[JPR] The balance of power in gaming

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Headfoot

Diamond Member
Feb 28, 2008
4,444
640
126
The obvious difference between JPR and Steam surveys, is JPR is telling you what is being sold in a particular small time frame. Steam shows what gamers are currently using, which may span many years worth of ownership and doesn't track GPU's being used for other purposes.

I really doubt the opt in and opt out aspect of Steam plays a significant role in results. It just shrinks the data a little.
Opt-in vs unbiased sampling is a very large determinant in the quality of a survey. This is well established in data science. http://blogs.abcnews.com/thenumbers/2009/09/study-finds-trouble-for-internet-surveys.html. It's due to selection bias - people who opt in may feel they have something to say and that influences their decision to opt-in in the first place.

Again - fine for its designed purpose (a general overview of Steam user's hardware for low-precision decisionmaking). Not at all fine when people are debating tenths of a percent.
 
  • Like
Reactions: Bacon1

bystander36

Diamond Member
Apr 1, 2013
5,154
132
106
Opt-in vs unbiased sampling is a very large determinant in the quality of a survey. This is well established in data science. http://blogs.abcnews.com/thenumbers/2009/09/study-finds-trouble-for-internet-surveys.html. It's due to selection bias - people who opt in may feel they have something to say and that influences their decision to opt-in in the first place.

Again - fine for its designed purpose (a general overview of Steam user's hardware for low-precision decisionmaking). Not at all fine when people are debating tenths of a percent.
I never said they were 100% accurate. You guys arguing over a tenth of a percent are just taking all this too seriously, unless you are investors. Just because it is less accurate, does not mean it's not accurate at all.

They also show very different results. Steam does not show purchase info. It shows what people are gaming on. It's very possible that there are groups who purchase new GPU's often, while another group holds on to them a long time and those groups may even have different brand preferences. These stats show different things. Trying to compare them is futile.
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
640
126
I never said they were 100% accurate. You guys arguing over a tenth of a percent are just taking all this too seriously, unless you are investors. Just because it is less accurate, does not mean it's not accurate at all.

They also show very different results. Steam does not show purchase info. It shows what people are gaming on. It's very possible that there are groups who purchase new GPU's often, while another group holds on to them a long time and those groups may even have different brand preferences. These stats show different things. Trying to compare them is futile.
I'm not arguing about tenths of a percent. I'm saying that if people here want to do that, Steam surveys are not a valid source for that argument. It's pretty clear research targeted at discovering number of units shipped is going to be way more accurate regarding number of units shipped than the Steam survey is.
 

AtenRa

Lifer
Feb 2, 2009
13,772
2,932
136
You guys know JPR works for AMD, right?
Seriously? It's been at least a decade. Whitepapers produced by JPR paid for by AMD are even on AMD's website. They aren't difficult to find at all.
There is a difference between paying JPR to research and create a market report for you(AMD) and another think to say that "JPR works for AMD"

JPR is an independent company, it doesnt work for AMD.
 

AtenRa

Lifer
Feb 2, 2009
13,772
2,932
136

bystander36

Diamond Member
Apr 1, 2013
5,154
132
106
I'm not arguing about tenths of a percent. I'm saying that if people here want to do that, Steam surveys are not a valid source for that argument. It's pretty clear research targeted at discovering number of units shipped is going to be way more accurate regarding number of units shipped than the Steam survey is.
Just because it has a larger range of accuracy, does not make them invalid. You constantly act as if their results mean nothing, but they are one of the best ways we can examine what people use on their gaming machines.
 

AtenRa

Lifer
Feb 2, 2009
13,772
2,932
136
Just because it has a larger range of accuracy, does not make them invalid. You constantly act as if their results mean nothing, but they are one of the best ways we can examine what people use on their gaming machines.

Only for STEAM users that took the survey.
 

[DHT]Osiris

Lifer
Dec 15, 2015
10,604
7,360
146
You better have a second look .........
Does developing for $consoleplatform translate to development for AMD though? What % of time developing for a PS4 game is doing AMD-specific stuff, and what of it is PS4 specific stuff? When you develop for a console, you develop for whatever's there. You're tweaking stuff for the console, increasing efficiency/fps/whatever for the console. That may be specific to AMD for a given generation, or nvidia for a given generation, but it isn't up to the dev to decide that. If they're developing a PC game however, they have to make a decision as to what to focus on (assuming they don't have time to focus on both to the extent they would like). That's what I was getting at when I was talking about focus.
 

