Getting new PC

What brand and latency???

1600mhz ram would be clocking in at a base rate of 200mhz and would need a bus multiplier of 8 to reach 1600mhz..
Peak transfer rate of about 12800MB/s

To work it out use this equation...
DDR3 SDRAM gives a transfer rate of (memory clock rate) × 4 (for bus clock multiplier) × 2 (for data rate) × 64 (number of bits transferred) / 8 (number of bits/byte).
(FROM WIKIPEDIA!!!!)

Edit..
Link for you to check..
http://en.wikipedia.org/wiki/DDR3_SDRAM

@occusoj
BTW i didn't mean multiple FSB's on one board i meant that there are many boards out there with varying FSB's some boards are 1333 some boards are 1600, some boards are settable from 1333 up to 3000 (OC'd ram only not the whole board!)

http://www.reddit.com/r/buildapc is a pretty good place to get some build advice. I find people don't always answer your questions directly, but if you read through the various threads on there and the stuff in the sidebar you should get a decent idea of what people are doing in general.

It would seem that currently the AMD FX6300 processor is one of the best deals for budget builds. Definitely want those 6 cores for games such as Planetary Annihilation or BF4. I personally have an FX8350 and a nVidia GTX 660Ti and am able to enjoy the game, although it does lag a bit once there are loads of units (but I'm lead to believe this is the case for everyone). The people over on reddit are very keen on the AMD video cards, as they tend to have better price/performance ratio. Some people seem to have problems with them, but my understanding is that the drivers are pretty good now. My friend built with an i5 3570k and AMD 7950 and we get very similar performance in games (i. e. we can play pretty much anything), he doesn't seem top have any problems with his AMD card.

Here in France in a price range similar to yours I advised a friend to get a FX6300 (he went with the slightly more powerful FX6350, but you pay more for a small performance gain) and r9 270x, he is able to run everything on High settings currently, for example BF4, which is a fairly resource hungry game, he doesn't play PA though.

Hope these pointers are useful to you. One thing you have to bear in mind is that there can be pretty big fanboy wars between AMD/Intel and AMD/nVidia, but ultimately you are better off getting whatever gives better price/performance for your budget (although I would advise checking reviews of your components before buying).

So even more people obviously copy/paste that, wow.

Thats from ancient times, look at whats written there.

But it hasnt to do so anymore. Its just not the case.

Next wrong statement in that post. My gosh.
Assuming its a classical system with CPU-FSB-NB-MC than you DONT want the bandwidth of the FSB to match the one of your RAM-MC.
Why?
Because the FSB is S H A R E D with a lot of other stuff, it is not exclusive to CPU-MC communication. Now theres PCI/AGP/... whatever talking simultaneously as the CPU communicates with the MC. So the link CPU to rest-of-world needs quite a bit MORE bandwidth than MC-RAM.

You stole even the calculation from there. Do your own ffs.
And its not even of any importance for a Haswell CPU.

You should have really read that stuff through, because later on that guy start on IMC (AMD had them already back then).
Here it is:

Intel made the switch to IMCs shortly after.

IMC only bypasses the need for the northbridge to redirect memory data to the cpu and vice versa, as there is now a direct link and a controller on the cpu..

It does not eliminate the need for having to match the mainboards FSB to accommodate for the FSB of the ram !!!
It does not eliminate the ram from using the systems bus speed to run.
The ram still uses the FSB speed hosted by the northbridge, it just doesn't need to send data through the bridge as there is now a direct link to the cpu...
THATS IT!!!!

Maybe i should just refer to it as BUS not FSB as the term FSB is out-dated since intel and amd renamed it..
It may make picky people happier .. lol

@occusoj
Would referring to it as just BUS be more relevant ?
Personally i still think getting a decent mainboard with a bus speed that can handle what your puttin on to the board ram and gpu wise is the best first choice when building a pc..
Does it sound better like that?

I tried finding refurbished units, but kept ending up being sent towards Alienware.
That CPU and Graphics card are coming to ~450USD even before the rest of the bits. For some reason, everyone likes to tack on 25% when selling to the UK.

I notice it lists the Ballistix RAM. I'd found that and been moving around numbers on sys3 to try and get the RAM done cheaper. Unfortunately their "Great" build, once you add DVD and OS would come out higher than on CCL. I've already bookmarked it for later.


Edit: Wha? Alerts said 3, not 3 pages. Incidentally, Max, you can edit posts rather than post twice in a row.

Alienware arent exactly the cheapest .
Refurbished, well, thats going to be a bit more expensive, I just bought straight from the previous user.

No because its still wrong.
First Intel/AMD didnt rename it, they redisgned it from the ground up.
And if were talking about QPI or HyperTransport, strictly speaking these arent busses. IMC-CPU most likely is some kind of ptp link too.

You dont "have to match" anything, and certainly not any form of bus speeds.

[Edit]Not clearly worded stuff removed[/Edit]

So you are right when saying that the RAM has to match what the IMC/CPU can do and make use of, thats perfectly fine and ensures you neither waste money nor sacrifice performance.

This of course assumes that no overclocking is to be done. So instead of wasting 50 bucks on that DDR3-2133 instead of 1600, invest that money on a decent GPU and youll see much more performance increase.

In fact Im going to test RAM speeds with PA now.

Neither is there a northbridge anymore as everything has been either moved into the CPU or combined into one external chipset nor any FSB nor does this "host" the reference clock.


EDIT:
So, the performance variation with RAM.
Since PA doesnt feature a repeatable benchmark, all I could do was to load the same system, measure loading times and fps of initial planet view with nothing built or done.

Running my modules wtih 1333 9-9-9-24 resulted in 1:04 mins to load the system and 65fps initially.
Running them as 800 9-9-9-24, which is extremly slow, resulted in 1:06 to load and 57 fps initially.

The initial fps arent THAT much of importance as there could have been a different viewing angle on the planet and thus different rendering going on. Even if we assume it was the same: Thats 14% performance loss during rendering and 3 for loading. In practical gaming I didnt notice any difference.

And another EDIT:
Running it 1600 9-9-9-24.
Loading time, 1:04 and 66-67 fps.

SSX - that's for You

Thanks to mrqasq for running it on a 670.
OpenGL is 78% as fast - so the NVidia stuff experiences roughly similiar losses than AMD does (72%). Nice to see and once again, thx for testing.

Thank you, can you make one more test for me please?

Can you enable "Threaded Optimization" in Nvidia Control Panel and test OpenGL renderer once again?
It's OpenGL-only feature so it's not affect D3D as long as I know.

That result is kind a strange isn't it? Although minimal fps is higher - overall score is lower!

Not really. Threading optimizations work differently for different OpenGL apps. It's main reason why this option not enabled by default.

Thank you for testing, really interesting results!