Sign in with
Sign up | Sign in
Your question

Explain this to me in terms I understand.

Last response: in CPUs
Share
August 21, 2009 5:37:25 AM

TL;DR: Skip to the last paragraph

I like to play RPG games. I always have, ever since I got Shining Force for my Sega Genesis. I especially like games that allow you to customize not only your party as a whole, but also them as individuals, so having a skill and stat point system is important to me. I like to make sure my magic/tech characters get minimal tanking skills and points and make sure my fighters/dpsers maximize their damage potential, and there have been only a few games in which I truly appreciated the skill system, because most of them suffer from one universal flaw, and that is what I call an Imprecise Upgrade System.

I will use X Men Legends as an example of this, but you will be able to relate my woe to just about any rpg game.

In X-Men legends, you get 1 stat point and 1 skill point for every time you level up. The 1 stat point can go into one of 4 categories, which are Strike, Agility, Focus, and Body. Strike is your ability to deal damage, Agility is your ability to avoid damage, Focus is your ability to regenerate energy (mana), and Body is your ability to regenerate HP. Obviously, your melee fighters should invest more in ability to deal DPS, tanks should focus on blocking damage, and mages need mana to cast their spells. The only problem is that YOU ARE NOT GIVEN PRECISE MEASUREMENTS FOR WHAT EACH STAT DOES; you are only given general details.

I don't want to know that investing a point here will "increase my chance to hit;" I want to know that each stat you put here will "increase your chance to hit by x%.

I don't want to know that x spell will "deal a large amount of ice damage;" I want to know that it "deals ice damage equal to 2.9 times your intelligence level, and this spell has a chance to stun enemies equal to .15 times your intelligence level. Upgrading this spell increases the chance to .18 times your intelligence level.

In case you haven't guessed by now, this is all a metaphor.

Whenever I ask for help on selecting parts, very commonly I am given imprecise details that are expected to heavily influence my decision. Like, I asked one board to compare x processor to y processor, and 95% of the replies I got told me GET THIS ONE BECAUSE IT HAS BLAH BLAH WHICH MAKES IT FASTER.

But I want to know HOW MUCH faster it makes it, under what conditions so everything can be fairly compared and I can weigh all my options on the same scale.

I am not asking to simplify things; actually I am asking quite the opposite. I want to know PRECISELY what each different value means, especially in processors, cuz that's what I'm shopping for right now. I don't want to hear that "higher bus speed makes your processor faster." I want to know how much faster it makes it and what all it depends on, like other parts of the CPU or other parts of my computer... things like that.

http://www.tomshardware.com/forum/247939-28-buyers-guid...

(This is the TL;DR paragraph)This article was very helpful in netting me the basic idea of it all, but it didn't give precise ideas until the chart at the very end, which I like very much, and I can get those precise values by just reading the specs on different items. What I want to know is what each value MEANS, and how they relate to each other. Don't tell me that "higher clock speed and higher bus speed make the CPU faster;" give me math. 2+2 = 4; 2+2=/= ~5.

Thanks.
August 21, 2009 5:50:24 AM

"What I want to know is what each value MEANS, and how they relate to each other." ... what values?

in one sentence... what do you want to know? ( no references to link of any sort pls)
m
0
l
August 21, 2009 5:56:43 AM

To put it the most simple wayI can, theres 2 things needed, the cpu, and the app. Cpus vary in speed, and cache, which basically holds info much like your ram does, but for the cpu, when its needed.
The apps sometimes need speed only, sometimes cache helps.
Theres HT or SMT, or basically, this is to work on varying parts of an app, instead of 1 part of it at a time, and literally speeds up everything
Always keep in mind, the app plays as big a role as the cpu, and is why , for some things, some cpus are great at, while others are great at other apps, it all comes down to design
Theres a few basics
Im sure therell be others explaining more
m
0
l
Related resources
August 21, 2009 5:59:29 AM

Was I really that unclear?

OK, a cpu has b clock speed, c front bus speed, d core count, e node and f wattage. I want to know how all these values interact with one another to give me g output.
m
0
l
August 21, 2009 6:17:58 AM

CPU... ok well, thats where the calculations are made... all the following things are the parts that make the cpu "fast"

Clock speed... its the cycles per second measured in hertz... thats how fast it can toggle between a 0 and a 1

Front side bus... its the bus that communicates with the northbridge... rated in MT/s or MHz... it plays a role in calculating the clockspeed, FSB times multiplyer equals clock speed

Core count... well that how many cores there are... they share from the same L2 cash , but have each their own L1... pretty much means they are working for the same cause (keeping you happy), but one could be thinking about which song is on your play list while the other is figuring how far shrapnel from a grenade flies in crysis.

Lithographic Node... pretty much the manufacturing technology... the smaller your 'wires' are in the core, the less heat they produce, the more of them you can fit on a die and still cool efficiently.

Wattage... i think you talking about the thermal design... pretty much the higher the wattage the bigger and more efficient heatsink you will need to decipate the heat


what else? more detail?
m
0
l
a b à CPUs
August 21, 2009 6:41:10 AM

trepanation said:
Was I really that unclear?

OK, a cpu has b clock speed, c front bus speed, d core count, e node and f wattage. I want to know how all these values interact with one another to give me g output.

The problem with this is that it isn't quite that simple. Some applications benefit more than others from various improvements. One application might not care at all about core count, but scale beautifully with cache quantity and frontside bus speed. A different app will scale beautifully with core clock and core count, but will be somewhat indifferent to FSB speed and cache. Complicating this even further, different CPU architectures behave differently as well, so even if the cache quantity, memory speed, node, clock speed, and core count are identical, an i7 and a Phenom II will perform completely differently. Basically, it really isn't all that simple, so qualitative descriptions of overall performance is the best you can do.

