NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products

[kac77]

NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products​

Intel to design and manufacture custom data center and client CPUs with NVIDIA NVLink; NVIDIA to invest $5 billion in Intel common stock
NVIDIA (NASDAQ: NVDA) and Intel Corporation (NASDAQ: INTC) today announced a collaboration to jointly develop multiple generations of custom data center and PC products that accelerate applications and workloads across hyperscale, enterprise and consumer markets.

 

[Marees]

So for servers: nvlink with x86

Client ?? — maybe $$$$ surface tablets ??

 

[AVATARAT]

This is very interesting news.
The both companies need each other, so the product range will now be very strong. But prices will skyrocket...

 

[littlegreenmenfromhome]

This is very interesting news.
The both companies need each other, so the product range will now be very strong. But prices will skyrocket...

The hardware cost is nothing and CAPEX so deductable the real killer are licenses.
A server migh cost $40.000 but that is a one time fee that often include 3 years warrenty with 2 years of added support then you add eg. a VMware license on top og then and a Windows sever license and to top that off a SQL license and suddenly the hardare cost is not what is costing you when you run a couple of hundred servers and combine om-prem with the could.
Remember as a consumer you are a "second rate citizen", enterprise is the "nobility" in terms of profit.
But as a consumer today you have never gotten as much compute for so cheap as you do today but many people have lost the perpective.

 

[HockeyJon]

This is very interesting news.
The both companies need each other, so the product range will now be very strong. But prices will skyrocket...

I'm not sure how much Nvidia actually needs Intel. They've become the world's largest company without any involvement from them, and they design their own CPUs based on ARM anyway. Nvidia is dominating the datacenter already. Intel, however, absolutely needs Nvidia, and it also just so happens that the US government has recently taken a 10% stake in the company, and it also just so happens that Jensen Huang meets with Trump literally all the time to work out business deals, and it just so happens that Trump views the health of a domestic chip manufacturer as paramount for American national security (which I don't disagree with). I suspect this agreement was facilitated more by involvement of the US government with Nvidia than Nvidia genuinely needing anything from Intel. Intel's struggling across the board. Nvidia is not.

Prices will definitely skyrocket though, we do agree there. I also wonder what this means for Intel Arc GPUs. Nothing good, I imagine.

 

[AVATARAT]

I'm not sure how much Nvidia actually needs Intel. They've become the world's largest company without any involvement from them, and they design their own CPUs based on ARM anyway.

And where is the Nvidia's ARM CPU? Dropped by any (slow) reason?

For the other part I agree.

 

[kac77]

Not a fan of this deal at all. It's a bailout with extra steps. Nvidia doesn't need Intel at all. Allowing them to buy stake in Intel is even worse. This is the furthest thing from free market as one could get. First the government buys 9 billion, then nVidia comes in with 5 billion more. In exchange nVidia gets access to the coveted x86 license, which is a conflict of interest beyond words as AMD and Intel were supposed to be working on extending x86. That's probably going to die on the vine now.

 

[DukenukemX]

And where is the Nvidia's ARM CPU? Dropped by any (slow) reason?

For the other part I agree.

I don't think Nvidia's partnership with MediaTek is going as well as they planned. If you've ever dealt with MediaTek chips then you'll know they're the most Chinesium ARM chip that you can get. A lot of cheap Android devices would usually end up with these chips. Why Nvidia thought MediaTek was the best choice for an ARM CPU is beyond me? Unless Nvidia invests heavily into ARM like Apple and Qualcomm did then off the shelf ARM chips are not going to cut it.

 

[kac77]

I don't think Nvidia's partnership with MediaTek is going as well as they planned. If you've ever dealt with MediaTek chips then you'll know they're the most Chinesium ARM chip that you can get. A lot of cheap Android devices would usually end up with these chips. Why Nvidia thought MediaTek was the best choice for an ARM CPU is beyond me? Unless Nvidia invests heavily into ARM like Apple and Qualcomm did then off the shelf ARM chips are not going to cut it.

Because Mediatek is in no position to make demands. They can pretty much dictate the terms and they will take them.

 

[Marees]

Because Mediatek is in no position to make demands. They can pretty much dictate the terms and they will take them.

Now let's replace mediatek with Intel in statement above

 

[ChadD]

So how long before Intel announces they don't need a GPU division anymore.
Any bets on Intel going back to paying Nvidia for GPU tech again? Or will come down the line after Nvidia buys a few more billions of Intel stock.

 

[Aurelius]

To me the bigger news is NVIDIA investing $5 billion in Intel through shares.

It's another part of the story of Intel's transformation from a seemingly unstoppable, entitled monopolist to a faded star getting bailouts. I hope Intel turns things around for competition's sake, but at this point the violin I'm playing is very tiny.

 

[madpistol]

Begun… the AI Wars, has.

 

[LukeTbk]

To me the bigger news is NVIDIA investing $5 billion in Intel through shares.

I was going to say $5 billion is a lot, Nvidia will own around 4% of Intel, Softbank had $2 billion, US gov nearly 9 billion, that a lot of cash.

but I am not sure it is a bigger new than those:
For data centers, Intel will build NVIDIA-custom x86 CPUs that NVIDIA will integrate into its AI infrastructure platforms and offer to the market.
Intel will build and offer to the market x86 system-on-chips (SOCs) that integrate NVIDIA RTX GPU chiplets


Intel laptop APU with an Nvidia GPU.... that could turn the narrative around of the competition attempt to push out Nvidia of the Laptop market, with that collaboration potential maybe Nvidia (and American taxpayer previously) bought Intel at a really good price here. Seem like the market believe so, despite the dilution the stock is +28%!!!

I feel that kind of reaction is more about the markets it open for intel CPU than the 5 billions (even the 9B did not do that), imagine how easier Intel-Nvidia bid for a Sony console can become if they have just 2 generation of high performance Laptop SOC together

 

[HockeyJon]

And where is the Nvidia's ARM CPU? Dropped by any (slow) reason?

For the other part I agree.

They’re still actively developing ARM CPUs, but if this were about partnering with the best CPU company in the datacenter, you’re not partnering with Intel, you’re partnering with AMD. Intel isn’t the best at anything right now. That’s why I think this must be due to a deal going on in the background with the US government. Jensen has met with Trump A LOT, and Trump just bought 10% of Intel and wants to show a return for the US taxpayer (and a win for him). This boosts Intel’s prospects for success and growth immensely. It does not boost Nvidia’s at all. CPUs were not holding Nvidia back from dominating the datacenter space. Everything Intel has done for the past decade has held it back.

The big losers here are going to be TSMC (and Samsung to a degree), and AMD. The former is a foreign company, but the latter is an American company, and American companies being negatively impacted by the government taking a direct stake in a specific corporation was always going to be a negative consequence of that strategy.

 

[Lakados]

Nvidia not having access to an x86 license is was a big problem for them. ARM is good for many things, but there is still a lot it simply can not do nearly as well, and it forces many deployment scenarios to be then paired with either Intel or AMD CPUs. But due to platform integration shanagans, using AMD CPUs with Nvidia GPUs is more difficult when working at that scale, so that leaves many datacenters using Intel CPUs with Nvidia GPUs or going hard into the Nvidia platform, going with both their CPUs and GPUs, but that is ultimately limiting because you are completely tied to Nvidia. And while Nvidia usually treats its partners well, having all your eggs in one basket is not always a comfortable position to be in.

This could be very mutually beneficial for Nvidia and certainly Intel, as much as I would like OneAPI to take off, Intel just isn't getting the support around it that it needs, CUDA is still crushing everything in sight, and ROCm is just doing its thing.
I feel like Nvidia does see the writing on the wall for the closed source nature of CUDA, it's getting too much government heat, from all sides, be it the US or China, and partnering with Intel could long term greatly assist their transition over to an "open source" alternative to CUDA without too much pushback all while not actually creating a competitor because any Intel victory is an Nvidia victory.

This could be a pretty big issue for AMD, though...

 

[Lakados]

They’re still actively developing ARM CPUs, but if this were about partnering with the best CPU company in the datacenter, you’re not partnering with Intel, you’re partnering with AMD. Intel isn’t the best at anything right now. That’s why I think this must be due to a deal going on in the background with the US government. Jensen has met with Trump A LOT, and Trump just bought 10% of Intel and wants to show a return for the US taxpayer (and a win for him). This boosts Intel’s prospects for success and growth immensely. It does not boost Nvidia’s at all. CPUs were not holding Nvidia back from dominating the datacenter space. Everything Intel has done for the past decade has held it back.

