales are in fact rather close in parity between the competitors. If you want to trend an outcome in the context of mining sales having not been there for AMD then the information more closely supports that the Steam survey share would be higher than it is now.

 

Edit:

34 minutes ago, Tech Enthusiast said:

Again: This is for gamers only. I don't care about workstations or anything else, because we don't have any numbers backing up any claims we may try to make. We do have those for gamers only

We do which are total unit shipments and market share, data far more reliable than Steam survey. They are known data points, unknown usage of the sold device but a sale/shipment is that. You don't have infer anything from the data to know how things are stacking up in sales specifically. Steam survey is just a sample or a larger data set, sample size needs to be large enough to have confidence in the prediction. Make predictions from information outside of confidence is pointless, the difference could be 10:1 or 30:1 which is not a minor difference.

[Tech Enthusiast]

Well, the sample size will be in the tens of thousands at a minimum judging from the sheer number of entries and percentages that go to the 1/100th, which is enough to predict millions of sales.

 

So, the shown 30:1 ratio is not large enough to just say: "well, nothing we can argue away." then i guess.

I strongly disagree with that. Seeing a 0,49% share is not magically gonna jump to the 14% of a 1060 due to "whatever is being used to make the numbers hurt less".

I really can't disagree anymore on this point. No magic in the world and arguing can change the numbers enough to even come close to a 50/50 split. There would have to be so many errors, surprises and unpredicted events that i just flat out claim it is impossible.

 

Sure, we have no numbers for workstations and maybe i am totally wrong with those and companies just love to use low end GPUs for the work that is paying their rents, but again: I am strictly talking about gamer share here, because that is what we have solid numbers for. All the arguing against them is purely out of thin air, without a single proof of evidence at all. And i am baffled at how it is possible to just ignore what is right in front of us, just because we don't want to see it as it is.

[leadeater]

1 hour ago, Tech Enthusiast said:

I strongly disagree with that. Seeing a 0,49% share is not magically gonna jump to the 14% of a 1060 due to "whatever is being used to make the numbers hurt less".

That's not the issue, if the error margin is +/- 0.3% then showing a share of say 0.6% means it could actually be between 0.3% to 0.9%. This greatly effects if something is out selling by a factor of 30 or just 10 or 5.

 

If there are 50 million Steam users then a sample of 10,000 may not be enough to get enough confidence in those lower represented GPUs. A 0.3% error on a 15% share is not a big deal, it's a really damn big deal on a 0.3% share. And that's not even getting in to how Steam decides on who to send the survey to.

 

1 hour ago, Tech Enthusiast said:

No magic in the world and arguing can change the numbers enough to even come close to a 50/50 split.

They are talking about this point in time, it could be true for 1 week or 4 weeks but it might not be for a full quarter. Plus 50/50 was just an example sake, same with his 90/10.

 

Quote

Huang: My point is that we just don’t know — when we’re going through what we see as our own sales, and what the percentage of Bitcoin is in our own sales, what we don’t know is how much inventory AMD pushed into the channel. There’s no way for us to calculate that.

Nvidia doesn't know nor can calculate anything about AMDs actual sales information, they say as much. They can insinuate as much as they like about how much of AMDs products are mining sales or gaming but they simply don't know.

 

1 hour ago, Tech Enthusiast said:

I am strictly talking about gamer share here, because that is what we have solid numbers for.

When you don't know population size, sample size and error then they are not solid numbers.

 

The market share I posted is far more accurate and reliable, whether you want to believe 90% of their sales is mining is up to you but pre-mining their share was actually more and it's recovering to that point or has already. AMD is perfectly capable, and has in the past maintained around 30% market share with zero of that being mining. Good products sell, Geforce 9 and 10 series were in large part across the range better products so they sold more and AMD's market share dropped, now with pricing changes and slight product optimizations AMD has competitive products in the high volume section of the market which makes it very easy to pickup market share, this why market share wise Vega 56/64 is actually irrelevant.

[S w a t s o n]

5 hours ago, mr moose said:

Only becasue you don't want it to be true.   Seriously,  do you expect everyone who makes a claim to back it up otherwise you call them a liar?  Did you call Huddy a liar when he spoke about Gsync?  I assume because AMD had no evidence when they called out GPP that you also accused them of pulling it out of their ass?

What? You literally can't apply your argument to him and me and have it make sense. Either evidence matters or it doesnt make up your mind.

[S w a t s o n]

4 hours ago, Tech Enthusiast said:

You don't seem to want to understand.

I even said that we, the techy people, need to help AMD out by letting others know the RX 580 is a great product.

 

But just because we know it is a great product, does not mean it sells well.

How you can still insist on it being widely used by gamers, even after seeing its share of less than 0,5% is just mind boggling. The meme with a burning house and a dog comes to mind to be honest.

