OC'ing My PII X3 720BE Again......

jedihobbit · As of 02/01/11

Mid December, 2009

I have a water cooled mATX system (DreamCatcher v3.5) that Iâ€™ve gotten the itch to â€œplayâ€ with again. So lets get started (warning I tend to ramble!)â€¦â€¦â€¦â€¦.

The hardware set up is as follows:

CASE: SilverStone SOG1
PSU: Enermax MODU82+
MOBO: Biostar TA760G AM2+
CPU: AMD PII X3 720BE
HSF: Cuplex XT di
GPU: Evga 8800GTS w/AquagraFX 8800
MEMORY: TWIN2X2048-8500C5DF
Primary HDD: Seagate 7200.10 ST380815AS 80GB
Secondary HDDs: 2 x WD5000KS-00MNB0 400GB
Optical Drive 1: Sony NEC Optiarc AD-7200S

The following has taken place over the last several days and have it here to kind of show what is or isnâ€™t happening.

My initial attempt turned into a triple bust........couldn't find any notes, system wouldnâ€™t work at any higher FSB (COMOS is your fiend), and then it to do anything at it prior settings. Bang Head

Ended up running at 17.5 x 200 = 3.5GHz where as before I had Primed it at 3.55GHz.

Here is my basic setup and the items marked with ** are open to suggestions and questions for an increase of performance.

With the Biostar most of the oc changes were/are done in the T-Series subheading of the bios. The only other things I changed are:

1. Under Advanced PowerNow (Cool N Quiet) is disabled

2. Other than that the other subheadings do not affect OCâ€™ing or have been left stock.

Under the T-Series

1. Set to Manual

2. CPU/HT Clock = variable

3. CPU Config = Auto ** BTW what the H does this do anyway??

4. CPU Tuning
>> Custom P-States = Enabled
>>>> Core FID = Unlocked
>>>> Core DID = Divided by 1 (stock)**
>>>> Core VID = 1.3000V (stock)**
>>>> NB FID = 1600MHz (changeable)
>> HT Link Speed = 1.6GHz (changeable)
>> HT Link Width = 16 Bit**(Auto, 8, 16)
>> ACC = Disabled**
>> Core Leveling = Auto** BTW what the H does this do anyway??

5. Clock Control (stock)**

6. Voltage Configuration
>> CPU V +0.30 (+.10, .20, .30)
>> Memory Over Voltage = 2.15V (Factory)
>> Chipset Over Voltage = 1.25V (1.25, 1.35, 1.45, 1.55)**
>> HT Over Voltage = 1.30V (1.20, 1.30, 1.40, 1.50) **

7. DRAM Timing Configuration
>> Memclock Value = DDR2-800 (changeable)**
>> DRAM Timing Mode = Both (Auto, DCT1, DCT2, Both)**
>>>> CL = 5
>>>> TRP = 5
>>>> tRCD = 2-4
>>>> TRAS = 18
>>>> TRC = 24
>>>> tWR = 6
>>>> TRRD = 5
>>>> tWTR = 3
>>>> tRFC0, 1, 2, & 3 = 127.5ns

8. Memory Configuration = All stock

9. G. P. U. Phase Control = All Stock

Just a short note here that was discovered over the course of ocâ€™ing attemptsâ€¦Iâ€™m better off leaving DRAM mode in auto for some reason. Also there is a weird system issue that may be moboâ€¦â€¦â€¦.

While trying to overclock, and I get above a certain (havenâ€™t defined that yet) GHz or have a BSOD it seems that now when I try to reboot the monitor goes into kind of a sleep mode, hitting reset doesn't restart the system, nor does the power button turn it off. I have to kill the psu at its switch to kill the system. When I power up it up it goes through the boot up and runs, but again if I try to restart the system (even after setting the system to default) it will still "hang".

Christmas, 2009

Again Iâ€™m back at doing my shotgun approach over the last few days with lots of #@*%$#@. Notably is the fact I could no longer get the original stable 3.55GHz to work. Also along the way while in another ocâ€™ing thread at pcper, I was shown another quick and dirty stress test.......Hyper Pi, Super Pi on steroids. http://files.extremeoverclocking.com/file.php?f=211

This basically run Super Pi on all cores at 100%. My new toy!

However I had to re-flash the bios for the 5th time after a BSOD. It was at a max boot attempt and when it came back up it couldn't find my video drivers and had to reboot before they could be found again. Also another "sign" it is time to re-flash, besides not doing a reboot, is the bios menu would go to black and white as opposed to the stock "colors".

Somewhere Around December 29th

Another side note, I have re-flashed at least 10 times now as this keeps cropping up and have been real happy that the CMOS jumper is easy to get to.

Under S & Gs mode and max vcore (1.6v) I did manage to get the system to boot at 17 x 221 = 3.756GHz

I was able to get an 8M run in HPi at 17 x 219 =3.725GHz

Now back to getting my notes organized and posted while still trying to figure out the re-flashing issue.

Forget notes who needs notesâ€¦â€¦..man Iâ€™m the worst at recording what Iâ€™ve done! Smash

jedihobbit · As of 02/01/11

Yesterday

Okay nothing definate right now as I'll try to make a post later with the details of a 4 - 5 hour trip into flustration. Sad

Anyway quick question, while my inability to easily reboot to reset bios is really irritating, and there seems to be a wall at 3.85GHz :brick: :censored:, I've been having some wierd stuff with video. I have never had to do anything in that area when oc'ing a cpu before, however there is the random situation where after bootup the resolution has changed from max on the LG 19" I have to 800 x 600. If I go to the trouble of rebooting it will come up "normal". :shrug:

Ideas?