The big losers here are going to be TSMC (and Samsung to a degree), and AMD. The former is a foreign company, but the latter is an American company, and American companies being negatively impacted by the government taking a direct stake in a specific corporation was always going to be a negative consequence of that strategy.

Yes and No...
AMD has the fastest and most efficient CPU, for sure, hands down, no contest.

Intel has a better management platform for monitoring and managing systems at scale, while still offering better support in many aspects.

So it's not as cut and dry as AMD good, Intel garbage, when working at that scale, because the environment they create is basically a living organism and they have weird behaviours that simply don't exist at a smaller scale.

 

[LukeTbk]

. It does not boost Nvidia’s at all. CPUs were not holding Nvidia back from dominating the datacenter space. Everything Intel has done for the past decade has held it back.

It protect it and kind of boost it, Playstation 7, Laptop market, there was a risk of discrete GPU free world in those space or at least an AMD one. Nvidia stock is up 3.5% that 150 billions these days... for a reason. It can also remove quickly any advantage/entry AMD via x86 had in that space (AI space).

AMD would have not accepted that deal I do not imagine (having Nvidia gpu in AMD soc instead of AMD one, maybe on the datacenter side but even then chance they much prefer making new custom CPU for AMD accelerator in that space has well, not sure they would have wanted to dilute to sell to Nvidia either.

There is some US taxpayer "quiproko" will the tax payer ban the sales of those Intel-Nvidia system like they did in the past ? now that so much of the profit goes in their pocket directly ?

 

[HockeyJon]

Yes and No...
AMD has the fastest and most efficient CPU, for sure, hands down, no contest.

Intel has a better management platform for monitoring and managing systems at scale, while still offering better support in many aspects.

So it's not as cut and dry as AMD good, Intel garbage, when working at that scale, because the environment they create is basically a living organism and they have weird behaviours that simply don't exist at a smaller scale.

That’s fair. I have heard that Intel still has a pretty enthusiastic engineering integration win as well which would be an advantage in spite of product limitations. Still, I have no doubt there was a conversation on this involving the White House. Not trying to get political, just observing what I think is the most likely scenario for how this went down.

 

[HockeyJon]

It protect it and kind of boost it, Playstation 7, Laptop market, there was a risk of discrete GPU free world in those space or at least an AMD one. Nvidia stock is up 3.5% that 150 billions these days... for a reason. It can also remove quickly any advantage/entry AMD via x86 had in that space.

AMD would have not accepted that deal I do not imagine (having Nvidia cpu in AMD soc instead of AMD one, maybe on the datacenter side but even then chance they much prefer making new custom CPU for AMD accelerator in that space has well, not sure they would have wanted to dilute to sell to Nvidia either.

There is some US taxpayer "quiproko" will the tax payer ban the sales of those Intel-Nvidia system like they did in the past ? now that so much of the profit goes in their pocket directly ?

That is a fair point vis-a-vis the laptop market, and it is something I thought of as well, so perhaps I should rephrase when I said it does not boost Nvidia at all. That said, Intel is disproportionately the beneficiary in this arrangement, which is why I feel like there was a conversation between Trump and Jensen about how the US really needs Intel not to die and there may be something in it for Nvidia if they help bail them out. This reminds me a bit like when the US government had JP Morgan buy Bear Stearns. JP Morgan didn't NEED to do it. They would still be dominating US banking today if that deal didn't happen. The problem is the consequences to the US of Bear Stearns collapsing were immense, and so the US government asked Jamie Dimon to do them a solid (and then ironically sued JP Morgan over Bear's misdeeds, but that's another conversation). Similarly, no doubt Intel failing would be a strategic problem for the US, so now you have the US government AND Nvidia as shareholders in the company.

In terms of free markets, I'd say AMD has a big problem here. If AMD's bidding on a government contract against Intel, the government has a direct stake with Intel, I'd say it's much harder for them to win potentially lucrative business contracts as a result of this arrangement regardless of who has the better product. Nvidia leads in GPUs, no question about it, but a superior AMD CPU offering now has a greater chance of being overlooked.

