OLD SCHOOL 965 BE OC

rossotron · Feb 12, 2014

hello all,

i purchased a cooler master seidon 120xl a few weeks ago and decided to finally overclock my 965. i thought the upgrade from a tuniq tower will enable me to finally push my cpu to 4 Ghz. however, after an hour of running prime, my computer shuts down. i have to dial it back to 3.8Ghz for stability.

here's a list of my hardware:

asus sabertooth 990fx
amd 965 be
4 GB of cheap 1600mhz ram
cooler master seidon 120xl

at 3.8Ghz, i ran prime for over an hour and my core temps never went past 50 degrees Celsius. is my core voltage too high at 1.512? perhaps its the cheap ram i am using. it's a cendyne and it went out of business a few years ago lol

i appreciate any advice :attn:

RGone · Feb 12, 2014

Throw away CoreTemp for our purposes and use HWMonitor Free version from CPUID com. It shows much more information than CoreTemp. Then we can see the socket_Cpu temp as well as the Core_Package temp if you open the HWMonitor enough at least down to showing "package temp.

Also if you suspect the cheap ram why not dial it back to DDR3-1333 and "test" it to determine if that be the case?
RGone...

Johan45 · Feb 12, 2014

I would suggest that you have a read through Dolk's guide for the PhenomII, It looks to me that you have the NB frequency pretty low. He'll explain how the NB frequency and CPU frequency are tied together and roughly where to set it for certain speeds. It really makes a difference. I would also think your voltage is a bit high for a C3 965. That voltage gets me up to 5.0G.
I would also suggest setting the CPU and CPU_NB LLC maually to high for both. On auto the system will really kick that in the arse and give you some high voltages.

Mandrake4565 · Feb 12, 2014

Some chips may just not go past X Mhz. Adding to the above comments, you may want to set the memory command rate to 2T it's at 1T and could cause some instability.

trents · Feb 12, 2014

You would have a better chance of reaching your goal of you would overclock with a combination of the multiplier and the FSB. Right now you are using only the multiplier. Using the FSB is really the "old school" method as it was all we had before black edition CPUs.

And I agree with what RGone said about the RAM speed. Even with quality RAM many Deneb core CPUs simply will not be stable running the RAM at 1600 mhz. The IMC (Integrated Memory Controller) on that family of CPUs was kind of weak. It's only rated for 1333 mhz and in my experience does not like to play when the RAM frequencies are all the way to 1600 mhz. They seem to be good up to around 1500 mhz and that's another reason for using the FSB to overclock as you can get the RAM to run at those "tweener" frequencies.

Tweaked1024 · Feb 12, 2014

The guide will tell you the formula for nb frequency. My nb is at 2600 and change. My CPU voltage is about 1.45-1.475. I'm air cooled and I keep my temps 45 or below. I am not an expert on over locking, these guys are, just throwing some numbers out there for you to see.

rossotron · Feb 13, 2014

Thank you all for your advice. I played around with it tonight and here are my results so far:

CPU Ratio 19
CPU Bus Frequency 210
Memory Frequency 1399MHz
CPU/NB Frequency 2520MHz
CPU Load Line on Ultra High
CPU/NB Load Line on High
CPU offset voltage 1.524V on Auto

I ran prime until i passed the first test. i cut it short because my temp went up to 56 degrees C. I'm thinking the high temp is the result of high VCore voltage and CPU load line on Ultra High and CPU/NB load line on High settings.

trents · Feb 13, 2014

You may not need the vcore to be quite that high to achieve 4.0 ghz. What you probably do need to increase is the CPU/NB voltage now that your frequency for that component is 2536 mhz and your RAM frequency is 1409 mhz. Try bumping the CPU/NB volts to about 1.225 or 1.25 and cutting back some on the vcore. I get a little uncomfortable when max vcore exceeds about 1.5 under load. Do you have Cool N Quiet and C1E disabled? If not, you should. For that CPU the max core temp should be in the mid 50s c. and the max CPU temp should be around 60c. Higher than these typically brings on temp-related instability.

If I were you I'd cut back on the CPU multiplier a bit and keep pushing the FSB a little (to keep the CPU around 4.0 ghz) while making adjustments to the CPU/NB to keep it in the 2400-2600 mhz range and to the HT Link to keep it in the 1800-2100 mhz range.

Mandrake4565 · Feb 13, 2014

trents said:
You may not need the vcore to be quite that high to achieve 4.0 ghz. What you probably do need to increase is the CPU/NB voltage now that your frequency for that component is 2536 mhz and your RAM frequency is 1409 mhz. Try bumping the CPU/NB volts to about 1.225 or 1.25 and cutting back some on the vcore. I get a little uncomfortable when max vcore exceeds about 1.5 under load. Do you have Cool N Quiet and C1E disabled? If not, you should. For that CPU the max core temp should be in the mid 50s c. and the max CPU temp should be around 60c. Higher than these typically brings on temp-related instability.

