Where to start with mods and possibly OC

fish_86

Distinguished
Jan 21, 2010
98
0
18,640
Hi all. Just wondering where a good starting point would be for benchmarks and potential overclocking. Main thing I want to do is make the computer run the smoothest it can at an efficient speed. I have a new build and I seem to have a few hangups at times. I'd like to learn how to tweak it.

Benchmarks can tell you things about your system, but are there some that can recommend settings for a better build? There seems to be tons of benchmarking software out there that give lots of info on your comp, but what does all that info mean?

I have some basic knowledge, but once i'm in depth, I'm lost.
 
Well, first of all, what are your system specs?

As for benchmarks, well, there's plenty for graphics out there and a few combination cpu/gpu benchs. 3DMark Vantage and 3DMark 11 are both really good. They give you graphics and cpu scores, and you can definitely see the difference that overclocking makes.

Intel Burn Test or SuperPi are pure CPU computing stress test/benchmark programs.

Unigene Heaven is a great DX11 bench, specifically for tesslation.

Also a great program for monitoring GPUs is MSI Afterburner. I highly recommend that. For CPU, I like Real Temp and HWMonitor but there's plenty more.
 

fish_86

Distinguished
Jan 21, 2010
98
0
18,640
I think the thing I'm looking for most right now is how well the computer runs at different clock speeds with the ram. I've tried playing with different settings after looking at CPUZ but every time I mess with them the bootup give me an error and won't let me pass so I end up changing the settings back. I'm using a Gigabyte mobo 890FX.

Also I want to mess with the cpu in conjunction with the ram and get the settings right so everything runs smooth. possibly OC once I get familiar with things.
 


Well you should definitely start by reading at least a few different guides on overclocking your CPU. There's a few things that could cause the bootup error. Could be the RAM, could be not enough CPU voltage, could be the HT or NB are OCed more than they should...

As for the RAM, it should have a rated speed from the manufacturer. That's usually the fastest it'll run but sometimes you can go higher. Changing the FSB will change the RAM speed and as the RAM speed goes up you need to either increase timings and/or increase voltage.
 

fish_86

Distinguished
Jan 21, 2010
98
0
18,640
From what I've read before they say an Phenom X4 965 is basically a 955 OCed. Correct me if i'm wrong. Is RAM not the same way? Currently running a Athlon X4 640, and would like to get the 955 sometime.

You say ram comes at a max setting. If I have ram that is factory set to 9-9-9-24 is there a way to benchmark to see if it runs better at 8-9-8-24, how do I know. Isn't a computer like a car, if the timing is a little off you don't get the full power, effeciency, and life expectancy out of the machine? This is what I'd like to start learning. Set it up so i'm running in tune...
 
Timings affect latency rather than transfer speed. You can get the free AIDA64 and it will benchmark your RAM speeds and latency for you. Everest will do it as well.

I ran tests on my RAM. My RAM is rated 1600mhz CL8. I found that it actually runs slower at 1800mhz CL9, and even slower than that at 1400mhz CL7. I also found that I got a higher transfer speed with a higher CPU clock (I did 200x19, 200x20 CPU multipliers with the RAM at 1600 both times). For the latency, surprisingly, it was also the lowest at 1600 CL8, even lower than 1400 CL7 which you would think should be lower...

Now, I have seen another thread where the guy got gains with the faster the speed, so 1800 beat 1600, and he even did 2000mhz and it was fastest.

And yes a PC is kind of like that. I fooled around with running Intel Burn Test to get the highest possible GFlops, and I found even just tweaking voltages a tad made a small difference in it.

My absolute fastest speed is in my sig (CPUZ Validator) of 206 base clock, 4.13ghz. Your AMD CPU doesn't use baseclock, but with mine, my RAM speed is a multiplier of the base clock so I set it to 8x (10, 8, 6 being only options) for 1648mhz CL8 @1.65V. This provided the absolute fastest overall speed for the RAM with highest transfer rates and lowest latency. I had to exceed 1.65V though, I think I finally got it stable around 1.68V DRAM...

So anyway RAM, due to the latencies, has a specific "highest speed" possible per latency, if the speed gets too high for the latency then it will be unstable. In my case, at CL7 my highest speed was 1420mhz or so (1.65V), at CL8 it's about 200mhz faster 1620mhz or so, and again at CL9 it's a tiny bit over 1800mhz. You can get a bit more speed out of the same latency with more voltage. I didn't try, but theoretically I might be able to get 1700mhz CL8 at 1.75V or something like that although of course 1.65V is the "max".
 

fish_86

Distinguished
Jan 21, 2010
98
0
18,640
Thanks for telling about your testing and results with comparisons. I'll have to read a little closer and try some testing myself. When you say "stable"...what is "not stable"?
 


Fails a stress test. Specifically, for examples, Intel Burn Test is very sensitive to stability of both CPU and RAM. I gives you a readout in GFlops for computing power but also what the remainder is. If you can pass 15 high stress tests then you're ok, but if during those tests a remainder comes out wrong then it stops and gives you a warning. The other test people generally run is Prime95 which you can do CPU or CPU and RAM testing. It's not quite as sensitive so usually it's ran overnight for 8-10 (some people even 24) hours and if any of the threads stop/get an error, then it's not stable.
 

fish_86

Distinguished
Jan 21, 2010
98
0
18,640
I'm running windows 7 64 bit. I was looking at the benchmarkhq site and it shows that Intel Burn Test doesn't works for windows 7, only shows XP and Vista. Can I still run it or is there something else out there other then prime95.

Another issue I had awhile back. I tried running memtest86 and it wouldn't work. I think its because it said I had a cache problem. Was wondering if its because my athlon x4 doesn't have an L3 cache?
 

user 18

Distinguished
Apr 8, 2010
363
0
18,810
I've run memtest86 just fine on the Gigabyte GA-890FXA-UD5 with an Athlon II 620 (slightly lower clock speed than the 640, no other real difference). Had no problems with cache or any other settings - except for the RAM being bad :)