 

[ChadD]

Nvidia not having access to an x86 license is was a big problem for them. ARM is good for many things, but there is still a lot it simply can not do nearly as well, and it forces many deployment scenarios to be then paired with either Intel or AMD CPUs. But due to platform integration shanagans, using AMD CPUs with Nvidia GPUs is more difficult when working at that scale, so that leaves many datacenters using Intel CPUs with Nvidia GPUs or going hard into the Nvidia platform, going with both their CPUs and GPUs, but that is ultimately limiting because you are completely tied to Nvidia. And while Nvidia usually treats its partners well, having all your eggs in one basket is not always a comfortable position to be in.

This could be very mutually beneficial for Nvidia and certainly Intel, as much as I would like OneAPI to take off, Intel just isn't getting the support around it that it needs, CUDA is still crushing everything in sight, and ROCm is just doing its thing.
I feel like Nvidia does see the writing on the wall for the closed source nature of CUDA, it's getting too much government heat, from all sides, be it the US or China, and partnering with Intel could long term greatly assist their transition over to an "open source" alternative to CUDA without too much pushback all while not actually creating a competitor because any Intel victory is an Nvidia victory.

This could be a pretty big issue for AMD, though...

Intel is still Intel.
Incorporating Nvidia interconnect tech... doesn't magically make Xeons any better tomorrow then they were yesterday.
On the consumer side of things... a laptop with a Intel/Nvidia APU to end customers isn't really any different then what they are seeing now. Its just a more economical solution potentially for both companies and OEMs. (maybe... I mean NV isn't likely to be giving Intel APU GPU chiplets fo free).

This helps Intel a bit on the consumer side going up against things like Strix Halo maybe. On the other hand AMD has working parts and has 100% control over their chips. Intel and Nvidia still have to design and create something that works and is competitive. Announcing something today means nothing real for probably 3-4 years consumer side.

Server side incorporating better interconnects can be done pretty quickly... but we are still talking about a solution that will be overall more expensive then a Grace solution. I know the selling thing here is AI.
I wonder if this maybe helps them in other markets like XPU render farms. In the VFX worlds things have been moving to XPU compute rendering (CPU and GPU rather then either alone). As I understand it right now that generally means AMD + Nvidia. Maybe some version of Xeon+NV better competes for stuff like that? Maybe, I am not sure I see it... right now AMDs raw crunch is just superior, I'm not sure current gen Intel CPU tech with better interconnects is enough alone. Intel still needs to land their next CPU arch, and can't afford to have AMD hit a home run with Zen6.

 

[chameleoneel]

I think a medium-game here, is for Intel to finally figure out a decently competitive silicon process which has good yields: And then Nvidia gets a really great price, and moves away from TSMC for their chips.

 

[Lakados]

Incorporating Nvidia interconnect tech... doesn't magically make Xeons any better tomorrow then they were yesterday.

It would at least give Intel and Nvidia something to compete against the AMD MI platform which has evolved into a series of massive APU’s.

It might also work to either steer Intel off the OAM platform over to Nvidia’s SXM platform or it may allow Nvidia to slip some of their tech into the OAM platform through Intel.


Consumer side it’s meh… apparently some sites are saying “High End” Intel gaming GPU’s will get paired with Nvidia RTX iGPU’s which will certainly help both compete against the upcoming AMD lineups.

It might also get Nvidia more favourable terms on Intels fab and packaging processes.
TSMC has the better fabs right now, but their packaging situation is a little different, on paper Intel appears to have the better packaging facilities and TSMC is trying to catch up.

It could allow Nvidia to adopt some different tech not directly associated with X86.

Intel is a beast and has fingers everywhere.

Let’s also not ignore the possibility that this gets some government regulators and watchdogs off their ass.

 

[Lakados]

That’s fair. I have heard that Intel still has a pretty enthusiastic engineering integration win as well which would be an advantage in spite of product limitations. Still, I have no doubt there was a conversation on this involving the White House. Not trying to get political, just observing what I think is the most likely scenario for how this went down.

Yeah, 100% some backs got scratched there.

I also wouldn’t be surprised if it was the WH putting pressure on Intel to find a partner fast because they can only give them so many buckets before they need to let them sink.

My biggest issue is I can’t tell if this is ultimately pro or anti consumer…

 

[LukeTbk]

I feel with Nvidia future volume it is not about moving away from TSMC/Samsung necessarily, but making deal for price each time if they are possible (and maybe a mix used at the same time, for different generation), for Nvidia 4% of Intel is not that big of a deal

a laptop with a Intel/Nvidia APU to end customers isn't really any different then what they are seeing now.

if they make a single coherent memory a la Apple SOC (or amd MI300A), it is quite different from what an typical Nvidia dgpu -intel cpu laptop/device look like too right ?

Could be over enthusiast, but would have known this in advance I would have bought some intel when it was at under 110 billions..... they are a 14A node working well away to be quite the business at that price (and the US government position of considering it a cannot buy business creating a stopgap...).

 

[ChadD]

It would at least give Intel and Nvidia something to compete against the AMD MI platform which has evolved into a series of massive APU’s.

It might also work to either steer Intel off the OAM platform over to Nvidia’s SXM platform or it may allow Nvidia to slip some of their tech into the OAM platform through Intel.


Consumer side it’s meh… apparently some sites are saying “High End” Intel gaming GPU’s will get paired with Nvidia RTX iGPU’s which will certainly help both compete against the upcoming AMD lineups.

It might also get Nvidia more favourable terms on Intels fab and packaging processes.
TSMC has the better fabs right now, but their packaging situation is a little different, on paper Intel appears to have the better packaging facilities and TSMC is trying to catch up.

It could allow Nvidia to adopt some different tech not directly associated with X86.

Intel is a beast and has fingers everywhere.

Let’s also not ignore the possibility that this gets some government regulators and watchdogs off their ass.

All true.
I see this more as a cheap way for Nvidia to buy themselves a little protection... and maybe deflate one actual competitor.
I could very well see Intels new CEO deciding this is good cover to ax their GPU plans. Or maybe he holds out for one more round of stock purchasing first. Maybe right at the end of one of the next few quarters reporting periods to help bump Intel stock.

I honestly don't see how this is a win for Intel in anyway long term. Short term maybe it helps them get a few more Xeons shipped then they would have, same thing consumer side maybe in the next few years they have a few compelling Intel/NV APUs that compete well enough in the mid stack segments to keep the OEMs like Dell and HP happy for another cycle. Long term Jensen wins here, its not a real partnership we know that much. Though I could see Jensen being more then happy to let Intel handle the consumer stuff paying for the NV chiplets for their parts. The deal Intel gets server side is going to progressively get worse I'm sure. The only potential upside for Intel long term is that it may be easier to get NV to use Intel fabs more. (though I hardly expect NV to shower Intel in Margin on any deals in that regard)

This deal is a Intel life line... and Jensen knows it, he will exploit Intel as hard as he can. I'm not sure Intels current CEO or board are thinking far enough down the road to avoid the moment where he truly bends them over.

 

[ChadD]

I feel with Nvidia future volume it is not about moving away from TSMC/Samsung necessarily, but making deal for price each time if they are possible (and maybe a mix used at the same time, for different generation), for Nvidia 4% of Intel is not that big of a deal

if they make a single coherent memory a la Apple SOC (or amd MI300A), it is quite different from what an typical Nvidia dgpu -intel cpu laptop/device look like too right ?

Could be over enthusiast, but would have known this in advance I would have bought some intel when it was at under 100 billions..... they are a 14A node working well away to be quite the business at that price (and the US government position of considering it a cannot buy business creating a stopgap...)

Personally I would still say its a shaky bet at best. Long way between announcing and delivering. Also one of the partners really just wants to knee cap the other imo. Jensen isn't really looking for a partnership. I could be wrong I just think he is a lot more ruthless then that. Lakados makes a good point this gets the current gov of their back, it makes Intel beholden. Lot of upside for NV even if nothing ever ships its still a win for NV. If everything goes south at Intel fabs... Intel just admited NVs tech is superior in basically every segment. Intel needs this to work to win now.

Time will tell. You may be right, its possible we are all crazy for not having bought in when the stock was at its lowest. Hopefully 14a rocks has great yields and is very productive. If that works and NV and Intel integrate their stuff and don't have any major issues like latency, or caching issues or any of the various issues with stuff like they are talking about... ya they probably have some very solid SKUs.

 

[Lakados]

All true.
I see this more as a cheap way for Nvidia to buy themselves a little protection... and maybe deflate one actual competitor.
I could very well see Intels new CEO deciding this is good cover to ax their GPU plans. Or maybe he holds out for one more round of stock purchasing first. Maybe right at the end of one of the next few quarters reporting periods to help bump Intel stock.

