First Overclock: Phenom II x4 965 Black Edition

Mar 6, 2012
122
0
0
Using the main/gaming rig system in my signature, looking to overclock a little bit. Hoping for 3.7-4.0GHz. Does that seem attainable? I get 35 celcius max during Total War: Shogun 2 with atrocious thermal pasting (had to remove a few times, so lots of air bubbles). Any tips for me?

Also, out of curiosity, what is really measured by the frequency?
 
Last edited:

Saffron

Member
Nov 16, 2012
130
1
41
When I had my Phenom II x4 965BE I was able to hit 3.8GHz stable on stock voltages (this will vary from CPU to CPU even if they are the same). So yes it is easily obtainable. Once you start pushing the overclock more and increasing voltages the stock heatsink/fan won't hold up very well.

EDIT:

Some may disagree because everyone has their own way of applying the thermal compound. Usually what I do is apply enough to form a credit cards thickness evenly over the entire processor. That should be enough without spooge and without air bubbles.
 
Last edited:
Mar 6, 2012
122
0
0
Corsair H60 liquid cooler with arctic silver 5 in a cold room :D I dropped 12-15 degrees under load.

My thermal paste is perfect and covers every mm of my cpu, I just had to lift the thing off the cpu a few times so its got bubbles I thibk.
 
Last edited:

KingFatty

Diamond Member
Dec 29, 2010
3,034
1
81
Also remember, overclock the CPU-NB. It has a separate setting than the CPU itself.
 
Mar 6, 2012
122
0
0
Should I use BIOS or TurboV/Ai Tweak (both in Asus Suite II)? The BIOS is daunting and confuses me. Kind of feel pathetic saying that.
 

Face2Face

Diamond Member
Jun 6, 2001
4,100
215
106
I overclocked my Phenom II 960T to 4.0 Ghz with my voltage set to 1.40V. All you need to do is adjust the multiplier and change the voltage to 1.40-1.425 and you should hit 4.0Ghz on that chip. Keep temps under 65C and your good. Once that is stable you can work on the Northbridge.
 

sm625

Diamond Member
May 6, 2011
8,172
137
106
Usually what I do is apply enough to form a credit cards thickness evenly over the entire processor.

lol that is a LOT of paste. Credit cards are not exactly thin. If you were to melt a credit card and pool the liquid it into a droplet shaped glob of plastic, that glob would be rather huge, at least 5 times more than you need.
 

KingFatty

Diamond Member
Dec 29, 2010
3,034
1
81
Thanks, extremely easy to understand. What's the FSB/HTT that I've read about?

Ah please someone correct any errors, I'm a bit rusty with this but here goes:

The overall CPU performance/frequency is a result of multiplying two values, 1) core frequency, and 2) multiplier. You can increase either one. Sometimes you can have two different settings produce the same result, like 200*10 or 100*20 both produce 2000 overall, but use different settings.

The FSB/HTT is the speed of how the CPU talks to other parts of the motherboard. I think there is a generalization that when people use this term, they are approximating what the CPU core frequency is, because the FSB/HTT speed usually follows the core frequency (but now they are independently controllable). In older architectures, the FSB/HTT speed was more important because the CPU needed to send lots of info back and forth between another motherboard chip (called a northbridge) that handled memory communication. However, current architectures have the memory controller functionalilty right inside the CPU chip. So there isn't a pressing reason to increase FSB/HTT.

Instead, you would look to the CPU-NB, which can be independently overclocked and increase performance because that affects how the CPU can handle the memory control functions, speeding that up.

So you would ideally figure out the highest CPU core frequency you could obtain, and the highest multiplier, that gives you the best performance. Sometimes you use a slightly lower multiplier and that lets you achieve a slightly higher core frequency, resulting in an overall higher overclock.

Because you have an Asus board, see if you have their automated tool to overclock, called TurboV I think? Anyway, the neat thing is this tool will let you set the multiplier, and then it will adjust the BIOS by increasing the core frequency/voltage accordingly, and stress testing, then rebooting and incrementally increasing the core frequency and stress testing, etc. etc. until it reaches such a high setting that it detects instability. It will then reboot itself and display the final results before instability was reached.

