AMD® Athlon™ II Overclocking Thread

Big.Wayne · 20 Mar 2010 at 20:50

Ah I think that's the DDR3 version of the board I am using!

Big.Wayne · 20 Mar 2010 at 21:04

Take your time basically . . . get a feel for it, check out the "low" temps and then proceed to break it!

. . . . I did run my first AM2+ system at stock for a week and found that surprisingly fast . . . . although everyone has their own pace with these things . . . . I'd expect when your ready that 3.5GHz would be a good first overclock to aim for . . . . may need to help the vCore but apart from that no problems! :cool:

Big.Wayne · 21 Mar 2010 at 22:31

mymorph, what processor are you using?

Big.Wayne · 26 Mar 2010 at 01:50

Just thought I'd update this thread although I've not made much progress since my last post!

. . . just have been playing about trying to get the HT Link/CPU-NB MHz running a little faster . . while using [Auto] volts for both of them of course!

The HT Link didn't like running faster than 2128MHz when combined with the higher 3724MHz CPU frequency? bit odd as it is ok running at 2500MHz when combined with a lesser 3500MHz CPU clock? . . . not a biggie and probably something I can work through by manually boosting the HT Link voltage . . . . anyway I thought I would come back to the HT Link later and have a go getting the CPU-NB boosted, changed it's multi from [x8] to [x9] and that worked fine!

Click for stability . . . 24 hour "Blend"

Oi don't laugh!

. . . 2128 >> 2394 is an extra 12.5% CPU-NB Speed, not bad for a weeks work!

. . . . after excessive stability tests I left the machine folding for a few days . . . PPD not as good as I hoped but I suspect I may be getting some dodgy work units? . . . either that or Folding likes level 3 cache?

At 3.7GHz the PC is plenty fast, considering the minimal effort needed in the BIOS with most settings still on [Auto] and it's "utter" stability I think few will complain! . . . . really want a GPU now to play some games and do a few game benches!

So after getting 3724/2394/2128/1064 "validated" for stability I decided to see if I could get a bit more speed from the CPU-NB . . . 2394 >> 2660 . . . this was accomplished by changing the CPU-NB multi from [x9] back to it's default [x10], I also needed to manually boost the CPU-NB voltage to 1.2500v to even get Windows loaded properly!

Prime Blend didn't last long using just 1.2500vCPU-NB so slowly but surely that got boosted all the way to 1.300v then finally 1.325vCPU-NB to get "full" Prime stability . . . however, this is when I noticed something strange . . . . the CPU-Multi started dropping to [x4] while the processor was fully Prime loaded, the vCore would also drop down? . . . basically as if CnQ was kicking in except the chip was 100% loaded across all cores . . . It's not locked at this speed but appears to drop once or twice a minute and lasts 2-3 seconds?

Processor is at full 100% load in the above image . . .

I'm still not sure what is causing this, CnQ is disabled in BIOS, I don't have CnQ drivers installed and I also tried changing the CPU multi in BIOS from [Auto] to [X14] but still no joy? . . . the "Throttling" issue goes away if I reduce the CPU-NB clock and reduce the CPU-NB voltage to reasonable levels? . . .

I thought maybe this could be a safety feature of the processor or perhaps a "limitation" of the baby uATX board I am using? . . . It seems to be down to a combination of high vCore along with a high CPU-NB voltage, I did notice the core temps increase a little bit from the added CPU-NB voltage but not seem them rise above 49°C yet so I'm not convince this is a Thermal problem? . . . . checking in SiSoftware Sandra shows the Propus has it's Thermal protection set at 99.5°C

Deneb 965 vs Propus 630 (Cutout) . . . thanks to gareth170 for the pointer!

My personal Athlon™ II X3 Rana runs fine at 2.66GHz CPU-NB so I'm thinking that maybe the extra core of the *overclocked* Athlon™ II X4 Propus is pulling a bit more power than perhaps the motherboard can accommodate, it's the "Pro" DDR2 board below:

ASUS M4A785D-M PRO Webpage

Support CPU up to 125W
4+1 Phase Power Design

So maybe 4+1 phase is not enough to deliver a high *dual-plane* voltage to an overclocked quad-core?, I'll have to look into that but I notice some of the slightly beefier board state they have 8+2 Phase Power Design . . . hmmm

Another area I am suspecting is the PWM/Mosfet cooling . . or lack thereof!

