New Intel CEO Making Waves: Rehiring Retired CPU Architects (anandtech.com) 70
We're following the state of play with Intel's new CEO, Pat Gelsinger, very closely. Even as an Intel employee for 30 years, rising to the rank of CTO, then taking 12 years away from the company, his arrival has been met with praise across the spectrum given his background and previous successes. He isn't even set to take his new role until February 15th, however his return is already causing a stir with Intel's current R&D teams. From a report: News in the last 24 hours, based on public statements, states that former Intel Senior Fellow Glenn Hinton, who lists being the lead architect of Intel's Nehalem CPU core in his list of achievements, is coming out of retirement to re-join the company. (The other lead architect of Nehalem are Ronak Singhal and Per Hammerlund - Ronak is still at Intel, working on next-gen processors, while Per has been at Apple for five years.)
Hinton is an old Intel hand, with 35 years of experience, leading microarchitecture development of Pentium 4, one of three senior architects of Intel's P6 processor design (which led to Pentium Pro, P2, P3), and ultimately one of the drivers to Intel's Core architecture which is still at the forefront of Intel's portfolio today. He also a lead microarchitect for Intel's i960 CA, the world's first super-scalar microprocessor. Hinton holds more than 90+ patents from 8 CPU designs from his endeavors. Hinton spent another 10+ years at Intel after Nehalem, but Nehalem is listed in many places as his primary public achievement at Intel. [...]
Hinton is an old Intel hand, with 35 years of experience, leading microarchitecture development of Pentium 4, one of three senior architects of Intel's P6 processor design (which led to Pentium Pro, P2, P3), and ultimately one of the drivers to Intel's Core architecture which is still at the forefront of Intel's portfolio today. He also a lead microarchitect for Intel's i960 CA, the world's first super-scalar microprocessor. Hinton holds more than 90+ patents from 8 CPU designs from his endeavors. Hinton spent another 10+ years at Intel after Nehalem, but Nehalem is listed in many places as his primary public achievement at Intel. [...]
Re:FDIV here we come.... (Score:4, Funny)
F00F! that's a low blow.
First new product? (Score:4, Funny)
Introducing the all new 286 SX+ with optional math co-processor.
Re: (Score:2)
Take an array of 16 year old CPUs (now 27) and try to compete with GPUs. [wikipedia.org]
(does anyone remember how many cores Larrabee was going to have? the architecture supports modes for "less than 16 cores" and "16 or more cores")
It is cool that it comes from a Pentium(P54C) that was kicking around Intel for newer fabrication processes because of a military rad-hardened edition. I can only assume it's in some spy satellites running lots of Ada code.
Re: (Score:2)
Original Knights Corner (effectively a stepping of Larrabee) was ~60 cores, each 4-way threaded. Top-shelf Knights Landing parts were 72 cores.
Rumor was that the P54C design was re-imported back into Intel -- they had contracted out a re-implementation, I assume because they needed a fully synthesizable design rather than being tied to a specific die process. My old team had an FPGA socket 7 module that could run an instantiation of this code around 100MHz.
Re: (Score:2)
Although that was an obvious joke, it was funny. So why wasn't it modded up? My guess is that not only is slashdot full of non-nerds, the modders probably weren't born when the 286 came out.
Ex intel engineer Francois Piednoel has a video .. (Score:5, Informative)
Ex Intel principal engineer Francois Piednoel, best known for his work as the architect behind the dual-processor high-end desktop (HEDT) Skulltrail platform, has this interesting video [youtube.com] on Pat Gelsinger.
Intel has basically been sitting on their laurels for the past decade. AMD's Ryzen and Threadripper are kicking their ass. Intel forgot the wisdom of Andy Grove: Only the paranoid survive.
While I'm no fan on Intel's shenanigans [youtube.com] having healthy competition again between Intel and AMD benefits consumers.
Now if only we could get GPUs ...
Re: (Score:1)
at this point, their GPUs will be welcome competition in the market, but i cant see them making a top-tier GPU id be interested in purchasing any time soon.
More Patrick Gelsinger information (Score:3)
You can’t out Moore lifespans (Score:3)
Young is the old old. (Score:3)
President Joe Biden is 78 years old. He just got a new job.
Re: (Score:3)
Re: (Score:3, Interesting)
Would you want an inexperienced 20-year old running the country?
