Hilux SSRGShame on all the Nvidiots who bought Nvidia graphics cards rather than AMD or Intel when you had an option and they were price competitive/feature similar. Reap what you sow from your overlords.

Hecate91It isn't competition because the 5090 isn't a gaming card, it's an AI card disguised with gamer branding, most cards end up in AI workstations or servers.

Hecate91I'd rather not give anything to a company that will gaslight and decieve consumers

efikkanWhat we need is Intel and AMD to mass produce their products in much higher quantities, and fill in the gaps in their lineups

efikkanAs from the Steam Hardware Survey, RX 9000-series doesn't even make a presence:
RTX 5060 1.69%
RTX 5060 Ti 1.26%
RTX 5070 2.28%
RTX 5070 Ti 1.20%
RTX 5080 1.07%
RTX 5090 0.36%
There are far more people gaming on a RTX 5090 than on a RX 9070 XT (or any 9000 series for that matter).

I'm also glad to see that 5070 is more popular than 5060, as I believe it's the best value option in the current lineup.

Hecate91Nvidia: mere peasant consumers want more than 8GB VRAM? Pay for Geforce Now to get a 6080 in the cloud.

john_You are absolutely right. if GPUs become more expensive and more difficult to get, online gaming as a service will be pushed hard from companies like Nvidia that are already selling subscriptions AND have the monopoly in GPUs. Imagine if AMD's best discrete GPU today was at the levels where the best Intel discrete GPU is and imagine if the majority of games where using Nvidia's proprietary techs to display the majority of visuals on screen. People would had two options. Bad framerates and low quality visuals with an Intel or AMD GPU, or pay Nvidia subscriptions to get the best visual quality at high resolutions, with good framerates, meaning, having to pay for the subscription every month AND the best internet service available in you region.

Hecate91Intel and AMD mass producing cards would be effectively pointless as people will simply buy Nvidia cards anyway, TSMC fab space limitations is a thing too.

Hecate91In addition to that, Valve only polls randomly, and doesn't show metrics for how many were polled or stats for people who haven't previously received a request for the survey.

efikkanNonsense. AMD have much larger fan base.

efikkanMathematically, you only need a tiny sample from a population provided the samples are randomly selected

Hecate91Nvidia has the largest fan base, and over 90% of the market.

Hecate91but a sample size of 1000 people would be fine if Valve listed what their sample size is, however they don't so the survey is just an outlook at best of how many people are using Steam, Valve doesn't even list if people are polled repeatedly, and things end up in the poll like massive influxes of certain operating systems…

Hecate91or a single GPU SKU because gaming cafes get counted as valid surveys

Hecate91Not everyone uses Steam either, or has it set to open on startup to always get a poll, plenty of other game storefronts exist as well as games which don't require a separate launcher to run.

efikkanThis clearly showcases the problem when we don't teach people math properly in school.
Mathematically, you only need a tiny sample from a population provided the samples are randomly selected. Even 1000 should be plenty, and by looking at the numbers and the amount of data points it appears that theirs must be in the thousands, probably over 10000. There are some fluctuations month to month due to holidays etc., but over time it does paint an accurate picture. Considering Steam have a majority market share among PC gamers, both casual and dedicated gamers, it's far exceeds the requirements to assess accurately the mass of PC gamers, you don't need ~100% poll coverage to do that. If you can't grasp this, I suggest you go buy a good math textbook and get knowledge, instead of just having strong opinions.

john_If I have 10000 balls with 500 of them being blue and 9500 of them white and decide to pick randomly 100 and check the percentage of blue balls and white balls, statistics say that blue balls will be 5%. In reality there is a possibility to randomly keep selecting only white balls out of that huge pool of 10000 balls, concluding that there are no blue balls in there. The sample needs to be big and the sample needs to be completely random every time I am making the poll.

Hecate91Nvidia has the largest fan base, and over 90% of the market.