Now, you can get quantitative numbers for one specific application run on one specific system, but beyond that, it's pretty much hopeless.
m
0
l
a c 84 à CPUs
August 21, 2009 6:49:08 AM

I'm afraid it doesn't work that way. There is no way someone (that I know of at least) can sit down and in one post write out a formula for you to follow so that you know how much faster x will be over y. Things matter so much that it can't be done. A 1066MHz FSB P4 with 2MBs L2 cache will be slower then a 800MHz C2D with 1MB. Intel made so many changes to the Core2 line that any formula would fail right there. The same can be said about the process node. Intels and AMDs 90 and 65nm process gave each company very different results. This is because of the different materials AMD and Intel use.

I'm afaid that unless you are willing to put in a great amount of effort learning all this, you'll have to trust us. I wish there was some way that you could ask a question and be certain of the answer you get, but thats pretty much impossible. Either learn, or ask, I don't see any other option.
m
0
l
August 21, 2009 6:55:51 AM

And theres your real answers right above my words here
m
0
l
a c 84 à CPUs
August 21, 2009 7:23:57 AM

HardOCP has an article right now that might explain what CJ was talking about. They CF'd two 4770s and compared it to the 4890. Some games the CF setup won, while other times the improved memory of the 4890 allowed it to win. Applications would behave the same way.

I do urge the learning btw. The more you know, the less you can be lead away by fanboys and the dumb.
m
0
l
August 21, 2009 7:31:51 AM

Its not as complicated as the human body, but just like the body, theres lots of parts, and each part has its own function, and it interacts with other parts.
To get a better understanding, start with why, and not how. To be more precise, why does this cpu work with this app, instead of asking how it does, it may help
Usually asking why will get you more direct simpler answers than asking how
m
0
l
August 21, 2009 7:38:03 AM

I know the above statement sounds crazy, but think about it. If you ask why this cpu works with that app but that cpu doesnt do as well, youll start to break it all down, and itll start making sense
m
0
l
a b à CPUs
August 21, 2009 7:42:03 AM

Sorry, but those are the facts. CPU's by themselves are fairly complex beasts. Now take those millions of transistors, dump them into a motherboard, with a bunch of other hardware, AND have the whole lot interact with different software apps, not to mention operating systems, device drivers, etc. There are too many variables to reduce CPU performance to a simple equation with a clear answer.
m
0
l
a c 128 à CPUs
August 21, 2009 2:09:14 PM

All of the above. Which is why you can look at benchmarks all day, but other people will tell you the benchmarks are irrelevant, and cite examples. Unless your specific application has been benchmarked, you'll only get approximations, or a general "feel" for how things might improve if a certain variable is changed. Consider that two identical chips (same part number, same lot, etc) may not even overclock the same. I really agree with JDJ in that asking "Why" is probably the most useful thing you can do to figure something out.
m
0
l
August 22, 2009 7:10:17 PM

This post made me think about and realize how much I actually know.... neat.

But everyone here is right. There are many, many more variables with how a CPU/GPU/Computer will perform than you likely realize. Even after you understand how clockspeed influences things, and how bandwidth can influence a GPU, how cache, FSB, and memory all impact things, though sometimes are completely irrelevant (triple-channel memory brings lots of bandwidth, but generally does pretty much nothing), there's this whole other, very unpredictable layer most people call 'software.'

Individual programs of the same kind can vary significantly, most notably in whether or not they are multithreaded, and how far are they multithreaded. While a fourth core is a great thing to have, if your multithreaded app only uses two threads, the 3rd is largely irrelevant, and the 4th is probably completely irrelevant. So.... you *can't* get a 'p' processor is *exactly* R% faster than 'q' processor because of 'x' in 'y' situation. Too many variables.

Just realized... it's like chaos (I loves me chaos theory).
m
0
l
August 22, 2009 8:32:05 PM

It is all a matter of interactions. You either have to take the time to learn it or ask someone who has. Assuming you want to learn it here is an intro to Intel's processors.
http://www.tomshardware.com/reviews/prescott-reworked,9...
http://www.tomshardware.com/reviews/benchmarks-conroe-f...
http://www.tomshardware.com/reviews/Intel-i7-nehalem-cp...

After that, you can look into chipsets, video, amd, hdd, memory...
It really is rather interesting the only bad thing is once you do get it all figured out they will release the next generation of components and you start all over again.
m
0
l
a b à CPUs
August 22, 2009 10:51:28 PM

trepanation said:
Was I really that unclear?
Yes. You spent several paragraphs spouting off stuff about gaming that was completely superfluous to the actual question.
trepanation said:
OK, a cpu has b clock speed, c front bus speed, d core count, e node and f wattage. I want to know how all these values interact with one another to give me g output.
That's more like it. The problem is that there isn't just one "output". It's like asking how to crunch horsepower, cubic inches, tire size, wheelbase, and curb weight to give you a number that you can use to compare cars. The answer is: it depends on what you are looking for in a car.

In simple terms:
- clock speed controls how quickly a non-multithreaded CPU-bound task takes.
- front bus speed controls the maximum rate at which the CPU can move data to I/O devices and memory
- core count controls how many simultaneous threads or applications can run.
- node - ? not sure what you're referring to here
- wattage - in general you'll have more watts the more of the above you have, and this means higher temperatures and power consumption.

If you're looking for how all this relates to playing a game - well that depends completely on the game and how it's coded.
m
0
l
August 24, 2009 12:34:51 AM

Quote:
Yes. You spent several paragraphs spouting off stuff about gaming that was completely superfluous to the actual question.


It was entirely relevant and helped to place my question into perspective. Your lack of discerning is no concern of mine.

Nonetheless, thank you for your contribution to the discussion.
m
0
l
!