P5K-Dlx: CPU Voltage Reference & NB Voltage Reference effects on OC-ability

[XtremeTiramisu]

I've a specific question for all P5K-Dlx OC'ers:

I'm at the borderline of being stable - FSB470*8 @ 1.465v(Bios Set)
I left CPU Voltage Reference & NB Voltage Reference both at AUTO and CPU PLL voltage was at 1.5v.
Orthos stress cpu small fft failed after 7hrs and 30mins (Common rounding error, expected 0.5 and only 0.4 from one core.)

Right now, I've changed the CPU Voltage Reference & NB Voltage Reference both at x0.61 and upped the CPU PLL voltage from 1.5 to 1.6v. I'm trying to find out if I can get my current OC configuration stable without increasing the Vcore one notch up from the current 1.465v set in BIOS.

As you can see from the following detail configuration spec. My CPU PLL voltage is set at 1.6v (Options: 1.5, 1.6, 1.7, 1.8) and CPU reference voltage is set at x0.61 (Options: x0.67, x0.61, x0.57, x0.53). I'm wondering if the cpu reference voltage should be set to corresponding CPU PLL voltage? (ie. CPU PLL Voltage 1.5v = Auto/x0.67, 1.6v = x0.61, and so on. I've being gather infos from various members on various forums. I normally see CPU PLL voltage set at 1.5v and CPU reference voltage set at AUTO or x0.61 at higher FSB (470+)

My SB voltage is set at 1.2v and the corresponding NB reference voltage is set at x0.61. P5K-Dlx owners has found there's a correlation in SB voltage/NB reference voltage with Ram's overclocking-ability. My Rams seem to be stable under memtest for a few hours at current configuration (please see table below).

I'd think it's just the CPU compartment that's not 100% fully stable. Orthos small fft is currently running in green at 3 hrs 44 mins as of now after I've adjusted the CPU reference voltage/NB reference voltage. I hope this will stabilize my OC with the rest of my other configurations(please see table below) for 8hrs +

Please share your suggestions and give out some pointers that would aid in my current situation

JumperFree Configuration Settings
AI Overclocking: Manual
CPU Ratio Control: Manual
- Ratio CMOS Setting: 8
FSB Strap to Northbridge: 266
FSB Frequency: 470
PCI-E Frequency: 103
DRAM Frequency: DDR2-1175
DRAM Command Rate: 2T
DRAM Timing Control: Manual
CAS# Latency: 5
RAS# to CAS# Delay: 5
RAS# Precharge: 4
RAS# Activate to Precharge: 9
TWR: 3
TRFC: 25
TWTR: 3
TRRD: 3
TRTP: 6
DRAM Static Read Control: ENABLE

Transaction Booster: AUTO
Boost Level: (only displays when Transaction Booster is set to Disable or Enabled)
Clock Over-Charging Mode: AUTO

CPU Spread Spectrum: Disabled
PCIE Spread Spectrum: Disabled

CPU Voltage: 1.4650
CPU Voltage Reference: x0.61
CPU Voltage Damper: Enabled
CPU PLL Voltage: 1.6
DRAM Voltage: 2.2
FSB Termination Voltage: 1.40
North Bridge Voltage: 1.55
North Bridge Voltage Reference: x0.61
South Bridge Voltage: 1.2

Advance CPU Settings
CPU Ratio Control: Manual
- Ratio CMOS Setting: 8
C1E Suppport: Disabaled
Max CPUID Value Limit: Disabaled
Vanderpool Technology: Disabaled
CPU TM Function: Disabaled
Execute Disable Bit: Enabled
PECI: Enabled

USB Configuration
USB Functions: Enabled
Legacy USB Support: Disabled (need to enable it to detect USB flash drives)

[Speederlander]

DRAM Static Read Control: DISABLE

[davexl]

I agree with the advice above about disabling DRAM Static Read Control, but I admit I have no clue what the voltage reference settings do.

I presumed that the higher the number, the higher the voltage, so I have not tried anything other than Auto, or x0.67.

Why would you want to go from x0.67 to x0.61 for example? Assuming good airflow and cooling, aren't we trying to pump this up a bit?

If it is the opposite of what I think, and x0.61 is more voltage than x0.67, then I have a lot of potential to explore!

[XtremeTiramisu]

As for me, if I set DRAM Static Read Control and Transaction booster BOTH to enabled, the motherboard will boot but NO post and screen.

I find there are only two options:

1. Dram Static Read Control set to Enable OR Auto
Transaction Booster set to AUTO

2. Dram Static Read Control set to Dissable
Transaction Booster set to Enable 0 or 1

Either options will boot and post fine, np.


Performance wise, I'm still uncertain which configuration is the best.
We all have seen EVA2000 excellent personal findings and contributions.
Now there is another new "Ultimate Overclocking Guide" done by X-Bit Lab with the new official BIOS 0404. The guys over there have some very interesting and informative results to disprove the "FSB Strap Change" in our mobo. Apparently, FSB Strap Change is solely upon to the user and not the mobo itself.

please have a look at this article before you read my next post
http://www.xbitlabs.com/articles/mai...overclock.html

[XtremeTiramisu]

1.

