GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

[Darkasantion]

Right, just discovered this forum about Gigabyte hardware, and figured you guys could me able to get more ouf of my system :D .

My current setup is:
Q9550
2x2GB OCZ 1066MHx @ 5-5-5-15
GA-X48-DQ6
HD4870(soon HD5870)
Enermax Galaxy DXX 850 Watt

My current settings are:
Robust Graphics Booster___________ [Auto]
CPU Clock Ratio ____________ [8]
Fine Clock Ratio ____________[0.5]
CPU Host Clock Control_ [Enabled]
CPU Host Frequency (MHz)__________ [455]
PCI Express Frequency (Mhz)_______ [102]
C.I.A. 2__________________________ [Disabled]
System Memory Multiplier (SPD)____ [2.40]
DRAM Timing Selectable_______ SPD __ [Manual]

Performance Enhance = [Standard]

CAS Latency Time______________ 5
Dram RAS# to CAS# Delay_______ 5
Dram RAS# Precharge Delay______ 5
Precharge Delay (tRAS)__________ 15

ACT to ACT Delay (tRRD)________ [Auto]
Rank Write to READ Delay________ [Auto]
Write to Precharge Delay_________ [Auto]
Refresh to ACT Delay______________ [Auto]
Read to Precharge Delay__________ [Auto]
Static tRead Value_______________ [Auto]
Static tRead Phase Adjust________ [Auto]

CPU/PCIEX Clock Driving Control [800mV]
CPU Clock Skew Control [normal]
GMCH Clock Skew Control [Normal]

System Voltage Control____ [Manual]
DDR2 OverVoltage Control__ [+0.400V]
PCI-E OverVoltage Control_ [Normal]
FSB OverVoltage Control___ [+0.35V]
(G)MCH OverVoltage Control [+0.275V]
CPU GTL1 Ref Voltage Ratio [+3.00%]
CPU GTL2 Ref Voltage Ratio [Normal]
CPU Voltage Control_______ [1.33125V]

Already toying with memory subtimings, will enter them into the BIOS once I've reached a stable OC on the Q9550, and will get EasyTune 6 as this gives access to more voltages and options.

But what can I do to further increase my overclock, thinking I'm running into a FSB-wall on the MB, but wondering if there are ways to solve this.

[profJim]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

Q9550 and EP45-UD3P overclocks up to 4.20GHz is doable, but at a price.
Overclocks up to 3.80 - 3.90GHz aren't very hard, but pushing higher takes a lot more effort.

Which stability tests have you run? How long were the tests?

If you want specific bios suggestions, we need the following:

1) List your system components in your signature.
--> User CP tab (upper left corner of web page)
--> select Edit Signature
--> enter your hardware components (brand, model, etc.)
--> include your bios version
--> preview
--> Save when complete
Be sure to include your bios version, motherboard version, power supply, cpu heatsink/fan setup and the model numbers of all components.

2) Link to your ram's full specifications.

3) Include a screenshot that displays:
Memset using the SPD and ">>" tabs
cpu-z CPU and Mainboard screens (several instances can be run at the same time)
EasyTune6 --> TUNER tab --> ADVANCE tab --> VOLTAGE tab (only the Voltage tab is needed)

Following these recommendations will eliminate possible errors and allow us to better help you.

[Salty]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

Easy tune 6 on the x48 does not display anything like the number of voltages the EP45 does.
This is all it shows for x48, so not much use for voltages:

[profJim]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

Salty,
You're right.
I forgot that the X48 didn't display very many settings.
Thanks.

