Intel is Gunning for NVIDIA
video description
Date: 2024-04-13
Related videos
Comments and reviews: 20
stinkycheese804
Well what is intel going to say Sorry guys, we can't compete, sell your stock and move on No, they're going to use all the marketing they can muster to tell you about a great future.
Unfortunately, if nvidia's making most of their money from AI, it makes you hope the trickle down effect results in good video cards for gamers and video enthusiasts, instead of nvidia continuing to price GPUs like they just don't care about gamers, who are people, not just numbers.
Intel will happily pick up that slack, if they choose to put the development effort into the drivers and the control panel similar to the features nvidia offers. Of course they need more momentum, not pretending cuda meant nothing but instead, enticing values to gain some more GPU accelerated interest.
reply
Well what is intel going to say Sorry guys, we can't compete, sell your stock and move on No, they're going to use all the marketing they can muster to tell you about a great future.
Unfortunately, if nvidia's making most of their money from AI, it makes you hope the trickle down effect results in good video cards for gamers and video enthusiasts, instead of nvidia continuing to price GPUs like they just don't care about gamers, who are people, not just numbers.
Intel will happily pick up that slack, if they choose to put the development effort into the drivers and the control panel similar to the features nvidia offers. Of course they need more momentum, not pretending cuda meant nothing but instead, enticing values to gain some more GPU accelerated interest.
reply
Petch85
Sometimes I forget that the market value of Nvidia is more than 10x that of Intel.
AMD has a market value of about 1.7x that of Intel.
Intel's profitability and turnover over the last 2 years looks real bad, still about 2x that of AMD.
Nvidia has a surprisingly low turnover but there profitability looks grate.
AMD looks like they are still in the fight, but Intel looks like they are going through it, especially the decline in turnover looks bad.
And Nvidia looks unreal. There turnover has been increasing fast and that looks good. But there profitability is insane, especially 2022. And there market value looks way to high compared with there turnover.
reply
Sometimes I forget that the market value of Nvidia is more than 10x that of Intel.
AMD has a market value of about 1.7x that of Intel.
Intel's profitability and turnover over the last 2 years looks real bad, still about 2x that of AMD.
Nvidia has a surprisingly low turnover but there profitability looks grate.
AMD looks like they are still in the fight, but Intel looks like they are going through it, especially the decline in turnover looks bad.
And Nvidia looks unreal. There turnover has been increasing fast and that looks good. But there profitability is insane, especially 2022. And there market value looks way to high compared with there turnover.
reply
jackchoy6969
NVIDIA and CUDA will crush all those vying for 2nd place. AMD is screaming please buy our expensive MI300 or 7900XTX and learn ROCm. Intel is screaming please buy my ARC GPU that no one wants and learn SYCL and oneAPI. Meanwhile, poor AI students are buying old, out-of-production NVIDIA gaming cards from eBay or a current one from Best Buy and are up and running with CUDA or PyTorch and can transfer that code and knowledge to H100 and future gen GPUs. NVIDIA has been quietly cooking CUDA and making inroads into universities and ML/DL community for over a decade, while Intel and AMD are flailing around, still trying to figure it out.
reply
NVIDIA and CUDA will crush all those vying for 2nd place. AMD is screaming please buy our expensive MI300 or 7900XTX and learn ROCm. Intel is screaming please buy my ARC GPU that no one wants and learn SYCL and oneAPI. Meanwhile, poor AI students are buying old, out-of-production NVIDIA gaming cards from eBay or a current one from Best Buy and are up and running with CUDA or PyTorch and can transfer that code and knowledge to H100 and future gen GPUs. NVIDIA has been quietly cooking CUDA and making inroads into universities and ML/DL community for over a decade, while Intel and AMD are flailing around, still trying to figure it out.
reply
xlr555usa
Im getting ads for Nvidias H100 8x Gpu setup, they cost around $300,000. When the A100 released I remembered a price of $27,000 for 4x A100. Nvidia will keep raising prices and sell less and less and eventually some companies will catch up to them. The hubris of Nvidia will be their downfall. Also open standards are needed in order to dominate the market, Nvidia's CUDA will only establish them with a niche market, much like what Apple does with their walled garden. Finally Intel should have its foundry capacity raging by 2030 so they will be able to produce more hardware than Nvidia.