The obtained results are given in the table below. The CPU in these tests worked at 3.2GHz. The memory frequency was set as close to 1066MHz as possible. The timings were locked at 4-4-4-12-8-30-8-8-8. Transaction Booster was set to Auto, and DRAM Static Read Control – to Enabled.

2.

In order to find out what the situation looks like on Asus P5K Deluxe mainboard, we tested the memory subsystem performance with the FSB frequency increment of 10MHz. the memory frequency in this session was set with the 4:5 FSB:Mem divider (this is the minimal divider available in the entire FSB frequency range). The CPU clock frequency multiplier was set to 8x. The memory timings were set manually at 5-5-5-15-11-42-10-10-10, so that the memory could work stably at maximum frequencies. Transaction Booster was set to Auto, DRAM Static Read Control – to Disabled.

3.

In conclusion to our extensive discussion I would like to say a few words to those Asus P5K Deluxe owners who for some reason believe that it changes FSB Strap at 500MHz. In reality there is no Strap changing whatsoever and cannot be. All tests that allegedly prove this statement haven’t been performed correctly. The thing is that if the DRAM Static Read Control parameter is set to Auto, the mainboard changes its value at this frequency. That is why the performance drops, which is erroneously considered to be caused by Strap changing. To eliminate this unpleasant effect, you should set DRAM Static Read Control to Enabled in the BIOS Setup.

*Question: Performance wise, should we set AUTO, Enable or Dissable for the DRAM Static Read Control. There's no doubt that with Transaction booster enabled 0 or 1, there is a significant performance increase. However, from my personal findings and with some of others, we'll have a problem upon booting with no post and screen when both DRAM Static Read Control set to Enable or Auto and Transaction Booster set to Enable 0 or 1. Only by dissabling the DRAM Static Read Control will remedy the problem.

[XtremeTiramisu]

not reallly sure it's slightly different for each cpu i have used but testing cpu and nb ref volts, cpu pll volts, nb volts are key to squeezing out a bit more FSB.

for i find it comes down to these 2 settings at borderline max fsb

CPU Voltage Reference: x0.61
CPU PLL Voltage: 1.6

try x0.63 and 1.7 respectively

This is the response I got from EVA2000 over at I4Memory

[XtremeTiramisu]

Looks like it's stable after I changed CPU/NB reference volt config (please see the table in the first post for all detail configs)

[newls1]

Glad to see you got it stable! Nice turn out....

[Pretender]

XtremeTiramisu, try the following, instead of FSB Strap to Northbridge: 266 leave it in auto and use the same speed in memories. I have better performance in everest memory test(all of them) when using auto and 4:5 that when using 266 and 4:5.
Try and let us see your results.

[davexl]

Just confirming your final settings - so you ended up at x0.63 and 1.7?

Higher ref volts was better?

[XtremeTiramisu]

Glad to see you got it stable! Nice turn out....

I'm not having good luck getting Orthos Blend stable for the past couple of days, sometimes it will fail within the first hour and most of the time it'll even run up to 5~8 hours before it started failing.

I'm starting to suspect it's the Vcore that's still not enough all along, 1.465v Bios set.
Weird, I even had a screenshot of ~10 hours Orthos small fft stable at 1.6 CPU PLL and x0.61 CPU reference voltage in my post above.

Now I'm back running orthos small fft once again and if it's stable for 10+ hours, I guess it could be due to slightly too tight timings/sub-timings with the given voltage at 2.2/2.25v (5-5-4-7/8/9-3-25/30-3-3-5/6/8)

It has being taking me days to find the sweetspot at DDR2 1175 5-5-4-x-3-x-3-3-x and 2.2/2.25v...

XtremeTiramisu, try the following, instead of FSB Strap to Northbridge: 266 leave it in auto and use the same speed in memories. I have better performance in everest memory test(all of them) when using auto and 4:5 that when using 266 and 4:5.
Try and let us see your results.

Thanks, NB FSB Strap has being changed from 266 to Auto now and using the same 5:4 divider @ 1175 DDR2.

Just confirming your final settings - so you ended up at x0.63 and 1.7?

Higher ref volts was better?

This is the exact same question I had for EVA2000 over at I4Memory. He believe that higher number = higher settings.

Therefore, x0.63 CPU Ref Vol & 1.7v CPU PLL > x0.61 CPU Ref Vol & 1.6 CPU PLL. However, this is what he believes from his experience. Only we can conclude whether they make a difference in the OC-ability in our own configuration

[Nosfer@tu]

Informative thread

[impar]

Bump!

So, anyone knows for a fact what the CPU voltage reference does?

The explanation I found was:
The CPU Reference Voltage option configures the CPU Vtt voltage via preset ratios.

[vmsein]

I'm really glad I stumbled onto this thread. I've been curious myself as to what cpu pll voltage does in regards to stability when overclocking with a high FSB. Your settings, xtreme tiramisu, have helped me stabilize my ASUS
P5K3 Dlx running a QX6850 at 418x9. I know it's a different motherboard because of DDR3 support, but most of the settings under the jumper free configuration are the same. Thanks very much for sharing your settings