Bomby569There is a story behind all this, why was microsoft using bulldozer (i know not called bulldozer still same thing) that was a joke, and not intel or nvidia back in the day.

I think i detected a bit of Nintendo hate. Deal with it, they are very sucessufull.

for total system sales, its actually lower since they combine it with handheld console, and i hate it, they should make true successor to nintendo ds, i really love dual screen aspect of it.

R0H1TThat's simply not true post 2013(14?) Nvidia did nothing for Tegra & their cores were power hungry & ran hot as hell, Apple was miles ahead in terms of efficiency & even QC was beating them with their custom cores! Nvidia's demise came at their own hands.

Sorry, that's just not true. Nvidia's custom ARM cores weren't a smash hit, but they had some rather interesting characteristics still. As for efficiency, they were fine, neither more nor less. But at that point they had already been effectively forced out of the mobile market through QC's bribery - that they didn't have tons of money to spend on R&D is rather a given at that point.

But as to what I meant, that you seem to have missed: at any point since then they could have put out an SoC with the highest number of stock ARM cores possible, clocked high for tethered use, and their own GPUs and marketed it for console use, if they wanted to. They clearly didn't.

Bomby569There is a story behind all this, why was microsoft using bulldozer (i know not called bulldozer still same thing) that was a joke, and not intel or nvidia back in the day.

Jaguar is based on Bobcat, which was a redesign of the original K6 core (from 3 generations earlier).

Bomby569detected a bit of Nintendo hate

Oh, a faulty detector is quite in line with the rest of your post. :D

ValantarBut as to what I meant, that you seem to have missed: at any point since then they could have put out an SoC with the highest number of stock ARM cores possible, clocked high for tethered use, and their own GPUs and marketed it for console use, if they wanted to. They clearly didn't.

You put it as if anyone an their grandmas could just roll out a successful console.

How did that NV Shield thing go? :D
Well, it didn't quite fly.
And no prospect of it ever flying either.
So that's how it died.

Same chip appearing in Nintendo's console, doesn't change it much.
"It is a business to them, but it's not a business to us" (or something along those lines) by NV CEO in regards to Switch.

Keep in mind that Nintendo is the only of the major console makers that sticks to making money on console hardware, shrinking supplier margins even further.

medi01You put it as if anyone an their grandmas could just roll out a successful console.

How did that NV Shield thing go? :D
Well, it didn't quite fly.
And no prospect of it ever flying either.
So that's how it died.

Same chip appearing in Nintendo's console, doesn't change it much.
"It is a business to them, but it's not a business to us" (or something along those lines) by NV CEO in regards to Switch.

Keep in mind that Nintendo is the only of the major console makers that sticks to making money on console hardware, shrinking supplier margins even further.

... no, I was specifically responding the discussion of how/why Nvidia was bypassed for all previous and current gen consoles except for the Switch. I was in no way whatsoever arguing for whether Nvidia or anyone else could have made a successful console - that's another discussion entirely. I was simply saying that Nvidia never tried or offered to make a proper console SoC, despite being able to do so if they had wanted to. This unwillingness, in turn, is a significant factor in the console industry turning to AMD, who has been very willing to make semi-custom SoCs for essentially anyone who wants one for more than a decade. The Shield and Switch and their Tegra SoCs don't really impact this argument, as they are low-powered designs intended purely for mobile implementations, and are materially different from what either Sony or MS would have wanted for their consoles (which in turn just highlights Nvidia's inflexibility). Nvidia could absolutely have put out an SoC to rival the PS4 and XO SoCs in 2012-2013, in a comparable or lower power budget. They weren't interested - and they are still notorious for their unwillingness to offer made-to-order implementations of their hardware. As such it's no wonder that their inroads into the console market have been very limited since consoles started moving to SoCs.

medi01There is no way CPU+GPU with reasonable margins would ever beat an APU offering.

You are doing a small mistake in your thinking.
You forget the ARM platform.
The answer to your question is everywhere, in all those ARM based systems.

