Tech Things You Dont Know But Are Too Afraid To Ask.

[Hoppa]

Why hasn't Linus been in an NCIX TV commercial? seriously. Why Linus? WHY?!?!? :angry:

[Fetzie]

MG2R, on 29 Jun 2013 - 6:43 PM, said:

The PLL isn't only necessary for overclocking, it is always necessary ;)

Phase locking doesn't mean making sure there is x amount of ticks for every base clock tick. Phase locking makes sure the ticks have there flanks synchronised (see the dotted lines in the picture).

Part of the instability of an overclock can be because the PLL can't properly "lock on" the CPU clock to the base clock. In such a case, it might help to overvolt it. If that happens (no lock-on), there will be indeed issues with components getting or sending data and instructions on the wrong time.

Hope that clears it up :)

Thanks for the clarification :)

[FireFox]

Why hasn't Linus been in an NCIX TV commercial? seriously. Why Linus? WHY?!?!? :angry:

 

Thought he left NCIX because of his current projects

[TheRetiredSlave]

I have tried over and over to understand SSD caching but to no avail. I think I have seen the techquickie video about it. I still haven't digested it. 

 

Thats my question. :D

 

When you have the common SSD boot drive + HDD storage drive setup, you would install the programs you use the most on the SSD so they load faster. Anything that you don't plan on using that often, you would put on the HDD so that it doesn't use up the limited space of your SSD.

 

Caching basically does this automatically.. You install everything to the HDD, while the SSD is hidden from your view. As you start using programs, the controller software records what you use most often. When you use something a lot, it will load that program to the SSD, so that it loads faster than if it were still on the HDD. All this is done in the background without you seeing anything (other than the result that your most used programs load faster).. Since your programs are still kind of tethered to the HDD, they don't load as fast as if they were on a pure SSD, but caching is ideal for beginners because it doesn't require any complicated setup (or remembering which drive to install too).

 

That's the simple overview of what it is,, I hope you weren't looking for an in-depth technical analysis of how it works, because I don't really know that.

[Hoppa]

Thought he left NCIX because of his current projects

yeah but he still does the 'what's hot' videos for them, and he has done all the other videos on youtube for them so he's basically the face of NCIX.

I'm surprised he hadn't done one a long time ago :(

[Paul9631]

OK this is going to so sound utterly stupid, but I am totally confused with this as there are so many different ways and ugh everyone says the other is bad But How do you correctly apply thermal paste/compound. I have seen more than 10 different ways of doing it all claiming which one is better i trust Linus and Logan with the pea technique but is it really the best?

 

I was confused on this for a long time.  Absolutely the best source I have ever seen regarding thermal paste (both analysis of brands, and application techniques) was BenchmarkReviews.com's 80-Way Thermal Interface Material Performance Test.

 

I STRONGLY suggest that everyone read the entire thing, but TL;DR is that you should spread it evenly with a latex glove, or just make two parallel lines.

[Breakdown]

What are TeraFLOPS, and how exactly do they affect performance?

[Godlygamer23]

What are TeraFLOPS, and how exactly do they affect performance?

FLOP stands for Floating Point Operations Per Second. Tera means trillion, so therefore teraflop means one trillion floating point operations per second.

More information: http://en.wikipedia.org/wiki/FLOPS

[Breakdown]

FLOP stands for Floating Point Operations Per Second. Tera means trillion, so therefore teraflop means one trillion floating point operations per second.

More information: http://en.wikipedia.org/wiki/FLOPS

Okay then, thanks. I read through that link and I still have some questions.

 

Can you give me an examply of some instruction that uses or needs only one floating point operation?

 

Like say I were to use the calculator program and add 2 + 2. Does having one teraflop mean that I can potentially do that calculation one trillion times in one second?

[Vitalius]

`

 

Blasfemy!!! Well, it depends what you understand by gaming, but i think 500$ PC will be better, having the fact that you can multitask by there side.

If we talk strictly gaming, consoles are always better price-wise. 

Them slowly being able to multi-task (such as the PS4's ability to download in the background and both consoles generally being more "media friendly") makes it even moar apparent.

PC master race, but for the price, in the short term (1-3 years), consoles are better.

[Godlygamer23]

Okay then, thanks. I read through that link and I still have some questions.

 

Can you give me an examply of some instruction that uses or needs only one floating point operation?

 

Like say I were to use the calculator program and add 2 + 2. Does having one teraflop mean that I can potentially do that calculation one trillion times in one second?

http://www.keil.com/support/docs/1476.htm

[trentonjeffro]

What does a Hyper Threading do?

