CPU-NB overclock issue

[Nakkiel]

I overclocked it from 2400 MHz to 2700 MHz (0.5 multiplier) and at first it worked fine. I hadn't done any stress testing yet, but did a Prime95 and Windows Experience benchmark with no problem.

About an hour later after restarting a couple times I tried lowering the HT Link and benchmarking it against the higher HT Link. (2400 vs 2700) After I bench marked it and restarted my computer it wouldn't work on either HT setting with the 2700 CPU-NB, and I had to restart several times via plug pull and then load off saved profile in start up. It wouldn't even let me get into bios until it randomly gave me the option to do that one of the start ups.

Any idea what I'm doing wrong? I did try increasing the CPU-NB voltage by 0.05, that didn't help.




System settings before CPU-NB OC, if you need any more info let me know.
CPU-Z TXT Report
-------------------------------------------------------------------------

Binaries
-------------------------------------------------------------------------

CPU-Z version 1.57

Processors
-------------------------------------------------------------------------

Number of processors 1
Number of threads 4

APICs
-------------------------------------------------------------------------

Processor 0
-- Core 0
-- Thread 0 0
-- Core 1
-- Thread 0 1
-- Core 2
-- Thread 0 2
-- Core 3
-- Thread 0 3

Processors Information
-------------------------------------------------------------------------

Processor 1 ID = 0
Number of cores 4 (max 4)
Number of threads 4 (max 4)
Name AMD Phenom II X4 970
Codename Deneb
Specification AMD Phenom(tm) II X4 970 Processor
Package Socket AM3 (938)
CPUID F.4.3
Extended CPUID 10.4
Brand ID 14
Core Stepping RB-C3
Technology 45 nm
TDP Limit 124 Watts
Core Speed 4200.1 MHz
Multiplier x FSB 14.0 x 300.0 MHz
HT Link speed 2400.1 MHz
Instructions sets MMX (+), 3DNow! (+), SSE, SSE2, SSE3, SSE4A, x86-64, AMD-V
L1 Data cache 4 x 64 KBytes, 2-way set associative, 64-byte line size
L1 Instruction cache 4 x 64 KBytes, 2-way set associative, 64-byte line size
L2 cache 4 x 512 KBytes, 16-way set associative, 64-byte line size
L3 cache 6 MBytes, 48-way set associative, 64-byte line size
FID/VID Control yes
Min FID 4.0x
P-State FID 0xC - VID 0x07 - IDD 21 (14.00x - 1.462 V)
P-State FID 0xC - VID 0x10 - IDD 15 (14.00x - 1.350 V)
P-State FID 0x6 - VID 0x18 - IDD 11 (11.00x - 1.250 V)
P-State FID 0x100 - VID 0x28 - IDD 4 (4.00x - 1.050 V)

 

[GObonzo]

to get my NB from 2000MHz to 2400MHz I have to increase voltage from 1.1v to 1.25v.
Phenom II 955BE on AMD 770 board.
I get better scores in OverDrive and 3DMark running NB at 2200MHz 1.2v with CPU @ either 3.4GHz or 3.6GHz. @ stock 3.2GHz I get better scores with the NB @ 2000MHz.
HT Link should stay @ 2000MHz.

 

[Nakkiel]

to get my NB from 2000MHz to 2400MHz I have to increase voltage from 1.1v to 1.25v.
Phenom II 955BE on AMD 770 board.
I get better scores in OverDrive and 3DMark running NB at 2200MHz 1.2v with CPU @ either 3.4GHz or 3.6GHz. @ stock 3.2GHz I get better scores with the NB @ 2000MHz.
HT Link should stay @ 2000MHz.

Well that's great and all but doesn't really answer my question. My voltage is already at 1.2, and as I said 1.25 didn't make it more stable. Do I need to try higher? Why would it be working for awhile then stop working when i decrease a setting? (HT Link) Are there other voltages to consider for CPU-NB overclock?

And I got MUCH better scores with the 2700 CPU-NB, which is why I want to get that booting every time and stable. It was .1 boost on windows experience along with a .2 boost to RAM performance, and significant boost on Prime95 as well. Lowering HT Link did boost my prime 95 results very slightly though.

