- Joined
- Oct 23, 2002
- Location
- Muscle Shoals, AL
System Specs
MB: Gigabyte EP43-DS3L : F4 bios
Memory: G.SKILL 4GB (2 x 2GB) PC2 8500
CPU: Intel Core 2 Duo E7200 Wolfdale
PSU: OCZ PowerStream 520ADJ - 520w
GPU: Saphire X1950XT
Drives: 2 WD HDs, 2 optical
First let me start out by saying that the last thing I overclocked was a mobile Barton, so I'm pretty out of touch with some of the newer settings. That being said here's what I've done so far.
First I disabled all the power saving options (the C* settings, speedstep, smart fan), disabled Extreme Memory Profile and set Performance Enhance to standard. Next I manually set the memory timings and voltage to [email protected]. Next I set the PCI-X frequency to 100mhz. Finally, I set the other misc bios features (boot options, etc).
Stock, everything was fine, Windows installed without a hitch. DualPrime ran for 12hrs.
On to the overclocking. First thing I did was lower the multi to 6 and look for the FSB limits. (Note all settings tried were with memory set to run @533 [1066DDR] or the closest setting not exceeding 1070DDR unless otherwise noted) 333FSB worked fine. Didn't bother with Priming since it's a supported FSB frequency. Tried 400 that booted into Windows fine. 425 wouldn't post.
I figured at that point that 400 would leave me a lot of overclocking room so I should focus on getting that stable, then see what the chip could do. First try with 400FSB and 6x multi and prime failed after 5-10 mins. Upped the Termination voltage from 1.2 to 1.3. Prime stable for 12hrs (no failure, I just didn't want to wait a full 24hrs).
At that point I started raising the multi to see what the chip could do. 7x was Prime stable for ~12hrs. No big surprise there. I then went for 7.5x. Prime insta-failed (within 1-2 min). I tried upping the voltage in the available increments until I had hit 1.2v. Prime is still insta-failing and I was starting to think I've got a dud.
After doing some reading I learned a lot of the voltages are "intertwined" for lack of a better word. So it was possible that with the added stress of the higher CPU clock would need additional voltage increases. Not really wanting to jump voltages up haphazardly, I decided to go with a 333FSB and take some of that out of the equation. [email protected] and Prime still insta-failed. Now I'm really starting to think I've got a dud chip. I know they can take more voltage than 1.2, but they shouldn't need it for 3ghz.
As a last test I dropped the memory to 400 (800DDR). Booted at 9x333, started Prime, no failing. Ran it for a little over 11hrs with no failing. Forgot to try and lover the core voltage, so this was @1.2v.
So now I'd like to be able to get the clocks higher while running the memory at its rated speed. I'm sure there's something I'm missing not being at all familiar with the chipset/bios settings.
Also there's a couple of settings I'm not sure about and some quirks with the board. First is
What does that do? Any effect on stability? I left it on auto, would setting it to fast help or should I just leave it alone.
Another oddity is that the memory divider doesn't actually show a divider. What I mean is that it shows options like 2.00A or 2.66C where A-D= 200, 266, 333, 400 (if I remember correctly). The engineers were kind enough to display the final memory speed, but I feel like an idiot just randomly selecting dividers until I find one that sets the memory to the desired speed. Is that normal or something Gigabyte came up with on their own.
Next, the voltages don't seem to be quite right with the board. For instance setting 1.1v for the voltage core gets a reading of 1.054 in both the bios (PC Health) and CPU-Z. The memory set at 2.1 gets a reading of 2.06 in the bios and Speedfan (I think at least, it reports a Vcore2 of 2.05 so I'm guessing it's reading the mem voltage). I should note that voltage rails are all steady under load.
I'm thinking there seems to be a consistent .05 difference in the voltages and I'm wondering if that applies across the board (NB, SB, etc). If so, it would just be a matter of upping everything a tick and go from there. Is there any kind of software that reports those voltages (I haven't tried the gigabyte software yet).
Finally what do you guys use for quick and dirty testing? In the past, I would use SuperPI, but I tried the one from Xtremesystems and never could get a valid checksum on any machine I have (confirmed 24hr Prime stable rigs).
One final note. Temps are fine (under 45 load with Tj.Max set to 95 in Coretemp and under 50 load in Speedfan).