ValantarIt's an ongoing decision though - they've continuously had the ability to put out a high performance ARM-based mobile gaming SoC since then - as illustrated by the Switch, even if its SoC was old even at launch - but they've just shown zero interest in doing so. My guess is that Qualcomm's bribery and shady dealings in the Tegra era just put them off the consumer ARM SoC market entirely, which would be understandable, but given both the improved ARM core performance and explosion in handheld gaming in recent years I still can't understand why they haven't given it another go. Even if the automotive market and AI/HPC is a lot more profitable, it should be relatively easy for them to make something like this happen. Then again, they also have this really weird insistence on not making semi-custom designs, which ... yeah, I still don't get that.

They wanted FULL control over ARM first.
Nvidia's biggest power and the same time biggest disadvantage, is their own arrogance as a company.

ValantarBut at that point they had already been effectively forced out of the mobile market through QC's bribery

I believe they could withstand what Mediatek withstood all those years.

They just DIDN'T WANTED TO.
Then ARM ecosystem started growing and they realized what they where missing. Instead on focusing on how they will get back to the ARM ecosystem with a tone of products made for various categories, from smartphones to servers, they did what their arrogance dictated, once more. Tried to buyout ARM.

john_They wanted FULL control over ARM first.
Nvidia's biggest power and the same time biggest disadvantage, is their own arrogance as a company.

You might be onto something, but there's a significant span of time between Nvidia dropping out of the ARM SoC market and their ambitions towards acquiring ARM - more than half a decade. I sincerely doubt they were hatching those plans back in 2012-13 when it was becoming clear that QC's grip on the smartphone market was too strong.

john_I believe they could withstand what Mediatek withstood all those years.

Mediatek as a serious competitor to QC is very new - a couple of years at most. QC doesn't have the same type of freedom to pull of anticompetitive shenanigans now that they did a decade ago, as there's infinitely more regulatory attention on the mobile SoC market now compared to then. If QC tried to squeeze out MTK now like they did Nvidia back then, they would be eviscerated by regulatory agencies. But up until quite recently, QC had no reason to try and squeeze out MTK, as they didn't really compete in the same markets, and MTK rather filled out a low-end space that QC either didn't have the resources to or want to fill themselves.

john_They just DIDN'T WANTED TO.
Then ARM ecosystem started growing and they realized what they where missing. Instead on focusing on how they will get back to the ARM ecosystem with a tone of products made for various categories, from smartphones to servers, they did what their arrogance dictated, once more. Tried to buyout ARM.

There is something to this, but it's also more complicated than this - and it happened over a significant amount of time that gets seriously compressed in your retelling. Nvidia today is not the same as Nvidia ten years ago, just as Qualcomm and Mediatek today aren't what they were a decade ago. Neither is ARM, for that matter. From an outside perspective it seems that Nvidia started wanting to buy out ARM at the same time as it became clear that ARM cores could compete reasonably in the server, datacenter and HPC space - which is a central market for Nvidia, unlike mobile.

Could Nvidia have withstood QC's bribery and other anticompetitive shenanigans in the mobile market? Possibly. Trying to compete there cost them quite a lot, and their Denver cores while innovative and interesting underperformed in those markets, which would likely mean hundreds of millions if not billions invested as well as several years for Nvidia to develop a more competitive mobile core again. It makes perfect sense that they left the market behind when they did. They cut their losses and moved on. Whether or not they were planning a return to that market after acquiring ARM is impossible to tell - they could have intended literally anything with that acquisition - but there's no material difference in their competitive situation in that regard owning or licencing ARM designs. They could make a high performance ARM SoC today if they wanted to, which is what I've been saying all along. As such, you're right in saying that they didn't want to, but that's also what I've been saying all along in regard to Nvidia competing in consoles or handheld consoles - they just aren't especially interested.

ValantarNvidia never tried or offered to make a proper console SoC, despite being able to do so if they had wanted to.

The "being able to" is an arguable part of the statement, to put it softly.

It is not like AMD has created APUs for consoles. It just happened to have product that was easily customizable for certain needs.

NV investing into R&D of "something" that might not even end up in console, is not a realistic scenario and that is not due to lack of NV's interest to be presented in consoles.

ValantarThey weren't interested

This statement is based on? Right, nothing. Just speculation.

AMD was the only company who had things of interest from both GPU and CPU world.
It was weaker on the CPU side, but Intel's margins and Jaguar being good enough.

Last, but not least, the way NV handled things with Microsoft earlier harshly undermined relationships between the two. Microsoft didn't even bother to talk to NV for the previous gen, they just went straight to AMD.

