Pat Gelsinger: "Intel Has to be Better at Making CPUs Than That Lifestyle Company"

[First Strike]

Does one pay for such ninja-advertisment-for-overhyped-5nm-25W-ARM-chip-that-loses-to-7nm-yet-smaller-15W-4800u, does it work via Apple fans being so much into hyping it into ad nauseum?

"Reportedly" Jesus freaking Christ...

1. M1 measures at 119 mm2, 4800U measures at 156 mm2. 156 is not smaller than 119.
2. 4800U does not beat M1 in most single-threaded load. Unless you use x86 emulation.
3. M1's GPU trashes Renoir's. It's just an overkill.
4. Renoir also got 45W H-variant. But it doesn't help.

Conclusion: which parallel universe do you live in?

 

[micropage7]

"Intel Has to be Better at Making CPUs Than That Lifestyle Company"​

wait, wait... he should take a look at Apple, Razer first

 

[Am*]

Well said. I'm beginning to like this guy already -- unlike their last useless CEO spouting nonsense about growing margins and "leadership products", at least he's acknowledging the problems Intel has. That's already a start, at least. If I had some substantial pocket change laying around, I would definitely be buying some shares in Intel now. They seem to be in the same slump now that AMD were in back around 2013 -- so they have nowhere to go but up with the right leadership. The real questions are -- how long will it take for them to answer AMD, will their new arch be efficient enough to scale down to sub-5W without butchering performance to answer Apple (and will they compete with non-x86 designs) -- and do their future plans include keeping their fabs in the process? I sure hope so, but we'll have to wait and see -- I'm hoping they are back to being competitive by the end of next year (this year is already set in stone and is a lost cause for them for the most part).

 

[medi01]

1. M1 measures at 119 mm2, 4800U measures at 156 mm2. 156 is not smaller than 119.

M1 is on 5nm TSMC< that is 1.84 times denser than 7nm.
So it's "equivalent" 7nm size is 219mm2.

2. 4800U does not beat M1 in most single-threaded load. Unless you use x86 emulation.

It doesn't beat M1 in a handful of benchmark (as there is barely any code you can run natively on that fashion thing.
M1 simply is dysmorphic: it has 4 bigger and 4 smaller cores (while using 1.6 billion transistors, vs 9.8 on 4800u).
Oh, and it also consumes 25W....

3. M1's GPU trashes Renoir's. It's just an overkill.

Only in hidden ad reviews and bazinga benchmarks like geekbench.
Actual games show we are back to early 2000s, when Apple was claiming IBM"s chip was "fastest CPU in the world".

4. Renoir also got 45W H-variant. But it doesn't help.

It doesn't have to help. 4800u, despite being 1.6 times smaller and on inferior node and 15W vs 25W, beats M1 at non synthetics.

Is there evidence that Apple is blocking benchmarks?

Dude.
Have you seen a bad guy in a movie or series, with an iPhone?
Is there evidence that Apple is blocking anything to achieve that?
Could it be... just by incident?

All Apple needs to do is prove to its customer base that the M1 Macs are viable upgrades.

"Has to be better" has many ways of "better" when applied to fashion devices.

 

[DeathtoGnomes]

Mobile phones, tablets, computers and media players isn't hardware?

No I dont, they are convenience tools, computers take hardware, the rest do not. You cant upgrade the graphics card on a phone or a media player or tablets since their hardware is not replaceable.

What would you consider as hardware?

I have a hard time believing people consider a phone as hardware.

 

[Frick]

No I dont, they are convenience tools, computers take hardware, the rest do not. You cant upgrade the graphics card on a phone or a media player or tablets since their hardware is not replaceable.


I have a hard time believing people consider a phone as hardware.

So it's only hardware if it has end user replacable parts and is built on a system of slots and sockets? So a modern laptop with everything soldered is not hardware? What about the Fairphone, which is modular? I think I understand what you're getting at and that you're coming from the perspective of a traditional PC user but that is an incredibly arbritrary approach.

