Follow Slashdot stories on Twitter


Forgot your password?
Slashdot Deals: Cyber Monday Sale Extended! Courses ranging from coding to project management - all eLearning deals 20% off with coupon code "CYBERMONDAY20". ×

Comment Degree shows you can finish what you start (Score 2, Insightful) 62

Because its the degree that matters...ffs

The degree demonstrates two things:
(1) A base level of knowledge, which does not necessarily indicate talent in the field so its not the sole qualification.
(2) An ability to **complete** a long bureaucratic process that includes some uninteresting tasks. That may be the more important thing demonstrated.

Comment Re:Overclocking errors can be a simple wrong numbe (Score 1) 320

Its not just cost/performance. In addition to possibly wasting your time debugging something that is an overclocking induced error, there is the ethics of delivering possibly faulty code to your customers. Real bugs are troublesome enough, no need to plague your customers with a flipped bit in an indexed operation or something.

Also the fact that he believes that testing software indicates stability indicates he doesn't fully understand the failure modes of overclocking. You can gradually increase speed until the software indicates a failure and then reduce speed so that the software runs flawlessly for hours, and there remains a chance that your system has errors. All you may have changed is the prerequisites for an error. At the higher speed the instruction may fail universally, at the slightly lower speed it may only fail selectively depending on preceding instructions and data patterns. Instruction sequences and data patterns that the testing software does not include and can't really be predicted anyway.

Again for casual use not a problem, you only inconvenience yourself. The cost in the cost/performance tradeoff is only paid by you. But for software development you are possibly putting a cost on your customers too. That is unethical.

Comment Re:Overclocking errors can be a simple wrong numbe (Score 2) 320

I don't know where this myth comes from, but it's bullshit. If you properly test the system, you'll catch any error producing instability before depending on it. A bit flipped is a bit flipped. They happen on stock systems too. This is why any sane system handling critical data has built in sanity checks.

No, the BS myth is that you can test for instability. Every manufactured CPU is unique. Its weaknesses unique. The point and manner that it fails unique. Some failing with a modest amount of overclocking, other not failing until much greater overclocking. And again, the failure is not necessarily something resulting in a crash. The mildest of overclocking errors can simply be an erroneous result of a mathematical operation.

Where a bit flips will vary from one CPU to the next. How can you test for that? And in a specific case when a bit flips in an instruction may depend upon the preceding sequence of instructions and upon a specific data pattern. How can you test for that, the prerequisites are unknown and vary one individual CPU to the next?

Testing by a manufacturer can involve things far more probative than running a software test. So a CPU rated for a certain speed can be confidently run at that speed. However once overclocked the confidence is lost. How much tolerance there is for overclocking can vary from one design to the next, one manufacturing process to the next, and from one individual CPU to the next.

Comment Look for manufacturing process changes for quiet (Score 2) 320

For "quiet" video cards you sometimes have to look for a previous generation design that has gone through a new manufacturing process (same circuitry but laid out at a smaller scale, a process with fewer nanometers (nm)). These may get smaller or slower speed fans. Might even go passively cooled if it was a modest GPU to begin with. Of course you won't get the greatest performance but perhaps something good enough for play.

For what its worth I tend to install auxiliary low-RPM fans that blow directly on passively cooled motherboard chipsets (often under a heat sink) and on passively cooled GPUs.

Comment Overclocking errors can be a simple wrong number (Score 4, Informative) 320

Overclocking is fine if its only a gaming PC but if it is also used for anything serious, anything where the correct results are important ... do *not* overclock.

Overclocking errors are not some black and white easy to recognize situation. It is literally a progression through various shades of gray. At the lightest shades of gray, where overclocking errors begin, at perhaps quite modest overclocking settings, the errors are subtle. Literally it may simply give the wrong answer, the wrong numeric value, no crash or anything dramatic. And what instruction yields this simple wrong answer, and beginning at what overclocking setting, and what instructions must precede it if any ... are all variable and will change from one specific CPU to another. Hence the inability to reliably test for overclocking errors. The errors manifest different on every CPU, and the required conditions manifest differently, and these conditions may include being immediately preceded certain instructions or certain data patterns. Instructions and data patterns that also differ CPU to CPU.

So if a PC is just for gaming and other casual use, overclocking does little harm. However if the computer is also used for serious numeric work, software development, etc ... its best to avoid overclocking.

Comment Quality and compatibility ... (Score 2) 320

The only good way to get what you want is to build one.

Absolutely, you get the exact parts that you want. Pre-built always seems to involve some compromise in one part or another.

Careful selection of parts is also very helpful if you want to do something like dual boot Linux.

Building your own is no longer the money saver that it once was though. But quality and compatibility are reason enough.

Comment Re: Build one (Score 1) 320

Building a custom computer? All you do is put the parts together. The hard part is picking the right parts.

Which is the same thing that Dell, HP, etc do. Screw together some parts, at least for desktops. Laptops can involve a little bit of design with respect to layout, to get airflow cooling right.

Comment How does Pi compare to PDP-11 (Score 1) 235

Aren't they the size of a filing cabinet?

That was just the external hard drive, more a full sized rack in their 1st generation, plus another rack for peripherals. They eventually got down to desktop workstation size. Maybe palm sized now with the Pi. How does a Pi with a remote text terminal session compare performance wise to a PDP-11 :-)

If you can't learn to do it well, learn to enjoy doing it badly.