01:04 December 31, 2009

To day has been the day of the "Useless Overclocks"â€¦â€¦â€¦been attempting MAXIMUM OVERLOAD in the S & Gs category. Think most of my issues have been mobo limited. While my Biostar TA760G is an excellent soldier it is not special forces. Below is what was â€œaccomplishedâ€ and really would like some input as to what else might be adjusted. However I donâ€™t need to set records and have had my ocâ€™ing hunger with this board pretty much satisfied.

Iâ€™ve noted anything that is different from the prior settings as everything else was left alone or of minor consequence.

CPU/HT Clock = as noted in screenies or text

CPU Config = Auto ** AGAIN what the H does this do anyway??

CPU Tuning
>>>> Core VID = 1.3250V**
>>>> NB FID = 1600MHz
>> HT Link Speed = 1.6GHz (or on AUTO)
>> HT Link Width = 16 Bit**(Auto, 8, 16)
>> Core Leveling = Auto** AGAIN what the H does this do anyway??

Voltage Configuration
>> CPU V +0.30 (+.10, .20, .30)
>> Memory Over Voltage = 2.15V (Factory)
>> Chipset Over Voltage = 1.55V (1.25, 1.35, 1.45, 1.55)**
>> HT Over Voltage = 1.50V (1.20, 1.30, 1.40, 1.50) **

7. DRAM Timing Configuration
>> Memclock Value = DDR2-800 (changeable)**
>> DRAM Timing Mode = Auto (Auto, DCT1, DCT2, Both)**

Not sure if memory was an issue but it seems to be more consistent if left in Auto. I did on occasions changed the values but 800 was the best unless running only FSB = 200.

As mentioned prior the mobo seems to have some kind of â€œsettingâ€ that when you exceed certain values it will not allow one to reboot for bios changes, needing to kill the system with the PSU switch.

For all intents and purposes there is a wall at 3.8GHz (did it once) so even though I tried several FSB / multi combos I just wasnâ€™t all that excited about continuing. For these S & Gs I used HPi until that wouldnâ€™t work and then went to SPi just to say I tested it with something. Also I believe these were all done at maximum voltages in the bios (except mem that was factory 2.1V).

Vcore at +0.30 (1.58 â€“ 1.60v)
Chipset Over Voltage = 1.55V
HT Over Voltage = 1.50V

To start with Iâ€™ll show the â€œfastestâ€ that I achieved using SPi, 16.5 x 233 for 3.844GHz with an 8M run. If I tried to go any higher it would BSOD or not boot.

Now the fastest using HPi was a 1M run at 15 x 221 = 3.750GHz.

The best 8M run of HPi was with 17 x 217 for 3.689GHz

Because it tests all cores at 100% Hyper Pi will show some temperature increases for the cores for a quick ref. Here is an example during a run.

And it will show you the slowest and fastest cores. So here is my question concerning that what is an â€œacceptableâ€ difference before it may become a factor in overclocking?

Here are two examples, in the first which is a 8M run of HPi core #0 was 1.719 seconds faster than core #2. Then in the second (16M run) there is a whopping 15.072 second difference.

If this is critical then what is the solution?

Done for now and doing a prime run at 14 x 254 = 3.556GHz
Vcore at +0.30 (1.58 â€“ 1.60v)
Chipset Over Voltage = 1.45V
HT Over Voltage = 1.40V

It saddens me I'm having to use max vcore as I would have preferred 1.488v as my highest. However after 1.5 hours into the run core temps are between 46 - 48C (using HWMon). BTW does anyone know which might be the NB temp out of:

TMPIN 0 = 55 - 60C
TMPIN 1 = 35 - 38C
TMPIN 2 = 50 - 52C

jedihobbit · As of 02/01/11

10:00 AM 12/31/09

Oh well that was a bust againâ€¦â€¦..can not repeat my Prime run at 3.55GHz, Iâ€™ve dropped it back to

Done for now and doing a prime run at 17.5 x 2 = 3.500GHz
Vcore at +0.30 (1.58 â€“ 1.60v)
Chipset Over Voltage = 1.45V
HT Over Voltage = 1.40V
800 @ 5-5-5-18-24-2T @ 2.15v

Evening 12/31/09

Just had an epiphany for both of the overclockings that Iâ€™m doing and why I canâ€™t match the prior highest primed settingsâ€¦â€¦â€¦on the earlier settings I was not folding with the GPU but when I got around to doing a â€œfinalâ€ setting I did so as to see how that affected case and therefore core temps. Is that a viable reason?

Right now Iâ€™m at about the 7th hour of the prime run at 3.500GHz and the temps are running in the range of 38 â€“ 44C. It seems to have enough head room at this setting that I was able to take a screen shot, save it using Paint, and upload it to Photobucket.

BTW because of needing to do "other things" in my ususal impatience I stopped the Prime run after 8 hours and will just kept my fingers crossed! Laughing

jedihobbit · As of 02/01/11

Does anyone really know what ACC does??? Just for S & Gs I played with it to see if there was a remote chance of unlocking the 4th core. I had made the attempt in another system with a 720 and with it activated the system wouldn't even boot.

Auto, All, or Per Core didn't do jack but the system would boot (the TA760G has the 710). However I played with the percentage values "Per Core" just to see what they would do.........

I had noticed that on average there is a 15 second difference between the fastest and slowest core in several 16M runs of HPi. After several combinations leaving the fastest core at 0%, the middle one at +4%, and then cranking the slowest one to max (+12%) the average dropped to 3 seconds. Wonder if this really does anything?