I honestly don't see how this is a win for Intel in anyway long term. Short term maybe it helps them get a few more Xeons shipped then they would have, same thing consumer side maybe in the next few years they have a few compelling Intel/NV APUs that compete well enough in the mid stack segments to keep the OEMs like Dell and HP happy for another cycle. Long term Jensen wins here, its not a real partnership we know that much. Though I could see Jensen being more then happy to let Intel handle the consumer stuff paying for the NV chiplets for their parts. The deal Intel gets server side is going to progressively get worse I'm sure. The only potential upside for Intel long term is that it may be easier to get NV to use Intel fabs more. (though I hardly expect NV to shower Intel in Margin on any deals in that regard)

This deal is an Intel life line... and Jensen knows it, he will exploit Intel as hard as he can. I'm not sure Intels current CEO or board are thinking far enough down the road to avoid the moment where he truly bends them over.

Intel is in a bad spot, this stinks of desperation.

Intel is in last place across the board.

They are losing to AMD in terms of X86, behind AMD and Nvidia for GPU compute.

They are losing to AMD, Broadcom, Qualcomm, and Nvidia for Networking.

They are losing to TSMC for advanced nodes, and even Samsung and Global are doing better with the old nodes for 3’rd parties than Intel is.

Intel spends the most in R&D and gets the least to show for it.

Intel may still ship volume parts but the market they serve is shrinking as ARM, AMD, and Nvidia gain more ground.

Intel is now long past the Adapt or Die part of their journey and they are square in the “we need a lifeline” portion and Nvidia just tossed them a rope.

I bet Jensen is creaming his leathers over this, I remember an interview of his from years ago where he lamented on all the times Intel tried to kill Nvidia off. Being the person to potentially reside over their future has to give him some smug degree of satisfaction.

 

[ChadD]

I bet Jensen is creaming his leathers over this, I remember an interview of his from years ago where he lamented on all the times Intel tried to kill Nvidia off. Being the person to potentially reside over their future has to give him some smug degree of satisfaction.

He is enjoying a bottle of his finest tonight no doubt. My guess is 10/30 when the next quarterly is due, Tan will let things be. It won't be good most likely but he'll put a positive spin on as much as possible and cross his fingers. If things go really bad I wouldn't be shocked if days before the Q4 report drops, NV throws Intel yet another life line buying another few billion in stock... and Tan announces Intel has to focus on fab and cpu and leave GPU to their good partner Nvidia.

 

[Lakados]

He is enjoying a bottle of his finest tonight no doubt. My guess is 10/30 when the next quarterly is due, Tan will let things be. It won't be good most likely but he'll put a positive spin on as much as possible and cross his fingers. If things go really bad I wouldn't be shocked if days before the Q4 report drops, NV throws Intel yet another life line buying another few billion in stock... and Tan announces Intel has to focus on fab and cpu and leave GPU to their good partner Nvidia.

Or talk about how Intel and Nvidia will be further building on their GPU partnership….

Intel does have some interesting patents that could aid Nvidia.

The Intel virtualization system they have built for DX11 and older is really robust.

If Nvidia could build off that and optimize it I have no doubt they could utilize it for future GPUs. It would let Nvidia shave down on the size of their chips a small amount, there are also some interesting multi GPU applications of a virtualized GPU interface that Nvidia could build on especially if paired with their combined interlink technology.

I could see a world where a small GPU interface sat in the middle of an L4 cache lake 512MB or larger that was surrounded by GPU cores connected via high speed data interconnects.

The small interface is what the OS detects and that chip utilizing their combined interlink cache and data bus batches the jobs accordingly independent of the game or the OS.

 

[ChadD]

Or talk about how Intel and Nvidia will be further building on their GPU partnership….

Intel does have some interesting patents that could aid Nvidia.

The Intel virtualization system they have built for DX11 and older is really robust.

If Nvidia could build off that and optimize it I have no doubt they could utilize it for future GPUs. It would let Nvidia shave down on the size of their chips a small amount, there are also some interesting multi GPU applications of a virtualized GPU interface that Nvidia could build on especially if paired with their combined interlink technology.

I could see a world where a small GPU interface sat in the middle of an L4 cache lake 512MB or larger that was surrounded by GPU cores connected via high speed data interconnects.