 

[GObonzo]

with the NB @ 2600MHz (@ any voltage) if I boot at all it doesn't usually take long to crash but sometimes it would run through a few sessions. if you read overclocking guides they will all tell you there is a"sweet spot" you will reach with the different multipliers and voltages, whether some be higher or lower than you expected or want them to be. there's just that spot between our CPU, NB, HT that they need for stability and we have to find it.
I get this same type of problem sometimes when playing with the NB and HT. After certain NB and HT overclocks I'll become unstable even back at stock settings for a short time. sometimes it'll be 5-6 hard restarts and even will have to leave the pc off for a while and then it's fine again. I'm hoping someone will have an answer for us here, I'm not preaching to you. my point with my NB clock/volts is that if my 2400MHz needs 1.25v I would imagine yours @ 2700MHz would need way more than 1.25v.
helps to get an answer if you post your cooling methods. most posts I've read these type things end up being a heat issue if not voltage.

 

[Nakkiel]

It technically hasn't crashed. It just booted up fine several times then started refusing to boot. I have tried increasing the CPU-NB voltage since then, but have yet to get it to boot again at all with the 2700 CPU-NB. Also I read a couple product reviews where they had the CPU-NB slightly higher, roughly the same settings otherwise.

I don't think its heat. My cooler is fairly good.. and I really doubt it would have heat issues at idle.
http://www.newegg.com/Product/Product.aspx?Item=N82E16835181013

 

[Nakkiel]

I got it to boot by lowering the FSB and FSB : DRAM ratio (FSB lower, everything else adjusted to be the same), and managed to get the CPU-NB ratio to 2640. Not sure if its stable yet, but if it is any clue on how to work my way back up to the higher FSB?

 

[Nakkiel]

For the Phenom II’s however, you need to wait all the way until you can be sure the IMC (NB) is stable. The NB is stressed the most during 512K FFTs. This happens about 2-3 hours into Blend, so I suggest running for at least 3 hours.
n order to achieve 4GHz 100% stable, you need 32-bit windows and you will have to back off memory and NB clocks maybe even going below stock. However, it would net you greater performance with a 3.8GHz and 2600-2800MHz NB.

Right now I'm running large FFTs to see how hot it gets. Seems alright so far, but I'm not even an hour in. Its been fairly consistent 55C, highest 56 (briefly hit during the 1024K), lowest 54 (Briefly during 896K).

Another question I have (Sorry for so many) is with the CPU-NB higher I am noticing significantly higher ratings on memory from windows experience index. I don't know the mechanics of how that works.. should I be loosening ram timings to compensate or is it essentially just the cpu performing better, "allowing" them to do more?

Edit: definitely not stable yet, just got an error on Prime95, going to try running memtest to see if its actually the CPU that needs more voltage or the ram.

 

[Nakkiel]

Best answer selected by Nakkiel.

 

[Nakkiel]

n order to achieve 4GHz 100% stable, you need 32-bit windows and you will have to back off memory and NB clocks maybe even going below stock. However, it would net you greater performance with a 3.8GHz and 2600-2800MHz NB.

64-bit windows, 8 gig memory running at 1600 8-9-8-24 (that is it's stock though), 4.2 GHz processor speed, 2640 MHz NB atm and stable as far as I can tell. (Ran large FFTs and memtest for several hours each.) Just had to increase the CPU-NB voltage by .05 and decrease the FSB (with everything adjusted to be the same frequency) My cpu is getting considerably hotter, but its worth the speed boost i guess.

I still want to get the FSB back up to 300 if i can (adjusting all multipliers so everything stays the same otherwise) without lowering the NB, but I'll start a new question for it. Unfortunately its looking like i hit my limit on CPU-NB, if i push it further it will probably get too hot, and the multipliers only go in increments of x1, no x.5

 

[GObonzo]

"to achieve 4GHz 100% stable, you need 32-bit windows" Never heard anything like this before. If anyone can verify, please do. I've read many AMD configurations that were at or past 4GHz that said they were running WinXP, Vista, and 7 x64.

 

[green300]

With all four dimm's populated try increasing NB voltage

 

[Nakkiel]

"to achieve 4GHz 100% stable, you need 32-bit windows" Never heard anything like this before. If anyone can verify, please do. I've read many AMD configurations that were at or past 4GHz that said they were running WinXP, Vista, and 7 x64.

From what I've seen its very common to get this chipset past 4k on 64-bit.. I dont know if its "100% stable" in the escoteric 24 hour+ continuous stress test sort of way, but stable enough that it has yet to create any problems.