. . . . I opened the case and had a good nose and yes indeed the PWM/Mosfets are running "Wickedly" hot while the machine is under heavy load, couldn't hold my finger-tips to them for much over a few seconds! . . . I've got a few spare BGA heatsinks knocking around so I may slap them on the PWMs and deploy some spot cooling to see if that helps rid me of the "Hokus-Pokus!" :cool:

Big.Wayne · 26 Mar 2010 at 05:24

Hey CAT, theres enough room to fit in a 80mm/92mm/120mm I believe, just been thinking of a way to attach it . . . do you think that the "throttling" problem could be connected with hot power circuitry? . . . I've been meaning to get some cooling down there but wanted to see how the board did at stock!

If on your travels you come across a uATX 785G AM2+ that features some Uber PWM cooling *and* 8-pin +12v ATX then let us know please!

Big.Wayne · 26 Mar 2010 at 17:55

Just whipped the machine open again to see what the dealio is . . . . the components highlighted in yellow below were really quite hot! . . . of course this is why most overclocking boards have a nice heatsink fitted here to keep the PWM/Mosfets from getting too toasty!

Didn't have enough BGA ramsinks so I thought as a test I'll just start with a fan and see how that cools, then if needs be go back and add some heatsinks . . . As the AM2+/AM3 motherboards have the CPU socket placed a little more centrally there was enough room to fit a 92mm fan totally flat or a 120mm fan at a slight angle . . . so went with the 120mm! . . . fits great, rests on part of the case/northbridge and is secured with a single cable tie . . .

It seems the fan alone is all thats needed to keep the Power-Delivery components nice and chilly . . . . Set the CPU-NB volts and Multi back to [Auto] giving 2.66GHz and has been Blending without issue for a few hours!

. . . I'd say the added Ghetto cooling has improved things a lot, I couldn't get it Blend stable before without manually setting 1.325vCPU-NB and that weird issue with the CPU "Throttling" down in speed appears to have gone away? . . . . fingers crossed this added cooling may have been the fix! :cool:

Big.Wayne · 26 Mar 2010 at 20:53

Nice one!

. . . I'd imagine it's a bit harder to clock the AIIX4 620 as it's got a X13 CPU multi? . . . Have you tried 266MHz Ht Ref. Clock= 3458/2660/2660/1064 . . . not a huge boost from where you are now but every little helps! :cool:

Big.Wayne · 26 Mar 2010 at 22:29

What "ram stability problems" have you been having? . . . I'm waiting to see your results!

Big.Wayne · 29 Mar 2010 at 20:19

Hi iTek,

35°C Processor load sounds good!

. . . so lets see where you are

AMD® Athlon™ II X4 620

2600MHz (13x200) - Memory 400MHz [1:2] DDR2-800

+20% [Auto] Overclock

3120MHz (13x240) Memory 480MHz [1:2] DDR2-960

Well considering you have DDR2-800 then that's pretty good going!

. . . If you wanted to make life a bit easier on yourself then it should be possible to lower the Memory-Multiplier in BIOS . . . that will take the strain of your RAM and let you concentrate on the CPU & Ht Ref Overclocking!

. . . . I'm not sure about your mobo or the BIOS layout but once you lowered the Memory-Multiplier you should see it in CPU [Memory] Tab FSB: DRAM [3:5]

Let me know if that works and if so we can work out your overclock a bit more! :cool:

Big.Wayne · 30 Mar 2010 at 14:21

iTek said:
Chipset ---> DRAM Configuration ---> Timing Mode / Memory Clock Frequency

Hi Gary,

that "Memory Clock Frequency" you mentioned above sounds like the setting you need to change? . . . I think the default is DDR2-800 [2:1] so it would help if you can lower that to DDR2-666 [3.2]

The ASUS M2A-VM HDMI motherboard looks quite nice, bit "dinky" but we will find out soon enough how it copes . . . I suspect it won't enjoy running a huge overclock as the power circuitary doesn't look that strong! . . . we will proceed with caution! :cool:

Big.Wayne · 30 Mar 2010 at 16:35

Hi Gary,

I just had a quick look at your motherboard specs and I can't see any listed support for the Propus?

