First Attempt at Overclocking (Expert Advice Appreciated)

Evening, I recently upgraded my computer to a DFI LANPARTY SLI-DR (Yeah I know its built for higher end OC'ers), AMD Athlon 64 X2 +4800, BFG 7800GTX, and 1GB Corsair XMS. The +4800 is already fast, but I know it could flex more then it is. So I broke open CMOS, and gave it a go. I played around with Multipliers on my last computer setup, I knew they were one option in raising clock speeds. This time, I played around with the FSB Frequency, increasing the FSB by 10, I saw an increase in clock speed.

Now the +4800 runs at 2.4, just playing around with the FSB, I got to here:

That is the 1st core


2nd core

My question is, how do I make sure this is as stable as can be? Do I just run it over an extended usage period and see how well things go? Or is there a program I can run that can really work the CPU and get the kinks out?

And my next question, taking Temps into consideration (CPU idles at about 35oC and active at around 40+oC at this speed) how much farther do you experts out there think I can go? My goal is to at least make 2.6 (FX55) and eventually 2.8 (FX57) per core. What do you all think?

Run Prime95 for, maybe, 12 hours at least. See what kind of temperatures you get and see if it errors out (doesn't complete the benchmark, but finds an error).

Also, to fully test the thermal stability of this chip, run Prime95 on Small FFT (Options-> Torture Test and select small FFTs before you start). This will ensure that it puts the CPU at it's fullest temperature. If it doesn't restart, or error, then its stable. If it restarts, ease back and consider a better heatsink. If it errors, don't panic, its not a big deal. But you should look into better cooling.

I would also recomend increasing the FSB more. The corsair XMS you have should push far (I have some PC4400 that goes to like FSB 300 )and it wil give you more memory bandwidth which is great.


Not that I am a professional...

I would also recomend that you look into

Welcome to the forums.


SuperPI 32M is a relative "quick" way to test CPU and system speed and stability (post 235)
Stability testing using memtest, SuperPI (32M) and Prime95 (post 242)
Stability testing using Stress Prime 2004 (SP2004) (post 281)

Ok, I recently attempted to adjust the FSB to 220, but to no avail. It never made it to the Windows loading screen. Am I doing something wrong? I mean my CPU is cool enough, at this current OC of 5% its running at 31oC idle and 37+oC under heavy load, so its not like it is to hot. What must I do to increase the FSB to my desired range so as to attain my goal? I heard something about Voltage, and how the CPU wont work OC'ed properly if it doesnt have enough power to support the change, could that be affecting it? Or is it something with the dual-core setup? Anyone have any ideas how I can continue?

When you increase the speed or the FSB, generally the CPU needs more power to reach it. This option is available in most BIOSes. Now, don't increase it a whole lot. 1, maybe 2 steps, and make sure that your temps are still decent (they are pretty good right now actually). Don't worry too much about heat though. On most boards, especially if it is name brand, they generally have a hrdware CPU protection in place that shuts down the PC before the CPU gets too hot. And these CPUs also have a clock down thingy for when they get too hot.

Anyway, up the voltage a tad and it should give you more. If it still doesn't work, you might have to mess with multipliers or up the ram voltage. Make sure you have a good Power supply and whatnot too. And while you are at it, post some specs up here dude.

When the CPU is at default x12 multiplier and HTT (FSB in your text) at 220 MHz, CPU frequency = 12 x 220 = 2640 MHz. It should be able to boot into Windows at that frequency for a CPU rated 2400 MHz.

Usually between Windows boot frequency and Prime stable clock, there is around 200-250 MHz difference.

To diagnose the problem, try to isolate components one by one.

Make sure PCI frequency is fixed at 100 MHz.
HT multiplier set to x3 or AUTO. You may set it to x4 or AUTO when testing is done (if HTT < 250 MHz).

You may also try to set CPU multiplier to 10 or 11 instead of 12, so HTT can go higher to test the limit of the memory modules, without the constraint limited by the CPU alone.