The small interface is what the OS detects and that chip utilizing their combined interlink cache and data bus batches the jobs accordingly independent of the game or the OS.

Well if Intel has tech Nvidia wants... I imagine instead of a buyout its something along the lines of. Nvidia has agreed to give us preferred partner pricing for Intel CPU GPU needs going forward; and have agreed to a cash sale of the Intel GPU division and patents for X billion dollars cash.

 

[Lakados]

Well if Intel has tech Nvidia wants... I imagine instead of a buyout its something along the lines of. Nvidia has agreed to give us preferred partner pricing for Intel CPU GPU needs going forward; and have agreed to a cash sale of the Intel GPU division and patents for X billion dollars cash.

Yeah but I think Nvidia and Jensen would be put off by that. There are only really 3 players in this race and the top guy buying out the bottom would get more lawyers up their ass than even their AI’s could track.

Some form of strategic partnership works best here because given the shitstorm NVidia faced buying ARM imagine the pushback they would face buying Intel anything.

 

[ChadD]

Yeah but I think Nvidia and Jensen would be put off by that. There are only really 3 players in this race and the top guy buying out the bottom would get more lawyers up their ass than even their AI’s could track.

Some form of strategic partnership works best here because given the shitstorm NVidia faced buying ARM imagine the pushback they would face buying Intel anything.

Intels GPU division? Na most people would laugh, and if Jensen acts quick the current gov is easily swayed with golden gifts.

 

[littlegreenmenfromhome]

Nvida cloud also be eyeing Intel's Fabs think long term.

 

[Lakados]

Intels GPU division? Na most people would laugh, and if Jensen acts quick the current gov is easily swayed with golden gifts.

Intel owns many of the patents that AMD uses for their Chiplets technology and it’s held by the GPU division.

The Intel/AMD cross license includes a lot of tech used in the DPU and FPGA space and that is also held under the Intel GPU division.

If Nvidia were to buy those it would completely mess up the existing AMD lineup unless AMD were to then make an agreement with Nvidia or retained access through Intel. It would get messy and I can see it getting nearly instantly tanked by the EU, most likely the US, and probably Japan too.

 

[ChadD]

Intel owns many of the patents that AMD uses for their Chiplets technology and it’s held by the GPU division.

The Intel/AMd cross license includes a lot of tech used in the DPU and FPGA space and that is also held under the Intel GPU division.

If Nvidia were to buy those it would completely mess up the existing AMD lineup unless AMD were to then make an agreement with Nvidia or retained access through Intel. It would get messy and I can see it getting nearly instantly tanked by the EU, most likely the US, and probably Japan too.

AMDs contracts are solid I am sure. Regardless patents that are part of a reporting unit today can be transferred to another tomorrow. I'm sure NV will have the pick of the patents when it comes time to make Jensen a deal.

 

[Lakados]

AMDs contracts are solid I am sure. Regardless patents that are part of a reporting unit today can be transferred to another tomorrow. I'm sure NV will have the pick of the patents when it comes time to make Jensen a deal.

I’m sure it could all be worked out, but I’m not sure that Nvidia would want to go through the process again where’s there a high likelihood that it gets shot down and leaves them a few extra $B in the hole with nothing to show for it.

 

[ChadD]

I’m sure it could all be worked out, but I’m not sure that Nvidia would want to go through the process again where’s there a high likelihood that it gets shot down and leaves them a few extra $B in the hole with nothing to show for it.

They would have something to show for it though. The entire point would be to end Intels GPU division. IMO Jensen would gladly throw them another 4 or 5 billion in cash to end Intel GPU development. Intel will probably need the cash sooner rather then later, and NV has tons of cash. It would make Intel 100% reliant on Nvidia forever more.... cause AMD isn't selling Intel GPU tiles. (Well I don't think they would but I guess why not for the right price lol)

 

[DukenukemX]

So how long before Intel announces they don't need a GPU division anymore.
Any bets on Intel going back to paying Nvidia for GPU tech again? Or will come down the line after Nvidia buys a few more billions of Intel stock.

Intel can't afford not to have a GPU.

 

[Lakados]

Intel can't afford not to have a GPU.

Yeah but at the same time I don’t think Intel wants to be in the Gaming GPU market. Business, Workstation, Server, Datacenter yeah. But gaming consumer…. I’m not so sure. I just don’t see how their margins justify the development and expense time.