PDA

View Full Version : 4Ghz and beyond: A PII X4 965 Overclocking Adventure


Madpistol
12-17-09, 11:18 PM
I will be using this thread to report results of what I'm able to do with this CPU. I believe I can reach 4Ghz fairly easily now that I've got my TRUE 120 bolted on to it.


Processor: AMD Phenom II X4 965 BE (125w)
Motherboard: Gigabyte UD3H 790GX AM3 motherboard
Memory: G.Skill PC3 12800 (DDR3 1600) 2x2GB

Testing program: Orthos (2 instances) stressing 4 threads.
Test time: 10 minutes per run w/ periodic overnight testing for full stability.



Test 1: 3.4Ghz (17x multiplier, 200mhz FSB) @ 1.2Vcore, RAM: 1333mhz (7-7-7-20 2T) - PASSED Temps (idle/load): 27/40

Test 2: 3.485Ghz (17x multiplier, 205mhz FSB) @ 1.2Vcore, RAM: 1366mhz (7-7-7-20 2T) - FAILED

Test 3: 3.485Ghz (17x multiplier, 205mhz FSB) @ 1.25Vcore, RAM 1366mhz (7-7-7-20 2T) - PASSED Temps: 28/42

Test 4: 3.570Ghz (17x multiplier, 210mhz FSB) @ 1.25Vcore, RAM 1398mhz (7-7-7-20 2T) - FAILED

Test 5: 3.570Ghz (17x multiplier, 210mhz FSB) @ 1.3Vcore, RAM 1398mhz (7-7-7-20 2T) - PASSED Temps: 29/45

DiscipleDOC
12-17-09, 11:50 PM
I made this thread a sticky for now.


Good luck with your tests! :thumbsup:

CaptNKILL
12-17-09, 11:50 PM
Nice job man. :D

Madpistol
12-18-09, 12:15 AM
Thanks for the temporary sticky! :D


Now the fun really begins...

I pushed it to 215mhz FSB, and it looked stable, until I tried to run a total system performance test... :headexplode: Now that I've seen that this processor can't take leaps on blind faith, I've decided to increase the CPU Vcore to 1.4V and leave it there while I take baby steps up the ladder. It looks like I'm starting to hit some RAM limitations too (low voltages) so I'm going to keep tabs on that as well incase it's not fully stable.

Onward and Upward!



Test 6: 3.655Ghz (17x multiplier, 215mhz FSB) @ 1.4V (RAM 1431mhz 9-9-9-24 2T) - PASSED Temps: 32/52 (BONUS: Ran Windows 7 performance test: 5.9 (stupid Hard Drives!!!! :fu: )

Test 7: 3.689Ghz (17x multiplier, 217mhz FSB) @ 1.4V (RAM 1445mhz 9-9-9-24 2T) - PASSED Temps: 32/52

Test 8: 3.736Ghz (17x multiplier, 219mhz FSB) @ 1.4V (RAM 1458mhz 9-9-9-24 2T) - PASSED Temps: 32/52 (BONUS: Cinebench 10 64-bit - 1cpu: 4069 Xcpu: 14025)

Test 9: 3.768Ghz (17x multiplier, 221mhz FSB) @ 1.4V (RAM 1471mhz 9-9-9-24 2T) - PASSED Temps: 32/52

Test 10: 3.806Ghz (17x multiplier, 223mhz FSB) @ 1.4V (RAM 1485mhz 9-9-9-24 2T) - PASSED Temps: 32/52 :D

Test 11: 3.860Ghz (17x multiplier, 227mhz FSB) @ 1.4V (RAM 1498mhz 9-9-9-24 2T) - PASSED Temps: 32/52

Test 12: 3.902Ghz (17x multiplier, 229mhz FSB) @ 1.4V (RAM 1525mhz 9-9-9-24 2T) - FAILED :(


Overnight Test 1: 3.860Ghz (17x multiplier, 227mhz FSB) @ 1.4V (RAM 1498mhz 9-9-9-24 2T) - FALIED

Now, this could have been for many reasons. All I know is that I woke up this morning and my computer had been restarted. However, I did not get any sort of "reports" that the computer had crashed, which tells me that either a power blip knocked it offline (it's raining really hard right now) or a windows update caused it to restart on its own. Either way, it was unavoidable.

Thank you for the hints about increasing the Northbridge clock frequency. I'll start trying to slowly push that up as well. :)

bob saget
12-18-09, 01:37 AM
wow nice results.

Viral
12-18-09, 01:42 AM
Looking good man. Remember, once you get your max clock, try overclocking the NB, it should make it to at least 2.6GHz if not somewhere near 3GHz up from the standard 2GHz. It makes a huge difference.

Johnny C
12-18-09, 10:32 AM
Looking good man. Remember, once you get your max clock, try overclocking the NB, it should make it to at least 2.6GHz if not somewhere near 3GHz up from the standard 2GHz. It makes a huge difference.

