Newest Overclock - W/ screenshots - Suggestions?

Greetings Everyone,

I have the current setup, and trying to ensure stability. Right now, its doing Sisoft tests and super pi.



Right now my volts according to SmartGuardian is seen in pic, is there anything in there, that you might think would cause instability? Under pressure?

Anyone else using this ram? Maybe some suggestions? Anyone have experience with Powerstream psu? is it bad if the light in the back is red? or is that alright still?

I know you can go even further, but want to make sure.

Everything seems ok - the OC is more than decent, and the vCore is at a safe level.

Suggestions:

Superpi 1M isn't really a thorough stability test - I'd try at least one 32M run.

Prime95 is a renowned stability test that especially stresses the CPU. Should run for many hours to ensure total stability.

Your memory timings are a loose. Winbond BH/CH based RAM runs at tight timings - you don't really gain much clock frequency from loosening the latencies. I recommend you use active RAM cooling (e.g. 120mm fan). Remember that your sticks are rated @ 2-2-2-x @ 250 MHz.

Did you test how high you could go with your memoryfrequency if you use a slower timing (CL3, CR2) and NO divider (1:1) ?

> you don't really gain much clock frequency from loosening the latencies
But often the gain in performance is bigger with this extra-frequency.

His memory should do 250 MHz at 2-2-2-5. Whilst he may get, say... 260 MHz at 2.5-4-4-10, the tighter timings more than make up for the ten extra MHz. With Winbond based RAM, it's very rare that the extra frequency you get with loose timings gives better performance than slower speed but tighter timings.

> His memory should do 250 MHz at 2-2-2-5
In this case I'm pretty sure that the system would be faster without a memorydivider. At least real1st should run some benchmarks (other than Sandra) to compare this configuration with his current one.


> the tighter timings more than make up for the ten extra MHz
Can you prove this?

I've read articles about what's more important, frequency or timing, and better timing usually resulted in about 5% better performance (of course it depends on the test and on the timings you compared. I think the 5% were an optimal CL2 compared with a medium CL2.5 timing). But I'm talking about the performance of the whole system, not only the memory, since the CPU also benetifts from a higher clock.

I'd suggest to simply try it out (isn't this the fun-part anyway? ). Should he gain only 5-10 Mhz, then probably you're right. But maybe he gets more. It depends on his RAM.

In very few cases, Winbond BH/CH RAM run at anything but 2-2-2-x - those RAM chips are customized for high voltage and tight timings. 2-2-2-x @ 250 will certainly yield better overall performance than 2.5-4-4-x @ 260.

Dividers don't hurt the performance of A64-platforms. As long as the memory clock is the same, a bigger gap between CPU and RAM clock (high CPU-clock and use of divider) actually improves memory efficiency because the CPU <-> IMC-communication is faster.

> In very few cases, Winbond BH/CH RAM run at anything but 2-2-2-x - those RAM chips are customized for high voltage and tight timings. 2-2-2-x @ 250 will certainly yield better overall performance than 2.5-4-4-x @ 260.

I don't have these RAM, so I can't say anything about them. If they behave as you say then you're right. I still suggest to try it out

> Dividers don't hurt the performance of A64-platforms

I know, but that's not what I'm talking about.

My suggestion (to try 1:1) is based on two thing:
1.) the assumption that the system-performance is bigger with CPU250/RAM250 than with CPU270/RAM225. The problem here is, what benchmark you want to trust. (I know that Sandra will prefer CPU270). And maybe 250 will not be the limit.

2.) In my case I was able to clock the memory _much_ higher with 1:1 than with any other divider. I have no idea why and I have never read anything about it. Maybe you can explain this strange behavior: http://www.ocforums.com/showthread.php?t=449990

Nice OC. There is a typo in your sig btw, 1.7ghz X2 eh?

real1st said:

Anyone have experience with Powerstream psu? is it bad if the light in the back is red? or is that alright still?

Click to expand...

Correct me if im wrong but I dont think that the red light is a good thing. Means your lines are either to high or to low, might want to look into that.

go to insanetek.com ... they compare BH5@ like 500 mhz and tight timings with a tccd@ 600 mhz with loose timings(or something like that)...bh5 won some and tccd won some, in general they concluded the divider had no major effect at all.

i think its ocz4000 review section (plantium or gold) i can't remember

- Update

Alright, odd situation.

I ran PRIME95 / Small FTT on FIRST Core, 6 hours, no problems.

I ran PRIME95 / Memory Focus, 6 hours, no problems.

I ran PRIME 95 / Small FTT on SECOND, 2 hours and it err'd out.

I ran memory tests/sisoft no problems.

I have tried upping the LDT and NB voltages to compensate for this, hoping it will fix the instability for the second core.

Anyone ever encounter this, where the first core is stable, but the second not?

real1st

Did you run both simultaenously?