reply
Im getting ads for Nvidias H100 8x Gpu setup, they cost around $300,000. When the A100 released I remembered a price of $27,000 for 4x A100. Nvidia will keep raising prices and sell less and less and eventually some companies will catch up to them. The hubris of Nvidia will be their downfall. Also open standards are needed in order to dominate the market, Nvidia's CUDA will only establish them with a niche market, much like what Apple does with their walled garden. Finally Intel should have its foundry capacity raging by 2030 so they will be able to produce more hardware than Nvidia.
reply
MarkBarrett
Those bots definitely fall short when it comes to understanding the nuances of human communication. Here's why:
Limited Programming: Spam bots are designed to follow simple rules and scripts. They can identify keywords and respond with pre-written messages, but they lack the ability to truly understand the context or meaning behind the words.
No Real-World Experience: Unlike humans who learn and adapt through experience, spam bots operate in a limited digital world. They haven't grasped the complexities of social interaction or the ability to interpret different situations.
reply
Those bots definitely fall short when it comes to understanding the nuances of human communication. Here's why:
Limited Programming: Spam bots are designed to follow simple rules and scripts. They can identify keywords and respond with pre-written messages, but they lack the ability to truly understand the context or meaning behind the words.
No Real-World Experience: Unlike humans who learn and adapt through experience, spam bots operate in a limited digital world. They haven't grasped the complexities of social interaction or the ability to interpret different situations.
reply
hypogogix9125
AI is something that really needs a halt in development. Humanity is going to get this so wrong its unreal. The pace with which its developed will increase the likelihood and pace with which it goes very wrong. The race between competition is the most irresponsible thing that could happen with this. Governments and companies should work as a unit on this. We will be sorry we didn't. Eventually AI will be designing the chips it runs on. That will be the first major problem because the complexity will exceed human understanding more quickly than most would like to admit.
reply
AI is something that really needs a halt in development. Humanity is going to get this so wrong its unreal. The pace with which its developed will increase the likelihood and pace with which it goes very wrong. The race between competition is the most irresponsible thing that could happen with this. Governments and companies should work as a unit on this. We will be sorry we didn't. Eventually AI will be designing the chips it runs on. That will be the first major problem because the complexity will exceed human understanding more quickly than most would like to admit.
reply
WXSTANG
1) nothing proprietary has ever stood the test of time. beta, mini disk, laser disk, glide api, and cuda is the current fad. Keep in mind AMD does have a superior compute GPU, and GPU / CPU / AI heterogenious interfaces. There is no other company that can do all three as seemlessly than AMD. To me... Jensen is just a superior marketer. Remember when DLSS was AI upscaling that couldn't be done on other GPUs Well now Intell has their 1.3 upscaler that works with even AMD, and is better than DLSS. No AI needed. Marketing, marketing, marketing.
reply
1) nothing proprietary has ever stood the test of time. beta, mini disk, laser disk, glide api, and cuda is the current fad. Keep in mind AMD does have a superior compute GPU, and GPU / CPU / AI heterogenious interfaces. There is no other company that can do all three as seemlessly than AMD. To me... Jensen is just a superior marketer. Remember when DLSS was AI upscaling that couldn't be done on other GPUs Well now Intell has their 1.3 upscaler that works with even AMD, and is better than DLSS. No AI needed. Marketing, marketing, marketing.
reply
-INFERNUS-
Intel can't even make a CPU that's efficient as AMD's Ryzen, or even have something like their 3D V-Cache. I won't even get started (again) on their joke GPU's, and now they want to compete with Nvidia with AI When AMD came out with Ryzen Intel was going downhill slowly, I was with Intel for 6yrs (3770k) before switching back to AMD back in 2018. AMD was in a very bad position before Ryzen and I feel Intel is going in that direction maybe not as bad ,but it's not good.
reply
Intel can't even make a CPU that's efficient as AMD's Ryzen, or even have something like their 3D V-Cache. I won't even get started (again) on their joke GPU's, and now they want to compete with Nvidia with AI When AMD came out with Ryzen Intel was going downhill slowly, I was with Intel for 6yrs (3770k) before switching back to AMD back in 2018. AMD was in a very bad position before Ryzen and I feel Intel is going in that direction maybe not as bad ,but it's not good.
reply
matthewbond375
I'm not usually one to rise to the defense of a behemoth like NVIDIA, but they did spend years in the trenches helping to develop LLMs and other learning technologies, while other companies focused on other things.