So there was basically no more land for NV to stand on. "But if they wanted" assume they could magically undercut AMD, who shares R&D costs of its APUs with other products. It is not even remotely imaginable.

john_You forget the ARM platform.
The answer to your question is everywhere, in all those ARM based systems.

"All those systems" that are in... phones eh? I recall there also was that "Tegra" thing... :)

Remind me of a single ARM based APU with TDP of 100W+.

medi01"All those systems" that are in... phones eh? I recall there also was that "Tegra" thing... :)

Remind me of a single ARM based APU with TDP of 100W+.

Does M1 /Ultra needs to have over 100W TDP to count as an example of a highly performing ARM based SOC/APU?
That's just an example of a SOC/APU that is based on ARM and performs great in comparison with a SOC/APU that is based on X86.

There are plenty of server based CPUs out there for servers that go far beyond 100W.
Those are examples that the ARM architecture does not self destruct if you feed it Mooooooar Powerrrrrrr!!!

The GeForce in Switch and RDNA2 in the latest Exynos are examples that you CAN in fact use what AMD and Nvidia created for desktops in SOCs with ARM cores.

Are the above examples enough for you? Because this is what it is strange. That companies haven't tried to build SOCs for laptops and desktops based on ARM. The above examples prove that X86 is NOT an absolute ingredient for an APU. It seems that no one was feeling it was financially viable to try to challenge the X86 platform in the consumer space. Even Qualcomm's collaboration with Microsoft seems to miss the necessary excitement needed to move forward and have a chance.

While you are right that a combination of two chips, CPU + discrete GPU, will always be more expensive, the fact that companies haven't tried to build REAL ARM based APUs for laptops(and desktops), doesn't mean it is impossible or that it will be expensive to manufacture them. ARM based APUs will probably be cheaper to build, but the cost to market them against an esptablished X86 market, much higher. And the last 10 years, I can only think of Nvidia having an interest to try it. Qualcomm's "try" is a joke. Apple did it because it control's it's ecosystem. And it's customers are loyal and go where the company points. Nvidia probably wanted that security, of full control of the platform, before investing dozens of billions of dollars. Now they will have to decide if they will try it, even without having full control of the platform. If they try it, yopu'll get your "100W+ TDP" example in a few years.

tegra soc was slow, hot and power hungry, there a reason why nintendo underclock it, to save power and have sustained performance

medi01The "being able to" is an arguable part of the statement, to put it softly.

And that is precisely what I'm arguing: if they wanted to, they could have created a high-power, higher clocking version of a standard ARM core and marketed it for console use. They have been designing ARM SoCs since pre-2010, so they have the engineers and licenses.

medi01It is not like AMD has created APUs for consoles. It just happened to have product that was easily customizable for certain needs.

... that's what semi-custom design is: you offer your various products and ways of combining them to customers and let them specify the combination they want. And Nvidia has a long and well documented history of refusing such collaborations.

medi01NV investing into R&D of "something" that might not even end up in console, is not a realistic scenario and that is not due to lack of NV's interest to be presented in consoles.

... this is just circular reasoning. If you plan to offer a product to a market, you need to start R&D before you have contracts in place - that's how causal reality works. If they had wanted to pre-8th gen consoles, they could absolutely have pitched themselves to both Sony and MS as a viable SoC vendor for consoles. But they didn't want to, and thus it didn't happen.

medi01This statement is based on? Right, nothing. Just speculation.

Speculation, the non-existence of any semi-custom Nvidia SoC, their storied and well-documented resistance to collaborative design.

medi01AMD was the only company who had things of interest from both GPU and CPU world.
It was weaker on the CPU side, but Intel's margins and Jaguar being good enough.

And Nvidia could have designed an ARM-based CPU that could have competed with Jaguar at similar power levels even back in 2012. Would it have been a bit of a challenge? Absolutely. But if they were interested in that market, they could have done it.

medi01Last, but not least, the way NV handled things with Microsoft earlier harshly undermined relationships between the two. Microsoft didn't even bother to talk to NV for the previous gen, they just went straight to AMD.

