Yeah, let’s ignore that Meteor Lake barely keeps up with Phoenix in AI and is already being beatend by 60% by Hawk Point with Strix in the horizon, they want to go after Nvidia…
The only luck they had is the actual boom happening before the MI300 launch so they were the only offering for most consumers.
SpringFuzzy
Does he want his CEO parachute?
Stupidest statement from a CEO I’ve heard all year. Luck has nothing to do with Nvidia’s performance, incompetence has something to do with Intel’s performance.
EducationalCamel1043
Intel is just looking more and more bad these days. These boomers have no idea how to appeal to a younger audience.
whoever is in charge of intel marketing is driving the company into the ground.
I mean calling out AMD for the same shit Intel does is crazy.
Material_Policy6327
Intel slept on another tech innovation once again.
Hattix
AI – at least as LLMs – is a bad fit for CPUs.
AI doesn’t give a rat’s ass about RAM latency or caches, it wants bandwidth, bandwidth, and more bandwidth. The actual calculations it does are pretty easy for anything to do, you can do it in SIMD, and both CPUs and GPUs have hefty SIMD execution resources, easily enough to become bandwidth limited.
The biggest bandwidth machine I can think of at the moment is something like the EPYC 9004 family, which uses twelve-channel DDR5-4800 for 460 GB/s of bandwidth.
Or, you know, the same bandwidth a GTX 1080 Ti was sporting six years ago or the R9 Fury had reached seven years ago.
CPUs are latency optimised. A decently configured CPU based system can give you latency below 1 ns on memory access, thanks to its massive caches, and still around 10 ns if you have to go out to RAM and start stepping columns. A GPU might as well not have a cache (though RDNA3 is arguinmg with that), but it runs its RAM at 20-50 ns.
Nvidia isn’t “lucky”. Nvidia spent years working on machine vision (which directly resulted in Volta’s tensor accelerators), on denoising algorithms for that machine vision (which directly resulted in RTX) and on training convolutional neural networks to do that denoising (which directly resulted in DLSS). Nvidia invested and developed. Nvidia had a demonstration CUDA-RTExtension (RTX) at on Kepler hardware, which could path trace basic 640×360 rendering at a low FPS, but not at a seconds-per-frame.
Intel did fuck all while Nvidia did all this. Even AMD’s powerful matrix units aren’t as well supported by software libraries as Nvidia’s technically lesser tensors are.
I-LOVE-TURTLES666
Meanwhile getting most its money from the US military
lordfappington69
Intel was extremly “unlucky” sitting on its hands for 13 of the last 20 years, printing check selling the same CPUs until Ryzen comes along.
Johnny_C13
Man… Pat Gelsinger has some pretty bad takes.
Highlander198116
And Dougray Scott (you are probably saying “Who?”) would have been one of the most famous actors in the world, instead of Hugh Jackman if he didn’t dip out of playing Wolverine in X-Men.
It’s such a stupid thing to even posit. “We would have been the leader in AI if we had the foresight to know it was gonna be huge with GPU’s. Nvidia was just lucky they happened to focus on making GPU’s despite not knowing the same thing we didn’t know at the time.”
and Hugh Jackman was lucky he took the role Dougray Scott thought would do less for his career than playing the villain in Mission Impossible 2. Hugh Jackman didn’t know that movie would do what it did for his career. For all he knew it would tank.
I’m sure if Dougray Scott could go back in time he would have told Tom Cruise to go screw himself and took the Wolverine role.
speedx10
Looks at movidius. Looks at realsense. What a shame intel. get bent.
ShittyLivingRoom
If it wasn’t AMD we’d still be using cpus with 4 cores!
FIWDIM
Well, the previous Intel’s CEO was under the impression that he has a true monopoly, so as a true non-tech bean counter he killed all research and just kept ramping up prices.
If it wasn’t for AMD’s Ryzen CPU’s that came out of nowhere, we would be now paying at least $2.5K for quad-core processors “because of covid” or something similar.
morgichor
Heard him speak on a graduation once. Came off like an absolute twat. His head was wayyyyyy up his own ass.
JamesMCC17
They tried 14nm AI chips for a while, didn’t go well.
