4070 reviews thread

[SMOGZINN]

Thing is, the 4070 Ti appears to be selling at least somewhat okay.

I think it's just evidence that releasing new Entry Level and Mid range cards is just a bad idea right now.

No, it is that they did not release an entry level or mid-range card. To everyone not posting on a tech forum that designation is a price point, and they expect a performance to match. NVidia gave us a mid-range performance card priced at the lower end of the upper range price, so few people are buying it because if you are willing to shell out for high end, you are buying a 4090, if you are looking for an entry level or mid-range, you are not looking to spend $700 for just the card, since you are probably looking to spend $1000 for the entire computer.

The main problem with the 4070 is its price point does not make sense to the average consumer.

 

[Heartbreaker]

Sorry but all the steam hate just sounds like a lot of empty rage, mostly based around forum users wanting AMD gpu's to be more popular then they actually are.

That seems to be the crux of it. Don't like the results so attack the messenger. As noted, it's also echoed by the tech analysts like JPR (IIRC, the same people attack these results as well).

 

[DAPUNISHER]

Speaking of astroturfing. Always the same people, always. I leave it to the reader to discern why that is.

It is logically consistent that issues with how the survey is conducted, existed well before the massive 25% misreporting of language, and almost 10% misreporting. This incident was such a huge red flag, that it couldn't be ignored. That is the only reason it was addressed.

Any conclusions that states all is well, the survey is totally legit, nevermind about that massive SNAFU, is viewed through green tinted glasses. I have Nvidia GPUs, so that AMD user accusation falls flat. Or do I have to only buy Nvidia for my opinion about the survey to be OK? GTFO with that garbage. A massive issue was just discovered, rectifying it to the way it was does nothing to dismiss the circumstances that allowed it to occur. Or absolve it of being inherently nerfed. That's a fact, that a massive SNAFU over how the data is collected occured. Attempting to validate the survey's results in its aftermath is where the astroturfing is.

 

[Ranulf]

Daniel-san compared it vs the 3080 12GB. 4070 is impressive in performance per watt. The 3080 is often using almost twice the power. Speaking for myself, that's a massive advantage. Cool and quiet are at the top of my checklist for parts shopping. One power connector regardless of which one, is appealing to me as well.

Perf per watt is the only real positive the card has, because of the high price. The power usage doesn't really impress me because that is what a 70 class card usually runs at.

 

[moonbogg]

I bet Jensen wishes he could shove one transistor up my rear at a time and charge me a nickel for each one.

 

[maddie]

I bet Jensen wishes he could shove one transistor up my rear at a time and charge me a nickel for each one.

Dime, don't forget inflation.

 

[coercitiv]

No one said it was perfect, but other then one fixed bug you haven't shown why it's so broken.

Some food for thought: the survey provides a breakdown by VRAM size. Cards with 11GB represent 1.8%. Going back to the breakdown by card type, we find the 1080 Ti @ 0.61% and the 2080 Ti @ 0.4%. These are the cards I could find with 11GB.

Now the question is, what other cards have 11GB VRAM buffer? We need enough of them to fill 0.79% of the total GPU count in the survey.

 

[MrTeal]

Most likely clumsy 3060 users who accidentally broke off half a ram chip installing their card.

 

[Mopetar]

No one said it was perfect, but other then one fixed bug you haven't shown why it's so broken.

I swear that the people who keep suggesting it's some kind of gold standard neither understand statistics or are capable of reading and comprehension of criticism of the survey by anyone who does.

It's not anyone else's job to show why the survey is broken. It's Valve's job to show why it's valid. Since they don't publish information on how they sample, the response rate, and other important factors, the data isn't particularly useful for making any kind of definitive claims.

This is the same standard used elsewhere in the world. A drug company doesn't get to say, "This is safe, prove me wrong!" to the FDA and any scientist that tried to argue the burden of proof rests on someone else to disprove their hypothesis would be laughed out of the room. You couldn't even publish results that were based on a survey without including data about the sample size, response rate, and trying to identify potential threads to the validity of the results for other reasons.

Every single anti steam survey thread ends up with the same "well it never surveys me" echo chamber and no quantifiable proof of anything.