With CPU multiplier to 10 or 11, the max memory clock can be set at 200 MHz running memory_bus and HTT at 1:1.

What is the rated specification of your memory module (Corsair XMS)?

What is the memory timing you set, such as tCAS, tRCD, tRP, tRAS, command rate. Typically something like 2.5-3-3-6 1T should be fine at 220 MHz at least.

What is the heat sink (stock heat sink?) and what is the power supply?

I received an error using PRIME95. The error was so small in terms of rounding error. It says I have damagd hardware. I assure you my hardware is fine. Should I be worried?

apopilot said:

I received an error using PRIME95. The error was so small in terms of rounding error. It says I have damagd hardware. I assure you my hardware is fine. Should I be worried?

Click to expand...

No, lots of people do not put much faith into prime95 anymore, maybe you should give this thread a look at before worrying about what prime has to say http://www.ocforums.com/showthread.php?t=399577

The specs are as follows:

CPU Fan: Zalman 7000B Cu LED
Powersupply: Thermaltake PurePower TWV500W (Modular)
RAM: XMS Corsair 1GB

I used CPU-Z to check the timings on my memory, I got 3-3-3-8 Its also running at 210Mhz Frequency.

http://www.mwave.com/mwave/viewspec.hmx?scriteria=BA21006

Above is the link to the exact part I ordered from Mwave, its rated at 400Mhz but running at 210? Is that why my times are so different?

Also, using the Pheonix-Award bios that came with the DFI LANPARTY SLI-DR (CMOS) there isnt any straight forward way on increasing or decreasing the Multiplier.

hitech, curious as to why its not stable in the slightest with a FSB set to a measly 220? There has to be something I am not doing that causes the instability.

Increasing the voltage is what can seriously cause harm to components, I want to make sure I know exactly what I am increasing the Volts to before I actually attempt to play around with it.

My goal is still to OC to 2.8Ghz per core.

For DFI NF4 board with X2, you may want to try later bios, at least the 0623 one. Look around for more info on that, 6.23-1 or 6.23-2 (for BH5/UTT) or 6.23-3 (for TCCD). You may want to try the 6.23-1 one without knowing what memory chips are used in the Corsair XMS.

Many BH5/UTT modules can handle 2-2-2-x 1T to ~250 MHz or so with 3.3 - 3.5 V. For TCCD based modules, many can do 2.5-3-3-x 1T to ~270-280 MHz with ~2.8 V.

From your Corsair XMS link, it says 400, what it means is DDR440 which is same as memory frequency 200 MHz. I am not sure what DRAM chip is used in your Corsair XMS. So I would say it should do at least 220 - 230 MHz 2.5-3-3-6 1T with enough voltage 2.8 - 3.0 V (guessing some what here since I have not looked into more details about your memory modules).

Assuming 220-230 MHz is the limit of your memory, if you want to get the CPU higher, you may want to set max memory clock to 180 MHz (or memory_HTT_ratio 9:10), this would give a memory_CPU_divider of 14.

E.g.

HTT 220 MHz
CPU_multiplier x12
CPU_frequency = 12 x 220 = 2640 MHz
memory_frequency = 2640 / 14 = 188 MHz

HTT 230 MHz
CPU_multiplier x12
CPU_frequency = 12 x 230 = 2760 MHz
memory_frequency = 2760 / 14 = 197 MHz

HTT 240 MHz
CPU_multiplier x12
CPU_frequency = 12 x 240 = 2880 MHz
memory_frequency = 2880 / 14 = 205 MHz

Try these settings first to see whether you can get the CPU to run at high frequency and start from there before going to improve memory.


Sometime, to get the CPU to run stably at frequency higher than rated (say 20%), slightly higher CPU VID is needed.

- CPU VID Startup Value ..... "StartUp"
- CPU VID Control ........... "AUTO"
- CPU VID Special Control ... "Above VID * 104%" or "Above VID * 110%"

I've read for amd 64 X2 you might possibly want to run 2 instances of prime 95.