 

To make it 100% clear: I know the RX 580 is great. The reviews know it is great. The people buying gaming GPUs don't care and buy 1060 anyways. Facts are right there in the survey. Claiming Steam is not a valid source for gaming market share is just delusional.

I claimed steam hardware survey does not disprove my claim of Jensen lying about it being the 24th best selling gpu. You know that exactly what I said, again splitting 480 from 580 is disingenuous in the same way that I'm betting huang included 3gb and 6gb 1060's in his claim even though they are much more different than a 480 vs 580.

 

Really dont see why everyone is tryharding for huang here so badly. As if I'm claiming they arent selling GPUs or something. I'm just saying he's making the 580 look like it's selling much worse than it is and you're playing right into him. If you actually believe that we should be helping people understand the actual GPU market then you shouldnt be playing into his mindgames.

[leadeater]

34 minutes ago, S w a t s o n said:

I claimed steam hardware survey does not disprove my claim of Jensen lying about it being the 24th best selling gpu.

Well I have a question, what are the other 23 GPUs that are selling? Not what people have but selling because I'm struggling to come up with that many active in production in network GPUs that you can buy.

[S w a t s o n]

1 hour ago, leadeater said:

Well I have a question, what are the other 23 GPUs that are selling? Not what people have but selling because I'm struggling to come up with that many active in production in network GPUs that you can buy.

This^

 

Like to put this situation into context. This is the CEO of the rival company in an effective duopoloy (discrete graphics, yes intel will enter soon maybe) giving us super unspecific sales number information about AMD's products. That's not a statement you just immediately trust and give benefit of the doubt. If AMD makes a statement about Nvidia sales you should be skeptical too.

[Curufinwe_wins]

1 hour ago, leadeater said:

Well I have a question, what are the other 23 GPUs that are selling? Not what people have but selling because I'm struggling to come up with that many active in production in network GPUs that you can buy.

Likely almost all mobile gpus. Maxwell is still available in the market for laptops, and so are the real bottom barrel things like the 1030/730. Also AMDs mobile GPUs.

 

[S w a t s o n]

1 minute ago, Curufinwe_wins said:

Likely almost all mobile gpus. Maxwell is still available in the market for laptops, and so are the real bottom barrel things like the 1030/730. Also AMDs mobile GPUs.

 

You seriously think the 1030,730, maxwell mobile and amd mobile each outsell the RX580?

[Curufinwe_wins]

1 minute ago, S w a t s o n said:

You seriously think the 1030,730, maxwell mobile and amd mobile each outsell the RX580?

When the 580 hadn't been available at ALL for the hyper majority since release? Yes. Yes I do. Without a doubt.

 

Of course that no longer is true, but Huang literally says this quarter the 580 was 50/50 with the 1060, which directly implies it became a top 3 seller. Before that? Totally believable. 

[S w a t s o n]

Just now, Curufinwe_wins said:

When the 580 hadn't been available at ALL for the hyper majority since release? Yes. Yes I do. Without a doubt.

 

Of course that no longer is true, but Huang literally says this quarter the 580 was 50/50 with the 1060, which directly implies it became a top 3 seller. Before that? Totally believable. 

It wasn't available because it was selling......just because it sold to miners doesn't mean it didn't sell. If we're just gonna disagree on whether he meant gaming sales or sales in general then we should just agree to disagree. Btw, if you think nvidia and amd know where for example xfx's or asus gpus go to specifically or something you're crazy. Example: Newegg doesn't have a self reporting survey tool to collect buyer's intentions, you could correlate via size of purchase but newegg isnt giving that info to other companies.

[Curufinwe_wins]

12 minutes ago, S w a t s o n said:

It wasn't available because it was selling......just because it sold to miners doesn't mean it didn't sell. If we're just gonna disagree on whether he meant gaming sales or sales in general then we should just agree to disagree. Btw, if you think nvidia and amd know where for example xfx's or asus gpus go to specifically or something you're crazy. Example: Newegg doesn't have a self reporting survey tool to collect buyer's intentions, you could correlate via size of purchase but newegg isnt giving that info to other companies.

You are right. They don't "know". They do spend millions if not billions of dollars on market research and pay for every survey under the sun to try to get the best analytics out there, with far more detail than we get with the little summaries that entice people into buying past the paywall. Plus I'm sure Nvidia and AMD have most if not all of those distributor datametrics themselves.  

 

The market analytics the comments are based on could also be wrong. But defaulting to "he lied", when lying to shareholders itself is a crime, and we don't have any evidence that the 580 was a popular seller (all the evidence, albeit non-conclusive is to the contrary) among the non-crypto crowd is beyond "measured skepticism" or "reasonable doubt".

[S w a t s o n]

29 minutes ago, Curufinwe_wins said:

You are right. They don't "know". They do spend millions if not billions of dollars on market research and pay for every survey under the sun to try to get the best analytics out there, with far more detail than we get with the little summaries that entice people into buying past the paywall. Plus I'm sure Nvidia and AMD have most if not all of those distributor datametrics themselves.  