I'll ignore the personal attack, but a sample size of 1000 people would be fine if Valve listed what their sample size is, however they don't so the survey is just an outlook at best of how many people are using Steam, Valve doesn't even list if people are polled repeatedly, and things end up in the poll like massive influxes of certain operating systems or a single GPU SKU because gaming cafes get counted as valid surveys, and thats obviously not the same as individual users. The math simply doesn't math when Valve has too many variables and not enough metrics listed.
Not everyone uses Steam either, or has it set to open on startup to always get a poll, plenty of other game storefronts exist as well as games which don't require a separate launcher to run.

efikkanMarket share doesn't equate fan base.
Just look at every tech forum, reddit, tech channels on YouTube etc. You will rarely find someone who is an Nvidia fanboy or even clear favoritism. On the AMD side though, it's clear as day.


Doesn't matter, as long as it's sampled randomly, all of those will even out.
Just look at the list of GPUs and the decimal precision of each listed. The underlying data must be in the several thousands at least.


They have tens of millions of active users, probably >100m total users. Even if gaming cafés are counted "incorrectly" it still wouldn't skew the numbers by .1%.


You still don't get it.
Once again, representative selection isn't about querying everyone. Steam's market share is more than large enough to get a representative selection.

john_If I have 10000 balls with 500 of them being blue and 9500 of them white and decide to pick randomly 100 and check the percentage of blue balls and white balls, statistics say that blue balls will be 5%. In reality there is a possibility to randomly keep selecting only white balls out of that huge pool of 10000 balls, concluding that there are no blue balls in there. The sample needs to be big and the sample needs to be completely random every time I am making the poll. If the sample isn't big enough and Valve is keeping the sample in a high degree stable, to avoid big variations in numbers from month to month, meaning checking let's say 75% of specific accounts to see if there is a hardware change in those accounts and only randomizes the rest 25% of the sample, the results will be biased.

Now considering Valve is an online service that can check your system specs at any time, Valve could be checking the hardware of every account in real time. It doesn't need to do a poll, when it could just check every system connected at the last 5 days of a month, for example and get a much more accurate result based on millions and not just a few thousands of systems.

efikkanNonsense, the chance to randomly select only white balls is decreasing.
The law of large numbers tells us the repeated randomized sampling of data will get closer and closer to the actual distribution, and it's becoming astronomically unlikely to repeatedly do an imbalanced selection if you keep doing it at random, even if there should be an odd error. This is basic math, and it's concerning that so many can't grasp these fundamental concepts of statistics.

efikkanNonsense, the chance to randomly select only white balls is decreasing.
The law of large numbers tells us the repeated randomized sampling of data will get closer and closer to the actual distribution, and it's becoming astronomically unlikely to repeatedly do an imbalanced selection if you keep doing it at random, even if there should be an odd error. This is basic math, and it's concerning that so many can't grasp these fundamental concepts of statistics.

Youseethist is not impossible to skew data despite the large sample but it is indeed, unlikely, because steam penetrated the asian markets as well (though... they love nvidia there :roll:)

Valve know every pc. It slipped through on the steam machine interviews that they know everything. I mean... how couldn't they?

efikkanYou will rarely find someone who is an Nvidia fanboy or even clear favoritism.

AusWolfReally? :roll:

I could give you a bunch of names just here on TPU, but I promised myself to be a good boy today. :rolleyes:

john_There are no Nvidia fanboys. You have to understand that there are AMD fans on one side and people with valid arguments on the other side :D

Dr. DroNot necessarily. If the volume is not high enough, the process of research and development, manufacturing and distribution basically becomes not worth it. With the low dGPU market share AMD currently holds plus the incentives you brought up, MSI has likely determined that continuing to work with AMD products is not worth the effort. You have to remember that MSI isn't just the MSI brand, but Micro-Star is a massive OEM that manufactures hardware for a megaton of OEMs and ODMs as well. It's about money and business sense, after all.