Hyper-threading adds another thread per core on your cpu, so If you have a dual core i3 lets say a 2105 it has 2 cores, but 4 multitasking threads. Helps alot in multitaking applications

[Breakdown]

http://www.keil.com/support/docs/1476.htm

Ah. I get it now. Thanks again.

 

That's all for now.

[MG2R]

Hyper-threading adds another thread per core on your cpu, so If you have a dual core i3 lets say a 2105 it has 2 cores, but 4 multitasking threads. Helps alot in multitaking applications

Well, that's the short answer. The longer (actual) answer:

 

Hyper-threading is Intel's trademark for CPUs which have the capability to share part of their execution pipeline between two threads at the same time. This does not mean you can fully execute two threads at the same time in every scenario (you need multiple physical cores for that). Let me illustrate it with an example:

 

Processors nowadays have many functional units (arithmetic logical units (ALU), floating point untis (FPU), etc...). Consider a (hypothetical ) CPU that has one ALU and one FPU. If it isn't hyper-threaded, you could run either a simple calculation on the ALU or a floating point calculation on the FPU. During that operation, one of the functional units would be active and the other one would be doing nothing. If the CPU is hyper-threaded, it would be capable to fetch two instructions and data sets at the same time, if one instruction is a simple operation and the other one a floating point operation, both executing on a different dataset. That way, you can have both functional units utilized at the same time and thus process two instructions per instruction cycle.

 

Of course, if both operations need to be performed on the same data set, one of the operations depends on the result of the previous operation, or both operations are floating point operations, the multi-threading would not be able to perform both operations at the same time, thus effectively rendering the CPU a simple single core unit. That's why you often hear that a hyper-threaded CPU isn't equal to a dual core CPU. It can perform two instructions at a time, but only if those instructions are completely independent.

[Glenwing]

1. Why Prices of Computer Components and Accessories much cheaper in US than other countries?

 

2. How does having a smaller die size (eg. 14nm) helps in improved performance of processor?

 

32nm/22nm/14nm etc. isn't the size of the die, it is the distance between the gates on the transistors that make up the die.  Smaller transistors can switch between states a little more quickly, and require less voltage to do so.

[Nexus]

1. Why Prices of Computer Components and Accessories much cheaper in US than other countries?

...

Taxes, and the fact that many companies just substitute the $ sign for the £ sign for example. Its unfair buts its just the way things work out.

[atrash]

how many fans can you practically connect to one single motherboard fan header?(using splitter cable)

[MG2R]

2. How does having a smaller die size (eg. 14nm) helps in improved performance of processor?

32nm/22nm/14nm etc. isn't the size of the die, it is the distance between the gates on the transistors that make up the die.  Smaller transistors can switch between states a little more quickly, and require less voltage to do so.

Changing the manufacturing process doesn't inherently come with an increase in performance. What Glenwing says is very true, but isn't the most important aspect in increasing processor performance.

 

I don't know if you have ever heard of the tick-tock development process CPU manufacturers use? It's basically a system where the development is a two stage process. In one stage, the architecture is changed, making the processor x% faster than its predecessor. In the other stage, the manufacturing process is changed, making the chip smaller (if the architecture remains the same). The last stage doesn't really increase preformance a whole, except for a few tweaks here and there.

 

For example: sandy-bridge was 32nm, after that followed a die shrink to 22nm (ivy-bridge), after that followed an architectural change (haswell).

 

The most important benefits of a smaller lithography are:

• Smaller chips
• Higher yields (smaller chips means that bad spots in the silicon affect a smaller percentage of chips)
• lower power (due to the lower voltage that is required, as mentioned by Glenwing)

The most important disadvantages of a smaller lithography are:

• The heat output is concentrated in a smaller area, meaning that heat transfer to the IHS is less efficient, meaning the chip runs hotter. This is in part countered by the fact that the chips consume less power and thus output less heat.
• Lower breakdown voltage
• Lowered resistance to electromigration

 

how many fans can you practically connect to one single motherboard fan header?(using splitter cable)

As much as that fan header supports. I know it sounds ironic, but it isn't. Every fan header is different in the amount of power it can provide. The more power it can provide, the more fans you can hook up to it.

[Lachrymal]

Can someone give me a rundown of Nvidia's and AMD's model numbers and what it all means? For example: Gtx 660 and 660ti, 690, 780, Radeon 7750, 7770, 7770 Ghz, 7870 XT....

 

 

What am I looking at?? I can watch Linus's graphics card videos for days and not retain most of it because of those numbers. And it's the only thing I haven't decided on in my first build.

