Many of which are extremely self-serving, btw. I am sure that NSA and GCHQ have some very good people, but to be as diplomatic as possible I have heard that morale at GCHQ at least is not of the highest. I am merely venturing a guess that Google might well have beaten the agencies to it. Apart from anything else, Google wants a share of Wintel’s desktops. Finding a serious weakness in Windows that doesn’t exist in Chrome or Android or is easily mitigated isn’t exactly doing them any harm.
I just had a nice thought-- could this be used to extract media decode keys?
It took a while, but oblig:
He’s wrong about one thing though; we don’t suck at computers. Compared to just about every other technology we’ve developed, computers have been rather defect free given that they’ve been around for 70 years. More than a hundred years after the development of steam engines boilers and engines were blowing up with depressing regularity, and over a hundred years after the Diesel cycle was made practical, the things were literally poisoning people in cities.
Having read, reviewed, debugged, and written code for many years, I’d say given the code that we write, it’s kind of a miracle that they do anything successfully. A double miracle when they do what we want them to. It’s ridiculously easy to overlook some tiny detail that could result in catastrophe but might not be noticed for years until just the right combination of circumstances triggers it.
A confession here.
Years ago, early in my career, I had to design a piece of equipment intended to carry out accurate resistance measurements at 170C +/- 0.025 degrees. The measurement was carried out in fluorinert. The fluorinert was agitated by a big paddle and stationary blades; it was heated by a PWM 450W DC power supply driving an element immersed in the fluorinert. It was all controlled by a TMS9995 microprocessor (that long ago) running at 12MHz - the performance being why we didn’t use a more conventional CPU.
The code was entirely written in assembler, carefully checked. It was quite a short prgram - it fitted into 2 32K Eproms. That’s bits, not bytes.
The thing ran for 6 months perfectly 24 hours a day 7 days a week, and then one morning the fluorinert had boiled dry, all $2500 worth of it.
As you can probably guess, there was a single bad instruction, a jump to a label which was one instruction away from the correct label. Under an error condition the thing was supposed to shut down. It should have jumped to the label that wrote an I/O instruction that killed power to the heater, then went into the main shutdown sequence. Instead, it went into the shutdown sequence without turning the heater off. And the error? The PWM controller had failed hard on, and so when the shutdown sequence set power to zero nothing happened, leaving the heater on full power. So, the tank boiled dry.
Amazingly the company forgave me, partly because my boss, who was a good guy, said the electrical engineer who built in, and who had left, had failed to include a simple temperature trip - which was duly added. But I have never forgotten how disastrous a single bad instruction can be, even in a system which is supposed to have failsafes.
Now think of a modern multi cpu system running a complex multitasking operating system, such as a phone,with gigabytes of storage and of running code, and consider my phone currently has an uptime over 1 month.
Well the first round of kernel updates are out from Debian
https://www.debian.org/security/2018/dsa-4078
https://www.debian.org/security/2018/dsa-4082
I am not noticing any difference in performance
With Windows 10 on older silicon (2015-era PCs with Haswell or older CPU), some benchmarks show more significant slowdowns, and we expect that some users will notice a decrease in system performance.
looks at his 4690 equipped imac , and cries
Microsoft’s apparently having issues making the patches work on older AMD machines, and is blaming AMD’s documentation. Which is odd, considering that the Meltdown patches aren’t supposed to be needed for AMD processors, and the Spectre ones shouldn’t be causing the kernel to crash…
No you need to draw a pentagram on it with thermal paste.
(note: it’s been a long time since I’ve delved deeply into hardware tweaking, so take this with a grain of salt)
I wouldn’t expect tweaks to the GPU to be needed. Most of the performance overhead involved with the patches seems to involve I/O-intensive operations, like large amounts of disk accesses. I’ve seen databases mentioned a lot as being impacted by it… that might also hit things like video editing or 3D rendering. There may be a framerate hit on gaming, but for the most part games don’t appear to do as much of the operations that are hit.
Here’s an article I found on a quick search with some real-world testing:
I have a lot of trouble with opening larger jpx-based pdfs, such as many internet archive pdfs, with navigating them, ocring them, and processing them so they are faster on the Mac and readable on the Kindle. I expect the update will give me more trouble.
For scans, I usually use k2pdfopt, but I often get seg faults with large files. For others, I get better results by converting to grayscale on an older version of os x.
The NSA is bleeding people like crazy. If you read the article, one of the cited reasons is the pay compared to…silicon valley.
It’s not even that there haven’t been defective computers (remember the Cyrix coma, Pentium FDIV and F00F bugs?). It’s just that there’s rarely something so catastrophic that it couldn’t be worked around with clever software tricks. Spectre’s breadth across numerous procesor architectures is definitely an anomaly.
No one has said the the NSA didn’t have talented people unless I mis-read the thread above. The point that was made is that they aren’t the best (at least any longer). Patriotism and sense of mission only gets you so far, and Snowden didn’t help either since it exposed their hypocrisy.
Cheltenham is a good place if you are looking for somewhere to bring up your kids. Before then it’s the last place you would want to live. There’s not a lot happening there.
Maybe the answer is to work for the dots until you are around 30 (when they lose interest in you because of ageism) then move to GCHQ.
Everybody I have known to have worked for GCHQ left in their 20s.
Don’t forget; the view of the Civil Service is that scientists are on tap, arts graduates or ex military are on top.
Getting rid of the dust would be a good first step. That and checking the fan, I have seen GPUs where the fan had enough dust stuck to it to reduce the air movement considerably.
I was already planning to go with Ryzen for my next upgrade when I have the money for it… from the technical details, that’s feeling like an even better idea.
At the end of the day, this is going to be expensive for the enterprise.
Just replacing any “hardware that will not have firmware updates made available” is a LARGE number.
Plus man-hours replacing, patching, reverting when we find the patch breaks Application X, adding more CPU when the performance hit Application Y sees is too great, is another big number.
And the amount of Bourbon cannot be calculated at this time. But even if I stick with the mid-shelf stuff, I’ll need to see if I can expense it.