Overclock.net › Forums › AMD › AMD CPUs › Problems Overclocking Athlon II X4 640
New Posts  All Forums:Forum Nav:

Problems Overclocking Athlon II X4 640

post #1 of 21
Thread Starter 
I am having a bit of trouble out of this CPU. I can overclock it to 3.42Ghz (228Mhz FSB) and it is stable for 10+ mins of Prime95 (blend). Any higher results in a BSoD long before 10 mins (usually in about 5). Here are some settings of the failed overclocking, and what I have tried to improve stability:

FSB: 230Mhz
Multiplier: 15x
NB Freq: 2312Mhz (have lowered the multiplier, while keeping NB Freq above HT Link Freq with no improvement)
HT Link Freq: 1850Mhz
Vcore: 1.385v (have taken it to 1.42 with no improvement)
CPU>NB Volt: 1.20v (taken to 1.24 with no improvement)
Ram Freq: 616 Mhz (666 Mhz is stock, timings are set to stock 9-9-9-24)
Ram Volt: 1.51v (1.5v is manufacturers recommendation)
CPU Core Temps: 23c idle, 49c full load from prime95 (tested the same from several programs)
PCIE Freq: 100Mhz (have increased to 110Mhz, this has helped with other processors in the past, but it did not with this processor in this situation)

Please note that the step ups to vcore:1.385v and CPU>NB:1.20v were made to get stability at 3.42Ghz. Anything less than 3.42Ghz can run with stock (auto) volts for both.

I have some decent experience in overclocking, and judging by these numbers, I should have plenty of head room to overclock more. I figure I should be able to hit 3.6 Ghz, with acceptable temps, and possibly go further. However, I am stuck at 3.45Ghz.

I realize that 10 minutes isn't an ideal test length to determine stability. I use a 10 minute test to check stability between FSB increments. Once I have found what I consider to be the highest frequency I can get with acceptable voltages and temps, I do a longer test (overnight or while I'm at work), usually after dropping the FSB by 1-2 steps to insure stability. At this point, I have not found an acceptable overclock, so the long tests have not been performed.

I have also run MemTest for 9 hours with the failed overclock settings (3.45Ghz) with no errors.

Voltage Settings and other settings that have not yet been listed:
CPU VDD Voltage: Auto
CPU>NB VDD Voltage: Auto
CPU Voltage: 1.385v
CPU>NB Voltage: 1.201v
NB Voltage: 1.132v
HT Link Voltage: 1.241v

NVidia Core Calibration: Disabled
Auto Disable DRam/PCI Frequency: Enabled (by default, what is this for? it doesn't seem to effect anything including stability)

Spread Spectrum and Quiet and cool are both off, as is the C1E feature (I think that's what its called)

Some other things I have tried with no success:
-Increasing CPU>NB VDD voltage
-Increasing NB voltage
-Increasing HT Link voltage
I reset these settings back to what i have listed above after none of them worked.

Anyone have any suggestions on what to try to improve stability?


UPDATE:
I ran a 30 min Prime95 blend test successfully with these settings:
FSB: 228Mhz
CPU Multiplier: x15 (3420Mhz)
HT Multiplier: x8 (1824Mhz)
CPU>NB Multiplier: x10 (2280Mhz)
DRAM Multiplier: x2.66 (608Mhz)
Ram Clocks: 9-9-9-24-1T
CPU Voltage: 1.385v
CPU>NB Voltage: 1.201v
NB Voltage: 1.132v
HT Link Voltage: 1.241v
CPU Core Temps: 23c Idle, 47c Max
Edited by Kyrle - 1/20/11 at 9:49pm
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
post #2 of 21
I am in the middle of OC'ing a Phenom II X2 and am having similar issues. One thing I have found is that I am having to give the CPU a LOT of juice to even reach 3.6 or 3.7GHz (stock is 3GHz). In order to even run a Prime95 test for a few minutes at 3.7Ghz, I have to put the Vcore at a whopping 1.5v (which is too high for my taste). Anything less than 1.5v and I get constant BSOD's and rounding errors on Prime95. So, maybe you have a CPU like this -- one that needs a ton of Vcore. Yes it sucks, but it is what it is. I envy those who say they can get 700MHz with only one small bump in Vcore. That is not my experience at all. Perhaps you are in the same boat.

Right now the highest I can get the CPU is 3.5GHz with 1.425v. That is as high of a Vcore as I want to run 24/7. I checked AMD's site and it says the max safe voltage for the 640 is 1.4v. So I wouldn't go much over that unless you have massive cooling (water cooling). And even then you are going to shorten its life.

