Overclocking problems

[chinaman1472]

Running stock HSF at stock settings, no problems.

- RAM settings are at 5-5-5-15. Unfortunately, I'm still not to familiar with the BIOS yet, and can't get the 1:1 FSBRAM ratio; best I get is usually 4:5 according to CPU-Z.
- Just for precaution, I've added +0.1v to the RAM, and running 1.400v for the CPU; the RAM at default settings is running at 1.8v according to CPU-Z. Default for the CPU is 1.325v in the BIOS, 1.360v according to CPU-Z.

At 9x250, I ran Memtest86+ for an hour with no problems. Ran Prime95 for about 20 minutes no problems. Idle temps hang around 29-31°C, load temps were around 55-57°C. I'm pretty sure the RAM is okay.

At 9x275, I ran Prime95 for about 5 minutes and the errors started coming.

At 9x290, I can't get even get it to post .

What am I doing wrong?

EDIT: Seems I can boot 8x333 at 1:1 FSBRAM, so I'll start running Prime on these settings and see what happens.

 

[RavenGuard]

Sounds to me like you're not changing your RAM clock when you up the FSB. There should be a selector that lets you set the RAM multiplier, make sure you're keeping the RAM under the 800mhz it's rated for until you get the CPU running stable. At that point, you can relax the timings and up the clock, or simply get a 1:1 and optimize the timings.

 

[chinaman1472]

Yea, I was a bit confused in the BIOS with how to set the RAM multiplier; it's a little different my old DFi nForce4 board where I could just force it to 1:1. I had the RAM clock running *way* too fast lol. Think I had it at 420mhz at one point ^^;

8x333 right now for about 3 hours, no errors at 1:1 in Prime95. Load temps are hovering around 53-55°C from SpeedFan and CoreTemp. CPU-Z has 1.360v, CoreTemp has 1.325v, and I have it set at 1.400v in the BIOS.

I'll see if 9x333 can boot up, and I'll start Prime95 for the night.

 

[RavenGuard]

You're starting to get pretty hot there though, are you running the stock HSF? I feel safe up to about 60* with my Q6600 at 3.2Ghz

 

[chinaman1472]

Yea, stock HSF. I plan on getting a budget cooler (AF7P probably) in the next month or two when I scrap up some more cash.

9x333 was no good with my RAM at 667mhz. It posted, detected the hard drives, and it fails after that. I can't load myself into the BIOS without clearing the CMOS. Was hoping to get 3.0GHz, but the path doesn't look too good at the moment with the stock HSF. Enough tinkering for one day.

2.66GHz at 8x333 isn't too bad though. Lowered the BIOS voltage down to 1.375v, load temps are at 48-51°C now.

 

[nullpointerus]

You might need +0.1v on the FSB and MCH w/ 9 x 333.

Boosting RAM voltage may also help. My P35-DS3R runs the RAM at 1.8v fine when the CPU is at stock clocks, but, when overclocked, my RAM voltage takes a big dip from 1.78v to 1.75v sometimes, so I give it +0.1v to keep it always above 1.8v.

The 1:1 multiplier on these Gigabyte board is called 2.0, because that's what you multiply the FSB by to get the DDR2 RAM speed when it's running at 1:1.

Frankly, I prefer Gigabyte's multipliers because it makes calculations so straightforward.

To isolate CPU voltage as the problem (i.e. to determine if you have a poor chip), set everything back to stock, RAM to 1:1 (w/ loose timings), and keep increasing FSB until you can't boot anymore. Raise vcore until you can boot again, note the value on a chart, and then repeat the process. Don't worry about stability; the point here is merely to gather information.

You should end up with a voltage curve that increases exponentially as CPU clock rate increases. This indicates a processor limitation, and there's not much you can do to fight this.

On the other hand, if there's no curve, then some setting other than vcore is holding you back.

Don't go above 1.45v for temporary testing. Don't go above 1.4v long-term, especially with stock cooling.

To eliminate heat as a limiting factor, you could open your case and point a big desk fan right at the CPU/RAM/NB area. Even if the CPU is fine heat-wise, consider whether you have passive cooling on the NB and if so whether that area is getting adequate airflow.

 

[chinaman1472]

^ thanks for the pointers. I'll do some more testing tonight or tomorrow when I get some more free time.

8x333 is stable (Prime95 for 14h) with my BIOS voltage at 1.375v. I could probably go lower. I also keep a +0.1v on my RAM for precaution.

My room is pretty cold (thanks to hardwood floors); I consider 60°F+ warm in my room. But a fan on the side is something I'll consider. My video card has cooled about 4°C just because my side panels of the case are open.

 

[chinaman1472]

9x310 was unsuccessful at 1.400v. Is there any particular reason why? I'll be doing that test nullpointerus tomorrow night to see if I have a poor chip.

9x300 is successful so far. +0.1V to FSB, MCH, and RAM. BIOS voltage is at 1.38* I think, CPU-Z has it at 1.360v under load, temps don't exceed 55°C. Prime95 has been running for about 5 hours now.

 

[Tullphan]

Originally posted by: chinaman1472
Yea, stock HSF. I plan on getting a budget cooler (AF7P probably) in the next month or two when I scrap up some more cash.

Budget cooler that will outperform an AF7P.

 

[Duvie]

Cool exact setup I built for a friend....I spent the little extra dough (not much actually) and got some DDR2-800 stuff at the time (because of the low price upgrade)....The intent is when he needs the power I will come over and boost him up.

Still doing this with a stock cooler, huh? That bodes well. I was thinking of getting a cheaper artic 7 cooler or something. maybe 3ghz would be within reach...

 

[chinaman1472]

Yea, I saw that new Xigmate Cooler last night and I'm considering it ^^
The only real problem is size. My case is pretty small, so I'm not sure how well it would fit in my case. The AF7P is a bit smaller but even then, I'm thinking it'll be quite snug with that cooler. The Xigmate is gigantic

 

[chinaman1472]

Well, 9x310 is a no go even at 1.45 vcore. I got it to boot, but it errors within a minute on Prime95.

I'm a bit afraid to do +0.2V to the FSB and MCH on stock cooling and average airflow in the case.