Makes a huge difference on all phenoms....

Maverick123w
12-18-09, 11:10 AM
The Phenoms still have a hard time hitting 4ghz on x64?

Madpistol
12-18-09, 11:51 AM
Day 2 of testing.



First, I ran it at the overnight settings for 30 minutes to make sure it was stable, and sure enough, with me surfing the internet and rearranging files on my computer, the system remained completely stable through everything, so I'm going to continue to push forward. I'll put an emphasis on 24/7 stability once I get it to 4Ghz. I'm getting close now.


Test 13: 3.908Ghz (17x multiplier, 229mhz FSB) @ 1.425V, (RAM 1525mhz 9-9-9-24 2T) - PASSED Temps: 33/54

Test 14: 3.919Ghz (17x multiplier, 230mhz FSB) @ 1.425V, (RAM 1531mhz 9-9-9-24 2T) - FAILED - Left it on for 30 minutes, and it locked up (not BSOD)


Since I got a system lockup rather than a BSOD, It's apparent that the RAM isn't happy now. I've loosened the timings and I'm going again.


Test 15: 3.919Ghz (17x multiplier, 230mhz FSB) @ 1.45V, (RAM 1531mhz 10-10-10-26 2T) - FAILED - BSOD


Increase NB and ram voltages


Test 16: 3.919Ghz (17x multiplier, 230mhz FSB) @ 1.45V, (RAM 1531mhz 10-10-10-26 2T) - PASSED 35/57

Test 17: 3.933Ghz (17x multiplier, 231mhz FSB) @ 1.45V, (RAM 1538mhz 10-10-10-26 2T) - PASSED 35/57

Test 18: 3.972Ghz (17x multiplier, 233mhz FSB) @ 1.45V, (RAM 1551mhz 10-10-10-26 2T) - FAILED - BSOD

Test 19: 3.972Ghz (17x multiplier, 233mhz FSB) @ 1.475V, (RAM 1551mhz 10-10-10-26 2T) - PASSED 35/57

Test 20: 4.005Ghz (17x multiplier, 235mhz FSB) @ 1.475V, (RAM 1565mhz 10-10-10-26 2T) - FAILED - BSOD

Test 21: 4.005Ghz (17x multiplier, 235mhz FSB) @ 1.5V, (RAM 1565mhz 10-10-10-26 2T) - FAILED - BSOD

Test 22: 4.005Ghz (17x multiplier, 235mhz FSB) @ 1.525V, (RAM 1565mhz 10-10-10-26 2T) - PASSED - 38/63

Redeemed
12-18-09, 01:22 PM
The Phenom II's can take *a lot* of voltage... so don't be afraid to raise it higher if need be. You should be able to hit 4GHz with ease, and even higher. Just raise the voltage. ;)

Madpistol
12-18-09, 02:19 PM
The Phenom II's can take *a lot* of voltage... so don't be afraid to raise it higher if need be. You should be able to hit 4GHz with ease, and even higher. Just raise the voltage. ;)

I read it was able to handle something like 1.8V, so that's inspiring. However, I've still got to contend with the thermal limitations of this processor, so I have a feeling that 1.55-1.6V will be about the highest I'm able to push it without going to a WC setup. I'm trying for 4Ghz on air.

Maverick123w
12-18-09, 02:34 PM
Looking good!!! Keep at it

Madpistol
12-18-09, 04:03 PM
It finally passed, and I have 4Ghz on my Phenom II X4 965! Now I'm going to begin a series of stability tests where I will start dropping the voltages/RAM timings, and see if I can get the system to crash. To do this, I will be running 4 threads of Orthos and playing games as well as other tasks on my computer. Like I said, I want to get the system to crash so I will know what the limitations are for this system.


Let the battle begin!



Stability tests for 4.005Ghz @ 1.525V, memory (1.65V) @ 10-10-10-26 2T)

Stability Test 1 - Dropped NB to 1.275V - FAILED - Temperature threshold exceeded. :(



Now I've got a unique problem. My room at ambient temps is around 68-70F. However, once the heater comes on, it quickly rises to around 75F. This means that my CPU temperature also rises, and it goes beyond the thermal threshold and BSOD's. I hate it, but that's the price of having a gas heater.

I'll continue this more in a few days. I'm going to play some games @ 3.9Ghz and see how this baby performs. :)

williv
12-18-09, 05:06 PM
Congrats on the OC man, 4ghz can be a tough one to get stable but it looks like your definitely close.
A bunch of us over on the AMD forums have had our overclocking adventures up in an ongoing thread for a while now, some of the guys were fine @ 3.9 but ran into stability issues @ 4ghz and above.

From my own experiences you need a vcore of 1.5v to keep stable at 4ghz for everyday use (gaming, benching etc), also need to bump up the cpu-nb voltage in order to keep the nb stable at 2600-2700mhz.