Phynaz

Lifer
Mar 13, 2006
10,140
817
126
There is a difference between paying JPR to research and create a market report for you(AMD) and another think to say that "JPR works for AMD"

JPR is an independent company, it doesnt work for AMD.
Semantics.

Would you like it better if I said JPR does work for AMD?
(Which means JPR works for AMD)
 
  • Like
Reactions: Arachnotronic

AtenRa

Lifer
Feb 2, 2009
13,772
2,932
136
Does developing for $consoleplatform translate to development for AMD though? What % of time developing for a PS4 game is doing AMD-specific stuff, and what of it is PS4 specific stuff? When you develop for a console, you develop for whatever's there. You're tweaking stuff for the console, increasing efficiency/fps/whatever for the console. That may be specific to AMD for a given generation, or nvidia for a given generation, but it isn't up to the dev to decide that. If they're developing a PC game however, they have to make a decision as to what to focus on (assuming they don't have time to focus on both to the extent they would like). That's what I was getting at when I was talking about focus.
Both Consoles and AMD PC GPUs are the same architecture (GCN), you will need less time and spend fewer resources to optimize for the AMD PC GPUs than for the NVIDIA.
So if you are time and budget constrained you better optimize for the Consoles and AMD PC GPUs because of the same Architecture. Also those two will make you address and cover the higher gamer volume.

I have said this before but few really understood what this really means, there are more than 60 Million latest generation AMD GCN Consoles (PS4 and XB1) used by gamers today. Add all those GCN Graphics Cards used by PC Gamers and you really have more than 70-80 Million GCN based Gamers. This is a huge volume for a developer to start a new project even without considering optimizing for any NVIDIA GPU.

But lets say you want to optimize for the NVIDIA GPUs, here comes the ugly part. You have to optimize for 3 different architectures, Kepler, Maxwell, Pascal. That will take you more time and more resources.
 
  • Like
Reactions: Bacon1

[DHT]Osiris

Lifer
Dec 15, 2015
10,604
7,360
146
Both Consoles and AMD PC GPUs are the same architecture (GCN), you will need less time and spend fewer resources to optimize for the AMD PC GPUs than for the NVIDIA.
So if you are time and budget constrained you better optimize for the Consoles and AMD PC GPUs because of the same Architecture. Also those two will make you address and cover the higher gamer volume.

I have said this before but few really understood what this really means, there are more than 60 Million latest generation AMD GCN Consoles (PS4 and XB1) used by gamers today. Add all those GCN Graphics Cards used by PC Gamers and you really have more than 70-80 Million GCN based Gamers. This is a huge volume for a developer to start a new project even without considering optimizing for any NVIDIA GPU.

But lets say you want to optimize for the NVIDIA GPUs, here comes the ugly part. You have to optimize for 3 different architectures, Kepler, Maxwell, Pascal. That will take you more time and more resources.
Makes sense. What on earth is AMD doing with this market share then, if they still cannot compete with nvidia on the non-budget GPU front? (For the record, I consider the 480 to be budget). Their R&D should be through the roof.
 

bystander36

Diamond Member
Apr 1, 2013
5,154
132
106
Only for STEAM users that took the survey.
For starters, the article about how opt in surveys are less accurate, were using Facebook Opt in surveys. Steam's survey is more of an opt out option. I'm not sure how different that makes it, but there is a little difference. The article did not say it was inaccurate either. They said it was less accurate.

Steam surveys, while not perfect, still reflect what the average gamer uses, even those who opted out.
 

nurturedhate

Golden Member
Aug 27, 2011
1,696
530
136
Semantics.

Would you like it better if I said JPR does work for AMD?
(Which means JPR works for AMD)
This is just an attempt to diffuse the orginal porpuse of your original post. The purpose of the original post was an attempt to discredit the report. You'd be posting counter-point after counter-point if someone went into an Nvidia marketshare thread and stated "You guys know JPR works for Nvidia, right?". Nice try though.
 
Last edited:
  • Like
Reactions: Bacon1

esquared

Forum Director & Omnipotent Overlord
Forum Director
Oct 8, 2000
22,428
3,829
136
OK, enough of JPR. Any further posting/speculation about JPR and AMD and their exact relationship, in this thread, without viable/credible links, will be met with infractions.

No more trolling here, just to troll.


esquared
Anandtech Forum Director
 

Pariah

Elite Member
Apr 16, 2000
7,357
20
81
Both Consoles and AMD PC GPUs are the same architecture (GCN), you will need less time and spend fewer resources to optimize for the AMD PC GPUs than for the NVIDIA.
So if you are time and budget constrained you better optimize for the Consoles and AMD PC GPUs because of the same Architecture. Also those two will make you address and cover the higher gamer volume.