Another bad argument does not dismiss the first or excuse it. I have been asked to do a Steam survey before (I declined) but that scarcely matters because the issue is that Valve doesn't make the information about how they conduct their surveys available.

I have never been contacted as part of a political poll, but I do trust those results far more because they publish their methodology and are upfront about the amount of error that exists in the results based on the sample size, response rate, etc.

Sorry but all the steam hate just sounds like a lot of empty rage, mostly based around forum users wanting AMD gpu's to be more popular then they actually are.

Have you considered that the Steam results might overstate AMD's market share? You couldn't possibly know because Valve doesn't provide enough information to know what the error in their results might be.

If there's any frustration on my part it's towards people who don't understand statistics and refuse to accept that the onus of proof is on the person making a claim with some data.

 

[Heartbreaker]

I swear that the people who keep suggesting it's some kind of gold standard neither understand statistics or are capable of reading and comprehension of criticism of the survey by anyone who does.

It's not anyone else's job to show why the survey is broken. It's Valve's job to show why it's valid. Since they don't publish information on how they sample, the response rate, and other important factors, the data isn't particularly useful for making any kind of definitive claims.

This is the same standard used elsewhere in the world. A drug company doesn't get to say, "This is safe, prove me wrong!" to the FDA and any scientist that tried to argue the burden of proof rests on someone else to disprove their hypothesis would be laughed out of the room. You couldn't even publish results that were based on a survey without including data about the sample size, response rate, and trying to identify potential threads to the validity of the results for other reasons.

Don't make blanket insults to dump on everyone who disagrees with you. I understand statistics very well, I took stats for the first three years of my CS degree.

This isn't about proving that drugs have beneficial/dangerous effects. Which obviously has a MUCH higher standard.

This is simply to some kind of indicator of GPUs that get into the hands of gamers. It doesn't matter at all what the P value is here, or if some group in some country is over or under represented.

Steam is the largest accessible pool of gamers (100 million plus) and Steam samples them continuously reporting every month. That is an AMAZING resource to have.

The imperfections/flaws simply don't matter, because we aren't using the data for life and death policy decisions like the FDA.

It's just fodder for discussion about GPU distribution, and the Steam data is MORE than adequate for that.

It's by far the BEST tool we have. It may not be perfect but it's better than everything else used around here.

I wonder how many people complaining about using Steam data in discussing GPU sales, have commented about sales at one German retailer that seems to show favorable AMD results? Or proclaimed doom because a product didn't sell out? Or pointed at best seller list at Amazon? Or pointed to pictures of stock on a shelf? Tell me which of those do you think have better statistical significance than the steam data? Do your run around hounding and insulting everyone that posts all of that?

Bottom line: Get off your high horse. This is nerd forum. A large number of us understand statistics quite well, and the Steam Data is still the best data we have regardless of any flaws within it.

 

[blckgrffn]

Don't make blanket insults to dump on everyone who disagrees with you. I understand statistics very well, I took stats for the first three years of my CS degree.

This isn't about proving that drugs have beneficial/dangerous effects. Which obviously has a MUCH higher standard.

This is simply to some kind of indicator of GPUs that get into the hands of gamers. It doesn't matter at all what the P value is here, or if some group in some country is over or under represented.

Steam is the largest accessible pool of gamers (100 million plus) and Steam samples them continuously reporting every month. That is an AMAZING resource to have.

The imperfections/flaws simply don't matter, because we aren't using the data for life and death policy decisions like the FDA.

It's just fodder for discussion about GPU distribution, and the Steam data is MORE than adequate for that.

It's by far the BEST tool we have. It may not be perfect but it's better than everything else used around here.

I wonder how many people complaining about using Steam data in discussing GPU sales, have commented about sales at one German retailer that seems to show favorable AMD results? Or proclaimed doom because a product didn't sell out? Or pointed at best seller list at Amazon? Or pointed to pictures of stock on a shelf? Tell me which of those do you think have better statistical significance than the steam data? Do your run around hounding and insulting everyone that posts all of that?

Bottom line: Get off your high horse. This is nerd forum. A large number of us understand statistics quite well, and the Steam Data is still the best data we have regardless of any flaws within it.

