• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

JPRFor the 4th quarter of 2015, GPU shipments increased 2.4% from last quarter

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
AtenRa said:
Also is you see what GAMEs each country plays the most, and what is the top 5 dGPUs you will scratch your head.

Example, USA
http://steamspy.com/country/US

Most played games are

1) CS : Global
2) Garry's Mode
3) TF2
4) DOTA 2
5) Rocket League

According to Hardware Survey of January 2016 , GTX970 is the Top dGPU. Does people buy the GTX970 to play those 5 games ??

You will have to go to number 6 and Fallout 4 to really need a card like GTX 970.

That means that most registered Steam Users dont participate in the Hardware Survey each month.
Click to expand...

The most common gpu being used in the steam hardware survey is intel graphics. All five of those games are playable with integrated graphics.

USER8000 said:
Steam wouldn't be a better indicator. Firstly around 10% to 15% of all PCs don't run Windows.

Then take the Windows PCs,and what percentage of systems with dGPUs will be used for gaming under Windows??

Then look at some games like WoW,which don't use Steam.

Then,add the fact the Steam survey is optional,and people like me had a break of a few YEARS to be asked,when I changed most parts of my system.

Unless you really honestly think more GTX970 cards were sold than GTX750 or GTX750TI derived cards in ALL PCs and all GTX960 derived cards in ALL PCs.

You might want to see what many prebuilt PCs and laptops are using.

Loads of the ones under £1000 even in the UK or Germany are mostly using GM107 and GM206 based chips.

The GTX750 and GTX750TI were out 10 months before the GTX970 alone. Nvidia and AIB partners have consistently said its a very popular card to the extent that they have been even using GM206 chips to meet the demand.

Its even been proven before on this very forum,the Steam metrics are quite wonky in picking up marketshare changes.

Its a useful data point but in the end it is deeply flawed too.

I would rather trust units shipped over an optional survey which only a small percentage of actual gamers will answer.

Valve has not even told us the exact sample number too.

I still expect lots of dGPUs to be sold for non-gaming use on top of this.

Even the units ship method is flawed,so in the end we can only get some rough indications of where changes are happening.

Only Microsoft will have any decent reliable metrics of actual Windows PC dGPU marketshare.
Click to expand...

Lots of people are making the mistake of assuming that because the steam survey does not sample certain areas of the market the steam survey is meaningless. Thats only the case if the areas not being sampled have a different composition than the sampled areas.

The area of the market that steam very likely undersamples is the low end (prebuilt) which may contain gpus but are not used for gaming.
 
maddie said:
I guess that these vendors must be hoping to be 1st in line for Polaris. Imagine ordering GPU dies that never get sold. Polaris must be fantastic.

Just to be clear to everyone, I'm being partially facetious.

Is it really being implied that GPUs are ordered from AMD but never get shipped or sold. Oh my.
Click to expand...

I'm saying shipments are shipments. Who knows how many were sold? Sales could be up 40% if vendors had massive inventories, and sold all those inventories and ordered more. Sales could be even, sales could be down. Shipments doesn't exactly = sales. If we don't know how much inventory these vendors were holding of AMD products we really can't make much of shipments other than to say there is an increase in demand for AMD products from Vendors....
 
tential said:
It's a positive indicator, we need AMD sales numbers to verify really.
Click to expand...

Looking at the individual segments, the Computing and Graphics segment had revenue of $470 million for the quarter, up 11% since last quarter but down 29% year-over-year. AMD had more notebook processor sales compared to last quarter, but lower client processor sales compared to last year. This segment had an operating loss of $99 million, compared to $181 million last quarter and $56 million last year. The inventory write-down last quarter was the main reason for the improvement this quarter, and lower sales caused the year-over-year drop. One good nugget for AMD is that average selling price (ASP) increase sequentially for processors, although it is down year-over-year, but GPU ASP increased both sequentially and year-over-year.

this is as far as we would go since amd doesnt seem to want to brake the cpu department from the gpu any time soon..
 
Enigmoid said:
The most common gpu being used in the steam hardware survey is intel graphics. All five of those games are playable with integrated graphics.



Lots of people are making the mistake of assuming that because the steam survey does not sample certain areas of the market the steam survey is meaningless. Thats only the case if the areas not being sampled have a different composition than the sampled areas.