Haven't they earned a year or two of absolutely destroying everybody who slept on LLMs, and is now scrambling to catch up
AMD already has ROCm and an open CUDA alternative called ZLUDA, it's just that the performance is really poor. For now.
reply
I'm not usually one to rise to the defense of a behemoth like NVIDIA, but they did spend years in the trenches helping to develop LLMs and other learning technologies, while other companies focused on other things.
Haven't they earned a year or two of absolutely destroying everybody who slept on LLMs, and is now scrambling to catch up
AMD already has ROCm and an open CUDA alternative called ZLUDA, it's just that the performance is really poor. For now.
reply
gamersnexus
Basically everyone is talking about who has the biggest dick but no one talks about energy consumption. Yeah your hardware can do a lot with a huge electric bill and thermals. While all of them are into this, no one spotted what matters, doing more with less. Devices the size of a raspberry pi or even the size of a current low profile GPUs that use power in the range of 45 to 150 watts. That's where the big cake is, escalation with RISC computing.
reply
Basically everyone is talking about who has the biggest dick but no one talks about energy consumption. Yeah your hardware can do a lot with a huge electric bill and thermals. While all of them are into this, no one spotted what matters, doing more with less. Devices the size of a raspberry pi or even the size of a current low profile GPUs that use power in the range of 45 to 150 watts. That's where the big cake is, escalation with RISC computing.
reply
rottencorp666
The advances in technology are impressive and all, having useful applications, but what we really need (and for which there is NO current market) is Computer Assisted Integrity, Morals and Ethics to provide answers to CEOs, Presidents and Directors of both government and business. As much as it is needed, however, there is very little in the way of moving forward with such an initiative. CAIME is a long way off.
reply
The advances in technology are impressive and all, having useful applications, but what we really need (and for which there is NO current market) is Computer Assisted Integrity, Morals and Ethics to provide answers to CEOs, Presidents and Directors of both government and business. As much as it is needed, however, there is very little in the way of moving forward with such an initiative. CAIME is a long way off.
reply
davidstocker2278
I work/worked on the LLM in the RA3 FAB that was shown off. Although it may have been staged in the sense that we knew what we were going to ask it before the camera turned on. It does work in FAB to get content real time without the MET searching the four or five different data sources that a MET would normally need to review for an RFC. It's freakin Epic. I wish I had it when I was an L3E in the factory.
reply
I work/worked on the LLM in the RA3 FAB that was shown off. Although it may have been staged in the sense that we knew what we were going to ask it before the camera turned on. It does work in FAB to get content real time without the MET searching the four or five different data sources that a MET would normally need to review for an RFC. It's freakin Epic. I wish I had it when I was an L3E in the factory.
reply
klinky
The problem with open is will Intel/AMD make the capital investments required to create tooling developers will want to use, as well as be able to steer design by committee tech into products that deliver on performance and function. Nvidia got a head start by investing in the GPGPU ecosystem early on, and reaped the benefits of delivering solutions that, while not perfect, actually work and provide benefit.
reply
The problem with open is will Intel/AMD make the capital investments required to create tooling developers will want to use, as well as be able to steer design by committee tech into products that deliver on performance and function. Nvidia got a head start by investing in the GPGPU ecosystem early on, and reaped the benefits of delivering solutions that, while not perfect, actually work and provide benefit.
reply
mr_daihatsu
LLM is a joke. It literally waste space electricity money all for a cheap parlor trick. LLMs have done nothing. Nothing. And people’s answer is to add more hardware. That’s not a solution at all. Not to mention we don’t have enough data to analyze to ever have a real Ai that learns. No today’s Ai is a parlor trick that’s wastes everyone’s time money space and electricity.