While youth have the advantage of not being indoctrinated with dogma, old(er) age usually brings about wisdom. Youth tend to be liberal while Elders tend be conservative.
Intel bringing back experienced chip architects means they are probably getting serious about rethinking performance since they can no longer rely on 14++++ anymore. If AMD can increase IPC by ~15% then why can't Intel?
-- /. with your lameness filter and your bullshit
Fuck off
Re: Young is the old old. (Score:3)
"Intel bringing back experienced chip architects means they are probably getting serious about rethinking performance"
Experienced, but out of date, and responsible for Intel's former flawed designs? That is serious?
Re: (Score:2)
I admit that I did not find "leading microarchitecture development of Pentium 4," to be a ringing endorsement. I'd have to ask what did he learn from his mistakes on that chip.
Re: Young is the old old. (Score:2)
You're making an assessment now, almost 2 decades later. Everybody can be smarter *now*. The trick would have been to have been smartet then - which you weren't. Nobody was, apparently.
Here's an article (German, sorry, but you surely can use a translation service) about what P4 actually meant, and what an ingenuous leap of innovation it was for that time.
https://heise.de/-4967087 [heise.de]
Re: (Score:1)
Re: (Score:2)
Everybody can be smarter *now*. The trick would have been to have been smartet then - which you weren't. Nobody was, apparently.
Well, except for the people at AMD who designed the Athlon and the Duron. I felt pretty clever overclocking my Duron 600 to a 1 GHz back in the day...
Re: (Score:2)
Did you even read the article?
I'm not saying P4 was the better processor. It's also not about rekindling old Intel-vs-AMD flame wars. It's only that the P4 had some ahead-of-it's-time improvements, and dismissing the guys who invented it is not good manners. For starters, it could go up to several GHz, which AMD couldn't do until later. AMD did do very nice hardware, but, like everybody else, they, too, were standing on the shoulders of giants. Maybe they would've done the same mistakes as the Intel guys fi
Re: (Score:2)
The P4 was a steaming pile of dogshit. It's extremely long pipeline led to frequent branch mispredictions, crippling performance. It was abandoned, and they went back to the P3 processor architecture with a die shrink. It was damn near fraudulent to sell the thing. It wasn't ahead of its time at all. If you know jack shit, maybe you should stop making assertions.
Re: (Score:2)
If you know jack shit, maybe you should stop making assertions.
Why? To take a good example in you and be a jerk instead, too?
In case you didn't realize: you still haven't refuted one single argument from the -- well written -- article. So far, I have a person who took the time and explained step by step what was novel and innovative about the architecture (please mind my choice of words) on one hand, and a slashdot dimwit who's throwing words around like "fraudulent" and "dogshit" on the other. And the latter somehow expects me to STFU because he's... right? She claims
Re: (Score:3)
Re: (Score:2)
That was more down to business shenanigans and their competitors making mistakes though. Like Cyrix went all in on integer performance but Intel's FPU performance became a key benchmark, not least thanks to games like Quake.
Re: (Score:2)
Re: (Score:2)
Experienced, but out of date, and responsible for Intel's former flawed designs? That is serious?
Most things he developed were _not_ flawed. And P4 was based on politics - Intel demanded high clock speed at any cost, so he gave them clock speed at any cost, which _was_ a flawed approach, but probably not something he could prevent.
Re: (Score:2)
Re: (Score:3)
I don't see how your default position of someone challenging one extreme is to counter it with another extreme. No, I don't want a 20-year-old running the country, but I also wouldn't want a 78-year-old in declining health to be running it either. You can certainly find a better line where experience vs. liability is more balanced. The majority of world leaders tend to be in their 50s, by comparison, though many in their 30s have also been effective leaders.
There's nothing wrong with bringing in experienced
Re: (Score:1)
Would you want an inexperienced 20-year old running the country?
While youth have the advantage of not being indoctrinated with dogma, old(er) age usually brings about wisdom. Youth tend to be liberal while Elders tend be conservative.
Intel bringing back experienced chip architects means they are probably getting serious about rethinking performance since they can no longer rely on 14++++ anymore. If AMD can increase IPC by ~15% then why can't Intel?
