New to OC, AMD Athlon II X4 635 help

[Neutrolizer]

Ok I just did that. I ran 3 scans and the first scan there was an error in one of the cores. The second scan was for an hour and completed. The 3rd scan was for 6 or 7 hours and completed. Does that mean anything for stability if it failed the first time but succeeded the next couple times?

 

[trents]

Well, failing any one stress test suggests instability, even if its subtle. A core dropping out typically suggests need for a little more vcore but I'm concerned you are already at a pretty high vcore. You may want to consider dropping the fsb down just a smidgen, 3-4 mhz and retesting. Do you have your new ram installed yet? How about your mosftet sink?

 

[trents]

How's about some new screenies of the three CPU-z tabs?

 

[Neutrolizer]

Yeah I ordered the stuff over the weekend so it should be coming in sometime next week. Here are the screen!

So in a situation like this, just keep decreasing by 1 FSB?

 

[trents]

Yeah I ordered the stuff over the weekend so it should be coming in sometime next week. Here are the screen!

So in a situation like this, just keep decreasing by 1 FSB?

Yes, drop the fsb a little. Your CPU is the C2 stepping and not the C3. I have an AII X4 630 C2 and it only takes me 1.4125 (plus a hefty amount of load line calibration added in) to be stable at 3.5 ghz but it takes 1.485 to be stable at 3.6 ghz. As you can see from that, the C2 frequency does not scale with the voltage very well beyond 3.5 ghz.

 

[Neutrolizer]

hello! Just got my mosfet sink and ram but one of the stick was DOA... Replies have been slow because I've been running couple several overnight OCCT to make sure my comp is running fine. Went down to 249 FSB but an error occured again so I'm down to 248 now.

With the mosfet sink, am I able to crank the Vcore up a bit?

 

[RGone]

With the mosfet sink, am I able to crank the Vcore up a bit? = Don't have any idea. You got any before and after HWMonitor captures when underload to see if VRM area temps went down? If the temps are no lower then nothing has really changed.

 

[trents]

Neutrolizer, it occurs to me that in your case you do not know whether its the CPU frequency or the fsb that is causing instability, or even the ram. My suggestion would be to start by checking the ram. Reset the cmos to put all frequencies and voltages at stock. Then download install Memtest86+ on a CD and run it overnight to test the ram. If that turns no errors then run Prime95 blend for a couple of hours. The blend test works the memory hard. If you get errors in either case, take out all but one stick of ram and retest each stick in turn to find the culprit.

To see if the fsb is causing the problem, turn down the CPU multiplier to a very safe level, say 10x so that you know for sure that the CPU isn't the source of instability and set the ram to low starting frequency, say 400/800. Then start turning up the fsb 10 mhz at a time until the computer won't boot or isn't stable loading windows. Then set it back 10 mhz to the last stable setting. Then stress it with Prime with the ram and CPU at low speeds. This will give you an idea of where the fsb max frequency is.

 

[Neutrolizer]

Hey been busy for awhile, hope you guys are still there. Anyways I did the memtest overnight and passed it 14 times.

Did as you said trents and lowered my multiplier to 10x and my ram ratio to 4. Then I continued to increase the FSB by 10 and I reached 280. At 290 I had trouble booting up. Ran OCCT for couple hours and it passed fine.

What is the purpose of finding FSB Max?

 

[keny]

The purpose of finding you're max fsb is to make sure you're board is not limiting you by reaching its max fsb, seeing as though you made 280 stable, you know that the board will go that high without a problem, so now you know it's not you're board limiting you @249 fsb.

 

[trents]

I think you back off on your CPU multiplier a little too far and the CPU is now running slower than stock speed at 2800 mhz. At the same time your fsb at 280 mhz is probably pushed to the limit. I'd bump up the CPU multiplier .5x at a time and run the OCCT test after each increase. When you start failing the test, increase your CPU core voltage a little to restore stability. The core voltage is pretty low at 1.344 and your core temps are only in the low 40s so you have considerable room to bump up that core voltage and go for a higher overclock.