[Darkasantion]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

1) Done, see below :)

2) I have this RAM:
OCZ Technology | Products | Memory | OCZ DDR2 PC2-8500 Platinum 4GB Edition Dual Channel

<table border="0" cellpadding="0" cellspacing="3"><tbody><tr><td colspan="2"></td></tr><tr><td colspan="2" bgcolor="#0070bc"></td></tr><tr><td valign="top"></td><td valign="top">1066MHz DDR2 </td> </tr><tr><td valign="top"></td><td valign="top">CL 5-5-5-18
(CAS-TRCD-TRP-TRAS) </td> </tr><tr><td valign="top"></td><td valign="top">Available in 2GB Modules and 4GB (2x2048MB) Dual Channel Kits </td> </tr><tr><td valign="top"></td><td valign="top">Unbuffered </td> </tr><tr><td valign="top"></td><td valign="top">Platinum-Mirrored XTC Heatspreader* </td> </tr><tr><td valign="top"></td><td valign="top">OCZ Lifetime Warranty </td> </tr><tr><td valign="top"></td><td valign="top">2.1 - 2.2V </td></tr></tbody></table>
3)



Hope this is enough info :) If you need more, say so and I'll do my best to provide it :D Thanks for replying sofar however

[Salty]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

Darkasantion,
I cant help you because you've gotten further fsb wise than I, and with a quad. I noticed your higher fsb voltage and use of gtlref. Have not grokked that yet.

There are a lot of posts here. So many that it is easy to drop out of sight. I'm interested in the answers u might get, so asking what will usually be asked out of general interest and perhaps for clues;

What programs are you using to test stability?
How long do you run them for?
What temperatures do you get?

[Darkasantion]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

1) I use IntelBurnTest for stability testing. First I do 1 or 2 runs to check if a new OC is a bit stable, and if it is stable, I increase the number to 10, 20 and 200 eventually to test for a long time.
After that I trow in GTA IV, who appears to be very sensitive in terms of stability, and if I pass that aswell, it's a stable OC.
2) My temps during IntelBurn test never go above 57 degree, so that's good, but no idea about the NB, it's reported to be cold at ~35 degree, but not sure, as the heatsink of the NB is quite warm during heavy workload.

[profJim]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

Running 200 IBT test loops is way too many.
If you test with Threads = 4 and using the highest available memory, you'll probably see your cpu core temperatures rise by 8 - 10 degrees, which would still be OK.

Performing a 12 - 24 hour stress test using Prime95, OCCT, or Orthos is better than running an excessively long IBT test.
I've had many cases where IBT passed 10 - 20 loops, only to have Prime95 or OCCT fail, sometimes in less than 30 minutes.

[Darkasantion]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

Hmm, Prime 95 marks it unstable at 3.85Ghz :( Any tips/tricks to make it stable, and maybe even further increase the FSB?

[Chike]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

What test did it fail (type and FFT size)?
You can try go down with the ratio and see if it's the FSB or CPU or it could be the memory
Was the FSB overvoltage needed so high? It seems a bit high.

[Darkasantion]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

I did run a test of Blend, assumed that would be the best test to test stability. FFT size was between 8K and 4096K.

Furthermore, I know it's the FSB beeing the limiting factor, did lower the multi already, together with the memory on a 2.00B setting, and there aswell I couldn't get much higher on FSB.

FSB overvoltage was that high as I was just testing it a bit, and as I didn't know exaclty what's causing the instability at 455+ FSB, I just added more voltage on it to ensure I'm not lacking there ;) .

[profJim]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

When Prime95 fails, the worker thread that had an error will display the last FFT size that failed. It helps if you tell us how long the test ran and the FFT size that failed. You can post a screen-shot like your post #5 settings, and also include the Prime95 screen. It helps if we can see the maximum monitored temperatures, especially the cpu core temps.

Generally smaller FFT sizes indicate a cpu related error and larger FFT errors point to memory related errors.
The Prime95 small FFT tests primarily stresses your cpu with little memory testing and generates more heat.
The large FFT stresses memory more than cpu.

[Darkasantion]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

Sorry for not replying, have been busy with some other stuff, but will get to OC-ing asap, so will rerun PRIME then, and post the results.

[Salty]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

I thought fsb overvoltage (vtt) should generally be between .o5 and .1 less than vcore?

[Chike]

Re: GA-X48-DQ6 Pushing the Q9550 beyond 4Ghz

In page 3 of Global, second paragraph it is explain why vtt should be less than vcore, as for how much less I don't know.