Here's a shot of cpuz with my latest everyday clocks. It's a really bad ass little cpu.

http://img692.imageshack.us/img692/6618/33223003.th.png (http://img692.imageshack.us/i/33223003.png/)

I do have the HT set too high though, I'm gonna drop it back down to 2200-2400, no real need for it to be any higher.

As for temps, I had the stock cooler on there for the first 2 weeks, temps were not nice, not nice at all. I picked up a Corsair H50 and both Idle and Load temps dropped 20 + degrees, so far I'm extremely happy with it.

Keep at it and I have no doubt you'll be able to lock her in at 4.

Viral
12-18-09, 10:28 PM
I'd be trying 18x 223 and lower the RAM timings back to 9-9-9, lower timings mean more than slightly faster memory IMO.

shaolin95
01-02-10, 01:45 PM
Good job there mate!
My previous Phenom 955 C2 was able to hit 4Ghz prime stable but only on XP, on W7 64 bits I was able to get 4Ghz but only OCCT stable not prime95. Still 3.9 was a piece of cake on 64 bits.

With my new 965 C3 I am able to do 4Ghz prime stable on W7 64 bits with only 1.44 vcore at NB2600.
I can also get it to do 4.1 Ghz prime stable but it takes 1.53vcore for that jump.
Tried going for 4.2 but even at 1.56vcore I was getting the computer to restart after 15-20mins of prime95...so it seems 4.1 is the limit on my chip with non crazy voltages at least. :headexplode:

Johnny C
01-02-10, 04:48 PM
Good job there mate!
My previous Phenom 955 C2 was able to hit 4Ghz prime stable but only on XP, on W7 64 bits I was able to get 4Ghz but only OCCT stable not prime95. Still 3.9 was a piece of cake on 64 bits.

With my new 965 C3 I am able to do 4Ghz prime stable on W7 64 bits with only 1.44 vcore at NB2600.
I can also get it to do 4.1 Ghz prime stable but it takes 1.53vcore for that jump.
Tried going for 4.2 but even at 1.56vcore I was getting the computer to restart after 15-20mins of prime95...so it seems 4.1 is the limit on my chip with non crazy voltages at least. :headexplode:

W7 is kinda ghey for overclocking I was running 3.2 in xp, 3.0 in vista, and now I'm running 2.86 in w7.....due to random BS blue screens....only benefit is that I'm currently running 1.25vcore @ 2.8 which runs pretty cool.

shaolin95
01-02-10, 04:59 PM
Sorry but that shows me that your current OC is not as stable as you think.
From my 3 tries at OCing phenoms the biggest difference in OC on all 3 between XP and W7 64 bit was 100 Mhz...

Drolfrawd
01-02-10, 05:52 PM
Great thread mates, i'll be following this closely

Johnny C
01-02-10, 09:59 PM
Sorry but that shows me that your current OC is not as stable as you think.
From my 3 tries at OCing phenoms the biggest difference in OC on all 3 between XP and W7 64 bit was 100 Mhz...

I guess you don't consider 12hrs prime...to be stable...cause those were my stable overclocks...

It's possible that the chip is aging somewhat....I could probably have maintained the 3.2ghz o/c with vista...but I didn't want to add more volts....3.2ghz was @ 1.325vcore in xp...and there was quite a while between xp and win 7

shaolin95
01-02-10, 10:05 PM
Could be voltage/temp "aging" the chip indeed then. Yeah 12 hour prime is perfect in my book.

Johnny C
01-02-10, 10:20 PM
Could be voltage/temp "aging" the chip indeed then. Yeah 12 hour prime is perfect in my book.

I usually run a bunch of quick benches....then 1hr of OCCT then prime overnight.....that's my stress test.

I run Memtest86 off a thumb drive...then run it again in windows for at least an hour..

Edit:

Ok folks I'm back to 3.0ghz in Windows 7, I actually took the side panels off today and noticed that my poor HSF was full of woolies....so after a liberal hosing with compressed air...I'm back up to running 3.1ghz 225X13.5 @ 1.3vcore..Idle is 34'C full loaded OCCT or Prime Quad is 51'C

Booyeah!

Johnny C
01-17-10, 04:57 PM
@ MadPistol,


Could you post your timings and voltages also your HTT and NB frequency as well as volts.

I'm working on getting 4.0 Stable in Win7 x64.

I've noticed so far that things seem more stable with the NB clock higher than the HT link speed. I knew that the HT speed should never exceed the NB clock....but I've been having stability problems even if they are the same speed. Clocking the NB higher seems to be working....just looking for some additional info.

I can boot to Win 7 @ 4.2 @ 1.55vcore (1.51vcore after vdroop)

So far I've been fairly successful with this O/C. I had 3.8ghz dialed in last night @ 1.45vcore(~1.4v after droop) I ran OCCT for 6.5hrs overnight. I'd really like to get 4.0 dialed in.