The area of the market that steam very likely undersamples is the low end (prebuilt) which may contain gpus but are not used for gaming.
Click to expand...

It's sad that people can't understand how to use a sample/survey.

We make national assumptions based on very very small sample sizes to national population. We do it because we have a diverse and large enough sample size to make a confident (confidence interval), assumption about the whole population.

It's all basic statistics which I'm surprised many people can't seem to wrap their heads around.

Steam is a far larger sample size than any other survey we would do.

Do people in this thread really think when they see national polls for anything that they polled at least 60%+ of the nation just to be safe? They poll a very very tiny portion.....
 
Of course they do also control those surveys in all sorts of ways, but yes some of this has been surprising me a little 🙂

Systemic issues like in Hitman's post a little bit above would be another kind of issue of course.
 
maddie said:
I guess that these vendors must be hoping to be 1st in line for Polaris. Imagine ordering GPU dies that never get sold. Polaris must be fantastic.

Just to be clear to everyone, I'm being partially facetious.

Is it really being implied that GPUs are ordered from AMD but never get shipped or sold. Oh my.
Click to expand...

Stuffing the channel doesn't mean its paid for yet. Hence you dont see the revenue for AMD that's supposed to follow.
 
tential said:
It's sad that people can't understand how to use a sample/survey.



We make national assumptions based on very very small sample sizes to national population. We do it because we have a diverse and large enough sample size to make a confident (confidence interval), assumption about the whole population.



It's all basic statistics which I'm surprised many people can't seem to wrap their heads around.



Steam is a far larger sample size than any other survey we would do.



Do people in this thread really think when they see national polls for anything that they polled at least 60%+ of the nation just to be safe? They poll a very very tiny portion.....
Click to expand...


Yes and no, but good points none nonetheless. Most political polls are national, in that there is more than one or two states participating and usually those who identify has 50/50 republican and democratic split. The steam survey is useful, but would not show the entire picture related to the OP. That's really the conclusion. Its useful for yearly trends and what not but nothing in the timeframe everyone is talking about.
 
tential said:
It's sad that people can't understand how to use a sample/survey.

We make national assumptions based on very very small sample sizes to national population. We do it because we have a diverse and large enough sample size to make a confident (confidence interval), assumption about the whole population.

It's all basic statistics which I'm surprised many people can't seem to wrap their heads around.

Steam is a far larger sample size than any other survey we would do.

Do people in this thread really think when they see national polls for anything that they polled at least 60%+ of the nation just to be safe? They poll a very very tiny portion.....
Click to expand...

Lets say we have 100 total steam users

60 of them use iGPUs (60%)

30 of them use NVDIA dGPUs (30%)

10 of them use AMD dGPUs (10%)

January 2016 hardware survey,

30% of iGPU users participated on the survey = 18 iGPU users

70% of NVIDIA dGPU users participated on the survey = 21 NVIDIA dGPUs

60% of AMD dGPU users participated on the survey = 6 AMD dGPUs.

January 2016 results

45 total users participated (45% of the total users).

NVIDIA dGPUs = 46,66%
iGPUs = 40%
AMD dGPUs = 13,3%
 
Last edited:
AtenRa said:
Lets say we have 100 total steam users

60 of them use iGPUs (60%)

30 of them use NVDIA dGPUs (30%)

10 of them use AMD dGPUs (10%)

January 2016 hardware survey,

30% of iGPU users participated on the survey = 18 iGPU users

70% of NVIDIA dGPU users participated on the survey = 21 NVIDIA dGPUs

60% of AMD dGPU users participated on the survey = 6 AMD dGPUs.

January 2016 results

45 total users participated (45% of the total users).

NVIDIA dGPUs = 46,66%
iGPUs = 40%
AMD dGPUs = 13,3%
Click to expand...

Seriously, can we please start learning how sampling works?

https://en.wikipedia.org/wiki/Sampling_distribution

You just made my point that statistics is being BUTCHERED in this thread and people do not understand how sampling works...

Edit: To clarify, you can't just throw in random participation rates.

With your logic, I can say, what if 100% of nvidia users people participated, 0% of igpu user and 0% of AMD users.
You see, there is also this thing called probability of a sample being drawn. So the probability of your sample is low... hence why we don't see it happen.
 