While you don't want a 20 year old running the country, I certainly think that 78 is quite a bit too late to start a 4 year term. Not that the other front runners were any better in that regard, Sanders is even older and while Trump is a couple of years younger he sounds a bit senile - not to mention that he just spent four years as the worst US president, personifying dishonesty, narcissism, lack of integrity, lack of knowledge, lack of skills of listening people with knowledge, corruption, nepotism and
Re: (Score:1)
Re: (Score:1)
playing the corpse in Weekend at Bernie's
Re: (Score:1)
Re:You can’t out Moore lifespans (Score:5, Insightful)
There's something to be said for people with long experience in a particular technology. They may not have the most up-to-date knowledge, but they also probably have stacks of ideas that weren't feasible 20 years ago that might be useful now. Let's remember, a lot of theory about microprocessor designs dates back as far as the 1960s, but never were fabricated simply because of manufacturing limitations. A lot of the "advancement" as it were isn't in design, but in far more advanced fabrication technologies
25 years to go from For Dummies to cutting-edge (Score:5, Insightful)
25 years ago I was studying the basics of my field.
I was reading the For Dummies books.
Learning how to use Linux 1.2.
That prepared me learn programming Linux 20 years ago.
I was reading programming books from intermediate to advanced.
That prepared me to contribute to the kernel 10 years ago.
I was reading the source code of the kernel and read and writing RFCs.
Experience in the kernel prepared me to find and help fix obscure security flaws in Linux 5 years ago.
I was reading research papers on technologies that may become commonplace in 2026.
That prepared me for today. Today I'm hacking compiled binaries. I read research papers on provably secure operating system architecture, ideas that might be used to design the next generation of operating systems. Today I'm reading (and modifying) binaries in hexedit.
You don't go from PHP For Dummies to writing software as machine code in hexedit in two years. Becoming an expert takes 25 years.
Re: (Score:3, Funny)
Becoming an expert takes 25 years.
That's why I put 25 years of React experience on my resume.
Re: (Score:2)
but they also probably have stacks of ideas that weren't feasible 20 years ago that might be useful now
Take the Wallace multiplier, which was invented in the 1960's, but would have been impossible to implement back then. (It is a multiplier that needs about twice the time of an adder, but requires thousands of full adders to implement. )
Re:You can’t out Moore lifespans (Score:5, Insightful)
Old talent can train fresh blood. That's the real goal - the architects are retired, they're not going to need more than a part time job, so they're going to have to train the new guys up and give them exposure.
The CEO doesn't expect the old guys to produce the next gen chip, but expects the old guys to train up the new guys so the next gen chip comes out as an application all the tribal knowledge that was built up and underappreciated.
Many of the new guys didn't get to benefit from the wisdom and have to relearn everything the hard way
Re:You can’t out Moore lifespans (Score:4, Interesting)
Many of the new guys didn't get to benefit from the wisdom and have to relearn everything the hard way
This is a double edged sword. A lot of the old guys may be hampered by the wisdom that the world has changed around them as well. The success of this will depend highly on the attitude the people bring to the table. It's important to filter through the bullshit. Sometimes "We tried this 20 years ago and it failed miserably, it won't work" can be insightful. Sometimes that same phrase is uttered by people who don't realise that the thing they tried 20 years is now the perfectly functional normal way of doing something because while they weren't paying attention that the industry changed.
My most comical case of this was someone complaining that a certain instrument wouldn't work in a particular service because he's tried it. When I told him it's standard practice these days he called bullshit and asked for examples because he's adamant it wouldn't work. I walked to the window, raised the blinds and pointed to a vessel built 5 years ago by a different project this guy wasn't on, same service, same conditions, same instrument and told him it works flawlessly there, why is his metal container so different.
Knowledge alone doesn't bring success, it needs to be paired with the right attitude.
Re: (Score:2)
Intel didn't just make a strategic decision to not move past 14nm. They had problems moving to 10nm.
That they've been able to accomplish what they have on 14nm is no small feat.
There are certainly ways to criticize Intel for sitting on their laurels in some areas, and certainly they did something wrong if the 10nm process they were working on didn't work out (though it's easier to criticize than to succeed at hard problems). But being on 14nm isn't just because they didn't want to compete.
Re: You can’t out Moore lifespans (Score:2)
People often forgot, that switching to EUV meant basically re-inventing ALL the things. Even mirrors and windows stop working. And are hard to develop. Everything has to be under a special inaccesible athmosphere.