MB: Gigabyte EP43-DS3L : F4 bios
Memory: G.SKILL 4GB (2 x 2GB) PC2 8500
CPU: Intel Core 2 Duo E7200 Wolfdale
PSU: OCZ PowerStream 520ADJ - 520w
GPU: Saphire X1950XT
Drives: 2 WD HDs, 2 optical
First let me start out by saying that the last thing I overclocked was a mobile Barton, so I'm pretty out of touch with some of the newer settings. That being said here's what I've done so far.
First I disabled all the power saving options (the C* settings, speedstep, smart fan), disabled Extreme Memory Profile and set Performance Enhance to standard. Next I manually set the memory timings and voltage to [email protected]. Next I set the PCI-X frequency to 100mhz. Finally, I set the other misc bios features (boot options, etc).
Stock, everything was fine, Windows installed without a hitch. DualPrime ran for 12hrs.
On to the overclocking. First thing I did was lower the multi to 6 and look for the FSB limits. (Note all settings tried were with memory set to run @533 [1066DDR] or the closest setting not exceeding 1070DDR unless otherwise noted) 333FSB worked fine. Didn't bother with Priming since it's a supported FSB frequency. Tried 400 that booted into Windows fine. 425 wouldn't post.
I figured at that point that 400 would leave me a lot of overclocking room so I should focus on getting that stable, then see what the chip could do. First try with 400FSB and 6x multi and prime failed after 5-10 mins. Upped the Termination voltage from 1.2 to 1.3. Prime stable for 12hrs (no failure, I just didn't want to wait a full 24hrs).
At that point I started raising the multi to see what the chip could do. 7x was Prime stable for ~12hrs. No big surprise there. I then went for 7.5x. Prime insta-failed (within 1-2 min). I tried upping the voltage in the available increments until I had hit 1.2v. Prime is still insta-failing and I was starting to think I've got a dud.
After doing some reading I learned a lot of the voltages are "intertwined" for lack of a better word. So it was possible that with the added stress of the higher CPU clock would need additional voltage increases. Not really wanting to jump voltages up haphazardly, I decided to go with a 333FSB and take some of that out of the equation. [email protected] and Prime still insta-failed. Now I'm really starting to think I've got a dud chip. I know they can take more voltage than 1.2, but they shouldn't need it for 3ghz.
As a last test I dropped the memory to 400 (800DDR). Booted at 9x333, started Prime, no failing. Ran it for a little over 11hrs with no failing. Forgot to try and lover the core voltage, so this was @1.2v.
So now I'd like to be able to get the clocks higher while running the memory at its rated speed. I'm sure there's something I'm missing not being at all familiar with the chipset/bios settings.
Also there's a couple of settings I'm not sure about and some quirks with the board. First is
What does that do? Any effect on stability? I left it on auto, would setting it to fast help or should I just leave it alone.
Another oddity is that the memory divider doesn't actually show a divider. What I mean is that it shows options like 2.00A or 2.66C where A-D= 200, 266, 333, 400 (if I remember correctly). The engineers were kind enough to display the final memory speed, but I feel like an idiot just randomly selecting dividers until I find one that sets the memory to the desired speed. Is that normal or something Gigabyte came up with on their own.
Next, the voltages don't seem to be quite right with the board. For instance setting 1.1v for the voltage core gets a reading of 1.054 in both the bios (PC Health) and CPU-Z. The memory set at 2.1 gets a reading of 2.06 in the bios and Speedfan (I think at least, it reports a Vcore2 of 2.05 so I'm guessing it's reading the mem voltage). I should note that voltage rails are all steady under load.
I'm thinking there seems to be a consistent .05 difference in the voltages and I'm wondering if that applies across the board (NB, SB, etc). If so, it would just be a matter of upping everything a tick and go from there. Is there any kind of software that reports those voltages (I haven't tried the gigabyte software yet).
Finally what do you guys use for quick and dirty testing? In the past, I would use SuperPI, but I tried the one from Xtremesystems and never could get a valid checksum on any machine I have (confirmed 24hr Prime stable rigs).
One final note. Temps are fine (under 45 load with Tj.Max set to 95 in Coretemp and under 50 load in Speedfan).