 

The market analytics the comments are based on could also be wrong. But defaulting to "he lied", when lying to shareholders itself is a crime, and we don't have any evidence that the 580 was a popular seller (all the evidence, albeit non-conclusive is to the contrary) among the non-crypto crowd is beyond "measured skepticism" or "reasonable doubt".

 

They see numbers of gpu moving to sellers but not to the end user unless you're Sapphire and partner with AMD to directly seed the mining market for very large operators in china. Other than situations like that all they see is massively increased demand as the retailer orders more from the aib and the AIB order more gpus from nividia. they can easily say it's crytpo causing the demand but they can't easily seperate out "gaming sales". There's literally no chance Huang knows the "gaming sales" of the RX580, I'm not sure AMD does. But the point is he definitely lied.

That is not how lying to shareholders is illegal. This was an interview with venture beat that he said this in. He can definitely lie about the RX580 sales numbers just like he can "lie" about turing not coming soon like 2 months before launch.

 

Quote

When reporters asked when the new GeForce GPUs will launch, Nvidia CEO Jensen Huang responded, “It’s a long time from now.”

^ Jun 4th
 

Aug 13

[Tech Enthusiast]

4 hours ago, leadeater said:

That's not the issue, if the error margin is +/- 0.3% then showing a share of say 0.6% means it could actually be between 0.3% to 0.9%. This greatly effects if something is out selling by a factor of 30 or just 10 or 5.

 

If there are 50 million Steam users then a sample of 10,000 may not be enough to get enough confidence in those lower represented GPUs.

Even if it was just a factor of 10:1, it would still be pretty clear cut. That is my point.

Also error margin does not simply get added, unless you only have two options to pick from. With about 50+ Options you can't just assume all answers that deviate from the real result can be added to the option you want to "win".

 

Check this calculator to see how easy it is to get VERY accurate data.

If we assume your numbers with 50 million Steam users and a sample size of 10.000, we get an error margin of 0.98%. Which would make RX 580 share "jump" to about 0.5 due to over 50+ possible outcomes. So it would not change at all. The confidence rating is well above 99% as well.

 

Basically the steam survey is accurate, statistically speaking.

We can insist on not believing what they say, but why in the world would steam fake these statistics?

 

[Tech Enthusiast]

4 hours ago, S w a t s o n said:

If you actually believe that we should be helping people understand the actual GPU market then you shouldnt be playing into his mindgames.

I do.

But I am more sick of people accusing everyone of lying left, right and center.

You have zero data that proves your point, but there is plenty of evidence that he may actually be correct. Unless you can prove he is lying, id suggest you just don't. You would not do that if you would be standing right in front of him. Well and he would probably sue you if you indeed would have the balls. You are playing a very risky game here. There has been suing for forum comments in the past and you really don't even try to sound reasonable at all.

[S w a t s o n]

1 hour ago, Tech Enthusiast said:

I do.

But I am more sick of people accusing everyone of lying left, right and center.

You have zero data that proves your point, but there is plenty of evidence that he may actually be correct. Unless you can prove he is lying, id suggest you just don't. You would not do that if you would be standing right in front of him. Well and he would probably sue you if you indeed would have the balls. You are playing a very risky game here. There has been suing for forum comments in the past and you really don't even try to sound reasonable at all.

He would sue me? You really have lost it

[Curufinwe_wins]

1 hour ago, Tech Enthusiast said:

Even if it was just a factor of 10:1, it would still be pretty clear cut. That is my point.

Also error margin does not simply get added, unless you only have two options to pick from. With about 50+ Options you can't just assume all answers that deviate from the real result can be added to the option you want to "win".

 

Check this calculator to see how easy it is to get VERY accurate data.

If we assume your numbers with 50 million Steam users and a sample size of 10.000, we get an error margin of 0.98%. Which would make RX 580 share "jump" to about 0.5 due to over 50+ possible outcomes. So it would not change at all. The confidence rating is well above 99% as well.

 

Basically the steam survey is accurate, statistically speaking.

We can insist on not believing what they say, but why in the world would steam fake these statistics?

 

In fairness to @leadeater... it can be accurate (rather precise... but still) without being representative. Or rather... it can be representative of the sub-group that includes steam-users eligible for survey, without being holistic of the entire GPU market space...

[TechyBen]

3 minutes ago, Curufinwe_wins said:

In fairness to @leadeater... it can be accurate (rather precise... but still) without being representative. Or rather... it can be representative of the sub-group that includes steam-users eligible for survey, without being holistic of the entire GPU market space...

This. People using an Intel Laptop for work only, won't be in the Steam survey. People using Oculus only, won't be in the Steam Survey.

 

It's a cherry picked dataset. And only really tells you what gamers/Steam gamers use. Not what general gamers (Many many MMOs out there that don't touch Steam).