Re: It is all about the taxes
The Oregon fabs are R&D with low scale production ramp up to about 2000 wafer starts per week when worked there. The fabs in Arizona, Ireland, Israel and new Mexico were for the volume production
24 posts • joined 28 Jan 2011
I for one and tired of Linus' over the top histrionics about careless developers crossing some imaginary line in imaginary head when he can't keep his own garden free of weeds. For FRICKIN YEARS I have dealt with the SND_HDA_INTEL module get screwed up during suspend and needing to do the "horror of linux horrors", reboot the computer to resolve it. It would be fixed with one kernel update and then screwed up yet again with the next. I've got a few choice and sweary words along with two big, fat fingers for him if Linus can't manage to keep fixes in place during kernel updates.
I can attest that Intel does indeed wall the garden much tighter after you give notice. I worked in the Oregon development fab 11 yrs ago before moving on. At least in the fab, most employees who gave notice would be escorted to the door and have their cube packed up & shipped to them, especially if they were going to work for any company deemed a competitor. I wasn't going to any such company so I was allowed to linger and tie things up for the couple weeks. However, my email was restricted to outgoing only and the fab badge was collected. It was like sort of like being a ghost.
At a former company I'd worked, maintenance turned off the wrong machine when attempting to perform some routine work. The unfortunate machine was a 20yr old Applied Materials etcher used in chip making. When they realized the mistake and turned it back on, it wouldn't boot. The company's contract with Applied had been allowed to lapse and along with that all the detailed knowledge of coaxing the beast through troubling computer issues. A contractor had to be brought in and ended up replacing the floppy drive with a SD card reader and card formatted to look like a 1.44Mb floppy disk imaged with the boot files. After 2 weeks of reconfiguring the machine, it was back up again and the company out many $100k.
"Scaleway is running a farm of Apple M1 Mac Minis..."
Apologies for the dense question, but why would anyone want this service? App development? Compiling binaries? If you can vnc or ssh into the server farm, then you clearly have a computer already and one that is likely better than any m1 mac mini. These units are not exactly speed demons with only 8gb ram / 256gb ssd.
I'm sure there's an obvious purpose for this service. Next round is on me.
Swan is taking the fall for the manufacturing issues at Intel. However, the development cycle from Rev99 base initial process to a production process rolled out to Intel's "virtual factory" runs about 4 years at the minimum. I know because I worked there.
There are more fundamental problems with the development team and Intel's divisional cpu architecture. These have not been addressed by the head swapping.
Yes, it is intentionally to confuse the consumer. I'm a former fab engineer at Intel's development site in Oregon (10 yrs ago). I asked one of the product engineers about the convoluted naming system at the time. He confirmed it was intentional as a means of focusing on the "core2" name and away from the raw MHz that had been the previous convention.
If I counted the layers on their sketch correctly, there are like 20 metal layers on these chips. That's a boatload of deposition, polish, via litho and etch, metal litho and etch, ALD dep at lower metal layers, copper electroplate at higher metal layers and polish again. With each step introducing defects. Hats off to getting a manufacturable process.
"Tragically, an Apple engineer was killed in the same year when his Tesla Model X accelerated into a roadside crash barrier."
No, not tragically. This is Darwin thinning the herd of those who blindly bet their life that every possible scenario has been preconceived, properly programmed, and correctly implemented with fully functioning sensors. Honestly, assuming ALL possible scenarios can be programmed into the software to allow autonomous driving is the definition of hubris. Of all people, this Apple engineer should have known better. And as an engineer myself, sorry, I'm not trusting my life or my family's to the rest of the lot.
Reminds me of when I worked in the fab at Intel. Some of the operators were found to be playing solitaire on one of the optical character reader tools intended for reading the wafer serial numbers and orienting the wafers. The machines were running some flavor win98se or NT. Management then had all the games removed from every Windows machine across every fab. Note that this is the same management that didn't allow clocks in the fab either!
Looking at the CPU lineup, it looks like the 2-core Cannon Lake line of chips is essentially a shrink of the 2-core Kaby Lake line. They would be designed as a dual-core chip only with the 4-core version using a completely different mask set and be a completely separate product.
Intel scraps wafers or lots with excessively high defect levels because it is likely the "good" die that didn't fail are actually marginal and will fail at a much higher rate. Basic Q&R. The die testing hasn't been functional for 15 years and is based on test patterns.
I suspect they will continue to resolve the manufacturing issues. Intel runs with 2 concurrent development teams, one that works on the current node (10nm) and the other working on the next (7nm). Process improvements developed under the 7nm node could certainly be ported backwards if it makes sense.
I used to work at Intel's development fab. This middling chip release reminds me of the problems they had when transitioning to the Pentium-4 90nm (Willamette chip) process that was the first to used EPI for strained PMOS performance. In that case, the PMOS leaked current like a sieve and they couldn't keep the chip in the spec'd power envelope. They had to back off the poly gate CD to drop the power consumption but that hurt performance and made the chips only usable for the entry level devices.
With the cost of developing a new process, it is certainly desired to have the new chip come in for the performance segment where the profit is the highest. These i3 chips are commodities and not terribly profitable.
"There's not much else that one can do with a USB Stick besides shoving it into a computer... "
Baldrick: It's a bit of a tiddler, ain't it?
Black Adder: Yes, but size isn't important. It's not what you've got, it's where you stick it.
I set up my mom with Ubuntu and she's been running it for the last 10 years. When she had to recently use Windows 7 for a couple classes, she couldn't wait to get back to linux. Even she commented how much more stable it is. That really speaks volumes for how easy linux can be to use. I plan to set her up with Mint next.
I've been essentially unemployed (< 1/4 time employed) for the last 18 months, in spite of my PhD in electrical engineering. Guess I picked an inopportune time to quit and relocate to the midwest!
The difficulty I have encountered has been with trying to break into a new field in this economy and job market. While I have over 8 years of semiconductor manufacturing experience, this seems to count for little when applying for an engineering position in a different manufacturing area. Employers are demanding very specific experience in their own manufacturing field and the extensive job pool is able to supply it.
In addition, employers are reluctant to hire. Though I was chosen as the top candidate for a position following a lengthy interview process, the company's management indefinitely froze hiring.
Having a doctorate may better insulate a worker from being sacked or made redundant, but it doesn't necessarily improve the chances of obtaining a position.
Biting the hand that feeds IT © 1998–2022