Hehehehe, I see what you did there.

he "did" "what" ?

Ryze. To graph. Is AMD's and nVidia's relative percentage... ?

Hmmmmm, Ryzen fanboy....there 😉

Caesar:

Hmmmmm, Ryzen fanboy....there 😉

everyone should be...heck Ryzen even got intel off its azz...8core's for the masses.

Fox2232:

To graph. Is AMD's and nVidia's relative percentage... ?

I was wondering the same thing. I don't see how that could be absolute marketshare, because Nvidia's should otherwise be much larger.

schmidtbag:

I was wondering the same thing. I don't see how that could be absolute marketshare, because Nvidia's should otherwise be much larger.

No actually, AMD's total GPU marketshare--excluding sales for console graphics--took more than nVidia last quarter for the first time in several years. https://www.tomshardware.com/news/amd-nvidia-gpu-market-share-report,40266.html Here's a slightly different take on it at Tom's...! It's amusing to see the reactions of the various pundits and the rationalizations they entail when they find the facts to be a bit different than what they had supposed they were. Here's the weird thing about these Peddie estimates: Intel doesn't make or market a discrete 3d GPU of any kind--so Intel's market share in that segment is 0%. Absolute zero. But instead of that being shown, we see all of Intel's iGPUs being counted as if they are the same thing as the discrete GPUs and CPUs that AMD and nVidia are selling--but they aren't. Another fact is that many laptops have both an Intel iGPU and an AMD or nVidia mobile GPU onboard. Just the fact that such solutions exist and are common seems proof enough that Intel's iGPUs are somewhat lacking when it comes to gaming API support and capability, when compared with solutions by AMD and nVidia in the same markets. Was the overlap counted accurately--or I should say "estimated" accurately? We don't know. The fact is that these days consumers are much more savvy than they were at the turn of the century, imo. And while it is true that $800 GPUs don't sell well at all, it's also true that $1400 GPUs sell much, much worse than $800 gpus, and so on. But what do people see in the tech site reviews? The rarer-than-hen's-teeth $1400 GPUs, which leads to entirely false expectations as to just how widespread these $1400 GPUs must be in the hands of the gaming public. Here's another anomaly from the story as written at Tom's Hardware: Those are the figures after AMD's console GPU sales are removed from the statistics, TH says--the idea being that they only want to count AMD's GPU sales to x86 Windows PC customers--not to AMD's console customers. OK, but the thing is that what is being sold today in the consoles is pretty much exactly what AMD sold in yesterdays x86 PCs, jaguar, etc. It's very close to that, indeed. So it's difficult to take Peddie and the various interpretations of Peddie's estimates that we see every time Peddie issues an estimate, all that seriously. "Let's penalize some of AMD's GPU sales and pretend they don't exist, but let's count everything Intel makes, even though, unlike both nVidia and AMD, Intel makes no discrete 3d accelerators at all." Nah, let's don't..;) I think the take away from this is that apparently AMD's GPU sales are doing much better than commonly believed by pop-culture pundits, and so if the trend continues I think it will go a long way to demonstrating that the students have eclipsed their teachers--at least in the application of common sense, eh?

waltc3:

So it's difficult to take Peddie and the various interpretations of Peddie's estimates that we see every time Peddie issues an estimate, all that seriously. "Let's penalize some of AMD's GPU sales and pretend they don't exist, but let's count everything Intel makes, even though, unlike both nVidia and AMD, Intel makes no discrete 3d accelerators at all." Nah, let's don't..;)

Intel makes a GPU. Period. As JPR is focusing on the PC market specifically, it's reasonable for them to exclude the XBox and PS4 as they are not PCs. If they add GPUs from consoles, they would then have to add GPUs from mobile phones, tablets, smart TVs, etc. Point is....if you're going to look at figures for TOTAL PC graphics market, you HAVE to include Intel.

schmidtbag:

I was wondering the same thing. I don't see how that could be absolute marketshare, because Nvidia's should otherwise be much larger.

For TOTAL PC graphics, NVidia can't compete with Intel, nor can AMD. The "absolute" market share is impossible to determine as market share is calculated by sales. The "absolute" market share would also include owners of previous generations.

sykozis:

For TOTAL PC graphics, NVidia can't compete with Intel, nor can AMD. The "absolute" market share is impossible to determine as market share is calculated by sales. The "absolute" market share would also include owners of previous generations.

Question is, how many of those GPU-bundled-with-CPU intel sells are actually in use.

Caesar:

he "did" "what" ?

Total madlad.

Navi gpus seems to be selling well. Thats great.

waltc3:

No actually, AMD's total GPU marketshare--excluding sales for console graphics--took more than nVidia last quarter for the first time in several years.

Then when do you make of this, also done by JPR: https://marketrealist.com/2019/06/nvidia-loses-some-discrete-gpu-market-share-to-amd/ That is more in-line with what I was expecting to see. This also lines up with the Steam hardware survey.

So it's difficult to take Peddie and the various interpretations of Peddie's estimates that we see every time Peddie issues an estimate, all that seriously. "Let's penalize some of AMD's GPU sales and pretend they don't exist, but let's count everything Intel makes, even though, unlike both nVidia and AMD, Intel makes no discrete 3d accelerators at all." Nah, let's don't..;)

I don't understand how this supports your point.