But as I said, in the end the distinction is meaningless.

 

[Vayra86]

Apple revenue streams/segments by quarter 2024 | Statista

Apple generated 69.7 billion U.S.

www.statista.com

However, the initial assertion is not about whether Apple is a hardware or a services company. The original assertion is that the CPU is not a central product in Apple's lineup, while it's supposed to be Intel's bread and butter.

This.

But hardware or software company really isnt a question either. Its clearly hardware. Does iOS or MacOS want to run on anything non Apple?! Only if you hack it. So Apple's software solutions are tailored to sell their hardware. Alwsys have been too.

The only reason their iphone sold and got big was becsuse they werre the only hardware phone with good touch, AND backed by a unique app store model - an ecosystem built for the hardware. The software development is aimed at efficiency and highly controlled 'customer journeys' - it enables their extreme margin on selling underpowered hardware. They always got away with it, because the user experience was well guarded. The M1 is the epitome of profit margin maxed out - they can now fully control that experience and offer the tightest possible hardware setup to suit the way they organize workloads. A big part of that is user experience: as long as the user has the impression of speed, mission accomplished. GUI smoothness is a big one for that.

Going back further, older Macs also were hardware made unique by software. Its the integration with the hardware that makes up Apple's unique selling points. Not the software alone - even if you emulate it on something else it wont usually work quite as well and hassle free.

I have a hard time believing people consider a phone as hardware.

A phone has more versatile hardware than a desktop PC... and ticks all the boxes: cpu, gpu, storage. It has better connextivity and potentially can also access high performance computing though the cloud.

Ironically, a phone is however much closer to being a CLIENT. Or a terminal for the mainframe. Whatever you call it: it depends on external providers of data to be worth anything over a desktop PC. The desktop PC can be both client and server. It can process locally without burning itself to a crisp or run out of power.

In the end we are talking about dependancy. A PC can do without. A phone exists because of it.

 

[watzupken]

I think what Pat said makes sense. But to turn that into action with results is another. I feel Intel has been asleep for too long, and given the corporate culture from what we read, it will take a lot of effort to do a U turn for this huge truck. So the question is whether it is too late for Intel to get back to the leading position where they used to be? Competition is not just coming from Apple, AMD, but also from all other big corporates that are building their own ARM based processor. With Qualcomm buying out Nuvia, Nvidia buying out ARM, Samsung partnering with AMD on their own SOC, etc, Intel is neck deep in trouble.

 

[bug]

No I dont, they are convenience tools, computers take hardware, the rest do not. You cant upgrade the graphics card on a phone or a media player or tablets since their hardware is not replaceable.

Using your own definition, if a phone, media player or tables takes hardware, what does that make Apple?

 

[mtcn77]

Using your own definition, if a phone, media player or tables takes hardware, what does that make Apple?

A premium brand, instead of a commodity manufacturer(something Intel has been accusing AMD of in the past, look how the tables turn!).

 

[mechtech]

Exactly. Intel indeed "has to be better", but can they beat a fresh, unencumbered design while dragging compatibility features dating back to 8086 (that's about 45 years ago)? Is that all that silicon that goes to AVX and heats up like crazy under load worth it?
And if you want a "clean slate", I figure it will be nearly impossible to come up with another in-house instruction set when AArch64 and RISC-V are already around.
I'm glad I'm not tasked with solving this problem, but if this guy can, hats off.

Instruction set is only half, it's the OS that decides things. If it's not familiar windows that 99% of the earth is familiar with than your cpus will only be for specific functions like RISC, etc.

As for Pat, definitely a good person to have at the helm, but he has 40 years in, reminds me of a saying, give someone else a shot at the reins.

 

[Darmok N Jalad]

Dude.
Have you seen a bad guy in a movie or series, with an iPhone?
Is there evidence that Apple is blocking anything to achieve that?
Could it be... just by incident?