Then don't be surprised when other people call it non-standard garbage. Because from a data collection standpoint, that is what it is. It's non-auditable, there's nothing comparable to measure it against. You know. Etc, etc, etc.

It might be better than nothing. Maybe. But diving into any specific outcomes outside of incredibly high level ones is silly. Making nuanced arguments based on it? That's folly and you understand why.

 

[DAPUNISHER]

@Mopetar

Thanks for making such a cogent argument.

Those defending the survey, accuse those that question its accuracy, as being upset about their prefered vendor doing so poorly in it. If we limit the debate to this schoolyard level, which you refused to engage in with your response? Then that sword cuts 2 ways. E.G. it reflects their prefered vendor as thoroughly dominating, which pleases them, and plays into their confirmation bias. Therefore, they conclude the survey is the best tool for knowing what products people are buying. It's a weak debating tactic used either way. And dumb brand loyal thinking. Resorting to whataboutism is when the debating became pitiable. So thanks again for elevating the discussion, and being the adult in the room.

I consider your position as virtually unassailable, because it cuts to the heart of the problem: The burden of proof.

"It sucks but it's all we've got", is a terrible position to defend/hill to die on. Why do we even need it? Whom does it benefit? Having just witnessed how broken it is, why should anyone trust or accept its results?

My own confirmation bias, is that it is no coincidence this massive SNAFU has taken place right on the heels of my raising concerns about the survey. Concerns founded on my personal experience. I play with a lot of hardware. Frequently setting up a fresh test platform. It has been ages since it asks me to submit when using AMD graphics. On my Nvidia setups, I bat over 50 percent. Despite AMD having about a 4 to 1 usage ratio advantage. I found that odd. Then I see someone else complaining about the same thing i.e they never have it ask on their AMD setups. But it asks every time they turn on their Nvidia laptop. They claimed they knew others that shared that anecdotal experience. Then, the very next survey turns out to be so drastically divergent that it raised eyebrows everywhere. Turns out it's as nerfed as it looks.

The odds and probability that a debacle like this would occur so shortly after sharing my concerns about its accuracy, have to be absurdly remote. For me, not placing much stock in coincidence, the evidence trail led to a smoking gun. And if the massive misreporting issue has not shaken your faith in its accuracy, ask yourself why that is. Why be so willing to accept results that are mysteriously collected, and demonstrably susceptible to a wildly varying degree, which serves only to further indict its accuracy?

 

[Mopetar]

The imperfections/flaws simply don't matter, because we aren't using the data for life and death policy decisions like the FDA.

No one on these forums would accept benchmarks that are posted without an explanation of system configuration or other settings that cannot be reproduced by anyone but the person who published them.

If someone tried to use your excuse to defend those benchmarks you'd cut them down in response.

It's just fodder for discussion about GPU distribution, and the Steam data is MORE than adequate for that.

It really isn't though. Suppose I post the TNMU (Totally Not Made Up) usage statistics which say AMD has a 15% share and NVidia an 85% share. Then I post the DAFR (Definitely Accurate For Realsies) status that claim AMD has 40% share and NVidia has 60%. Both of these post as much information on their methodology as Steam.

All you can say about them is that they don't agree. It's entirely possible that with the actual data on their sampling methodology and response rate that the error bars overlap and neither of them are guilty of anything beyond posting data that has a wide margin of error and not reporting the information necessary to determine that.

You can't even use the Steam information to refute either of those surveys, because it's just as useless. Any argument between which numbers are better or more accurate would necessarily have to rely on other sources for support.

It's by far the BEST tool we have. It may not be perfect but it's better than everything else used around here.

No, it's the worst. Financial data from either company is far better as is any report from a market analyst. You'd probably need to pay them several thousands of dollars for the detailed report that describes how they arrived at the nice bar graph they put out in a press report, but they'll at least tell you their methodology. Companies won't give you a full breakdown on all of their sales figures, but the won't lie about what information they do provide because the cost of doing so is so much worse than any potential gains. Even just looking at shipment data for a single chain of stores like Microcenter is more useful than a survey that doesn't give any methodology.

I wonder how many people complaining about using Steam data in discussing GPU sales, have commented about sales at one German retailer that seems to show favorable AMD results?