 

[Neutrolizer]

I think you back off on your CPU multiplier a little too far and the CPU is now running slower than stock speed at 2800 mhz. At the same time your fsb at 280 mhz is probably pushed to the limit. I'd bump up the CPU multiplier .5x at a time and run the OCCT test after each increase. When you start failing the test, increase your CPU core voltage a little to restore stability. The core voltage is pretty low at 1.344 and your core temps are only in the low 40s so you have considerable room to bump up that core voltage and go for a higher overclock.

Ok so I got my vcore up to 1.425 with my cpu multiplier at 12.5 and that was when I was able to run OCCT for 3 hours without my comp hard resetting on me. The core temps when running prime95 was up to 45C. Should I be doing anything about my ram? or can I be doing anything about OCing my memory? Its at 400 right now.

Also, will it be fine if I bump up the NB a bit since its sitting at like 1960. Its set on automatic in the bios.

The purpose of finding you're max fsb is to make sure you're board is not limiting you by reaching its max fsb, seeing as though you made 280 stable, you know that the board will go that high without a problem, so now you know it's not you're board limiting you @249 fsb.

I see that makes sense, seems easier to do it this way. thanks!

 

[trents]

Your ram frequency is still low. I would change the ram divider (may be expressed as a ratio) to start the ram at a higher frequency, back off on the fsb and increase the CPU multiplier.

 

[Neutrolizer]

Alrighty. I increased it by another .5 and decreased my fsb by 5 but it was still crashing when booting. So i changed the vcore up to 1.475 which was stable to run OCCT for 7 and a half hours. The core temps didn't exceed 55C.

I also changed the ratio for the ram, but i'm not sure if the timings are good enough as I just followed the numbers listed in CPUz.

Also, how can I find out what the best NB frequency should be? Looking at the guide it tells you how to calculate it but then on other cases, the number calculated isn't always the best.

Do you think I can go higher? I remember it was crashing before around 3.6. But I suppose I can try to and maybe lower the HT frequency as that is getting a tad high

 

[trents]

Do you have Cool N Quiet and C1E disabled? Do that if not and go into Windows Control Panel and set Power Options to High Performance. That might help your boot up crashes.

Your DRAM timings are fine. Leave them as they are.

Give your DRAM voltage a bump to 1.55.

Your HT Link is too high at 2199. That may be causing your boot up crashes. Set it to between 1800 and 2000.

Bump up your CPUNB voltage to 1.225 an the CPUNB frequency to 2500.

 

[Neutrolizer]

Alrighty will do. ALready have the power options set to high.

So what is the purpose of bumping my dram voltage to 1.55? I thought the reason to get 1.5v ram was because the mobo likes it at that voltage and anything higher might be problematic?

What is the purpose of disabling cool n quiet and C1E? I haven't touched those two at all.

 

[trents]

Cool N Quiet and C1e are throttle down technologies that lower the CPU frequencies and voltages way down and can create instability in a significantly overclocked CPU because the swings are so wide as it switches in and out of down throttle mode in response to changes in load. I think they might be disabled when you put power options on High Performance but not positive.

 

[Neutrolizer]

Ok I changed the HT link freq in the bios down to 1.6ghz which reflects 1925 MHz in CPUz. Disabled Cool N Quiet and C1E. Changed CPUNB voltage to 1.220 and the CPUNB Freq to 2475. The only thing that I could not do was change the DRAM voltage to 1.55. WHen I did do that, I was not able to get past the mobo start up screen. I ran a 1 hour OCCT test which passed. I plan on doing another one through the night to make sure!

 

[trents]

That's weird that it wouldn't boot on 1.55 vdimm!

 

[trents]

Better do a Prime95 blend test for couple of hours 'cause it will stress them memory subsystem more than OCCT and you are almost in the extended profile with your memory frequency.