So bad guys in movies not using iPhones = Apple blocking benchmarks? These are two entirely different things. One is regarding product placements in motion pictures, the other is software on a storefront. Typically, the bad guys in movies use unidentifiable hardware, because no corporation will sign up for that.

 

[TheoneandonlyMrK]

So bad guys in movies not using iPhones = Apple blocking benchmarks? These are two entirely different things. One is regarding product placements in motion pictures, the other is software on a storefront. Typically, the bad guys in movies use unidentifiable hardware, because no corporation will sign up for that.

Nah proprietary ie Ecorp .

 

[Vayra86]

I think what Pat said makes sense. But to turn that into action with results is another. I feel Intel has been asleep for too long, and given the corporate culture from what we read, it will take a lot of effort to do a U turn for this huge truck. So the question is whether it is too late for Intel to get back to the leading position where they used to be? Competition is not just coming from Apple, AMD, but also from all other big corporates that are building their own ARM based processor. With Qualcomm buying out Nuvia, Nvidia buying out ARM, Samsung partnering with AMD on their own SOC, etc, Intel is neck deep in trouble.

Pat was at the same Intel that he now supposedly leads. It feels a bit like trying more of the same hoping for a different result. Also, he isnt presenting solutions here, at all, he just FINALLY managed to say "Intel has a problem".

Pats problem is, silicon limitations are not the kind of problems where you can say "admitting you have a problem is half the solution"... Rather its "admittingyou have a problem means evaluating every dot and comma of your technology to hopefully find a way to make it better, or conclude you need a full new engineering process".

Intel has been screwing around now sticking to their old agenda even up until today and is still investing in 10nm which still doesnt perform as desired.

 

[First Strike]

M1 is on 5nm TSMC< that is 1.84 times denser than 7nm.
So it's "equivalent" 7nm size is 219mm2.

You must be completely new to semi process to do such an "equivalent" transformation. As always, it's a bad habit to think you've mastered something by scrapping through online forums.

I/O barely shrinks with process, and I/O alone is taking up a good part of modern SoC. Your 1.84x density only occurs for an ideal infinite gate array, a scenario only possible in CPU cache.

It doesn't beat M1 in a handful of benchmark (as there is barely any code you can run natively on that fashion thing.
M1 simply is dysmorphic: it has 4 bigger and 4 smaller cores (while using 1.6 billion transistors, vs 9.8 on 4800u).
Oh, and it also consumes 25W....

SPECINT2006/2013 and SPECFP2006/2013 is the industry standard and everything. They are written in C and compiles on everything that has a compiler support. They don't freaking care about how you think them.

Only in hidden ad reviews and bazinga benchmarks like geekbench.
Actual games show we are back to early 2000s, when Apple was claiming IBM"s chip was "fastest CPU in the world".

It's your fault to know only hidden ad reviews and geekbench, while there are credible and professional reviewers doing a lot more than geekbench. Go see Anandtech and Phoronix

Oh wait, are there even people using geekbench for GPU?

 

[InVasMani]

I don't pay any attention to benchmarks at all.
What impressed me is running emulated Windows X86 Witcher on that thing.
Just to put that in perspective we are not emulating OSX x86 to OSX ARM but straight from Windows x86 and it runs, not perfect but runs.
Witcher 3 is demanding game running natively let alone emulated.
We talking about RISC chips so of course we cutting corners here but the bottom line is I don't care if it's smoke and mirrors as long it performs.
Their MAC OSX x86 apps are usable on M1 unlike Windows for Arm where you can't even use Chrome X86 emulated on ARM running super crappy.
There is a lot of fat on X86 and AMD is already making efforts to trim it by cutting bunch of legacy instructions replacing with FPGA section to reduce transistor count and lower power consumption.