I assume you're talking about MindFactory. Drawing any bros conclusions from that would be just as bad. It's a German site so may not generalize to worldwide data and I've pointed out that Germany seems a bit more pro-AMD because they used to fab chips in Dresden. It's also only tracking the DIY market so it misses out on pre-built sales which are a far larger pet of the market. It's also entirely possible that AMD customers tend to prefer buying from MindFactory over other online retailers, so the results could be non-representative as well.

The only thing the results would be good for using is to compare against previous results from the same company. You could probably make a compelling argument about the popularity of certain products among MindFactory customers, but if you tried to claim that reflected the rest of the market you'd have to explain or find other data to support that hypothesis.

Bottom line: Get off your high horse. This is nerd forum.

It is a nerd forum which is why we don't accept bad data. Like I said before, no one around here would accept benchmark data so lacking. It would get chewed to shreds by nerds who wouldn't accept something so sloppy. People will spend pages arguing over RAM timings (or any number of other things) in benchmark data.

So quit crawling around in the mud and get up on the horse with the rest of us.

 

[maddie]

No one on these forums would accept benchmarks that are posted without an explanation of system configuration or other settings that cannot be reproduced by anyone but the person who published them.

If someone tried to use your excuse to defend those benchmarks you'd cut them down in response.



It really isn't though. Suppose I post the TNMU (Totally Not Made Up) usage statistics which say AMD has a 15% share and NVidia an 85% share. Then I post the DAFR (Definitely Accurate For Realsies) status that claim AMD has 40% share and NVidia has 60%. Both of these post as much information on their methodology as Steam.

All you can say about them is that they don't agree. It's entirely possible that with the actual data on their sampling methodology and response rate that the error bars overlap and neither of them are guilty of anything beyond posting data that has a wide margin of error and not reporting the information necessary to determine that.

You can't even use the Steam information to refute either of those surveys, because it's just as useless. Any argument between which numbers are better or more accurate would necessarily have to rely on other sources for support.



No, it's the worst. Financial data from either company is far better as is any report from a market analyst. You'd probably need to pay them several thousands of dollars for the detailed report that describes how they arrived at the nice bar graph they put out in a press report, but they'll at least tell you their methodology. Companies won't give you a full breakdown on all of their sales figures, but the won't lie about what information they do provide because the cost of doing so is so much worse than any potential gains. Even just looking at shipment data for a single chain of stores like Microcenter is more useful than a survey that doesn't give any methodology.



I assume you're talking about MindFactory. Drawing any bros conclusions from that would be just as bad. It's a German site so may not generalize to worldwide data and I've pointed out that Germany seems a bit more pro-AMD because they used to fab chips in Dresden. It's also only tracking the DIY market so it misses out on pre-built sales which are a far larger pet of the market. It's also entirely possible that AMD customers tend to prefer buying from MindFactory over other online retailers, so the results could be non-representative as well.

The only thing the results would be good for using is to compare against previous results from the same company. You could probably make a compelling argument about the popularity of certain products among MindFactory customers, but if you tried to claim that reflected the rest of the market you'd have to explain or find other data to support that hypothesis.



It is a nerd forum which is why we don't accept bad data. Like I said before, no one around here would accept benchmark data so lacking. It would get chewed to shreds by nerds who wouldn't accept something so sloppy. People will spend pages arguing over RAM timings (or any number of other things) in benchmark data.

So quit crawling around in the mud and get up on the horse with the rest of us.

I would even argue that bad data is worse than no data.

 

[Heartbreaker]

No, it's the worst. Financial data from either company is far better as is any report from a market analyst. You'd probably need to pay them several thousands of dollars for the detailed report that describes how they arrived at the nice bar graph they put out in a press report, but they'll at least tell you their methodology. Companies won't give you a full breakdown on all of their sales figures, but the won't lie about what information they do provide because the cost of doing so is so much worse than any potential gains. Even just looking at shipment data for a single chain of stores like Microcenter is more useful than a survey that doesn't give any methodology.