So myself, I set my bios to 16x, then ran the tool to find the highest stable frequency. I could walk away from my computer while it performed this tedious function for me. Wrote down the values (16x & 250, for example). Then I set my multiplier to 16.5x, and repeat. Interestingly, the highest combination was by using a lower multiplier than I thought, but I wouldn't have found out unless I did this tedious testing and the TurboV tool helps make it less work.

You can play with voltage settings too, that increases the values you can achieve. Also, you can see about disabling unused cores. I got a higher result using 2 of my cores with the other 2 cores disabled. So that would be helpful for games like Starcraft 2 where the extra 2 cores are not used anyway, but the higher frequency will be more valuable.
 
Mar 6, 2012
122
0
0
@KingFatty Wow thanks, that was perfect. Now I think I understand the different components of overclocking. Multiplier is, well, a multiplier of the core speed (CPU-NB), all of which relies on the voltage the socket is receiving. Cooling needs to be able to handle it too. I have a stress test running at 200 x 18 while I'm at work.

I did a little math while I ran Prime95 at stock to see temps (42°C max), and I'd like to know if it makes sense. HWMonitor told me my CPU power consumption was 119.7 and my max voltage was 1.4V.
119.7W / 1.4V = 85.5 Amps.
123.8W (max TDP - 0.2) / 85.5A = 1.448V
Does this mean I can get to 1.448V stable?

I do have TurboV, it's a nice tool. I didn't realize it could do all this testing by itself, though. I've heard doing things manually is more accurate anyway.
 
Last edited:
Mar 6, 2012
122
0
0
I got it looking good on at 3.6GHz, but I'll need to reapply my TIM, it's running pretty warm.
attachment.php
 

samule

Junior Member
Oct 17, 2012
22
0
0
my old phenom 2 955..well it wasnt a legit 955 it was an unlocked 555 hit 4.0 ghz with an h100, however it got super hot even with the h100. it could have something to do with the fact it wasnt a "real" 955, and i couldnt monitor temps because i had unlocked it. and it wasnt very stable in prime, so i upped the voltage and it blue screened and wouldn't turn on for a few hours.

this being said 4.0 ghz is attainable with proper cooling, and since you can monitor temps it should just take some tweaking, you might just have to settle with 3.8 or so though.
 

KingFatty

Diamond Member
Dec 29, 2010
3,034
1
81
@KingFatty Wow thanks, that was perfect. Now I think I understand the different components of overclocking. Multiplier is, well, a multiplier of the core speed (CPU-NB), all of which relies on the voltage the socket is receiving. Cooling needs to be able to handle it too. I have a stress test running at 200 x 18 while I'm at work.

I did a little math while I ran Prime95 at stock to see temps (42°C max), and I'd like to know if it makes sense. HWMonitor told me my CPU power consumption was 119.7 and my max voltage was 1.4V.
119.7W / 1.4V = 85.5 Amps.
123.8W (max TDP - 0.2) / 85.5A = 1.448V
Does this mean I can get to 1.448V stable?

I do have TurboV, it's a nice tool. I didn't realize it could do all this testing by itself, though. I've heard doing things manually is more accurate anyway.

I can't recall the specific upper safe voltage for day-to-day operation, but generally when overclocking you can use a higher voltage to make things more stable - generally you raise voltage to get more stable, the higher the voltage, the more stable... up to a point.

It's important not to let the temperature get too hot, so that might be your constraint.

Your limiting factor will be the temperature, and to some extent, the voltage. You can't use a crazy high voltage, it will wear out the chip (but do you care if it dies in a year or two?). Lower voltage allows lower temperatures, but higher voltage may be necessary to maintain stability which causes higher temps. So it's a balance.

One way could be to just pick a voltage like you've done, and then see how you can maximize performance at that voltage by raising speeds until you either become unstable or start exceeding safe temps. But better to verify what's a comfortable max safe voltage for that chip, assuming you have decent cooling.