So, my advice is to just step up the Vcore one notch when you see a BSOD or rounding error in Prime95. This seems obvious, but I spent a whole day testing my RAM (thinking it was the problem) only to find out that the problem was too low of Vcore. I assumed 1.45v was more than enough for my rather meager OC, but it wasn't. You said you have tried 1.425v already, so that might not be it.

The only other thing it could be is the CPU-NB voltage. I have my NB at 2400 and it needs a significant bump to run there (I think I have it at 1.275v right now). Some people get lucky and can hit 2400MHz on the NB without any CPU-NB increase. I am not so lucky. So, increase that CPU-NB voltage to 1.25v and retest.

Anyway, good luck.
Skylake Build
(12 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5-6600k Gigabyte Z-170 Gaming 7 Gigabyte R9 390  Gskill Ripjaws V DDR4 
Hard DriveCoolingOSMonitor
Samsung 850 Evo Corsair H115i Windows 10 Pro Asus  
KeyboardPowerCaseMouse
Generic EVGA NEX750 G1 Phanteks Eclipse P400 GSkill MX780 
  hide details  
Reply
Skylake Build
(12 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5-6600k Gigabyte Z-170 Gaming 7 Gigabyte R9 390  Gskill Ripjaws V DDR4 
Hard DriveCoolingOSMonitor
Samsung 850 Evo Corsair H115i Windows 10 Pro Asus  
KeyboardPowerCaseMouse
Generic EVGA NEX750 G1 Phanteks Eclipse P400 GSkill MX780 
  hide details  
Reply
post #3 of 21
Looks like the RAM is holding back your overclock. The memory itself may not seem erratic but everything else just might be. S'what you get for buying generics; you could've bought a 1600Mhz kit for the same price. The Silverline 1333 CL9 kit is one of the more notorious for not overclocking well.
post #4 of 21
Thread Starter 
Quote:
Originally Posted by xd_1771 View Post
Looks like the RAM is holding back your overclock. The memory itself may not seem erratic but everything else just might be. S'what you get for buying generics; you could've bought a 1600Mhz kit for the same price. The Silverline 1333 CL9 kit is one of the more notorious for not overclocking well.
Just to clarify. I'm not saying your wrong, nor that I'm right. Just sayin how its setup and why I never thought it was the problem.

My ram isnt even overclocked yet, In fact, when I first installed it, i did a 5 hour memtest (like i always do), before i install anything on the HD. It passed. Later I discovered that the MB set the timings too tight, at 8-8-8-21, I didnt change it. Now, I've changed that to stock. It's running at 9-9-9-24, at a lower clock speed.

I don't quite follow you on how my ram is holding me back when it ran fine when everything was set to stock with the advanced timings. I can still drop my ram multiplier further, if you think it will help, I can also increase the voltage a bit, as it is only .o1 above stock. I never considered them a problem since it has never failed memtest with any settings. Ive never seen an error.

On the reason i got this ram, It had good reviews and I got a steal on the price for 8GB. I knew about the lack of overclocking ability going into it, hence the stock timings and less than stock frequency.
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
post #5 of 21
Thread Starter 
Quote:
Originally Posted by thiussat View Post
I am in the middle of OC'ing a Phenom II X2 and am having similar issues. One thing I have found is that I am having to give the CPU a LOT of juice to even reach 3.6 or 3.7GHz (stock is 3GHz). In order to even run a Prime95 test for a few minutes at 3.7Ghz, I have to put the Vcore at a whopping 1.5v (which is too high for my taste). Anything less than 1.5v and I get constant BSOD's and rounding errors on Prime95. So, maybe you have a CPU like this -- one that needs a ton of Vcore. Yes it sucks, but it is what it is. I envy those who say they can get 700MHz with only one small bump in Vcore. That is not my experience at all. Perhaps you are in the same boat.

Right now the highest I can get the CPU is 3.5GHz with 1.425v. That is as high of a Vcore as I want to run 24/7. I checked AMD's site and it says the max safe voltage for the 640 is 1.4v. So I wouldn't go much over that unless you have massive cooling (water cooling). And even then you are going to shorten its life.

So, my advice is to just step up the Vcore one notch when you see a BSOD or rounding error in Prime95. This seems obvious, but I spent a whole day testing my RAM (thinking it was the problem) only to find out that the problem was too low of Vcore. I assumed 1.45v was more than enough for my rather meager OC, but it wasn't. You said you have tried 1.425v already, so that might not be it.