[Glenwing]

Can someone give me a rundown of Nvidia's and AMD's model numbers? For example: Gtx 660 and 660ti, 690, 780, Radeon 7750, 7770, 7770 Ghz, 7870 XT....

 

How do you mean?  Explain their naming conventions or give you a giant list of all their model numbers?

[joelthezombie15]

Holy crap, it was pinned. I just noticed, thanks what ever mod did this. 

[Lachrymal]

How do you mean?  Explain their naming conventions or give you a giant list of all their model numbers?

 

What order of quality and/or release they go in. What one model offers over the other generally. And the recipe behind the place of digits mean. Like the first 7 in 7990, the 2nd and 3rd number in 7990, and whats with the zero? Same for GTX.

[Vitalius]

Lachrymal, on 01 Jul 2013 - 11:31 AM, said:

Lachrymal, on 01 Jul 2013 - 11:31 AM, said:

What order of quality and/or release they go in. What one model offers over the other generally. And the recipe behind the place of digits mean. Like the first 7 in 7990, the 2nd and 3rd number in 7990, and whats with the zero? Same for GTX.

The first number for both AMD and Nvidia cards is the series number which is essentially what generation of GPU the card has in it. Higher numbers mean newer generation which means more power and better efficiency.

The 2nd number (and 3rd number for AMD) represent how powerful they are within that series.

For example, a 760 is weaker than a 770. A 7850 is weaker than a 7870 or a 7970.

The 0 in AMD's line-up is just to make it in the thousands as a marketing gimmick.

When it comes to comparing cards from different generations, usually you can subtract 1 from their power number (the 2nd number for Nvidia and AMD) for every generational gap.

For example, a 680 =/= 780 in power. A 680 is much closer to a 770 in power. That's more of a rough way to do it, but it more or less works.

One special case in all this is when the numbers end in "90". For example, 7990 and 690 have 2 GPU's on the card. Meaning a 690 is basically 2 680's on one card in SLI mode. Except it only takes up 1 PCI-e slot.

Same thing with a 7990 and 7970's.

This is not true of older cards (the 5XXX series had the 5970x2 meaning "times 2" and older generations followed that scheme).

One other thing to note is that the AMD 6XXX series is NOT the same GPU generation as the Nvidia GTX 6XX series.

The Generations go like this comparing Nvidia to AMD:

GTX 7XX - Radeon 9XXX (The 9XXX series isn't out yet and won't be for a while)

GTX 6XX - Radeon 7XXX (The 8XXX series was skipped over.)

GTX 5XX - Radeon 6XXX

And so on.

One other special case is the GTX Ti (short for Titanium) versions of cards. I have no idea how that works, however. I just know that a Ti version of a card is better than a non-Ti version (mostly) and that this usually only applies to GTX X60 power cards.

For example, a GTX 560 would have a Ti version. So would a 660 and (we assume) the 760 will as well, though some people think that's not going to happen. Me being one of them.

Here's a chart to show you relatively (this is a rough estimate) similar cards in power. Note that if they are in the same row, they are relatively in the same "tier" of power.

GTX Titan GTX 780

GTX 680 GTX 770 Radeon 7970

GTX 670 Radeon 7950

GTX 660 Ti GTX 760 Radeon 7870

GTX 660 Radeon 7850

GTX 650 Radeon 7770 (This is where GPU's stop being for gaming and start being for Home Theater's and Video Playback).

GTX 640 Radeon 7670

And so on...

And do note that is only power, not price tiers.

[Glenwing]

The first number in each model number represents the series.  The GeForce 600 series, 700 series, etc.  Or the Radeon HD 6000 series, 7000 series, and so on.

 

In GeForce model numbers, the middle digit represents the relative position in the lineup.  GTX 650, GTX 660, GTX 670.  NVIDIA also uses the "Ti" moniker, which stands for Titanium.  It doesn't mean anything, they just use it when they are running out of model numbers, since they usually have more than 8 graphics cards in each series, filling up the 610-680 positions (690 position is reserved for dual-GPU cards).  The GTX 660 Ti goes in between the 660 and 670.  The "Ti" doesn't mean the GTX 660 Ti is related to the GTX 660 in any way.  It just means they ran out of model numbers and had to make up some in-between ones.

[Patthecat]

When does intel turbo boost actually turn on? I have a i5 2500k and,  if I run it at stock speeds, I use program like prime95 and cpu-z, I see no increase in clock speed despite the supposed "turbo boost" that should occur when the processor is getting heavily tasked. Can someone explain?