I think the take away from this is that apparently AMD's GPU sales are doing much better than commonly believed by pop-culture pundits, and so if the trend continues I think it will go a long way to demonstrating that the students have eclipsed their teachers--at least in the application of common sense, eh?

Which pundits? What's the context of "common sense" here - are you implying AMD is the obvious choice? Because as much as I try to avoid Nvidia for my own needs, I strongly disagree.

sykozis:

For TOTAL PC graphics, NVidia can't compete with Intel, nor can AMD. The "absolute" market share is impossible to determine as market share is calculated by sales. The "absolute" market share would also include owners of previous generations.

Intel I get; since iGPUs are obviously included, Intel is bound to leap ahead like that. But when comparing AMD to Nvidia, something just doesn't make sense. When you look at the Steam statistics, Nvidia makes up roughly 3/4 of the entire GPU market. That's a very different story than what JPR is telling. Of course gamers are relatively low-revenue sales compared to servers, but I got the impression Nvidia has been doing much better in servers, primarily due to CUDA. Almost every time I hear about a GPU-powered supercomputer, it's Nvidia.

schmidtbag:

Then when do you make of this, also done by JPR: https://marketrealist.com/2019/06/nvidia-loses-some-discrete-gpu-market-share-to-amd/ That is more in-line with what I was expecting to see. This also lines up with the Steam hardware survey. I don't understand how this supports your point. Which pundits? What's the context of "common sense" here - are you implying AMD is the obvious choice? Because as much as I try to avoid Nvidia for my own needs, I strongly disagree. Intel I get; since iGPUs are obviously included, Intel is bound to leap ahead like that. But when comparing AMD to Nvidia, something just doesn't make sense. When you look at the Steam statistics, Nvidia makes up roughly 3/4 of the entire GPU market. That's a very different story than what JPR is telling. Of course gamers are relatively low-volume sales compared to servers, but I got the impression Nvidia has been doing much better in servers, primarily due to CUDA. Almost every time I hear about a GPU-powered supercomputer, it's Nvidia.

Just a small note. Graph in OT is based on shipments. Steam stats and that graph you refer to are from installation base that run/share their HW data on given platform. So, while thing you posted is to show what is, and therefore help developer to see current situation. Shipment data show what is to be in future if situation persists or trend continues. If we speculate that any GPU becomes obsolete and replaced in 6 years, then short term development may focus on what is, while games that will ship in 2~3 years should focus on expected situation of HW in future down the line. If you plan for development of Game engine, then focus should look even more into future. So if we imagined that some of big Game development Engines said: "AMD is insignificant, we will not optimize for their GPUs, nor will we bring technologies developer by them." They would be in for rude awakening in 3~4 years.

schmidtbag:

Of course gamers are relatively low-volume sales compared to servers, but I got the impression Nvidia has been doing much better in servers, primarily due to CUDA. Almost every time I hear about a GPU-powered supercomputer, it's Nvidia.

Gamers are low volume compared to GPU accelerated servers? Nvidia's gaming revenue is double, sometimes nearly triple it's server revenue and has been the past 4-5 years. The margins on gaming cards are way, way less than server cards. I don't know why you'd think gamers are low volume compared to servers - it's completely the opposite. Honestly the main difference here just seems to be APUs. From 2010-2015, 75% of AMD's CPU sales were APUs. I don't know how many CPUs AMD was selling during that time but I'd assume a fair number of them compared to just Nvidia's discreet GPUs - the volume there is way more than even low end discreet. Edit: Yeah - https://images.anandtech.com/doci/10857/jpr_q3_2016_GPU_SHIPMENTS_NVDA.png https://www.techpowerup.com/img/12-11-19/146b.jpg I can only find 2012 for AMD but look at the APU units sold for that year. It's by far the bulk of shipped GPUs for AMD - where as Nvidia is all discreet but shipping nearly the same number of units overall. Which is why there is such a massive revenue difference.

Denial:

Gamers are low volume compared to GPU accelerated servers? Nvidia's gaming revenue is double, sometimes nearly triple it's server revenue and has been the past 4-5 years. The margins on gaming cards are way, way less than server cards. I don't know why you'd think gamers are low volume compared to servers - it's completely the opposite.

Whoops volume was the wrong word. I meant revenue, because yeah, there are a LOT more gaming GPUs sold than server GPUs. I fixed the wording in my post. Though even still, where are you getting your data from? I actually tried finding it myself but couldn't find anything direct. Generally speaking, servers have always been the big money-makers. That's why AMD's GPU architecture was so heavily revolved around compute tasks. Take the Summit supercomputer for example: that has 27648 Nvidia Tesla V100s. I'm not sure if they're 16GB or 32GB, but to average out the price, that's about $11000 a piece for their MSRP. That means Summit is comprised of $304,128,000 worth of Nvidia GPUs. Lately, Nvidia's yearly revenue is on average roughly $3 billion. That means Summit alone accounted for about 10% of that. Summit isn't the only Nvidia-based server. Stuff like the Tegra series, automotive chips, and products that aren't discrete GPUs account for roughly 15% of Nvidia's revenue. Nvidia, to my knowledge, is the most popular choice among workstations too. So, although I have no direct numbers of gaming GPUs, I have a hard time believing they make up the majority of Nvidia's revenue. I know that isn't exactly what you were arguing against per-se, but just thought I'd point all that out. Don't forget that Nvidia profits more from stuff like server and workstation GPUs too. I know that doesn't really have anything to do with how much products are sold, but, at least when you account for net revenue, I'm sure that dramatically tilts the scales in favor of non-gaming GPUs.

