AMD Athlon 2.8 64 X2 Overcklock limits

[QuietIce]

I agree. Better to save your money for the next system. But you've done a great job on this one ...

 

[trents]

Wow, that ram really would do DDR450 mhz! Corsair makes good stuff, though, so I shouldn't be surprised. My next piece of advice was going to be to lower the CPU multiplier very low and see how fast you could get the ram to go before encountering instability. That way you would know it wasn't the CPU speed causing instability. But you got it done yourself your way. Good job! You have a good motherboard and memory to base a better system on. Now you can work toward a newer CPU and a case with better air flow as you can afford it.

 

[VeRDe]

I'll give it a try soon!!

 

[trents]

I wasn't trying to discourage you from taking your ram past 400 yesterday but I just wanted you to focus on finding out what the CPU speed limit was first, a truly stable CPU speed limit that would pass a long Prime95 test. One of the basic principles of overclocking is to work with one variable at time and find its limit before you start changing other things. If you work with more than one variable at a time your are not sure which one caused the instability. Does that make sense?

 

[QuietIce]

Wow, that ram really would do DDR450 mhz! Corsair makes good stuff, though, so I shouldn't be surprised.

I agree, you shouldn't be surprised!

My XMS2's (DDR2-800 @ 4-4-4-12) will run clear up to 460 MHz and maybe just a hair more before I have to loosen the timings. It wasn't the RAM I was worried about running that fast but the rest of the system. Those sticks are great ...

 

[trents]

I was just concerned that VeRDe learn the discipline as an inexperienced overclocker. Had he not had that good ram there would have been more confusion because he would have been working with two unknowns. So I was advising him as if he were using no name ram that wouldn't overclock past its rated speed.

 

[QuietIce]

I understand and agree. I was advising the same direction and some caution when pushing the RAM that high. Like I said, I knew the RAM would do it but the system may not have been able to do it. Most good RAM will run 440, and most system will support that much OC, but once you're above that it's anybodies guess. In this case it was the CPU I was unsure of ...

 

[VeRDe]

I wasn't trying to discourage you from taking your ram past 400 yesterday but I just wanted you to focus on finding out what the CPU speed limit was first, a truly stable CPU speed limit that would pass a long Prime95 test. One of the basic principles of overclocking is to work with one variable at time and find its limit before you start changing other things. If you work with more than one variable at a time your are not sure which one caused the instability. Does that make sense?

Absolutely. I know that some times i am a little (or more than....little) bit rash!

i understand that if you overpass steps on overclocking, then you dont know the source of the problem and you have to guess..

it seems that RAM is very good and i'll try step by step to push it to see its limits.

i am very confident that i will pass 533MHz...

i'll see if the system would be stable with FSB274x11.5 and then i will try to see if RAM will pass it.

i'm sure that no over danger is with this freq becouse i had it as i said for 3 years like this. of course you can never be sure...

i hope that it will pass the Prim95 first the CPU and then the RAM.

45nm CPU's need less voltage and have less temperatures?

 

[VeRDe]

Great news. the following setup tested for 1 hour and 6minutes with Prime95.

Stable!

FSB: 300
Mltplr: 10.5
HT Link: 900MHz
vCore: 1.35
DRAM: 525MHz
RAM v: 2.0v
FSB: DRAM: CPU/6

it could be stable with less RAM voltage. i think i have to give it a try. do you agree?

could you explain me how HT and RAM multiplier affect DRAM Freq so i can calculate HT link and DRAM without CPU-Z?

 

[QuietIce]

The HT Link doesn't effect RAM frequency at all but you should keep the HT Link below 1000 MHz. HT Link = "FSB" x HT Link multiplier


For the RAM frequency calculation see post #5 in this thread: http://www.overclockers.com/forums/showthread.php?t=555058

 

[VeRDe]

Thanks!

So if i try it with 10 Multiplier RAM will have FSB 620MHz!!

hehe...

I'll try and if you hear about an explosion to Greece, that would be me!!

PS i have near me an extinguisher!!! (i'm kidding)

 

[VeRDe]

Update:

1) there is no way to work with FSB more that 300MHz so i cant go to Multiplier 10 x FSB 315 (3.15GHz)

2) There is no way RAM to work with 620MHz.... whatever voltage i give it!!

in first case windows go to blue screan saying to "set FSB settings to default".

in second case BIOS is alarming and setting FSB to default by itself.

So i'll stay as i am now with previous settings (CPU:3150MHz DRAM:1050/2MHz)

Not bad....

 

[trents]

Thanks!

So if i try it with 10 Multiplier RAM will have FSB 620MHz!!

hehe...

I'll try and if you hear about an explosion to Greece, that would be me!!

PS i have near me an extinguisher!!! (i'm kidding)

I got a good belly laugh out of this! Thanks, I needed the comic relief!

 

[VeRDe]

PassMark PT 7.0

My RAM Mark: 560
My CPU Mark: 1928

AMD Athlon 64 X2 Dual Core 5600+ Average Mark: 1536 (http://www.cpubenchmark.net/cpu.php?cpu=AMD+Athlon+64+X2+Dual+Core+5600+)

 

[trents]

No explosions from Greece, huh? What did you finally settle on for your overclock settings?

 

[VeRDe]

So i'll stay as i am now with previous settings (CPU:3150MHz DRAM:1050/2MHz)

Not bad....

FSB 300MHz
Multiplier 10,5
vCore 1,35v
HT Link 900MHz
DRAM Freq 525MHz
RAMv 2.0v

my last step is to lower RAMv as much as possible.

In which cases do someone have to increase North and South bridge voltages?

 

[VeRDe]

Very very strange things happen in my PC..

I decreased voltage of RAM till 1.8v which is default and the RAM remained stable with Prime95!!

plus that i ran Passmark PT 7.0 again and RAM scored better that before with 2.0v!!

RAM Score this time was 576!

 

[QuietIce]

It's good the RAM voltage is lower!


The test scores are still within the normal limits of error for benchmarks, which (IMO) is ±2%. 576/560 = 1.029 or ±1.5% ...

 

[VeRDe]

Ok nice then,

i would like to experiment with latencies of RAM.

but i want some advices.

i dont know in every +1+1+1+3(or +6) step i do, what i have to do with tRC? do it have to change?

my defaults are 5-5-5-15-19-2T

 

[trents]

If you want to experiment with your latencies just for the experience of it that would be good but realize if you lower your latencies you may have to lower your ram speed to remain stable and the net performance result may not be positive. If you had a CPU with an upwardly unlocked multiplier and didn't have to overclock with your ram you could realize some performance increase by lowering latencies without also lowering ram speed. The "memory" tab of CPU-z will give you the most important latencies to adjust. There may be many more choices in your bios but most of them can be left on auto as they have very little effect on performance. Match the latency header names in CPU-z the best you can with bios. Unfortunately, terminology is not consistent in the industry. After you change something and you are not sure if is what you wanted to change, go back and check what changed in CPU-z "memory" tab.