So I just completed designing my new PC and am now impatiently waiting for it to all get here in the mail so I can throw it together and build my first PC from scratch. While hunting around online doing research I came across a ton of information and numbers that were slightly different from device to device. I was just wondering what other folks opinions are on how big a performance difference you really get from jumping from one item to another.

For example, say a Corsair 670 vs Corsair 680. I mean sure, the core clock speed is a few hundred Mhz bigger, and sure it has an extra 100-200 CUDA cores but when you actually plug the bad boy in and start playing games with it is it really that noticeable or is it just a few extra FPS? What about a stick of RAM that's exactly the same as another, but one has a 7 latency instead of a 9? Sure anyone can do some research and find out what numbers being higher/lower then another means it's "better", but how much do you feel it really needs to increase to be a noticeable performance gain even ignoring if it's worth the price jump from one version to another. Personally I would focus on devices with a quieter decibel level when confronted with similar devices around the same price as my PC is right next to where I sleep so silence is very important to me. What are some other things you tend to take into consideration when confronted with devices that are pretty much the same? I would imagine brand loyalty and warranty would be the other big ones.

The faster the machine is, the longer you can wait before upgrading.

It depends. Some numbers make a difference while some don't when we're talking about small increments. We can't really generalise them all here.

Quality really matters when it's pc components so numbers shouldn't be the only focus. I'd rather get a corsair tx 750 - 850 instead of some no-name 900 or 1K PSU for instance.

Most of the technical specs are there for boasting. Numbers matter, but just like TV specs, console specs, camera specs it's just manufacturers trying to confuse the consumer with numbers. You have to determine which of them actually matter.

Clock speeds do matter. In a GPU a few hundred Mhz isn't going to mean the difference between 30 and 60fps, but it might give you like 3fps extra in some games, 10 in another.

I forget the term for it (diminishing returns?) but there's a certain point where more numbers and power aren't worth it. For instance, buying a GTX 690 is something you'd only do for bragging rights; by the time games come out that can utilize its full power, something else will have come out that's half the price and twice as powerful.

Of course those are rough estimates but you get the picture. If you've got the money for the newest and best processor, a few large SSD's, and two GTX 690's, then chances are you will never utilize all of that power because a year (or less) from now you'll be building a new computer with that year's newest and best parts. No one needs all of that for home computing, not even if you want to play Battlefield 3 across three monitors. or something crazy like that.

EDIT: And besides, technology seems to have hit a point where it's not getting tons and tons better. We're not making the same leap that Crysis took from Far Cry. There were only three years between those games, and it's been five since the original Crysis came out and, while I would say that game has been surpassed, it hasn't been surpassed by much. The argument could be made that consoles are holding PC's too far back, but I don't think that holds water when Battlefield 3 and Metro 2033 are both, on their max settings, light years ahead of what consoles can do.

Definitely don't concern yourself with memory latency numbers and crap like that unless you plan on becoming a hardcore memory overclocker, which, from my admittedly complete lack of experience OCing RAM, seems totally insane.

Just buy the good stuff that's reviewed well by reputable reviewers and is within your price range. That's about it.

@believer258: I don't know, man. Have you played Crysis recently? That game really doesn't look all that fantastic anymore, even on max settings, when compared to BF3 at max. It is still hard to run super smooth, though, and I think part of that is honestly just bad optimization, or a bad DX10 update or something. I agree though, the jump from Crysis to BF3 isn't as great as the leap from Far Cry to Crysis, but I do think this extremely long console generation does have something to do with that. When PS4/720 are finally out, I think we're gonna start seeing more impressive visuals on PC.

Definitely don't concern yourself with memory latency numbers and crap like that unless you plan on becoming a hardcore memory overclocker, which, from my admittedly complete lack of experience OCing RAM, seems totally insane.

Just buy the good stuff that's reviewed well by reputable reviewers and is within your price range. That's about it.

That is one thing I'm worried about and it seems like @PillClinton offered some good advice. This is the first time I heard/read anyone talk about latency and that other crap, which just seems like a way for companies to jack-up the price of their products.

@PillClinton said:

Definitely don't concern yourself with memory latency numbers and crap like that unless you plan on becoming a hardcore memory overclocker, which, from my admittedly complete lack of experience OCing RAM, seems totally insane.

Just buy the good stuff that's reviewed well by reputable reviewers and is within your price range. That's about it.

@believer258: I don't know, man. Have you played Crysis recently? That game really doesn't look all that fantastic anymore, even on max settings, when compared to BF3 at max. It is still hard to run super smooth, though, and I think part of that is honestly just bad optimization, or a bad DX10 update or something. I agree though, the jump from Crysis to BF3 isn't as great as the leap from Far Cry to Crysis, but I do think this extremely long console generation does have something to do with that. When PS4/720 are finally out, I think we're gonna start seeing more impressive visuals on PC.

Maybe, but I was mostly getting at the idea that we won't really see quite the same kinds of visual leaps we've seen between Half-Life 1 and Crysis 1. Think abut that - that's only NINE YEARS. That's nuts when you consider how slow pretty much every other technology, ever, has advanced - but since Crysis 1 it has slowed down a good bit and, honestly, Watch Dogs and Star Wars 1313 didn't really look that much better than Battlefield 3. I just think it's beginning to level off, that's all.

But maybe I'm completely wrong, new consoles will come out, and we'll go back to needing a thousand dollars or so to build a decent gaming computer instead of 800 for a great one.

Those numbers are useful, but really only if you are doing really intensive calculations such as 3D modelling or running a very busy server.

Here are some that are often touted but don't really matter: 

• Number of physical CPU cores (as long as you have at least three or four, speed per core beats number of cores)
• Speed of RAM (amount of RAM beats speed of RAM)
• Quality of the PSU beats Wattage of the PSU (I have a 460W PSU, my system doesn't need half of that under full stress)
• Version numbers of Drivers and Software. Dudes who install every shoddy alpha and beta they can find 'just cause' are nutjobs.

It's all about the spirit that lies inside the PC, not the brutal numbers. Like the Karate Kid.

@crusader8463: I know this is an old post, but never look at those numbers on the box, find a reliable site that benchmarks EVERYTHING (I use tomshardware.com, just because they benchmark anything worth benchmarking, and then some probably) and compare those numbers. If it performs better, it performs better. If it doesn't, it doesn't. Some numbers are a good place to start, like the number jigabertz a CPU runs at, but with video cards, it's generally more important to look at the base model, the cooling, the amount of memory. RAM is all GBs, and with HDDs it's pretty simple.

@MordeaniisChaos: Ya, that's what I generally do. I look at the numbers to get a general idea to see what's what but then I go to reviews to see how they stack up in benchmarks. I made this out of curiosity of what others do. I'm happy with my build that I got. I could have probably saved a few hundred by getting a slightly lesser model here and there that performed around the same, but I'm still happy to be getting what I'm getting.

regarding ram, todays ram is so fast that its not an issue, everything else in the computer is the bottleneck.

so unless you are running some special application that is notoriously heavy on the ram, you wont notice a difference, if you ran such an application, you'd already know this fact and have purchased apropriate ram.

when it comes to graphics cards, they are probably the biggest issue today, pretty much any single gpu card wont be enough if you are picky about the framerates.

but dont try to guesstimate how good they are by reading the specs, they are of course meaningful, but not if they are incomplete, and they are always incomplete.

instead, just head over to a site like techpowerup, and look at their graphs. BAM, any question you have about two cards performance, answered instantly.

You really have to research every component individually to work out exactly what kind of difference it will make, as opposed to guessing by the number on it.