For clarity sake, discrete GPU market share (as opposed to the other chart which includes AMD APUs): https://articles2.marketrealist.com/?load=aHR0cHM6Ly9tYXJrZXRyZWFsaXN0LmltZ2l4Lm5ldC91cGxvYWRzLzIwMTkvMDYvQTRfU2VtaWNvbmR1Y3RvcnNfTlZEQS1EaXNyZXRlLUdQVS1tYXJrZXQtc2hhcmUtUTExOS5wbmc_dz01ODcmZml0PW1heCZhdXRvPWZvcm1hdA

schmidtbag:

Whoops volume was the wrong word. I meant revenue, because yeah, there are a LOT more gaming GPUs sold than server GPUs. I fixed the wording in my post. Though even still, where are you getting your data from? I actually tried finding it myself but couldn't find anything direct. Generally speaking, servers have always been the big money-makers. That's why AMD's GPU architecture was so heavily revolved around compute tasks. Take the Summit supercomputer for example: that has 27648 Nvidia Tesla V100s. I'm not sure if they're 16GB or 32GB, but to average out the price, that's about $11000 a piece for their MSRP. That means Summit is comprised of $304,128,000 worth of Nvidia GPUs. Lately, Nvidia's yearly revenue is on average roughly $3 billion. That means Summit alone accounted for about 10% of that. Summit isn't the only Nvidia-based server. Stuff like the Tegra series, automotive chips, and products that aren't discrete GPUs account for roughly 15% of Nvidia's revenue. Nvidia, to my knowledge, is the most popular choice among workstations too. So, although I have no direct numbers of gaming GPUs, I have a hard time believing they make up the majority of Nvidia's revenue. I know that isn't exactly what you were arguing against per-se, but just thought I'd point all that out. Don't forget that Nvidia profits more from stuff like server and workstation GPUs too. I know that doesn't really have anything to do with how much products are sold, but, at least when you account for net revenue, I'm sure that dramatically tilts the scales in favor of non-gaming GPUs.

Revenue by Segment: https://thumbor.forbes.com/thumbor/960x0/https://blogs-images.forbes.com/moorinsights/files/2018/05/Picture1-2.jpg But I'm not sure what you're asking for - you originally said Nvidia's marketshare should be much larger but when the marketshare is simply measured by units sold it's easy to see why Nvidia and AMD are even. Nvidia receives a ton more revenue for it's products but it's only shipping around the same number of them per quarter as shown in the two links I posted above. Also large purchases like that of a super computer or any large organization aren't paying MSRP. The prices for contracts like that are highly negotiable to the point where they might be paying near half for the actual GPUs because there is a bunch of other stuff Nvidia is providing as part of that contract.

schmidtbag:

... Don't forget that Nvidia profits more from stuff like server and workstation GPUs too. I know that doesn't really have anything to do with how much products are sold, but, at least when you account for net revenue, I'm sure that dramatically tilts the scales in favor of non-gaming GPUs.

Definitely! The GPU co-processors used purely for computation cost roughly ten times a typical video card. Here is something new (I think) they are selling now and that is the GPU stand-alone, not on a card. I saw several companies at the GTC in March with motherboards for servers with GPUs on them. Here's a picture of one. Look at the size of the power wires in the back. Wow! https://i35.servimg.com/u/f35/17/98/38/10/img_0810.jpg Lots of companies sell server boxes with lots of cards - I saw one with sixteen cards in the box, but this is the first I have seen or heard of with GPUs in them and not on a card.

schmidtbag:

But when comparing AMD to Nvidia, something just doesn't make sense. When you look at the Steam statistics, Nvidia makes up roughly 3/4 of the entire GPU market. That's a very different story than what JPR is telling. Of course gamers are relatively low-revenue sales compared to servers, but I got the impression Nvidia has been doing much better in servers, primarily due to CUDA. Almost every time I hear about a GPU-powered supercomputer, it's Nvidia.

Steam statistics are not an accurate measure of market share. Steam counts every GPU in use. If Intel's iGPU is disabled due to a dedicated card being installed, it doesn't get counted. Market share, as reported by JPR and other firms, is determined solely by quarterly sales. It's actually the share of quarterly sales for the market.

Fox2232:

Question is, how many of those GPU-bundled-with-CPU intel sells are actually in use.

It doesn't matter if they're in use or not. They're still counted as a GPU sale. No marketing firm has access to accurate consumer product usage information. These firms only report "quarterly sales market share". They don't report consumer product usage nor do they report accurate product ownership figures.