Scholar_Erasmus
Intel has been on a fire streak of putting their foot in their mouth and/or embarrassing themselves publicly lately, it’s impressive
MrOphicer
Lucky? I hate nvidia as much as most, but they need to be given credit where its due… Both CUDA and Tensor cores were amazing when the were launched. I work with 3d rendering and being able to render on GPU is almost a night and day difference, speed wise. Not to mention all the other productivity programs that benefited imensly from it. I guess Tensor cores were as revolutionary for ML engineers.
Nvidia invested heavily in R&D, they inovated, the implemented their tools in many pipelines, they did the work. It wasn’t an overnight success. Intel just sounds butthurt will all the finger pointing, and snake oil accusations.
inqisitor_euro
Nothing lucky about it, He’s massively downplaying the years and years of very hard work by the engineers at Nvidia and the dizzyingly massive amounts of money that Nvidia constantly invest into R&D.
Luck plays no role in this, It’s effort + time = success.
imaginary_num6er
Coming from a guy who declared “AMD in the rearview mirror” I don’t think it was luck, Pat.
Yodplods
Intel are a bit late the GPU game, if it could have been an intel dominated market… why didn’t they just do it?
EZ
NighthawK1911
That’s what resting on your laurels get you.
Should’ve invested your advantage properly when you were ahead of everybody.
DogeMLG420-Doge
coping
gnocchicotti
Sounds like a skill issue git gud
UraniumDisulfide
Even ignoring everything else that nvidia has deliberately done that Intel hasn’t…
Power efficiency lmao
Revo_Int92
It’s ironic how Intel acted as the shittiest company in this industry for many years (talking about the “morality” side of things), but at the same time, nowadays they provide the best cost x benefit GPUs in the market. It shows how capitalism can be so volatile and bizarre
ecktt
Even Intel has sour grapes now?
tiramisucks
Nvidia recognizes the point. They will start transferring all the AI technology to Intel asap, with apologies. /s
difused_shade
Lucky huh, when intel achieved the same level of dominance nvidia has on GPUs we had 4 core CPUs for a decade.
Love it or hate it, NVIDIA has been pumping out new technologies even without a real competitor for the last 15 years
ShaidarHaran2
Just to be clear, Pat said *Jen-Hsun directly said to him* he couldn’t believe how lucky they got with AI. The article title has tried to propagandize this as arrogance for some reason. They do work together a lot, especially with Foundry Services ramping and test products going back and forth you know.
monkeybanana550
Dafac is intel CEO smoking?
tonynca
I bet my 401k in Intel to be the masters of silicon. I guess I didn’t get lucky.
Astigi
Incompetence blames luck everytime.
Apple is lucky too, *should* be running on Intel
32 Comments
Aww shucks
Yeah, let’s ignore that Meteor Lake barely keeps up with Phoenix in AI and is already being beatend by 60% by Hawk Point with Strix in the horizon, they want to go after Nvidia…
The only luck they had is the actual boom happening before the MI300 launch so they were the only offering for most consumers.
Does he want his CEO parachute?
Stupidest statement from a CEO I’ve heard all year. Luck has nothing to do with Nvidia’s performance, incompetence has something to do with Intel’s performance.
Intel is just looking more and more bad these days. These boomers have no idea how to appeal to a younger audience.
whoever is in charge of intel marketing is driving the company into the ground.
I mean calling out AMD for the same shit Intel does is crazy.
Intel slept on another tech innovation once again.
AI – at least as LLMs – is a bad fit for CPUs.
AI doesn’t give a rat’s ass about RAM latency or caches, it wants bandwidth, bandwidth, and more bandwidth. The actual calculations it does are pretty easy for anything to do, you can do it in SIMD, and both CPUs and GPUs have hefty SIMD execution resources, easily enough to become bandwidth limited.
The biggest bandwidth machine I can think of at the moment is something like the EPYC 9004 family, which uses twelve-channel DDR5-4800 for 460 GB/s of bandwidth.
Or, you know, the same bandwidth a GTX 1080 Ti was sporting six years ago or the R9 Fury had reached seven years ago.
CPUs are latency optimised. A decently configured CPU based system can give you latency below 1 ns on memory access, thanks to its massive caches, and still around 10 ns if you have to go out to RAM and start stepping columns. A GPU might as well not have a cache (though RDNA3 is arguinmg with that), but it runs its RAM at 20-50 ns.
