I'm looking to buy a rig with under 800euros. Here's what I've come up with and pretty convinced that it would be epic.
Asus ATI Radeon HD 6950 DirectCU II 1024MB GDDR5
Intel Core i5-2500K 3.30GHz
Asus P67 Sabertooth Intel P67
Corsair Vengeance 4GB (2x2GB) DDR3 PC3-12800C8 1600MHz Dual Channel Kit
Antec High Current Gamer 750W Quad Rail
I need opinions and other options, if anyone has them. Two things, I wan't good overclocking features, and I don't want MSI shit etc.
1. Maybe, but it costs more money. Actually 10euros more on one website but out of stock No hurry though, I will get the 2gb if it's there in the moment of buying.
2. I will get other pair as soon as possible, is 8gb ok?
E: There's same memory but CL9 for lower price, I could probably buy 8gb of them and upgrade it to 16gb whenever I need to as it is 4gb in one, not in pairs.. Dunno what's the difference between CL9 and 10.
i currently have 8gb of ddr2 ram. i've never used more that 3gb of that since i bought it nearly 3 years ago. i have no problem playing the new games either. 4gb will be fine imo.
8 GB will be perfectly fine unless you're using very specific professional software. tl;dr for gaming it will do.
CL means CAS latency, think of it like ping - the lower it is, the better (but only in same frequency range. IE 1600 MHz CL9 is better than 1600MHz CL10 but worse than 1066 MHz CL7)
Windows shows wrong RAM usage figures if pagefile is enabled, even if it's not being used (no idea why that happens). Disable it and check again
i'll rephrase what i said. i have 8gb of ram, if i remove half of it, i will not see any difference in performance. i have friends who have very similar setups as my pc only with 4gb(of the same ram DDR2 OCZ gold). there's no performance difference.
imo ram numbers are bullshit. you could get SD ram with a CL as low as 2. DDR1 ram had low CL too. from what i read, the CL was the number of cycles per read/write so a CL2 is two cycles per one read/write and a CL9 is 9 cycles per read/write.
i remember having this discussion before about ddr2 ram over ddr1. yes the frequency is higher, but the CL is also higher. for the sake of the example (i can't be bothered getting the numbers again)the DDR3 you mentioned (1600Mhz CL10) is no faster than DDR2 800Mhz CL5 which is no faster than DDR1 400Mhz CL3.
Also, your real RAM usage was over 4 GB quite often. You just don't noticed it because, when pagefile is enabled, Windows shows RAM usage wrong - for example, when I had 4 GB and was playing GTA IV, with even 16MB pagefile Windows showed 2.8-3.0 GB RAM usage, while, with pagefile off, I was getting "low on memory" alerts and 3.5+GB used quite often.
CL factor determines CAS latency and this latency is the same, but that doesn't make CL10@1600 memory no faster than CL5@800. If you're reading more data, or know where it is long enough in advance, the bandwidth will depend more on speed than CL. Say you need to read 4 words/pages of memory (or whatever is the unit), the first one will take the same time, but the next 3 will come 2x quicker for 1600MHz.
As for OP, the build looks ok. 8Gb would be nice since it's cheap, but if it's for gaming, it won't matter as much since most games are 32-bit, thus limited to 2GB of RAM.
divide the frequency by the CL and you'll see that the read/write cycles are the same(or close to). but thanks for confirming it for me.
1600MHz=1600 cycles per second. / CL9 (9 cycles per read/write) = 177.7 read/writes per second.
800MHz=800 cycles per second. / CL5 (5 cycles per read/write) = 160 read/writes per second. (200 if the CL is 4)
400MHz=400cycles per second. / CL2 (2 cycles per read/write) = 200 read/writes per second.
if i'm wrong then i've fluked building computers for 10 years and my luck is bound to run out.
i'm currently running a duel core phenom (socket AM3)with 8gb of DDR2 ram (socket AM2+ board)with a GTX470 and there isn't a game on the planet (for the pc) that i can't play. even with my 8800GT i could play GTA4 with the graphics set at full/ultra.
i even did a large scale battle on Arma2 (75 units per side fighting)to test the performance. my pc still gave good frames.
I'm thinking about the future here. I will save money next time when upgrading a computer.
Why would I get a 600w one if I get 750w with just a tiny little bit of more money?
when it's comes to power, is there really such a thing as overkill? i would much rather have too much power than not enough. and the 750 would give you a nice bit of head room. and it wouldn't be running as hot as a 600.
Getting powersupply that's (even) 2x as powerful as your system maximally uses is, from some point of view, very practical - cause power supplies are most efficient at 40-60% load.
First, 1600MHz isn't 1600 cycles per second, it's 1,600,000 :P
This is not what CL means, you cannot compute bandwidth (reads/writes per second) in such a naive way:
1600 MHz = 0.625ns per cycle
The time to read amount of memory whose address is known @ CL9:
1 word = 9*0.625ns = 5.625ns (effectively 177M reads/s)
4 words = 9*0.625ns + 3*0.625ns = 7.5ns (effectively 533M reads/s)
16 words = 9*0.625ns + 15*0.625ns = 15ns (effectively 1067M reads/s)
Even though it should be slower by your logic, 1600@CL9 will generally be faster over 800@CL4, because you often read more than just one chunk.
Memory is very slow compared to CPU, so modern CPUs use various mechanisms to avoid memory bottlenecks like L1/L2/L3 caches, out-of-order execution, Intel HT also helps etc. So in real life faster memory won't matter so much in most cases. But there are such cases: take Llano APU, it's graphics part scales very well when you go from 1066MHz to 1866MHz RAM even if higher CL makes this CAS latency about the same, because it cares a lot more about mem bandwidth than CL.
No hurry though, I will get the 2gb if it's there in the moment of buying.
illepall :dunce: