Hell has frozen over, Fudzilla Both Microsoft and Intel trying to acquire AMD

[ShintaiDK]

That is where the volume comes into place, specially at over 125 million. Making your case completely irrelevant in statistics.

Fact is. dGPU sales are in free fall. Gamers gaming on IGPs is in a rapid rise. So no matter if you like it or not, this is what we have to relate to. Not some alternate world.

 

[Genx87]

I could see MS buying them at a cheap enough price to use their GPU tech in consoles, phones, and tablets. Intel would want them for the GPU technology, dropping the x86 division obviously. I am not sure the govt would allow Intel to buy AMD. I could see them allowing Microsoft. If Intel got a hold of AMDs GPU tech the next company on the chopping block would probably be Nvidia.

 

[Blitzvogel]

I have 2-3 desktops and perhaps 1 or more Laptops.

One month the Steam survey get me when im using my Laptop, then the next moth Steam Survey got my using the SFF desktop on my TV, then the other month the Steam Survay gets me using my Gaming PC.

no need to say more.

I have wondered about this issue. My fiancee and I have two desktops and a laptop, though if I get a request for the survey on my Yoga, I decline it, or at least I think I do.

 

[AtenRa]

That is where the volume comes into place, specially at over 125 million. Making your case completely irrelevant in statistics.

You are joking right ?? how many have more than a single desktop nowdays ??

Fact is. dGPU sales are in free fall. Gamers gaming on IGPs is in a rapid rise. So no matter if you like it or not, this is what we have to relate to. Not some alternate world.

dGPU volume is decreasing because OEM PCs are not selling with a dGPU anymore.
But here we are talking about dGPU upgrades that those with older iGPUs will have to make sometime in the life of the PC.

 

[AtenRa]

I have wondered about this issue. My fiancee and I have two desktops and a laptop, though if I get a request for the survey on my Yoga, I decline it, or at least I think I do.

If you decline, you will only be able to take the survey the next month.

 

[cytg111]

...
And just to show that you need to stop thinking traditionally.
http://store.steampowered.com/hwsurvey/cpus/

Dualcores up, quadcores down.

This is reality.

Are you sure that specific statistic is supporting the point you're trying to make? You should know the total installation base of steam installations over time if you want to say anything meaningful, scrap that, active steam installations over time. On top of that, what installations is the moneymaking installations? http://steamspy.com/ <- these games in general is not dual core friendly, much less so dual-core-igp friendly. I could easily believe that given all the free content and extremely discounted content(for older hardware) on steam, that older machines, lesser hardware gets a steamclient, cause, hell why not. Given the facts given here I dont see any evidence for "Dualcores up, quadcores down".

 

[Dresdenboy]

If this interposer option was really that good why isn't Nvidia trying it now? Or even AMD? Given the critically low volumes of AMD today, that could be an option for them.

An out of context question? Ok. As long as these 600sqmm dies with their yields and per design costs make them enough money, there is no need dor that option. It will be more feasible if creating separate big die designs don't generate enough ROI. Together with optional HBM it would be cheaper to combine smaller dies, sharing design costs.

 

[mrmt]

An out of context question? Ok. As long as these 600sqmm dies with their yields and per design costs make them enough money, there is no need dor that option. It will be more feasible if creating separate big die designs don't generate enough ROI. Together with optional HBM it would be cheaper to combine smaller dies, sharing design costs.

So using an interposer is not more profitable than using monolithic dies. Got it. And I would not have much hope for HBM in the first place. In anything it should be more expensive than GDDR5. With Fury AMD made the miracle of transforming a $650 dGPU in a low margin product.

 

[Dresdenboy]

Going onwards to lower nodes is well documented in terms of cost. And its no news that companies like nVidia publicly complain that there is no cost savings on a transistor/cost basis. Improving the chip designs also cost money. It all needs to be payed from somewhere and its anything but cheap. And the amount base this money comes from is on a rapid retreat.



So it was just some guess, ok.

Ok, how much smaller than this estimation is it then? Your point is valid, if it's just <300 sqmm.

And what will the size of the gamer market be after iGPUs squeezed out the smaller mobile, desktop and OEM dGPUs? And how will growing markets like BRIC (temporary crises aside since we are talking a multi year future) influence or even offset this? If some parts of a market are reasonably shrinking due to obsolete HW, this doesn't mean doom for all market sectors at once.

Dont try and make it personal just because you dont like the message. I am simply looking at it from a financial standpoint. Because I know it doesnt matter what I want and dont want. Its the market and financials that dictate the progress. And currently they are voting for IGPs big time with a slight twist of gaming GPUs. The problem is then that gaming GPUs cant stand on their own legs in the long run.

If you look at nVidias financials. While you see the PR thunder about Geforce GTX revenue increase, you also see a in drop in OEM revenue.

Yes, financials and market demands dictate this. But while this is true, it doesn't automatically prove your underlying assumptions.