I've been saying it for awhile move the less utilized instruction sets to another chiplet or core makes a whole lot of sense. Save that transistor space for other stuff not every core needs it just placing a bunch of legacy stuff one the last two cores would be sufficient. They don't all need to have the same cache size and structure for that matter either which is another area where things could be adjusted. Core designs don't need to be perfectly symmetrical in their entirety. Some of it can be parallel and symmetrical, but across all cores not so much and it's exactly optimal to have all cores trying to be a jack of all trades that's the opposite of efficient and minimal. Flexible yes efficient and powerful not so much.

 

[Caring1]

"Intel Has to be Better at Making CPUs Than That Lifestyle Company"​

wait, wait... he should take a look at Apple, Razer first

Pretty sure the lifestyle company he referred to is Apple.

 

[jrocket]

M1 would be a good product if they actually sold the SOC for other companies to build products on instead of tying it in with their own closed ecosystem hardware/software. They won't do that because they are anti-competitive and anti right to repair.

Apple now requires that applications be code signed to run on native m1. How much longer till Apple requires apps from their mac store like they do with iOS?

I can acknowledge that the m1 is fast while also not going anywhere near an Apple product for these reasons- no matter how fast it is.

 

[zlobby]

Someone's going to eat their words.

 

[medi01]

So bad guys in movies not using iPhones = Apple blocking benchmarks? These are two entirely different things. One is regarding product placements in motion pictures, the other is software on a storefront. Typically, the bad guys in movies use unidentifiable hardware, because no corporation will sign up for that.

I was merely asking whether there is an official source that confirms that Apple is blocking bad guys from having iPhones...
As there ain't any (or how could a company accomplish that, right? Do you think is it by incident?

Since I see similar situation in 99% of reviews involving Apple, good half of them I find vomit inducing.
Is that orgasmic tone towards the said company just an incident too?

You must be completely new to semi process to do such an "equivalent" transformation. As always, it's a bad habit to think you've mastered something by scrapping through online forums.

You are so kind to bother answering the likes of me, oh the knowledgeable one!

SPECINT2006/2013 and SPECFP2006/2013 is the industry standard and everything.

Yeah, it's a bit dated standard, ain't it? I mean, there is a version of the spec that has "2017" in, which I'd expect to be good 11 years younger, but surely an expert like that would know better.

see Anandtech

Oh, you mean the anandtech I've linked or some... other anndtech?

 

[evernessince]

In the end the distinction is meaningless to be honest. They make hardware. Are hardware companies - whatever that means - morally superior than software/services companies?

Not sure about any moral superiority BS but the point is the company's focus. Apple's focus isn't hardware, it's using that hardware for an end goal. That's fundamentally different from AMD or Nvidia, which only make hardware to be used by others for whatever end goals they may have.

AMD and Nvidia's sole goal is making the best hardware. Apple's hardware is a means to an end, a minor detail in the broad picture. It's not the main focus. This is the number one problem for bigger companies, in that they expand to the point where lack of focus causes them to output subpar products. Intel's a good example of that.

What Intel is saying here is that they have forgotten their roots. If Intel can't even beat Apple when Intel is supposed to be THE CPU manufacturer, they can't rightfully call themselves much of a CPU focused company anymore.

 

[pavle]

Being stuck at 14nm for years has got to be some ironic punishment for Intel indeed.

 

[mtcn77]

Being stuck at 14nm for years has got to be some ironic punishment for Intel indeed.

They took the ugly approach.
The time code isn't working; listen from 1:32 to 2:00.

 

[DeathtoGnomes]

Using your own definition, if a phone, media player or tables takes hardware, what does that make Apple?

apples and oranges.

So it's only hardware if it has end user replacable parts and is built on a system of slots and sockets? So a modern laptop with everything soldered is not hardware? What about the Fairphone, which is modular? I think I understand what you're getting at and that you're coming from the perspective of a traditional PC user but that is an incredibly arbritrary approach.

But as I said, in the end the distinction is meaningless.

you're right I am thinking in the traditional sense that a video card or memory board or thumb drive is the hardware not the laptop itself.