• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Can't get unlocked 550 BE stable, help please.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

trents

Senior Member
Joined
Dec 27, 2008
By the way, your CPU is a C2 stepping (Look at CPU-z tab: "CPU"). It will not overclock as high as the later Callistos with the C3 stepping but you still should get 3.8+ out of it.
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
Thanks for the advice on the TMPIN0 as it reassures what I thought was correct. TMPIN0 - 8C = Core Temps when OC'ing all 4 cores later on.

Pity I got a C2 core considering I just got the processor a weak ago, figured they would have been weeded out by now, oh well.

I will be able to increase the VCORE and ratios later this evening or tonight. I will report back with screenshots of what I found to be the maximum stable clock ratio and lowest VCORE value for that clock. Also, my mobo won't go above 1.55 VCORE anyway, so that should be good, not that I'll need that value without getting into the other stuff like RAM and NB.

Yet again, thanks for the continued help.
 

trents

Senior Member
Joined
Dec 27, 2008
You're welcome. Note that I recommended not to exceed 1.55 vcore under load. That's not the same as what you have it set to in bios necessarily as it appears your board adds some vcore to compensate for the voltage drop off that naturally occurs when loading an electronic component. The natural vcore fall off is called vdroop and when the reading is reported as higher under load it's called negative vdroop. A lot of enthusiast boards have a setting in bios called LLC (Load Line Calibratiion) which enables you to adjust the amount of offset. Looking at the maximum core voltage reported by HWMonitor it looks like you may have negative vdroop of about .025 volts. Now, all this assumes HWMonitor is reporting vcore accurately but better safe than sorry. Go with whichever is higher, what you have it set to in bios or what the reporting software says under load. Someboards do not compensate so the voltage reading under load is higher in bios than what is reported under load.
 
Last edited:
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
Okay I think I understand. Theoretically if I have my VCORE at 1.5V and HW Monitor reports at CPU idle that VCORE is 1.49V then that is regular vdroop. Then underload if it were to go up to like 1.52V or so that is negative vdroop. Correct?

If so negative can cause overheating and in extreme circumstances a fried processor, while vdroop can cause instability due to reduced voltage.
 

trents

Senior Member
Joined
Dec 27, 2008
Yeah, I think you have it right although I would consider the difference between 1.5 and 1.49 so small they would essentially be the same. But if you have it set to 1.375 in bios and its showing 1.400 in HWMonitor under load then I would call that .025 vdroop. Yes, the import of vdroop is that it can cause instability and the design of the bios compensation is to counteract that. Through the LLC adjustment you can sort of even things out so there isn't too little or too much compensation and the swings aren't out of reason. Does your board have LLC?
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
I just looked through the BIOS during a reboot and I didn't see anything that resembled LLC. Bringing up the BIOS however...

I originally had the latest BIOS version 18.3 and was having OC issues of course. I read one user that said to downgrade to 18.1 as it was a better OC candidate.

Version 18.2 added
- Support AM3+ CPU.
- Update M-Flash module.
Version 18.3 added
- Update CPU AGESA code.
- Update M-Flash module.

One difference I did notice however is that in the voltages area....Version 18.3 had 6 voltages I could change while the version I am on 18.1 only has 4 voltages I can change.

I am not against going back to 18.3 as I was obviously not OC'ing right to begin with so I am sure that was part of the problem.

Anyway the 4 voltages on 18.1 are:
CPU VDD Voltage
CPU-NB VDD Voltage
DRAM Voltage
NB Voltage

The 6 voltages on 18.3 are the previous 4 listed and:

CPU Voltage
CPU-NB Voltage

Are those extra two voltages important? All they do is confuse me as they are identical to two other voltages except "VDD" isn't there.

Should I proceed on 18.1 or flash to 18.3 to proceed? I don't have an AM3+ processor and not too worried about the mflash updates. However the CPU AGESA code sounds like something important that they either improved or screwed up.

Thoughts?

PS: grinding through the ratios and voltages still
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
I found this about the voltages, almost the same board by MSI, just the GD70 version. Albeit I am still not 100% sure how I would handle all those voltages if I were to use them...

http://www.bit-tech.net/hardware/cpus/2009/05/11/overclocking-amd-s-phenom-ii-x3-720-be/3

CPU Vdd Voltage: This is the positive voltage supplied to the FETs. MSI states (on another 790FX board with Phenom II) that 1.3 to 1.4V is typical safe value, however we've increased it here because we've added additional cooling to the CPU socket area with two large fans.

CPU-NB Vdd Voltage: This can also help with increasing the integrated memory controller overclock - we've not increased it here, but considering AMD claims 1.4 to 1.5V for the CPU-NB core for optimum results, the CPU-NB Vdd should also be ~1.3 to 1.4V.

CPU Voltage: Typically the 45nm SOI Phenom IIs can withstand quite high voltages so 1.5V should be OK if it's properly cooled. However it's worth mentioning that this is wrongly labeled, and is actually the voltage offset, as increasing the Vdd above adjusts this value as well.

CPU-NB Voltage: Like the above setting, working in combination with the CPU-NB Vdd setting will give you an additional voltage. Work with mostly Vdd and start with a small setting here.
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
Ok I continued as we discussed and found the proverbial wall based of stock settings and just adjusting the CPU voltage and ratios.

I was able to reach 3800 MHz and complete a 20 minute blend. The lowest voltage I was able to achieve this on was 1.5V as set in the BIOS.