BTW:

 

[DrMrLordX]

Wrong. Performance/watt is just one of these examples where AMD is horrible behind. Having to use 600mm2 dies to compete with 398mm2 shows exactly whats wrong.

Dunno what you're going on about. The Nano is an excellent product, better than Fury X from a perf/watt perspective. When it came out, the 290X was also an excellent product . . . it took the 970/980 to really put it behind. AMD just flogged that thing for too long, and it's mystifying that the 290X didn't become a market sensation especially with the virtual currency mining craze. They should have been able to outsell Nvidia like mad, and yet they didn't.

Nano should also sell like hotcakes. Will it? Time will tell. A lot of that's going to come down to the availability of the 65nm interposer, though.

Those two guys were the public face of two AMD debacles (Bulldozer, Ambidextrous) but by no means they were alone conceiving it let alone executing it, and no strategy as big as these two could be executed without express authorization of the BoD. No one with billions on the account would be dumb enough to buy these sacrificial lambs while preserving the rest of the BoD.

Then we'll see exactly how "dumb" investors are between now and . . . whenever AMD threatens to go insolvent. Again, looks like Silver Lake found something in there they liked. It's quite possible that they plan to simply override the current board of directors wherever necessary, though how they plan to do that with only a 20% stake is beyond me. I'm sure they have their ways.

If this interposer option was really that good why isn't Nvidia trying it now? Or even AMD? Given the critically low volumes of AMD today, that could be an option for them.

The interposer is the alleged reason for poor availability of Fiji parts. AMD can't get enough of the things for larger shipping volumes. That being said, AMD *is* trying an interposer, and Fiji is their test case. If it works out, yes, theoretically, they could use interposers and try die stacking on APU/CPU products as well. I'm not sure how the thermals will work out for them. GPUs tend to run hot, so if it works for Fiji, I don't see why not use it for something on AM4 (for example).

 

[Dresdenboy]

So using an interposer is not more profitable than using monolithic dies. Got it. And I would not have much hope for HBM in the first place. In anything it should be more expensive than GDDR5. With Fury AMD made the miracle of transforming a $650 dGPU in a low margin product.

HBM needs some time to mature. And that mem BW needs some strong GPU to provide enough of an advantage as the costs if adding 2 stacks are fixed, but offset at the OEM's side (PCB, testing, mem chips). There is some crossover point.

 

[ShintaiDK]

Yes, financials and market demands dictate this. But while this is true, it doesn't automatically prove your underlying assumptions.

BTW:

You can see OEM shipment drops. And thats even after nVidia completely destroyed AMD. (Btw nVidia is in FY2016).

So what happens when nVidia cant cannibalize on AMD anymore because there is nothing to cannibalize on?

 

[ShintaiDK]

Dunno what you're going on about. The Nano is an excellent product, better than Fury X from a perf/watt perspective. When it came out, the 290X was also an excellent product . . . it took the 970/980 to really put it behind. AMD just flogged that thing for too long, and it's mystifying that the 290X didn't become a market sensation especially with the virtual currency mining craze. They should have been able to outsell Nvidia like mad, and yet they didn't.

Nano should also sell like hotcakes. Will it? Time will tell. A lot of that's going to come down to the availability of the 65nm interposer, though.

Why do you think it will sell like hotcakes?

And the Nano is essentially what you could get a year ago and cheaper. See the issue? Even Bulldozer would have been a good chip, if it wasnt for the time period.

 

[Dresdenboy]

You can see OEM shipment drops. And thats even after nVidia completely destroyed AMD. (Btw nVidia is in FY2016).

So what happens when nVidia cant cannibalize on AMD anymore because there is nothing to cannibalize on?

I'm short on time as I'm answering on my way home (also reason for some typos). Please find the FY2016 chart. Where are your calculations showing that Nvidia's growth only exists due to taking mss from AMD and having no other base?

 

[shady28]

I have 2-3 desktops and perhaps 1 or more Laptops.

One month the Steam survey get me when im using my Laptop, then the next moth Steam Survey got my using the SFF desktop on my TV, then the other month the Steam Survay gets me using my Gaming PC.

no need to say more.

That's ridiculous. Most analysts freely admit that Steam overrates "hard core" gamers; people with iGPUs are far far less likely to be using Steam. They're a lot more likely to be playing Zynga games or some Chrome game.

For example, the systems in my sig - exactly one of them has Steam installed. Can you guess which one?

And these are all dGPU boxes - none of them except the desktop capable of playing games decently (because they're old). That being the case, I don't stick Steam on them to pointlessly consume resources. The logic follows that many, many iGPU users do not play dedicated games like those on Steam, and hence are under-reflected.

But even within your (bizarre) use case of only using one system per month, all of your machines at least get counted. Multiply your use case by 10 people doing the same thing and you'll find that statistically the ratio of dGPU:iGPU is dead spot on.

Steam is the only massive sample size of real, hard information available. It may not be 100% accurate, but it's going to be a heck of a lot more accurate than a bunch of people trying to pass their opinion off as factual data in a forum somewhere...