The thing was, that everything up to this has been marginal evolution compared to this.
Then again, TSMC *did* manage it.
Ane Intel's management didn't become any less of a bunch of assholes everybody wanted to stab in the back either.
Re: (Score:2)
Actually, it was ASML who managed it, now the sole supplier of EUV lithography equipment to all of TSMC, Samsung and Intel.
Re: (Score:2)
As nearly as I can tell, the 10nm fiasco was caused by marketing demanding 10% finer feature size than physics permitted. Then their attempts to fix it made it more bulky and power hungry while yields never got into the zone where they could compete with the old fabs.
Re: (Score:1)
True. But so far, the diversity hires haven't worked out so great. So, maybe some new hires can learn from the old hands.
Connect the dots (Score:2)
Cannot shrink the process? Bring old men instead!
Re: Connect the dots (Score:2)
Don't you get it?
Shrinking the process . . .
magic . . .
wolves! . . .
night . . .
moon . . .
shoes! [youtube.com] . . .
walking . . .
old men!
Re: (Score:2)
I was hoping somebody could do it for me. Slashdot keeps being amazing. Thank you!
Re: (Score:3)
Better yet, bring back the old blue men! https://www.youtube.com/watch?... [youtube.com]
Meet New Boss (Score:3)
Same as the old boss. Literally!
new version of Celeron 300A ? (Score:2)
I loved my Celeron 300A, got it to 464 Mhz stable or 503 Mhz on cold mornings.
Re: (Score:2)
I had mine clocked up to 700Mhz, got lucky with the malaysian silicon on that one. Kept that guy running up for the longest time; I wish I would have kept it to be honest.
Re: new version of Celeron 300A ? (Score:2)
The first Athlons/Durons. Bought 533 ones, ran them at 933 for half a decade, no problems, on $10 worth of cooling.
(Remember: The ones where you could bridge a connection with a pencil.)
Re: new version of Celeron 300A ? (Score:2)
I remember but why would you do that? It was only needed to unlock them into MP processors for multiple cpu boards.
If I remember right they were already unlocked multiplier-wise. The pencil trick just turned it into a more expensive version of the same chip XP to MP.
Sounds desperate. (Score:2)
10 years is a lot of time in that field.
And if they pull an AMD with him, I hope they remember to send a Thank You cake to AMD for waking them up from their complacent and ultimately deadly slumber.
Re: (Score:2)
10 years is a lot of time in that field.
Your ignorance is showing again. 10 years is current tech in the field. Ever wonder why the current intel CPUs are called 11th Gen? Here's a hint: It's about 11 years old and based on an architecture identical to that was developed 14 years ago. Even AMD's "new" CPU architecture is the result of work that is now close to 6 years old.
Another ignorant post brought to us by BAReFO0t.
So the guy 10 years out of date (Score:2)
Methinks I'm selling Intel stock and buying AMD. If I wasn't retired with a long term horizon of "die peacefully".
Now do that with Boeing (Score:2)
Fire the MBAs and bring the old farts back, I think there is still time.
No Fabs = No Intel (Score:1)
That sends a message... (Score:3)
That's the way to kick off a new direction. You can say "chips are number one", but bringing back the best engineers to help get it done is a whole other ball game.
Let Intel keep some tax breaks if they keep fabrication here and kindly let Apple, AMD and others know that there will be increasing costs in not investing in plants in the US. I say we turn Detroit into Chip City. Chips, batteries and electric motors are the new infrastructure and we can built it all here effectively.
Netburst, really? (Score:3)
P4 almost killed Intel last time.
"Hertz, Hertz, Hertz!" said marketing and Intel complied, to the delight of cooler manufacturers everywhere.
Meanwhile, an Intel Israel skunkworks kept working on P3 and after Intel was backed into a corner, facing doom from AMD x86_64, they started rummaging through the cupboards, found the "Core" design, and "realigned", kicking Netburst to the curb. The current Intel line is descended from "Core" nee P3, but with lots of added speculative branching for sidechannel attacks ... uh, performance.
So ... now being backed into a corner by AMD and facing certain doom from Zen ... they're bringing in the Netburst people?
Old != better. Where them Israelis at?
CEO worship failing again (Score:1)
intel made the first superscalar microprocessor? (Score:2)
That doesn't sound right
in my mind intel did not even exist when that happened