If I were you I'd cut back on the CPU multiplier a bit and keep pushing the FSB a little (to keep the CPU around 4.0 ghz) while making adjustments to the CPU/NB to keep it in the 2400-2600 mhz range and to the HT Link to keep it in the 1800-2100 mhz range.

I agree with Trents, 1.5+ Cpu V seems like a lot for 4.0 but then again some chips just need a lot of voltage to get to X mhz.

Out of curiosity, how did you get to that voltage? I usually find setting a starting point like 1.4 on the Cpu V and raising the Cpu Nb V to 1.2. Shutting all the power saving features off then start raising the multiplier and testing using prime works well. What I'll do is the above then raise the multi .5 test prime 20 minutes, keeping temps in check 60 on the socket and 62 on the cores max. If it passes prime bump the multi again until it fails. When it does back the multi off .5 and run prime blend 2 hours and add Cpu V if it fails, again keeping temps in check. When it finally passes 2 hours prime i start the process again by bumping the Cpu V to about .05 higher if I have the headroom temperature wise. I will continue to do this until I find my max oc for the chip whether it's having hit the temp threshold or the chips wall. Then what I'll usually do is back the multi off and see what I can do with the FSB.

Johan45 · Feb 13, 2014

Mandrake4565 said:
I agree with Trents, 1.5+ Cpu V seems like a lot for 4.0 but then again some chips just need a lot of voltage to get to X mhz. .

I on the other hand have a 965 C3 that needs 1.51v at 4.0 to be completely stable in P95 blend. So I don't think it's unreasonable TBH. The 955 like you have seem to be the easy clockers of the bunch from what I have seen. On the other hand there haven't been too many low volt 965s that I have witnessed.
I don't disagree that if you can lower the volt by all means do so, I just don't think there'll be much to gain.

trents · Feb 13, 2014

Yeah, but he's hitting 1.524 under load.

Rossotron, what is your vcore set to in bios? Have you tried dialing back some on the LLC?

Johan45 · Feb 13, 2014

trents said:
Yeah, but he's hitting 1.524 under load.

Rossotron, what is your vcore set to in bios? Have you tried dialing back some on the LLC?

I agree I would try High vs Ultra on the sabertooth that should keep it just about even with your bios setting. I also just saw this

CPU offset voltage 1.524V on Auto

Offset on auto takes the control out of your hands.

rossotron · Feb 13, 2014

trents said:
Yeah, but he's hitting 1.524 under load.

Rossotron, what is your vcore set to in bios? Have you tried dialing back some on the LLC?

it is currently set on AUTO

Mandrake4565 · Feb 13, 2014

Johan45 said:
I on the other hand have a 965 C3 that needs 1.51v at 4.0 to be completely stable in P95 blend. So I don't think it's unreasonable TBH. The 955 like you have seem to be the easy clockers of the bunch from what I have seen. On the other hand there haven't been too many low volt 965s that I have witnessed.
I don't disagree that if you can lower the volt by all means do so, I just don't think there'll be much to gain.

It does seem like the 965's need more voltage for X mhz then the 955's.

Johan45 · Feb 13, 2014

rossotron said:
it is currently set on AUTO

Try setting it to manual with 1.5v and see how it does. Mine would crap out right away with P95 at anything below that.
@Manny it was a big jump for mine at the 4.0 mark, I remember thinking is this thing ever going to stop taking the volts. That's why I left it between 3.8 and 3.9 with around 1.45v

Mandrake4565 · Feb 13, 2014

From what I remember my 955 did 4.0 with 1.42 on Air.

RGone · Feb 13, 2014

Mandrake4565 said:
From what I remember my 955 did 4.0 with 1.42 on Air.

Yes but you so good and fortunate. Hehehe. Some of us cannot roll cpu dice worth a h00t.
BobRGone...sterstignatorzzzzzzzzz

rossotron · Feb 13, 2014

trents said:
Do you have Cool N Quiet and C1E disabled?

yes

rossotron · Feb 13, 2014

Johan45 said:
Try setting it to manual with 1.5v and see how it does.

I manually set it at 1.5 and I got the Blue Screen Of Death as well.

rossotron · Feb 13, 2014

New figures:

CPU Ratio 18.5
CPU Bus Frequency 220
Memory Frequency 1466MHz
CPU/NB Frequency 2640MHz
CPU Load Line Ultra High
CPU/NB Load Line High
CPU Offset Voltage 1.524V on Auto
NB Voltage 1.2

I was running Prime while I got that screen shot. It passed the first test again. For some reason, I have to keep the VCore above 1.5V to keep it from getting BSOD.

OLD SCHOOL 965 BE OC

Registered

Senior DFI Staff

Benching Team Leader Super Moderator

Mr. Clean Senior Member

Senior Member

Member

Registered

Senior Member

Mr. Clean Senior Member

Benching Team Leader Super Moderator

Senior Member

Benching Team Leader Super Moderator

Registered

Mr. Clean Senior Member

Benching Team Leader Super Moderator

Mr. Clean Senior Member

Senior DFI Staff

Registered

Registered

Registered

Similar threads