AM2 Overclock / 780G board

[Lufusol]

I have a problem reaching a decent, bootable reference clock on my motherboard, and I've been following the isolation method prescribed here.

Here is my gear:
CPU: AMD Athlon X2 4850e, 1.25v, 2.5Ghz (+Arctic Cooling Freezer 64 Pro +AS5)
MOBO: Gigabyte GA-MA78GM-S2H rev. 1.0 (+60mm fan over the NB HS +AS Ceramique)
RAM: 4x1GB GeiL Ultra DDR2-800 (PC2-6400), 1.8-2.3v, 4-4-4-12
GFX: Sapphire Radeon HD4850 Toxic Edition, GPU: 675MHz GDDR3: 550MHz, Zalman VF900 + RAMsinks (all stock)

1)

I have followed the isolation method and dropped the CPU and RAM to minimum speed and loose timings, then increased the ref. clock 5 MHz at a time. If it doesn't boot AND run Orthos blend for atleast 2 minutes then I drop the HT multi or add NB voltage until stable before increasing the reference clock any further.

2)

At first, HT Multi drops had to occur around a low ref. clock of only 210-215MHz (x5 to x4), then at 250-255Mhz (x4 to x3 and +0.2v to the chipset), and then I got impatient, stopped doing it in the BIOS and started doing it in AMD OverDrive and hit the wall at 345MHz (tried x3 HT Multi, +0.3v[max] to the chipset, and still no stability). But then I started to discover that while I could go that high, certain lower frequencies wouldn't post without errors, seemingly at random.

3)

I found a setting in the BIOS under Advanced that says "NB Power Management" and "Dynamic clock gating of IOC/NT/MCU/CFG" which was set to Auto. I set it to Disabled.

4)

Now back to step 2, different results. I went back to upping the ref. clock 5MHz at a time as before, but this time using only the BIOS. With "NB Power Management" Disabled, HT Multi drops didn't need to occur until 250MHz (HT x5 Orthos Blend stable for 24h with no added voltage whatsoever). Rock solid with no voltage or HT multi drop up to 250MHz. But at 255MHz, POST errors occur. At 255MHz HT x4 POST errors still occur. At 255MHz HT x4 +0.1v, .2v, and .3v POST errors occur. Even another drop in HT multi to 3x with any voltage from stock to max, POST errors occur. I even tried jumping up to 275, 300, 315, 330 and various HT multis and voltages, but with NB Power Management disabled it simply won't boot above 250MHz without reporting that "POST errors have occured".

Before I do anything else, I'm posting this issue to see if anybody is familiar with overclocking this board, and what the heck am I doing wrong?


Sincerely,
Luf.

 

[Lufusol]

I forgot to mention the PSU but it shouldn't be anywhere near the problem. Here are my OC results (reference clock limited) at 250MHz FSB, no voltage:

CPU: 12.5x 250 = 3125 MHz @ 1.400v 24H Orthos Small FFT stable.
RAM: CPU /8 = 390 MHz (DDR2-780 effective) 4-4-4-12-18-2T @ 2.1v 24H Orthos Large FFT stable (damn no 1T or CPU/7=446 MHz at these timings due to mobo 2.1v limitation, ram can take up to 2.3v)
MOBO: HTT 250 MHz, HT Multi 4x, no voltage increase, 24H Orthos Blend stable.

I would like to try pushing my CPU but obviously the mobo is a problem. Any suggestions (other than "Replace with 790FX board," ha ha).

Thanks,

Luf

 

[cparker]

I've overclocked two of those boards. The first was with an le-1640 (2.6 ghz). It was real easy. The ram I have is rated at 667 (hp surplus 1gb sticks that have micron chips that really work well at 800) so I just kept the ram at auto and kept voltages at auto. Then I upped bus speed. I have it running now at 3.0 ghz at 230. I've had it higher, to around 3.1, but temps (using speedfan) seemed to be getting high when running prime 95, so I backed off and it's stable, normally temps are fine, and the pc flies for my usage, which isn't cpu intensive, just normal browsing/office/compiling kind of stuff on Vista 32 Home/premium. Cooling is stock in a qpack apevia case with stock case fans. PS: Antec Earthwatts 380. A very simple overclock. The pc is a pleasure, fast as a bullet with what I need to use it for and it's way improved over 2.6 ghz.

The second pc was more interesting. It was on a computer I just built for "grandpa". This time I had an le-1600 lying around (2.2 ghz chip) and I put it in a fancy Silverstone sg01. Same 2 gigs of ram with the hp surplus sticks. Vista Basic 64. Case came with 2 fans and one extra space for another fan. Originally I was going to just set it up for him to run at stock speed. It seemed to fly with no need for overclocking. I figured that it was either or both of the Basic vs the Home Premium with Aero, or maybe the 64 bit version was just faster. So all seemed good until it started dying on me after an hour or two. I tried lowering the ram speed and it made things better but now it would die after a few more hours. Not good. So I upped the ram voltage a notch, put in another fairly strong fan in the vacant position and now it ran without a problem. I then ran prime95 and it would freeze the pc in a minute or two. I figured it was a bad stick of ram. So I replaced a stick and reset everything to automatic and it worked fine. Then I started overclocking it, keeping ram and voltages at auto, like with the other build, and just kept upping the bus freq. I got it up to 260 or so, and it would run prime95 without a problem, but the temps were getting a bit high. So I backed it down to 250 which gives me 2.75 ghz speed. At that speed Vista 64 Basic is just amazingly fast with this chip.

As to the NB temps, well it's always pretty hot. With prime95 running it goes up 6-8 degrees or so I think based on what I believe the NB temp reading is on speedfan, which varies from 82 to 90 or so under full load. I have no idea as to the accuracy of the values. But with my (and "grandpas") usage, the nb temps will be maybe one degree higher than at stock. But the extra cooling fan on his pc makes a difference as the nb temp is a few degrees lower with the case cover on than with it off.

In your case, since you have 800 ram I would just lower it to 667 and keep the voltages the same and just raise the bus speed. I'd keep everthing else at auto and wouldn't fool with the nb settings. I would monitor temps with speedfan under normal conditions for your usage and also with stressor software.

I'm not an expert on any of this, by the way, but I have been very pleased with my results and while it would be great if the board didn't have that NB heating problem, it does seem to work fine at the OC levels I've used in these two builds.

Hope it helps.