. . . it does however show a few 95w TDP chips so that's cool . . . . I'm not sure how much pure clocking that board will alllow? . . . the power circuitry (PWM's etc) looks like it may really strain feeding an overclocked/overvolted QuadCore . . . . steady as she goes and make sure the area enlarged in the pic below doesn't get too toasty! :cool:

iTek said:
AMD® 690G/SB600

Big.Wayne · 30 Mar 2010 at 18:09

iTek, is the CPU voltage still on [Auto] ? . . . if you open CPU-z what figure is dispayed next to Core Voltage? . . . does it move about?, sag under load etc? . . . . you probably need to increase the VCore a little, stick with the lower memory multiplier for the moment!

The PWM's Mosfets, Capacitors etc all play a part in converting the current into CPU-Friendly format, the power flows from the 4-pin (8-pin) +12v cable then through the PWM/Mosfets/Capacitors/Phases etc before finally arriving at the processor . . . these components have an "optimal" load & operating temperature and they start to spaz out if pushed way beyond what there designed to run at!

. . . . Any disruption or fluctuation to the voltage stream can only mean one thing . . . invasion! of the BSOD!

What is new to me (having come from LGA775) is the way these chips have Dual-Plane voltage, one rail feeding the processor core and another rail feeding the CPU-NB/IMC/HT-Link . . . that makes it extra tough on the old power circuitry and also makes it doubly "prone" to errors caused by fluctuating voltages! . . . . Basically once you OverClock & OverVolt the CPU/CPU-NB/HT-Link and then load the machine to the max the power circuitry is gonna "explode" unless it is high spec and well cooled!

This enlarged ASUS sales blurb illustrates the Dual-Plane thing quite well! nice tasty 8+2 phase there with heatpiped PWM cooling, lush! :cool:

Big.Wayne · 30 Mar 2010 at 18:28

Put the "memory multiplier" back down so your DDR2 runs in spec and then increase the vCore a little and see if you can raise the CPU clock!

Big.Wayne · 30 Mar 2010 at 18:45

What is a "multiplier of 11" ??

Big.Wayne · 30 Mar 2010 at 19:09

iTek said:
Memory wont boot at 245 on a multiplier of 11.

this? . . . Memory multi or CPU multi?

Big.Wayne · 15 Apr 2010 at 19:17

frozennova: Nice to see your first results although I'm having problems reading the details of the clock? (midget screenie). . . is that 1.584vCore? . . . I wouldn't bother running Super-PI but instead try Fritz, Cinebench and WPrime . . . . I would have thought that chip could do 14x250 @ 1.400vCore?

iTek: Are you still using your DDR2-800 memory? . . . Nice new board you got there!

. . . have you worked out how to adjust the Memory frequency yet? . . . this will be a crucial setting for you if you intend to get some more overclocking progress as at the moment "if" your stick using memory rated at 400MHz (DDR2-800) along with a HT ref. Clock of 255MHz and 1 [1:2] memory multiplier your ram is now running at 510MHz (DDR2-1020) . . . any chance you can post a screenie like this please . .

CPU-z download unzip, doubleclick the CPU-z icon, when it's opened doubleclick the CPU-z icon to run a second copy . . . .

Big.Wayne · 15 Apr 2010 at 20:04

iTek: I'm not sure how much of the BIOS you have changed yourself? . . . is it just the vCore and CPU-NB/HT Link Multi? . . . I'm tempted to get you to reset the BIOS and start again?

Now you have a new mobo and Ram I would expect you should achieve a 250MHz Ht ref. Clock with little effort, to make life easier and less complicated for yourself at this early stage you could reduce both the CPU-NB/HT Link Multi to [x8], this means that when you dial in a 250MHz Ht ref. Clock you *won't* be clocking either the CPU-NB/HT Link and just the processor i.e

2600/2000/2000/800
CPU x13/CPU-NB x10/HT Link x10/2:1 Mem

3250/2000/2000/1000
CPU x13/CPU-NB x8/HT Link x8/2:1 Mem

For that clock the only voltage that may need taking off [auto] would be vCore . . . and at that speed I can't see the chip needing more than 1.280/1.300vCore? . . . if you can crack the above clock then you will be ready for the next step! :cool:

Big.Wayne · 16 Apr 2010 at 01:24

Just thought I'd make an update . . . going back to post #76 . . . adding the extra cooling on the power circuitry sorted out a lot of little niggles, that CPU throttling down to [x4] multi under extreme load has gone and I was "finally" able to get the CPU-NB/HT-Link speeds tweaked right up . . .