I have said this before but few really understood what this really means, there are more than 60 Million latest generation AMD GCN Consoles (PS4 and XB1) used by gamers today. Add all those GCN Graphics Cards used by PC Gamers and you really have more than 70-80 Million GCN based Gamers. This is a huge volume for a developer to start a new project even without considering optimizing for any NVIDIA GPU.

But lets say you want to optimize for the NVIDIA GPUs, here comes the ugly part. You have to optimize for 3 different architectures, Kepler, Maxwell, Pascal. That will take you more time and more resources.
Realistically, the opposite of what you are proposing is probably happening in the PC arena. If developers spent time optimizing code for GCN when developing for consoles, they will likely port as is and not spend any additional time optimizing for PC GCN considering the puny market share they have. Any time spent optimizing for the PC platform will be spent on Nvidia's architectures because 1) they control the majority of the market and pretty much all of the high end, (When you see marketing screenshots for games, they aren't done on low/midrange products) and 2) if it is a AAA title, Nvidia will donate developer time/money/code to the developers to assist them with optimizations. Something, AMD with no money does not do nearly to the degree that Nvidia does.
 

Krteq

Senior member
May 22, 2015
985
664
136
There is nothing like "GCN in console" and "GCN in PC". There is only GCN in few iterations across consoles (PS4/XBONE - GCN Gen1/Gen2 hybrid, PS4 Pro - GCN Gen4/Vega hybrid) and PC (GCN Gen 1/2/3/4). It's still the same GCN uarch with slightly different frontend/backend and minor changes in shadercore.
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
640
126
Just because it has a larger range of accuracy, does not make them invalid. You constantly act as if their results mean nothing, but they are one of the best ways we can examine what people use on their gaming machines.
I have never once said they mean nothing. In fact, I've repeatedly said they are fine for their intended purpose but not for debating market share to tenths of a percent.
 
  • Like
Reactions: Bacon1

Pariah

Elite Member
Apr 16, 2000
7,357
20
81
There is nothing like "GCN in console" and "GCN in PC". There is only GCN in few iterations across consoles (PS4/XBONE - GCN Gen1/Gen2 hybrid, PS4 Pro - GCN Gen4/Vega hybrid) and PC (GCN Gen 1/2/3/4). It's still the same GCN uarch with slightly different frontend/backend and minor changes in shadercore.
You agree that optimizing for a console is not the same as optimizing for a PC, so why are you arguing the semantics of how I label those differences?
 

Krteq

Senior member
May 22, 2015
985
664
136
Because with game developed for console there is almost no optimizitation needed for PCs with GCN cards.
 

Pariah

Elite Member
Apr 16, 2000
7,357
20
81
There are absolutely features the consoles have that the pc gpus don't have and vice versa. As well as significant differences between how they interface with the rest of the hardware in the system which you mentioned. Again it looks like you are arguing just for the sake of arguing.
 

ocre

Golden Member
Dec 26, 2008
1,594
7
76

In Desktop AMD gained share from 22,8% vs 25,5% Q2 to Q3 , while NVIDIA lost share from 77,2% vs 71,2% Q2 to Q3.

In order for AMD to have gained share, they needed to ship more GPU than NVIDIA, do you actually see this AMD increased GPU shipments in the Steam Hardware Survey above ??

There is not even a single Polaris GPU in that Steam survey for DX-12 cards.

Steam Hardware Survey doesnt mean Global GPU Shipments, you may only take the survey numbers for what GPUs people using on STEAM and nothing more. And in that context, without knowing how many people took the survey its even less reliable to use Steam numbers.
.
Absolutely not!!
That is not how it works. Not at all how it works.

Amd does not have to sell more cards than nvidia in order to gain marketshare in the quarterly report. For amd to even have a chance to have sold more gpus, they would have to have over 50% of the quarterly dgpu marketshare. And once again (as I said in my previous post), these reports are not chips sold to the individual. This is data collected on the front end. The number of chips shipped to AiBs and vendors in a quarter.

Its not the amount of chips sold to people.
So you must understand how wrong your analysis is here. In both counts, major misunderstanding in how it works.

So, the front end sales to AiBs and partners will eventually end up as user end sales. But that takes more time to trickle out. And there is the possibility that the expected sales don't match and the AiBs dont keep pace with what a quarterly result may suggest. This will result in a dip or adjustment in the following quarter. But, the future is needed to tell how many gpus land in homes from the Q3 shipments to AiBs.