Attemping to reach 3900 MHz resulted in Windows locking and such even after testing up to the max allowable voltage I could set in the BIOS of 1.55V.

Core temps are still looking pretty solid under load.
 

Attachments

  • OC2.jpg
    OC2.jpg
    300.9 KB · Views: 32

trents

Senior Member
Joined
Dec 27, 2008
This is your max under load, 1.53.

Okay, you are pretty much maxed out on vcore. Maybe one small bump would be okay. But let's save it because now . . .

we need to run a longer Prime95 blend test to confirm stability. I think two hours is enough and so that's what I recommend. Some overclockers on the forum insist on 8 hrs. or even more but I haven not ever encountered instability in running real apps if I can pass two hours of Prime. Just wanted you to be aware of the fact that there is a variety of opinion and advice in this matter. So put her in gear and grind away. If you can't pass the two hr. test we'll try one more increment of vcore.

Oh, yes, give your ram a voltage bump to about 1.55.

After this we will give attention to the NB frequency and voltage.

Boy, your temps are really holding cool! No problem there.
 

Attachments

  • Mazx.JPG
    Mazx.JPG
    36.8 KB · Views: 31

RGone

Senior DFI Staff
Joined
Apr 20, 2004
Location
JAX, Mississauna
"bp007" you are actually doing quite well with that C2 processor. I see peeps all the time that have very overclockable cpus that have not got the drift of what is happening as quickly as you are doing.

When you get time if you would, how about getting us a cpuz of the CPU and Memory at 3900Mhz even though we know it has not passed 20mins prime. I want to see what it shows for ram speed and set timigns. A HWMonitor would not hurt either.
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
RGone,

Thanks for the compliment, I think those hours upon confusing hours of reading are finally coming together with Trents help and definitely patience.

For the requested info the only things I have changed was 3900 @ 1.55V and I bumped up the DRAM voltage as Trent requested from Auto to 1.55V.
 

Attachments

  • 3900 test.jpg
    3900 test.jpg
    331.9 KB · Views: 35

RGone

Senior DFI Staff
Joined
Apr 20, 2004
Location
JAX, Mississauna
The last 5 or 6 boards I had to test were Intel and I could not remember if the odd multiplier and the half multiplier influenced the ram speed still. Seems CPU-z says nope.

Let me not get you and 'trents' off course. He has a few more things you guys can try. I have a suspicion or two as well. We will meet later if necessary.

You could always ebay that cpu and get another and the next might be the golden thing with 4.0Ghz in its' future if that were a needful thing. Hang with it man, you are getting it and not just plugging someone else's numbers in.
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
RGone,

Thanks for the support and encouragement!

I'm curious what your suspicions are, you can bring them up later or send me a PM. I'm saving my BIOS configs in this post as Trents while I tinker on the side a bit with alternate configs. And nah not gonna eBay this, it's a good chip to learn on and is not my "main" desktop per se, so it's not worth taking a loss just to get a C3 version for me.
 

trents

Senior Member
Joined
Dec 27, 2008
I have never messed with the vdd voltages. There is kind of a mystery surrounding them and most people just leave them alone. The best I can figure is that they offer finer tweaks to the main voltage adjustments as they affect the voltage going to the voltage chokes. Something like that. I actually have not owned a motherboard with vdd adjustments.

Does AMD really say that 1.4-1.5 is the recommended voltage for optimum results for the CPUNB? That surprises me. Most overclockers find that 1.225-1.3 does the trick. Speaking of CPUNB, after we settle on a final overclock for the CPU speed we need to increase the CPUNB frequency and voltage. You'll want to get it up to around 2600 if you can. It has a positive impact on memory performance.
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
Yeah what manufacturers say and recommend for voltages often is a conflict of interest on any overclocker forums that I have read. It seems all we are concerned with is max "safe" voltages and sometimes that is not even abided by.

As for the CPUNB, the rule of thumb that I have gathered is basically at least triple your base DRAM speeds. Or am I thinking of HT speed...anyway! The stability test seems to be going quite smoothly so far. I think I am just going to run it for the night and check on it at about 7 am EST. I will give an update then.
 

trents

Senior Member
Joined
Dec 27, 2008
As for the CPUNB, the rule of thumb that I have gathered is basically at least triple your base DRAM speeds.

3x1333 = 4000? Don't think that will work.

3x2000 (stock HT) = 6000? Don't think that will work either.

I think 2400-2600 is what the Callistos/Denebs will do. Thubans will go a little higher.
 
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
I wasn't sure what to call it, but i used "base" DRAM speeds.

What was intended was ~667 * 3 = 2000 or 800 * 3 = 2400.

Sorry for the confusion. Of course this was only a general rule and nothing specific, so I am sure there is a more accurate way of going about it. Either way, you're the boss of this OC regardless as your instructions have been clear and worked flawlessly so far. Much better than how I did on my own!
 

trents

Senior Member
Joined
Dec 27, 2008
I think the term is the "ram bus frequency". Its the base frequency of the ram before the DDRx transfer rate multiplication is factored in.
 
Last edited:
OP
B

bpatterson007

Registered
Joined
Sep 19, 2011
That definitely sounds more accurate. If only I knew this stuff as well as I know networking, haha. Have to learn somewhere :)
 

trents

Senior Member
Joined
Dec 27, 2008
CPU-z reports the ram speed in ram bus terms, not in DDRx terms. That throws a lot of people at first.