The only other thing it could be is the CPU-NB voltage. I have my NB at 2400 and it needs a significant bump to run there (I think I have it at 1.275v right now). Some people get lucky and can hit 2400MHz on the NB without any CPU-NB increase. I am not so lucky. So, increase that CPU-NB voltage to 1.25v and retest.

Anyway, good luck.
I have read that it is best to keep the NB between 2200 and 2400, which is what I have tried to do. What is a relatively save voltage for the CPU-NB? I could try increasing that in small bits until I get close to an uncomfortable voltage to rule that out as an issue.
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
post #6 of 21
Quote:
Originally Posted by Kyrle View Post
I have read that it is best to keep the NB between 2200 and 2400, which is what I have tried to do.
Nope. The higher the better. Various benchmarks you can find on Google show that a higher NB frequency = better performance (especially in memory intensive tasks). Indeed, some people prefer to keep their NB clock high and their memory speed at or below stock! What I mean is that when faced with either overclocking memory or overclocking the NB, most will tell you to choose the NB. A good NB overclock is 2600-2800. You might not be able to hit those speeds, however. I can only achieve a stable 2400 and that is with a decent increase to the CPU-NB voltage (everything above 2400 will either fail Prime tests or require too much voltage). Every chip is different. I have a uncooperative chip, maybe you will be more lucky.

Quote:
What is a relatively save voltage for the CPU-NB? I could try increasing that in small bits until I get close to an uncomfortable voltage to rule that out as an issue.
Some people will say it can be as high as the Vcore (i.e. 1.4-1.5v) while others say to keep it at 1.35v or below. Either way, 1.35 should be more than you need for 2400. Basically, get it as high as you can get it while staying at or below 1.35v.
Skylake Build
(12 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5-6600k Gigabyte Z-170 Gaming 7 Gigabyte R9 390  Gskill Ripjaws V DDR4 
Hard DriveCoolingOSMonitor
Samsung 850 Evo Corsair H115i Windows 10 Pro Asus  
KeyboardPowerCaseMouse
Generic EVGA NEX750 G1 Phanteks Eclipse P400 GSkill MX780 
  hide details  
Reply
Skylake Build
(12 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5-6600k Gigabyte Z-170 Gaming 7 Gigabyte R9 390  Gskill Ripjaws V DDR4 
Hard DriveCoolingOSMonitor
Samsung 850 Evo Corsair H115i Windows 10 Pro Asus  
KeyboardPowerCaseMouse
Generic EVGA NEX750 G1 Phanteks Eclipse P400 GSkill MX780 
  hide details  
Reply
post #7 of 21
Take your Vcore up more. I have this chip and in my experince, it takes 1.46Vcore to push past 3.45Ghz. Its a little dissapointing but Most of us are getting 3.6Ghz stable with this chip. I came down to 3.510Ghz because I wanted to bring my Vcore below 1.5.
BullDriver
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX-8350 Vishera 4.0GHz Asus 990FX Sabertooth MSI N465GTX 8gb Gskill Sniper 1600/9.9.9.25 
Hard DriveOptical DriveOSMonitor
64Gb Crucial SSD(os) +1tb Baracuda SataIII Asus OEM SATA DVD RW x64 Windows 7 Home Premium 25" Hanns G 
KeyboardPowerCaseMouse
Microsoft Flatty Tethered RocketFish 550w Modular Corsair 600t Logitech MX Revolution 
  hide details  
Reply
BullDriver
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX-8350 Vishera 4.0GHz Asus 990FX Sabertooth MSI N465GTX 8gb Gskill Sniper 1600/9.9.9.25 
Hard DriveOptical DriveOSMonitor
64Gb Crucial SSD(os) +1tb Baracuda SataIII Asus OEM SATA DVD RW x64 Windows 7 Home Premium 25" Hanns G 
KeyboardPowerCaseMouse
Microsoft Flatty Tethered RocketFish 550w Modular Corsair 600t Logitech MX Revolution 
  hide details  
Reply
post #8 of 21
Thread Starter 
Quote:
Originally Posted by thiussat View Post
Nope. The higher the better. Various benchmarks you can find on Google show that a higher NB frequency = better performance (especially in memory intensive tasks). Indeed, some people prefer to keep their NB clock high and their memory speed at or below stock! What I mean is that when faced with either overclocking memory or overclocking the NB, most will tell you to choose the NB. A good NB overclock is 2600-2800. You might not be able to hit those speeds, however. I can only achieve a stable 2400 and that is with a decent increase to the CPU-NB voltage (everything above 2400 will either fail Prime tests or require too much voltage). Every chip is different. I have a uncooperative chip, maybe you will be more lucky.