 

[sm625]

I would sure hope that Intel would not run into antitrust issues if indeed they actually wanted to buy AMD. Intel is far weaker than they ever used to be. It doesnt really matter much anymore if they had a total monopoly on x86. They already charge about as much as they possibly can.

 

[ShintaiDK]

Keep an eye on it next week. But this is money flowing into an - according to many world-influential forum members - doomed company.

How did the Monday turn out so far? -9.45% and regular volume.

Someone rich fooled the poor again.

 

[Dresdenboy]

How did the Monday turn out so far? -9.45% and regular volume.

Someone rich fooled the poor again.

Maybe even not the rich. It just needs a rumour site.

The smart ones could even make 10-20% on that move as direction doesn't matter as long as there is enough vola.

 

[ShintaiDK]

Maybe even not the rich. It just needs a rumour site.

The smart ones could even make 10-20% on that move as direction doesn't matter as long as there is enough vola.

So you dont believe in your "money flowing in" statement anymore?

 

[Chaosblade02]

The worst thing imaginable for the PC owners is for M$ to get their hands directly in the PC hardware market. Excluding their tablets. I hope the antitrust laws in the USA keep them from doing this, but I'm not optimistic at all.

 

[SlowSpyder]

I see dGPU's going the way of desktop CPU's. That is, they probably won't be designed for desktop use first, but we'll get the scraps of what they can make work. They'll design for mobile or workstation and scale up or down to make it work on as a desktop part, kind of like AMD's and Intel's CPU line up now.

 

[AtenRa]

people with iGPUs are far far less likely to be using Steam. They're a lot more likely to be playing Zynga games or some Chrome game.

Steam currently has 275 FREE games, of those the vast majority can be played with HD4000 and above iGPUs at 720p or above.

You can play the vast majority of them on a AMD A8-A10 APU at 1080p(especially with Kaveri APUs).

For example, the systems in my sig - exactly one of them has Steam installed. Can you guess which one?

And these are all dGPU boxes - none of them except the desktop capable of playing games decently (because they're old). That being the case, I don't stick Steam on them to pointlessly consume resources. The logic follows that many, many iGPU users do not play dedicated games like those on Steam, and hence are under-reflected.

At home I have a main Gaming PC with an i7 3770K and HD7950 with steam installed. I have an SFF HTPC with an AMD APU that also has steam installed to play games on my large TV. Car simulators and other less demanding free games that my wife and children of friends coming home play. Wife using a laptop with steam installed as well, fully capable of playing those free Steam games.

My main Gaming PC has took part in the Steam Hardware Survey ONLY two or three times the last 24 months or so. The HTPC was second with 3 or 4 times if i remember correctly and wife's Laptop took more than 5 (she plays more hours than me by far being home with the kid all day).

But even within your (bizarre) use case of only using one system per month, all of your machines at least get counted. Multiply your use case by 10 people doing the same thing and you'll find that statistically the ratio of dGPU:iGPU is dead spot on.

See above, my main PC hasn't participated on Steam Hardware Survey the last 6 months or more. There is no data from my main Gaming PC in the current Steam Hardware Survey.

Ohh one more thing,

NOT everyone with a steam account is taking the survey every month. That can happen because you might not open Steam that day the survey it takes place.
I really dont remember when was the last time i took the survey.

So Steam accounts may be 125M, but not everyone takes the survey every month and that also skews the results.

 

[frozentundra123456]

You can see OEM shipment drops. And thats even after nVidia completely destroyed AMD. (Btw nVidia is in FY2016).

So what happens when nVidia cant cannibalize on AMD anymore because there is nothing to cannibalize on?

So? Even in absolute dollars, OEM shipments dropped (obviously because of igps), while gaming increased by about twice as much as OEM shipments dropped.

 

[ShintaiDK]

So? Even in absolute dollars, OEM shipments dropped (obviously because of igps), while gaming increased by about twice as much as OEM shipments dropped.

All GTX cards account as gaming cards. And its not sustainable. What happens when AMD dont have anymore to lose in dGPU share?

http://files.shareholder.com/downlo...8-2346D1AABED8/Q216_CFO_Commentary_NASDAQ.pdf

Q2 2015 to Q2 2016 GPU sales is up 9% from 878M$ to 959M$. At the same time nVidia went from 62% to 81.9% marketshare. Or a 32.1% increase. See the point? 9% vs 32.1%. GTX sales up 51% and so on.

 

[VirtualLarry]

I see dGPU's going the way of desktop CPU's. That is, they probably won't be designed for desktop use first, but we'll get the scraps of what they can make work. They'll design for mobile or workstation and scale up or down to make it work on as a desktop part, kind of like AMD's and Intel's CPU line up now.

This is probably the most accurate assessment that I've seen. As IGPs grow in performance, they will need R&D, and that same R&D can be turned around and scaled up to produce dGPUs. If there's still demand for dGPUs, then I have no doubt that they will be produced. (At least, until PCI-E dies.)