Click for stability . . . 30 hour "Blend"

The Loadline Calibration (LLC) option has also been quite useful, on the ASUS board setting it to 19% has almost eliminated vDrop & vDroop, this meant I was able to dial in a lower vCore . . .

On the two AMD® Athlon™ II chips I have tested "both" have been happy running the CPU-NB & HT-Link at 2.5GHz using stock voltages . . . pushing the CPU-NB past 2.5GHz with standard volts resulted in a BSOD so I dialled in 1.225vCPU-NB and that sorted it! . . . I may try lowering it to 1.200vCPU-NB in the near future . . . . I've also "witnessed" a connection between HT-Link speeds and standard vCore in as much as a chip running any given frequency needs less vCore if the HT-Link isn't pushed too much above 2.0/2.5GHz? . . . anyone working on a CPU clock may find it easier if they lower the HT-Link multi to [x8] or lower (depending on HT ref. Clock) . . . once you got the processor frequency worked out then you can do more testing with a higher HT-Link multi/speed and if you run into stability issues then I'm sure they will go away with a little nudge of vanilla vCore!

So moving on . . . . with the new improved PWM cooling I thought I would have a crack at 3.8GHz . . . dropped the memory multiplier down from [1:2] to [3:5] and also lowered both the CPU-NB/HT-Link speeds to help isolate the testing, nudged the HT ref. Clock up from 266MHz to 272MHz and slammed the vCore up to 1.600v

Click for stability . . . 8 hour "Blend"

Lol, nice 24/7 voltage there, I'm sure I saw the street lights outside my house "Flicker" when I stress tested the PC!

So having basically maxed the chip I thought what next? . . . how about a bit of low voltage testing? . . . I was gonna begin at stock 2800MHz but a nice round 3000MHz seemed a more suitable start point . . .

Click for stability . . . 35 hour "Blend"

Next step up was changing the CPU-Multi from [x12] to [x13] . . .

Click for stability . . . 37 hour "Blend"

And finally I wanted to have another go at the 3.5GHz clock which I did back in Post #50 to see if having LLC enabled allowed me to shave a little vCore . . . . which it did!

Click for stability . . . 32 hour "Blend"

UnderVolting Results:

3000MHz @ 1.200vCore
3250MHz @ 1.280vCore
3500MHz @ 1.376vCore

I think I will be running either 3.0GHz or 3.5GHz clocks for 24/7 usage, will get around to doing some power readings soon to see how much wattage this machine is drawing . . . . I think I'm nearly done with testing now, just need to jiggle the final clock to max out the DDR2-1066 speeds and then get some benchmarks published, until then thanks for reading and I'll try to help out with peoples clocks as much as possible! :cool:

If anybody quotes this "entire" post they will be shot!

Big.Wayne · 17 Apr 2010 at 23:12

iTek said:
Whats the next step?

Do a stability test @ 3250/2000/2000/1000 and if that passes publish your results . . . . then return the CPU-NB multi to [x10] and test for stability @ 3250/2500/2000/1000 and publish your results and settings, then return the HT-Link multi to [x10] and test for stability @ 3250/2500/2500/1000 and publish your results and settings . . .

Big.Wayne · 18 Apr 2010 at 16:33

Hey frozennova, nice results from a sub £50 chip . . . I'm wondering though if the CPU-Clock is being limited by the Geil Value DDR3-1333? . . .

Have you tried switching the Memory-Multiplier down from [3:8] to [1:2] in an effort to remove the strain from the memory and help "isolate" the CPU overclock? . . . . I'm not sure if all AM3 boards have the [1:2] memory option though?

iTek: As suggested by frozennova, download Memtest86+ V4.00 here and make a bootable CD-ROM and test both sticks separately at their rated speeds, you can also test the sticks in different slots if you have time . . . . really like the look of your new board (as suggested by CAT-THE-FIFTH in post #73) . . . it appears to have all the overclocker friendly options like a decent PWM heatsink and 8-pin ATX +12v as well as a decent BIOS etc

How hot does the PWM heatsink get while the system is running a heavy load?