Hacker News Reader

Comment by nimish | Hacker News Reader

You could see this happen in real-time as Intel sold essentially the same processor between 2015 and 2020 stuck on 14nm while increasing prices and lowering quality -- replacing solder with thermal paste, sticking with 4 cores as long as possible to maximize $.

You can see the rather large change under Gelsinger: https://ycharts.com/companies/INTC/stock_buyback

View on HN

Those 2015 processors were some damn good processors though - at that point they were quite far ahead. I haven't counted Intel out of the game just yet like many here seem to - they have definitely hit a stumble though for sure.

Big companies have to stumble to see actual change. I also think they're going to recover from this. But it's going to take some time.

It's a risky bet to say intel is going to recover imho. A lot of good engineering talents have left intel. I dunno if they are capable of rehiring back those people. I suppose the new CEO is going to help.

Not many companies recover like this. Microsoft is one notable exception - and they recovered because they have a huge cash cow, and that gives them a tonne of runway.

AMD recovered well. As long as they don't do a Nokia / Elop they have a shot imo

AMD also recovered because they couldn't be killed, as Intel was never going to be allowed becoming a single source for x86. I.e. the US military said: We need more than one company doing the processors we need. Not sure if the same is true for Intel today. There are multiple providers for sufficiently powerful processors and from an enduser perspective the actual architecture doesn't matter much. I.e. it's far simpler to repurpose any other source today than it was 20 years ago.

Nokia also recovered, not only they own UNIX birthplace, the Android phones using their branding are actually the few Android ones that get proper updates.

However hiring Elop, with the contract bonus clause to sell Nokia Mobile, was a really dumb move.

>and they recovered because they have a huge cash cow

Intel also have huge cash cows in x86 market, for at least another 5 - 8 years.

You can always hire more people; my concern would more be the huge lead time required to actually bring a process to market

The 12000 series has already bought intel a lot of breathing room. It's outperforming AMD on gaming workloads at a lower price point, AMD have recognized this and lowered their prices.

My 2014 era i5-4670k is still running happy at 4.2ish GHz. Nothing really beats it without adding more cores.

Replacing solder with thermal paste? Thermal paste has always been the preferred conductor between CPUs and heatsinks. Since 486DX2 days. Unless I'm mistaken, what do you mean?

There is a metal heatspreader on top of the actual CPU which interfaces with the heatsink/fan the user installs. During the time, Intel made a change from soldering the heatspreader to the CPU to using thermal paste between the heatspreader and CPU, which decreased performance and led to users “delidding” their CPUs to increase thermal transfer and, as a result, cooling performance.

They're referring to the interface between the CPU die and the integrated heat spreader (IHS).

Inside the CPU package, between the actual chips and the integrated heat spreader. If I didn't explain it well: https://www.dvhardware.net/article70404.html

Ohhh.. Got it. Thanks! Wow, that IS a problem.

Yes and as the article says:

> Intel has also long preached the perils of solder TIM, which can lead to shorter lifespans due to microcracks in the solder TIM that form during heat-induced expansion and contraction.

It's not just about "cheaper solution". Yes, solder will be best for an overclock but that's not the primary objective here.

I worked alongside this team. Without going into too much details, PTIM vs STIM are primarily cost driven. Engineers are not stupid, which is what Linus, et. al. claim. In fact, they know and listen to what Linus has to say about shitty PTIM. I would love to tell you everything that goes into selection of TIM but for obvious reasons, I cannot. You'd be surprised the amount of engineering that is behind something as simple as thermal interface material.

As they say, good engineering is about optimal tradeoff between every aspect of your product. That includes cost (directly, we are responsible for it) and customer happiness (indirectly, since in a giant company, that gets handed off to the Marketing/PR team). These can get out of sync.

Here is some public information about PTIM: https://patents.google.com/patent/US20150279762A1/en

Reply on news.ycombinator.com