Nvidia isn’t “lucky”. Nvidia spent years working on machine vision (which directly resulted in Volta’s tensor accelerators), on denoising algorithms for that machine vision (which directly resulted in RTX) and on training convolutional neural networks to do that denoising (which directly resulted in DLSS). Nvidia invested and developed. Nvidia had a demonstration CUDA-RTExtension (RTX) at on Kepler hardware, which could path trace basic 640×360 rendering at a low FPS, but not at a seconds-per-frame.
Intel did fuck all while Nvidia did all this. Even AMD’s powerful matrix units aren’t as well supported by software libraries as Nvidia’s technically lesser tensors are.
Meanwhile getting most its money from the US military
Intel was extremly “unlucky” sitting on its hands for 13 of the last 20 years, printing check selling the same CPUs until Ryzen comes along.
Man… Pat Gelsinger has some pretty bad takes.
And Dougray Scott (you are probably saying “Who?”) would have been one of the most famous actors in the world, instead of Hugh Jackman if he didn’t dip out of playing Wolverine in X-Men.
It’s such a stupid thing to even posit. “We would have been the leader in AI if we had the foresight to know it was gonna be huge with GPU’s. Nvidia was just lucky they happened to focus on making GPU’s despite not knowing the same thing we didn’t know at the time.”
and Hugh Jackman was lucky he took the role Dougray Scott thought would do less for his career than playing the villain in Mission Impossible 2. Hugh Jackman didn’t know that movie would do what it did for his career. For all he knew it would tank.
I’m sure if Dougray Scott could go back in time he would have told Tom Cruise to go screw himself and took the Wolverine role.
Looks at movidius. Looks at realsense. What a shame intel. get bent.
If it wasn’t AMD we’d still be using cpus with 4 cores!
Well, the previous Intel’s CEO was under the impression that he has a true monopoly, so as a true non-tech bean counter he killed all research and just kept ramping up prices.
If it wasn’t for AMD’s Ryzen CPU’s that came out of nowhere, we would be now paying at least $2.5K for quad-core processors “because of covid” or something similar.
Heard him speak on a graduation once. Came off like an absolute twat. His head was wayyyyyy up his own ass.
They tried 14nm AI chips for a while, didn’t go well.
Intel has been on a fire streak of putting their foot in their mouth and/or embarrassing themselves publicly lately, it’s impressive
Lucky? I hate nvidia as much as most, but they need to be given credit where its due… Both CUDA and Tensor cores were amazing when the were launched. I work with 3d rendering and being able to render on GPU is almost a night and day difference, speed wise. Not to mention all the other productivity programs that benefited imensly from it. I guess Tensor cores were as revolutionary for ML engineers.
Nvidia invested heavily in R&D, they inovated, the implemented their tools in many pipelines, they did the work. It wasn’t an overnight success. Intel just sounds butthurt will all the finger pointing, and snake oil accusations.
Nothing lucky about it, He’s massively downplaying the years and years of very hard work by the engineers at Nvidia and the dizzyingly massive amounts of money that Nvidia constantly invest into R&D.
Luck plays no role in this, It’s effort + time = success.
Coming from a guy who declared “AMD in the rearview mirror” I don’t think it was luck, Pat.
Intel are a bit late the GPU game, if it could have been an intel dominated market… why didn’t they just do it?
EZ
That’s what resting on your laurels get you.
Should’ve invested your advantage properly when you were ahead of everybody.
coping
Sounds like a skill issue git gud
Even ignoring everything else that nvidia has deliberately done that Intel hasn’t…
Power efficiency lmao
It’s ironic how Intel acted as the shittiest company in this industry for many years (talking about the “morality” side of things), but at the same time, nowadays they provide the best cost x benefit GPUs in the market. It shows how capitalism can be so volatile and bizarre
Even Intel has sour grapes now?
Nvidia recognizes the point. They will start transferring all the AI technology to Intel asap, with apologies. /s
Lucky huh, when intel achieved the same level of dominance nvidia has on GPUs we had 4 core CPUs for a decade.
Love it or hate it, NVIDIA has been pumping out new technologies even without a real competitor for the last 15 years
Just to be clear, Pat said *Jen-Hsun directly said to him* he couldn’t believe how lucky they got with AI. The article title has tried to propagandize this as arrogance for some reason. They do work together a lot, especially with Foundry Services ramping and test products going back and forth you know.
Dafac is intel CEO smoking?
I bet my 401k in Intel to be the masters of silicon. I guess I didn’t get lucky.
Incompetence blames luck everytime.
Apple is lucky too, *should* be running on Intel