For-profit corporations do not spend money on charity, and they do enough charity even building these gaming cards. The truth remains, if Nvidia pulled out of the consumer space today and devoted their entire stock of GB203 and GB202 chips to AI products, it's quite likely their stock price would increase.

john_It depends by how they gather the data for their survey. If the data is collected randomly, OK. If the data is collected in a degree by the same users, because only those users are reminded to participate or care to participate, then the results are biased and changes that are happening either doesn't show up, because they might be temporary, or are seen happening in a much lesser degree than what is really happening.

yfn_ratchetI don't have much to add to the discussion as to why Radeon RX 9000 cards don't appear as present as they could be expected to in the hardware survey, but I would like to bring up that the actual sales of the GPUs seem to have outstripped what was expected especially at launch. The lack of their presence in combined statistics and the very odd presentation of Navi 48 models in the Linux satistics even to this day would indicate that there are problems in Valve's hardware survey that are only rearing their head now.

As is, the hardware survey isn't really a be-all end-all in the first place. It's viable for presenting macro trends in PC gaming and not much else.

Dr. DroThe DIY channel is very small. Once you bring in the largest and more famous system integrators known worldwide, you'll find that the statistics are more than likely accurate. No large system integrator works with Radeon GPUs because most customers don't want them.

AusWolfHow do you know that if almost none of the below examples come with a Radeon GPU? If I sell 10 systems with Nvidia, and 0 with AMD, then how do I conclude that nobody wants AMD based on my sales numbers? It's like saying that nobody wants a McDonalds steak based on the fact that McDonalds has sold 0 steaks in the last year, completely disregarding the fact that steak is not even on the McDonalds menu.

Dr. DroIt's a good thing that this is a computer part and all of these are computer manufacturers, not someone looking for steak at a fast-food burger restaurant chain instead of, you know, a steakhouse.

If there is strong demand, supply will be procured to meet it.

AusWolfHow do you know that if almost none of the below examples come with a Radeon GPU? If I sell 10 systems with Nvidia, and 0 with AMD, then how do I conclude that nobody wants AMD based on my sales numbers? It's like saying that nobody wants a McDonalds steak based on the fact that McDonalds has sold 0 steaks in the last year, completely disregarding the fact that steak is not even on the McDonalds menu.

sLowEndAMD has a track record of not being able to meet supply requirements for partners. They're not ready to sell a lot (relatively speaking, compared to Intel/Nvidia), even if potential clients are receptive.

A relatively recent example off the top of my head was a delay for GPD because AMD couldn't supply enough 7840u APUs in time
videocardz.com/newz/gpd-accuses-amd-of-breaching-contract-by-not-supplying-enough-ryzen-7-7840u-apus-on-time

Dr. DroThe DIY channel is very small. Once you bring in the largest and more famous system integrators known worldwide, you'll find that the statistics are more than likely accurate. No large system integrator works with Radeon GPUs because most customers don't want them. Just as a few examples:

Dell has zero gaming desktop computers with Radeon dGPU available
HP has only three gaming desktop computers with Radeon dGPUs available, only one of them with the 9070 XT. The other two are low cost models with substandard performance (7600 and 6400)
Lenovo has zero gaming desktop computers with Radeon dGPUs available
Maingear has zero gaming desktop computers with Radeon dGPUs available
Falcon Northwest offers the 9070 XT in some of their configurable builds, but they are not positioned as a premium offering and the default GPU is the RTX 5070 for their most premium Talon computer

Someone with an ungrounded tin foil hat could say Jensen's been paying all of them off (could have sworn that just 5 minutes ago Nvidia no longer cared about gaming or consumer market, really), but if you start to research smaller, less known OEMs, you'll find the same pattern. Laptops? Same pattern. The market overall has rejected AMD dGPUs and continues to do so. Only in techie circles where people are prone to be far more forgiving, budget conscious, apt and willing to deal with any problems that arise, do they have any meaningful foothold left, and their ardent fans do an absolutely stellar job of eroding that.