Financial data tells you almost nothing, except how much money a company made. Often you can't even tell the volume of graphics cards as the lump it together with other products. It tells you nothing about relative amounts of specific cards. Analyst data provides a bit more detail but still ZERO about specfic cards. What the analyst data does do though, is reinforce the overall Steam Numbers.

Recently someone said there was "market acceptance" of the 3090 Ti.

I disagreed and pointed out that reviewers were strongly against it, and I checked the Steam Survey.

3090 Ti, didn't sell enough to even make the survey 0.15% cutoff. Something that the 3090, 4090, and even heavily maligned 4080 has done with about double that cutoff.

No financial data would tell you that. Nothing else we have access too, would tell us that. So the data that Steam presents is totally unique.

While Steam Survey may contain some minor sampling errors, it most likely still has the relative standing of those cards correct, there is almost no reasonable kind of sampling error that would get the relative position of 3090 Ti and 4080 backwards give the scale of the difference between them in the results. Do you think the relative position of 3090 Ti and 4080 could be wrong? How?

So Steam Survey provides completely unique data unavailable elsewhere that almost certainly represents the relative amounts of cards in the hands of Steam users (one of the most important pools of PC gamers out there numbering over 100 million). It gives us unique look at things we don't otherwise have access to.

It is a nerd forum which is why we don't accept bad data. Like I said before, no one around here would accept benchmark data so lacking. It would get chewed to shreds by nerds who wouldn't accept something so sloppy.

This is blatantly false. Obviously much worse data was pointed out doesn't get chewed to shreds. Mindfactory results, pictures of a stock shelf, anecdotes, rumors...

There is massive confirmation bias in the forum, will to leap to accept anything however tenuous as long as it paints NVidia in a bad light.

There is no evidence Steam Survey has bad data.

It's a continuous sample, monthly reporting of the largest PC gamer database in the world, done for almost 2 decades straight. There is no reason to believe it is done in bad faith, and with a client on every Steam PC, it would be trivial to set up wide scale sampling, also since it's reading data directly from the computer it doesn't have the inaccuracy of just asking people. Any middling level competent Comp Sci staffer should have no problem doing this kind of sampling properly. Steam should have VERY good data.

 

[blckgrffn]

Financial data tells you almost nothing, except how much money a company made. Often you can't even tell the volume of graphics cards as the lump it together with other products. It tells you nothing about relative amounts of specific cards. Analyst data provides a bit more detail but still ZERO about specfic cards. What the analyst data does do though, is reinforce the overall Steam Numbers.

Recently someone said there was "market acceptance" of the 3090 Ti.

I disagreed and pointed out that reviewers were strongly against it, and I checked the Steam Survey.

3090 Ti, didn't sell enough to even make the survey 0.15% cutoff. Something that the 3090, 4090, and even heavily maligned 4080 has done with about double that cutoff.

No financial data would tell you that. Nothing else we have access too, would tell us that. So the data that Steam presents is totally unique.

While Steam Survey may contain some minor sampling errors, it most likely still has the relative standing of those cards correct, there is almost no reasonable kind of sampling error that would get the relative position of 3090 Ti and 4080 backwards give the scale of the difference between them in the results. Do you think the relative position of 3090 Ti and 4080 could be wrong? How?

So Steam Survey provides completely unique data unavailable elsewhere that almost certainly represents the relative amounts of cards in the hands of Steam users (one of the most important pools of PC gamers out there numbering over 100 million). It gives us unique look at things we don't otherwise have access to.




This is blatantly false. Obviously much worse data was pointed out doesn't get chewed to shreds. Mindfactory results, pictures of a stock shelf, anecdotes, rumors...

There is massive confirmation bias in the forum, will to leap to accept anything however tenuous as long as it paints NVidia in a bad light.

There is no evidence Steam Survey has bad data.

It's a continuous sample, monthly reporting of the largest PC gamer database in the world, done for almost 2 decades straight. There is no reason to believe it is done in bad faith, and with a client on every Steam PC, it would be trivial to set up wide scale sampling, also since it's reading data directly from the computer it doesn't have the inaccuracy of just asking people. Any middling level competent Comp Sci staffer should have no problem doing this kind of sampling properly. Steam should have VERY good data.

You’re conflating “lots of data” with “data collected and presented in a statistical analysis sound way”.