It's best to use manual settings for overclocking, agreed. But to get the highest overall chip performance, you have to test a lot of combinations of multiplier and base frequency. It's fairly safe to just keep the base frequency the same (say, 200) and then change the multiplier only. But, you may find that you'll hit, say, a multiplier of 19x max, and if you try 19.5x, it's unstable. So 19x200=3800 is ok, but you can't get to 19.5x200=3900. But it may turn out that you could have gotten your core frequency all the way up to 220, if you'd just tried that when you were set to 18x. That would be 18x220=3960. So that's where you can use TurboV to save a lot of time, b3ecause it's time consuming to go through each multiplier (17x, 17.5x, 18x, 18.5x, 19x, 19.5x, 20x) and with each one try to find the max base frequency. So really it's just to give you an idea of where to look if you are searching for the highest possible combination of multiplier x core freq. But really, I mean if you get to 3800 by just adjusting the base freq to 19x, is that good enough?

Probably, but if you want to really get into it, you can make a spreadsheet and everything and keep track of your progress (after a while the combination of numbers will be too numerous to remember).
 
Mar 6, 2012
122
0
0
My temps were a little higher than they will usually be on that first test. I usually only use this computer around 8PM-1AM, and I open the window to further cool it, but that first test was started at 4PM. After 20 minutes at 3.8 I'm only at 43.1C max right now.

Looking at TurboV, my base clock is bouncing around between 197.3and 202.7. Is this a good thing? Also, I just got a little warning telling me "+5V @ 5.078V"

As for chip life, I honestly might be getting myself an i7-3370K if it's really worth $400-$500 for the CPU and motherboard. So 2 years is plenty, and I'll be going off to college not this year, but the year after, so I'm not sure what's happening to this computer. Any tips for that? Haha
 

Vectronic

Senior member
Jan 9, 2013
489
0
0
Variance is normal (±1.5Hz), but that seems like a lot... but, haven't used an AMD in a long time.

ATX Power Supply

In short: ±0.05v is acceptable, so yours is 0.038 above.
 
Mar 6, 2012
122
0
0
Okay, my test at 4.0 GHz was running great until about 22 minutes in. I got an error which I have circled in the picture, and it only happened on core #3. What would be causing this? Otherwise, the whole test looked great to me.

attachment.php
 
Last edited:

Vectronic

Senior member
Jan 9, 2013
489
0
0
Voltage... always with the voltage.

You should be pretty close though, no BSOD, so maybe 1.37.

Lots of droop though... 1.425 down to 1.356 at load?
 

pyjujiop

Senior member
Mar 17, 2001
243
0
76
You're on the right track. Just give it a little more juice. If you have good cooling, and you clearly do, you can feed 1.5 or even 1.55v into a Deneb (although not a Thuban) and not damage the chip. It looks like you're at 1.425 already, so 1.45 or 1.475 might be just enough to stabilize it. Make sure you have LLC enabled in the BIOS, it helps.

Just about every 965 I see people buy these days seems to be able to hit 4 GHz. Nothing like an end-of-the-production-run CPU for a great overclock, and these are steals at the prices they sell them for now. The only thing AMD has better that you can still buy are the 8-core Piledrivers, at twice the price.
 

codyray10

Senior member
Apr 14, 2008
854
4
81
I have a similar setup as you (see sig below). My mobo has some pretty bad vdroop, but I run my 965 at 4ghz with 1.5v and NB @ 2600mhz on stock voltage. I have relatively good cooling in my antec 900 case. My computer is also in the coolest room in my place, with ambient temps of 66-68F.
 
Last edited:

mojothehut

Senior member
Feb 26, 2012
354
6
81
I'm running my 975 at 4.0 at 1.476v.
Haven't really tried to oc HTlink, still at 2000mhz
Think I'd have much of a performance increase jacking it up a little higher?
I'm running on an Asus Sabertooth 990fx mobo
 

mojothehut

Senior member
Feb 26, 2012
354
6
81
Also a quick question about CPUID hardware monitor
Under temperatures, it shows my CPU (running prime95 for 20mins) at 52c
Then there's another field that shows the temps of all four cores. They can be around 60-62c :confused:

Which temp should I be worried about, I know generally everyone says keep the Phenom II under 60. Well one readout is fine, 52c but the ones for each core seem high =\
 
Mar 6, 2012
122
0
0
I got it all stable at 4.0 with 1.45V, but than I got really greedy and went for 4.4 on 1.5V and BSOD. I think 4.0 is plenty for me though, and I also tightened up my RAM XMP so that it's 9-9-9-24-1T.

Aren't the eight-cores (8350 I think) just AMD's shot at a hyperthreaded quad core? That's what I've read.