Trying to get a stable 4ghz 960t BE

[Xaxinian]

Hello all, I am new to the forums and hoping to get some feedback from people with experience with my setup or similar setups.

I am hoping to get to 4ghz stable. but it seems no matter what I do I cant seem to get past an hour with OCCT (before system locks up). Anyone see anything I could possibly change or adjust? Something that looks wrong, etc. Thanks in advance.

AMD 960t BE 6 Unlocked
Gigabyte 970a-UD3 Mobo
Gskill Ripjaw DDR3-1600 4gbx2 9-9-9-24 stock
Coolermaster Hyper 212 plus PSU
Raidmax RS 530SS heatsink
Radeon XFX 6770 GPU

CPU: x16 4000MHz
CPU NB: x12 3000MHz
Core Performance Boost: Disabled
CPU Host Clock Control: Manual
CPU Frequency: 250
HT Link Frequency: x9 2250MHz
Memory Clock: 5.33 1332 mhz Dram voltage 1.51v

Voltages
CPU NB vid control 1.3v
Cpu voltage control 1.475

 

[keny]

Have you tried just using the multi on the CPU only ? as using the fsb brings loads of other stuff into the possible culprit for the crashes. If you want to stay with the 250 fsb then you could give you're nb chipset a bump to help with the higher fsb.what timings are you running on you're ram at 1332?

Edit: an easy oc on this chip should be
FSB 200 x 20
Everything on auto except for CPU vcore
Vcore@ 1.475
That should be a safe 4ghz as long as you're temps are good

 

[Xaxinian]

Actually I am at 1666 on the memory with timings at 9-9-9-24. As for just raising the multiplier and leaving everything else on auto everything I read of peoples OC, most at least change the FSB. Is there no advantage to the other changes or is it just my not understanding the setups.

 

[RGone]

On an unlocked BE type AMD cpu, the only reason to n0t completely use the multipier only is if one is trying to tweak ram speeds.

 

[keny]

Using the multi makes it easier for you to gain a decent oc with ease, using the fsb is for squeezing the last few MHz out of it and as RGone said tweaking you're ram

 

[manu2b]

On an unlocked BE type AMD cpu, the only reason to n0t completely use the multipier only is if one is trying to tweak ram speeds.

I don't know how it is with Thubans, but with my PhII x4 and my MoBo (see sig), with a 20 multi, power saving features don't work anymore. I have to play between fsb and multi.

 

[manu2b]

Well, one of my posts disapeared...

So, I was saying:

Try lowering the CPU-NB to 2800MHz, and bring the HT to the same freq. Thuban's like to have both synchronised.

What are your temps on load?

 

[keny]

My 960t runs out of multi scaling @22.5 @200 fsb but will go higher with a lower multi and a bit of a bump on the fsb but it's only stable enough to bench at those speeds and not 24/7 but 22@200 is rock solid, I know once I oc on my board it automatically disables a lot of the power saving functions if not all of them.

 

[Xaxinian]

Thanks for the Tip Manu, my temps under load dont seem to get past 40c. I will try the adjustment you mentioned. I just dont know what would be causing the lockup, only thing I can think of is the voltages, but again I am a novice at best and am really trying to learn.

 

[Xaxinian]

So I am still getting lockups about 45 minutes into the testing. Does this sound like a voltage issue? Anyone with a similar setup that has some insight into voltages? I dont want to go too high, but I want to be able to pass OCCT. I am not having any issues in the applications I use this for, but I dont want to have it running if its unstable.

Also the suggestion to stay at 200fsb and using the multiplier sounds ok, but People are also saying that you should get the NB up to 2800-3000 for stability. Anyone tried doing this at 200fsb and if so what voltages are you using to accommodate.

 

[manu2b]

When you say "lock up", do you mean freeze?

 

[keny]

You can raise the CPU-nb to anything up to 3000 on about 1.275 cpu-nb volts, but that's a separate entity in itself, you should plan on getting a stable CPU clock thn move onto the CPU-nb and then the ram(in that order) do one thing at a time or it gets messy and difficult to remedy a fault when one arises. Hard locks can be caused by almost anything and finding the source will be simple if you are just multi overclocking the cpu. Have you tried the 20x200 @ 1.475vcore , it should run that or you're chip is out of a bad batch? Or one of those disabled cores is actually not up to scratch

 

[DjArtifact]

You can raise the CPU-nb to anything up to 3000 on about 1.275 cpu-nb volts, but that's a separate entity in itself, you should plan on getting a stable CPU clock thn move onto the CPU-nb and then the ram(in that order) do one thing at a time or it gets messy and difficult to remedy a fault when one arises. Hard locks can be caused by almost anything and finding the source will be simple if you are just multi overclocking the cpu. Have you tried the 20x200 @ 1.475vcore , it should run that or you're chip is out of a bad batch? Or one of those disabled cores is actually not up to scratch

I'm a bit late to the party but I tried this overclock and it gets really hot, I wouldn't recommend going over 1.435vcore unless you are water cooling.

 

[trents]

I'm on 1.45 vcore with air cooling and CPU socket temps under full load never get above 54c and core temps 10c less than that with a 10c offset added to core temp because the Thuban sensors are calibrated on the cool side.

 

[trents]

OP, I think you need to work with your CPUNB frequency and voltage and possibly your LLC.

 

[BeepBeep2]

@op
Overclock one thing at a time.

If you want 4 GHz, try 4 GHz CPU @ 1.475v, 2 GHz CPU-NB stock voltage/ HT Link, DDR3-1600 9-9-9.
Pass?
Yes: Proceed with CPU-NB overclock, try 2.6 at 1.25v
No: Raise vcore, watch temperatures though, keep under 65c. (Phenom II X6 has a bugged CPU temperature sensor, so add 15c to temperature readouts, keep sum under 65c.)

etc.

 

[Dlaw]

My setup isn't much different than yours (see sig), so let me see if I can help you out a little.

What I did was this:
1) I tried the 20x 200 for 4GHz method that *normally* works with a BE CPU. Lockups/freezes and or BSODs on anything under 1.425v. Anything @ or above 1.425v was way too hot.
2) I tried lowering to 3.8GHz (19x 200), which took a step down from 1.4v and ran cool and stable enough to pass 2hrs of P95, and tried raising CPU-NB to 2400GHz and a bump on voltage. Passed, but ran too hot again.
3) I decided (after seeing an old post of trents') that I would try upping the FSB, setting the vcore at ~1.35 with LLC on auto (would have to look in BIOS for actual setting), leaving the bump on the CPU-NB voltage, and dropping the DRAM, and see what I could get out of it with a x17 multi.

I ended up with 17x240=4080MHz. Temps have yet to rise above 55C on a hot day with the AC off (about 80F weather). I think I could get even more out of it, and I know I can run stable at x6, but I am happy with where I'm at.

Bottom line, you may get a more stable system with a higher clock and lower temps by using the FSB. Just remember to keep everything else in check.