That’s a mistake. One does not make the other true.

Sampling the data is the easy part.

 

[Heartbreaker]

You’re conflating “lots of data” with “data collected and presented in a statistical analysis sound way”.

That’s a mistake. One does not make the other true.

Sampling the data is the easy part.

This is also fairly trivial data, to present. You are just reporting percentage of each category, that you recorded. It doesn't need to be adjusted for presentation, there aren't a multitude of confounding factors like if this was trying to present the impact of eating processed meat on health. This just straight up reporting of objective data.

Plus we aren't using it in a scientific paper, so we don't need to know confidence intervals, sample size. Though given the month to month consistency of results (with one exception noted below), the sample size is almost certainly adequate, and given the ease of data sampling, there is no pressure at all to under sample.

This is barely Stats 101 stuff, it's among the simplest, least confounding data to collect and present. Valve has been at this for near two decades. I think they can handle it.

The Steam Survey is going to be MUCH more accurate than most surveys that usually involve getting somewhat subjective, or memory dependent responses from humans, where how you ask questions can swing results significantly, where some questions may be sensitive and skew results.

This is just easy objective, automated, data collection, and presentation.

It's not perfect, but to pretend those imperfections make it useless is preposterous.

The main issues seem to be a mechanism to handle shared computers (Internet Cafe's), which it recently had a one month glitch with, and had a similar issue years ago. This is the only real snag in an otherwise trivial sampling and reporting effort.

Other than that it comes down to whether there is a reason to distrust Valve... I don't think there is.

You sometimes see a similar negative response to Steam Survey on Linux forums: Steam Survey is biased against them and under-representing them. There is ZERO evidence that is happening, it's just that some in the Linux community don't like the results and want to shoot the messenger. Valve is a massive Linux proponent, and they have been trying for years to increase Linux gaming, so if they were inclined to color the results, it would be over-representing Linux.

So again, it's just distrust based on disliking the results. Even if Valve published detailed methodologies, it would still be faced with the same distrust from those who don't like the results, because Valve could simply be lying and manipulating the output. Then what, demand third party auditors in to check everything? Only then would looking at Steam Survey make sense. Some still wouldn't because how can you trust the third party auditors, when we can't know Valve didn't bribe them after all...

 

[blckgrffn]

This is also fairly trivial data, to present. You are just reporting percentage of each category, that you recorded. It doesn't need to be adjusted for presentation, there aren't a multitude of confounding factors like if this was trying to present the impact of eating processed meat on health. This just straight up reporting of objective data.

Plus we aren't using it in a scientific paper, so we don't need to know confidence intervals, sample size. Though given the month to month consistency of results (with one exception noted below), the sample size is almost certainly adequate, and given the ease of data sampling, there is no pressure at all to under sample.

This is barely Stats 101 stuff, it's among the simplest, least confounding data to collect and present. Valve has been at this for near two decades. I think they can handle it.

The Steam Survey is going to be MUCH more accurate than most surveys that usually involve getting somewhat subjective, or memory dependent responses from humans, where how you ask questions can swing results significantly, where some questions may be sensitive and skew results.

This is just easy objective, automated, data collection, and presentation.

It's not perfect, but to pretend those imperfections make it useless is preposterous.

The main issues seem to be a mechanism to handle shared computers (Internet Cafe's), which it recently had a one month glitch with, and had a similar issue years ago. This is the only real snag in an otherwise trivial sampling and reporting effort.

Other than that it comes down to whether there is a reason to distrust Valve... I don't think there is.

You sometimes see a similar negative response to Steam Survey on Linux forums: Steam Survey is biased against them and under-representing them. There is ZERO evidence that is happening, it's just that some in the Linux community don't like the results and want to shoot the messenger. Valve is a massive Linux proponent, and they have been trying for years to increase Linux gaming, so if they were inclined to color the results, it would be over-representing Linux.

So again, it's just distrust based on disliking the results. Even if Valve published detailed methodologies, it would still be faced with the same distrust from those who don't like the results, because Valve could simply be lying and manipulating the output. Then what, demand third party auditors in to check everything? Only then would looking at Steam Survey make sense. Some still wouldn't because how can you trust the third party auditors, when we can't know Valve didn't bribe them after all...