Last edited:
ShintaiDK said:
Stuffing the channel doesn't mean its paid for yet. Hence you dont see the revenue for AMD that's supposed to follow.
Click to expand...
So as to have no misunderstanding.

Are you claiming that AMD is forcing [stuffing in you words] the vendors to take more GPU die that they would normally want and expect to sell?
 
Steam got 125-150 million users. Even if only 1% is sampled, the statistics are valid due to the huge volume alone. Since it will limit variance due to amount.
 
tential said:
It's sad that people can't understand how to use a sample/survey.

We make national assumptions based on very very small sample sizes to national population. We do it because we have a diverse and large enough sample size to make a confident (confidence interval), assumption about the whole population.

It's all basic statistics which I'm surprised many people can't seem to wrap their heads around.

Steam is a far larger sample size than any other survey we would do.

Do people in this thread really think when they see national polls for anything that they polled at least 60%+ of the nation just to be safe? They poll a very very tiny portion.....
Click to expand...


We can’t directly compare the Steam Survey to other survey methods such as national polls, etc. because we simply don’t have enough transparency around the Steam Survey methodology; what sample size, how are the users prompted for survey, what happens if a user has already taken a survey in the past, frequency of prompts, what triggers steam software to prompt a user for survey- a hardware change, etc.


Steam Survey also is not a predictive indicator unlike an election poll, as it only captures what hardware the user may have currently, but not what user intends to purchase next (so similar to how someone may vote next). This is what I said it is just one data point. In addition, the lack of transparency around the methodology also invalidates the survey to be used with a high level of statistical accuracy. This is why I said it is A data point, not THE data point.
 
Enigmoid said:
The area of the market that steam very likely undersamples is the low end (prebuilt) which may contain gpus but are not used for gaming.
Click to expand...
I'd word it differently. Steam is massively oversampling the USA, Russia and parts of the EU. At least if we take a napkin math comparison of countryGDP/worldGDP to countrySteamUsers/worldSteamUsers (data is public).
 
Piroko said:
I'd word it differently. Steam is massively oversampling the USA, Russia and parts of the EU. At least if we take a napkin math comparison of countryGDP/worldGDP to countrySteamUsers/worldSteamUsers (data is public).
Click to expand...

Not really. Do you really think there is a booming electronic gaming scene in South America and Africa?

http://www.bloomberg.com/news/2014-...pensive-countries-to-buy-a-playstation-4.html

3 of the top 4 most expensive countries for the PS4 are in South America because of absurd import taxes on electronics. Nintendo pulled out of the Brazilian market altogether last year because it was too expensive.
 
ShintaiDK said:
Stuffing the channel doesn't mean its paid for yet. Hence you dont see the revenue for AMD that's supposed to follow.
Click to expand...

Are you claiming that AMD basically advances GPUs on credit, and doesn't get paid until they are not only manufactured into cards but actually sold at retail? That seems odd. Do you have any sources to support this assertion?
 
JDG1980 said:
Are you claiming that AMD basically advances GPUs on credit, and doesn't get paid until they are not only manufactured into cards but actually sold at retail? That seems odd. Do you have any sources to support this assertion?
Click to expand...

Maybe they just give them away? 😀
 
JDG1980 said:
Are you claiming that AMD basically advances GPUs on credit, and doesn't get paid until they are not only manufactured into cards but actually sold at retail? That seems odd. Do you have any sources to support this assertion?
Click to expand...

Its done several times previously by them. Both for GPUs and APUs. When revenue and shipping doesn't match it raises the flag.

The Q1 outlook also tells its own story that will end below 900M$ revenue if true.
 
Last edited:
I think we can all agree its impossible AMD GPU sales improved. The only logical explanation I can see is that AMD discovered a large cache of pirate treasure and is purchasing their own product with it before dumping said product into the ocean to hide the evidence.
 
PingSpike said:
I think we can all agree its impossible AMD GPU sales improved. The only logical explanation I can see is that AMD discovered a large cache of pirate treasure and is purchasing their own product with it before dumping said product into the ocean to hide the evidence.
Click to expand...

That's much too logical for the anti-AMD squad.

But you are of course spot on: this has nothing to do with recent drivers/games/sentiments.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top