The chips take time to move to homes, so a surge in quarterly shipments on the front end would take time to have an impact on a surge in users on steam.
But its way more complicated because steam survey is current user marketshare. Chips sold last quarter, last year, and yrs ago all have an impact. This is data based on the actual machines being used to game with.

The shipment marketshare data, its completely different. It doesnt count a single gpu that is in someones house. Its the amount of chips sent to vendors and AiBs to become products that people will buy. And, amd cannot sell more chips than nvidia unless their marketshare on shipments surpasses nvidias market share in shipments. They did not ship enough chips to outsell nvidia, not even close. Its mathematically impossible
 

Det0x

Senior member
Sep 11, 2014
634
1,070
136
You guys know JPR works for AMD, right?
Seriously? It's been at least a decade. Whitepapers produced by JPR paid for by AMD are even on AMD's website. They aren't difficult to find at all.
So when Nvidia is using JPR,its AMD PR helping them?? I know AMD PR has been accused of many things,but that is a new one for me!

Nvidia has been using JPR since 2005:
https://www.nvidia.com/object/IO_18301.html

2007:

http://www.nvidia.com/object/io_1190676485302.html

Nvidia used them this year. An example:
https://blogs.nvidia.com/blog/2016/02/16/vulkan-graphics-api/

Nvidia PR has no problem using JPR. They have been using them for 11 years.

I think Nvidia knows better about what companies they use in their PR releases,than random people on a forum.

Edit to post.

Nvidia used them on Twitter this year too:
https://twitter.com/nvidia/status/715238319966302209
1. I didn't say there's anything wrong with using JPR. Or hiring them for their marketing assistance.
2. You're link is to a press release that quotes JPR. It does not contain any evidence that the report was commissioned by Nvidia.
I know i risk get a vacation for this, but this shouldn't be allowed to stand without a response.

Hypocrisy is thy name.




My warning stands. Not only an infraction. The next one gets a week off and the thread will be locked.


esquared
Anandtech Forum Director
 
Last edited by a moderator:

AtenRa

Lifer
Feb 2, 2009
13,772
2,932
136
Absolutely not!!
That is not how it works. Not at all how it works.

Amd does not have to sell more cards than nvidia in order to gain marketshare in the quarterly report. For amd to even have a chance to have sold more gpus, they would have to have over 50% of the quarterly dgpu marketshare. And once again (as I said in my previous post), these reports are not chips sold to the individual. This is data collected on the front end. The number of chips shipped to AiBs and vendors in a quarter.

Its not the amount of chips sold to people.
So you must understand how wrong your analysis is here. In both counts, major misunderstanding in how it works.

So, the front end sales to AiBs and partners will eventually end up as user end sales. But that takes more time to trickle out. And there is the possibility that the expected sales don't match and the AiBs dont keep pace with what a quarterly result may suggest. This will result in a dip or adjustment in the following quarter. But, the future is needed to tell how many gpus land in homes from the Q3 shipments to AiBs.

The chips take time to move to homes, so a surge in quarterly shipments on the front end would take time to have an impact on a surge in users on steam.
But its way more complicated because steam survey is current user marketshare. Chips sold last quarter, last year, and yrs ago all have an impact. This is data based on the actual machines being used to game with.

The shipment marketshare data, its completely different. It doesnt count a single gpu that is in someones house. Its the amount of chips sent to vendors and AiBs to become products that people will buy. And, amd cannot sell more chips than nvidia unless their marketshare on shipments surpasses nvidias market share in shipments. They did not ship enough chips to outsell nvidia, not even close. Its mathematically impossible
Q3 2016 shipments are counting three months, from July to September of 2016. Plenty of time for people to buy Polaris cards and take the October Steam Hardware survey.

RX 480 official release 29th June 2016

GTX 1060 6GB official release 19th July 2016

Interestingly, GTX 1060 first show up in Steam Hardware Survey DX-12 GPUs on August while RX 480 in October.

http://store.steampowered.com/hwsurvey/videocard/

So, AMD gained market share in Q3 (July to September) but that doesnt show in October Steam Hardware Survey.
 

Pariah

Elite Member
Apr 16, 2000
7,357
20
81
So, AMD gained market share in Q3 (July to September) but that doesnt show in October Steam Hardware Survey.
AMD lost .5% total market share in q3, and probably a substantial percentage of the dGPU market to Nvidia. Even ignoring the facts, a shift of a few percentages for a quarter is not going to be reflected in installed user base surveys like steam. The math doesn't work that way.
 

ASK THE COMMUNITY