That’s a lot of projection there.

 

[coercitiv]

This is just easy objective, automated, data collection, and presentation.

I'll repeat this. According to the survey:
11GB VRAM card share --> 1.8%
1080Ti share --> 0.61%
2080Ti share --> 0.4%

Where's the rest of 0.79%?

In this particular case there are no professional cards with 11GB buffer that could all stay bellow 0.15% while collectively amounting to 0.79% of the total sample size.

 

[Heartbreaker]

That’s a lot of projection there.

I reporting, not projecting.

I can link posts from Linux users claiming the survey is biased against them.

I've literally also seen AMD users go as far as claiming NVidia must be paying Valve to under-report AMD.

I have seen this for many years and it gets tedious after a while.

Some people don't like results, so they just attack the source.

 

[blckgrffn]

I reporting, not projecting.

I can link posts from Linux users claiming the survey is biased against them.

I've literally also seen AMD users go as far as claiming NVidia must be paying Valve to under-report AMD.

I have seen this for many years and it gets tedious after a while.

Some people don't like results, so they just attack the source.

Is that what was written? I saw a wall a text deflecting from the fact we don't know how the data is collected. And then attacks on the character of anyone who would doubt its accuracy when reduced down to sub percentage levels.

None of that addresses the fact that its a large source of data, but not a transparently accurate one. Which, by your own admission, you know the difference between. We don't even know what Valve's own confidence level is in the data, because we know they would probably put a huge qualifier on the accuracy of the results in many scenarios like any good data analytics company would. But then we go on and on and on about it, which makes me wonder.

Wishing it was accurate is different than it being accurate. Expecting others to share your faith - because that is what it is at this point - is exhausting.

 

[Heartbreaker]

Is that what was written? I saw a wall a text deflecting from the fact we don't know how the data is collected.

I am not deflecting from that. I acknowledge we don't know the full details of data collection.

But the collection process is trivial to get right, and it's not unreasonable to assume Valve is neither malicious or completely incompetent at trivial data collection.

 

[blckgrffn]

I am not deflecting from that. I acknowledge we don't know the full details of data collection.

But the collection process is trivial to get right, and it's not unreasonable to assume Valve is neither malicious or completely incompetent at trivial data collection.

Statistically accurate data is anything but trivial to collect and present.

That's the difference here. Valve has an easy, big pile of data. From all accounts, they get it lazily and we have to assume they think it's interesting to them but only useful when they use all sorts of internal filters that we aren't privy to. If they even use it all, and it isn't part of some other project that's been started and sorta half implemented as a way to ensure they stay relevant as a sort of PR investment.

 

[DAPUNISHER]

Is that what was written? I saw a wall a text deflecting from the fact we don't know how the data is collected. And then attacks on the character of anyone who would doubt its accuracy when reduced down to sub percentage levels.

None of that addresses the fact that its a large source of data, but not a transparently accurate one. Which, by your own admission, you know the difference between. We don't even know what Valve's own confidence level is in the data, because we know they would probably put a huge qualifier on the accuracy of the results in many scenarios like any good data analytics company would. But then we go on and on and on about it, which makes me wonder.

Wishing it was accurate is different than it being accurate. Expecting others to share your faith - because that is what it is at this point - is exhausting.

Ad hominems and whataboutisms are the least persuasive of all debating tactics. Tossing in argument to ignorance is the trifecta.

 

[maddie]

Just a tiny data massage.

Steam April 2023: discrete card GPU only, not including laptop models.

Nvidia ------ 62.31%
AMD --------- 6.75%

Nvidia = 923 % of AMD (means AMD powers 10.8% of all graphic cards since Polaris days)

Card sales since Polaris is reported by several independent market share charts as roughly:

Nvidia - 75%
AMD --- 25%

Nvidia = 300% of AMD

I can't understand this as agreeing with each other, as some have claimed in attempting to justify Steam stats accuracy. There is obviously a huge discrepancy here.

One possible explanation is that more AMD cards are used for other purposes than gaming. Unlikely, as mining & professional work seemed to favor Nvidia.

Anyhow, the Steam stats chart is suspect and should definitely be questioned.