reply
LLM is a joke. It literally waste space electricity money all for a cheap parlor trick. LLMs have done nothing. Nothing. And people’s answer is to add more hardware. That’s not a solution at all. Not to mention we don’t have enough data to analyze to ever have a real Ai that learns. No today’s Ai is a parlor trick that’s wastes everyone’s time money space and electricity.
reply
ghimmy47
As you clearly struggle with the hocus-pocus AI hype (which I think most realize is not very useful at this time), perhaps you feel a little of what I (a guy who knew what he was doing ten years ago) feels when I hear you spew mumbled acronyms I never heard of. Sure would be nice to just see a recent decent pc get put together with some discrete what it is and more importantly; why.
reply
As you clearly struggle with the hocus-pocus AI hype (which I think most realize is not very useful at this time), perhaps you feel a little of what I (a guy who knew what he was doing ten years ago) feels when I hear you spew mumbled acronyms I never heard of. Sure would be nice to just see a recent decent pc get put together with some discrete what it is and more importantly; why.
reply
gamers-generation
At least part of the issue is that a lot of development already leverages pytorch/CUDA so anyone not using Nvidia card has to jump through hoops to do the same thingsI think of it like the old days of tean green where you'd have to pay big green tax for gsync and game ready drivers or go budget and hope the games you want will actually run on team red.
reply
At least part of the issue is that a lot of development already leverages pytorch/CUDA so anyone not using Nvidia card has to jump through hoops to do the same thingsI think of it like the old days of tean green where you'd have to pay big green tax for gsync and game ready drivers or go budget and hope the games you want will actually run on team red.
reply
stop_tryharding
I'm excited for Intel to get involved in the AI space, and I really hope they do us enthusiasts and hobbyists a solid and sell a consumer grade card with 24 gb of VRAM for inference. It would be nice to have an alternative to either spending $1700 for a 4090, buying a used 3090 or trying to deal with whatever the hell AMD is doing with ROCm.
reply
I'm excited for Intel to get involved in the AI space, and I really hope they do us enthusiasts and hobbyists a solid and sell a consumer grade card with 24 gb of VRAM for inference. It would be nice to have an alternative to either spending $1700 for a 4090, buying a used 3090 or trying to deal with whatever the hell AMD is doing with ROCm.
reply
cfbmoo1
Has a sudden urge to find an adult because of the creepy guy with glasses on stage Mom! He's rubbing the silicon!
I can see AI being nice to make dynamic Zork text adventures with maybe AI image generation to help move the story.
It's dark, your wallet is likely to be eaten by a Jensen price point.
reply
Has a sudden urge to find an adult because of the creepy guy with glasses on stage Mom! He's rubbing the silicon!
I can see AI being nice to make dynamic Zork text adventures with maybe AI image generation to help move the story.
It's dark, your wallet is likely to be eaten by a Jensen price point.
reply
playeronthebeat
Honestly, I'd be happy if you'd focus a bit more on so-called AI due to the fact that I am very, very certain that games will include ML models soon, too.
At least for like NPCs, etc, I guess. It might still take a bit of time, but it'll get there rather sooner than later.
reply
Honestly, I'd be happy if you'd focus a bit more on so-called AI due to the fact that I am very, very certain that games will include ML models soon, too.
At least for like NPCs, etc, I guess. It might still take a bit of time, but it'll get there rather sooner than later.
reply
matsiv5707
Intel is not traditionally a silicon manifacturer...yeah, right. There are 3 companies that can be called silicon manifacturers with any seriousness, TSMC, Samsung and Intel. There aren't any other foundries that can produce at that level. Intel is a foundry first and foremost
reply
Intel is not traditionally a silicon manifacturer...yeah, right. There are 3 companies that can be called silicon manifacturers with any seriousness, TSMC, Samsung and Intel. There aren't any other foundries that can produce at that level. Intel is a foundry first and foremost
reply
Add a review, comment
Other channel videos