Which again plays into their lack of interest: their "interest" was in terms of "our way or the highway", which isn't so much an expression of interest as it is saying "well, if you really want to buy our hardware, I guess you can". If nvidia had a genuine interest in catering to the console market, they wouldn't be acting that way towards their potential partners. Assuming they're at least moderately competent, that is.

wahdanguntegra soc was slow, hot and power hungry, there a reason why nintendo underclock it, to save power and have sustained performance

The Tegra X1 in the Switch was several years old even when the Switch launched, so it's hardly surprising that it wasn't amazing - but it also demonstrates really well what can be done with low performance hardware if software is designed within those constraints. Of course theoretically Nvidia could put out a chip in the same power envelope as the Switch that would absolutely trounce it today - but unless Nintendo buys one of their automotive chips this sadly doesn't seem to be materializing. Weird, really - do Nvidia not want to sell millions more chips?

GarrusThis is a bit small when the Switch itself is already 7", but this is great news! Alder Lake and Ryzen 6000, that's what we need. Not the horrible Mendocino. Can't wait for the comparison videos.

Actually there's a Mendocino option as well, it's ubiquitous!
liliputing.com/2022/05/ayn-loki-handheld-gaming-pc-goes-up-for-pre-order-june-1st-for-299-and-up-with-intel-or-amd-processor-options.html
Mendocino will equip the AMD alternative for the Loki Mini and Loki Mini Pro, while on Intel's side there's a Celeron 7300 for the Loki Mini and a Pentium 8500 for the Loki Mini Pro. Both come with 128GB M2 2230 NVMe drives and 8GB of RAM (LPDDR4x-4266 for Intel and LPDDR5-6400 for AMD).

It's hard to find comparison reference points but I think the Intel option will get trashed by AMD :D

4 zen2 cores vs 1P+4e Intel on the very limited envelope of a mobile device like this, don't think it will end well for Intel
www.cpubenchmark.net/compare/Intel-Pentium-Gold-8505-vs-AMD-Ryzen-3-PRO-4450U/4775vs3816 (this is the 8505 that runs at a much higher power level, and an old Renoir using lpddr4 - not that different from lpddr5)

Valantar... this is just circular reasoning. If you plan to offer a product to a market, you need to start R&D before you have contracts in place - that's how causal reality works. If they had wanted to pre-8th gen consoles, they could absolutely have pitched themselves to both Sony and MS as a viable SoC vendor for consoles. But they didn't want to, and thus it didn't happen.

That is you simply brushing off an argument.

AMD didn't design APUs with sole purpose of putting them into consoles, they have other uses for it and R&D is justified.

On the opposite, NV investing R&D (after Tegra failing) with NO GUARANTEE that any of console manufacturers will pick it up, is too risky an investment for that to happen. It has nothing to do with what one "wants" or "is interested in that market".

ValantarWhich again plays into their lack of interest: their "interest" was in terms of "our way or the highway", which isn't so much an expression of interest

I recall Apple has dropped contacts with NV too, and that back in times where AMD was quite behind.

Not sure I'd call it "lack of interest", but rather "excess of arrogance". It rendered NV a not-worth-making-business-with partner.

medi01That is you simply brushing off an argument.

AMD didn't design APUs with sole purpose of putting them into consoles, they have other uses for it and R&D is justified.

On the opposite, NV investing R&D (after Tegra failing) with NO GUARANTEE that any of console manufacturers will pick it up, is too risky an investment for that to happen. It has nothing to do with what one "wants" or "is interested in that market".

They didn't design APUs in general/as a concept for consoles, but an APU is just an SoC, and hardly a novel concept even in the mid-2000s. The APUs used in 8th gen consoles on the other hand were absolutely made for that sole purpose, as they share essentially nothing with contemporary PC APUs.

Also, I think you have your time lines a bit mixed up here. Tegra failed a couple of years into the 8th console generation, so there would have been nothing from that failure stopping them from pitching themselves to console partners for that generation. They even could have done so for the current/9th gen, as they have continued making high performance Tegra chips for the automotive market (which are obviously not directly transplantable, but a good basis to start from, and which gives them a lot of experience in high performance SoC design). Hence my point of this not coming down to an inability to deliver a competitive solution.

medi01I recall Apple has dropped contacts with NV too, and that back in times where AMD was quite behind.

Not sure I'd call it "lack of interest", but rather "excess of arrogance". It rendered NV a not-worth-making-business-with partner.

That is literally exactly what I was saying. Nvidia was "interested" in the console market only if they could dictate their own terms, which no partner was willing to (for obvious reasons).