Some people will say it can be as high as the Vcore (i.e. 1.4-1.5v) while others say to keep it at 1.35v or below. Either way, 1.35 should be more than you need for 2400. Basically, get it as high as you can get it while staying at or below 1.35v.
You said that higher NB frequencies are better for more performance. That makes sense, but for the momment, performance is less of a concern than stability. Will having that high of a NB clock make it much harder to configure a stable setup?

For now, I am also trying to focus on the CPU frequency, while keeping everything else near stock until I have the processor setup and stable (then I'll come back and tweak the other settings to increase performance). The only reason I have allowed the NB frequency to overclock is like I said before, Many people say that when overclocking the CPU, having a NB of 2200-2400 helps to increase stability vs. leaving it around the stock 2000. If this is not true, I'd rather just put it near stock for now, or keep letting it rise beyond 2400 whichever will help stability.

Quote:
Originally Posted by cook View Post
Take your Vcore up more. I have this chip and in my experince, it takes 1.46Vcore to push past 3.45Ghz. Its a little dissapointing but Most of us are getting 3.6Ghz stable with this chip. I came down to 3.510Ghz because I wanted to bring my Vcore below 1.5.
I'll bump it up to 1.5 and see what that will let me clock to a little later. I'm fairly sure this will be the highest voltage I've ever applied to a CPU, which is part of the reason I'm still in the 1.3x range. While my temps are still reasonably low and not an issue, I am unsure as to how much voltage is too much in terms of any electrical damage that might happen.

I understand that 1.55 is the maximum voltage as stated by AMD (is this accurate? I haven't personally checked.), however, ignoring temperature, what is the maximum voltage I can expect to apply before I have to worry about damage (voltage damage, not thermal damage)? Usually the manufacturers ratings can be exceeded, since there is some amount of safety margin in place with most electronic components. How much of a margin is usually harder to determine from datasheets and recommended specifications.

Expected life isn't really a concern as long as it will last a reasonable time (6 months to a year maybe), considering Ive never had a chip die on me before it is replaced by an upgrade. I have had chips die, but only old ones that were not in my primary system.

I hear lots of people say that 3.6 is the upper limit to this chip. Then, quite often someone else will come in and say that is only true for the C2 stepping, Mine is C3, is there a trick to using the C3 stepping for higher clocks that I'm just missing? or are these 3.8Ghz+ clocks only for show with no usable value (as in being unstable)? What Stepping is yours?
Edited by Kyrle - 1/21/11 at 5:25pm
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
BlueBox
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon II X4 640 (Propus) 3.0Ghz (3.4Ghz OC) MSI NF750-G55 Asus ENGTS450 TOP Mushkin Enhanced Silverline 996770 DDR3 1333 2x4GB 
Hard DriveOSPowerCase
Seagate ST3500418AS (500GB) Windows 7 Ultimate (x64) RaidMax RX-730SS 730W RaidMax Quantum 
  hide details  
Reply
post #9 of 21
If you can run at 8-8-8-21 without any stability issues you should be running at 8-8-8-21. Why: it's not 9-9-9-24, which is not optimal for low-timing-favouring AMD.

Generic memory (aka 1333 9-9-9-24 rated) usually doesn't overclock well because generic memory isn't made to overclock; this memory is usually used in prebuilts and systems that cannot set the timings. The generic settings are at the absolute bottom end of DDR3 and any DDR3 chips/ICs can be made to run at those with 1.5V, so as a result, often lower-binned or worse-overclocking ICs are used, which often hold back your overclock if you use the ref. clock/FSB (and you have to because the 640 is locked). To sum it all up, there's usually always a 1600 9-9-9 memory kit for the same or lower price (which ~70% of the time overclocks better depending on the kit) which is why I'm hereditarily against 1333 CL9 and often facepalm at people using such memory.
post #10 of 21
Yes 1.55 is the highest. I have heard of a few people going to 1.57 or around there. I would be a little weary of that, but I have been running 24/7 with average constant load at 1.5v with no issues.

The major argument involved with using a higher Vcore is that it could lower your processor life span- which is supposed to be like 10 years. But seriously are you going to be using this X4 640 in ten years? That would be like using an early generation P4 processor today. These little propus processors need the extra voltage to go the extra GHZ mile.

The C3 stepping drew ohhs and ahhs on when these chips first came out, but for some reason they have the same 3.6 upper limit as the earlier C2's. There is this review that came out a while back when the 640 chips first released that has the 640 oced to 3.8 but the vcore was a ridiculous 1.6. That's insane! When these chips first came out they where locked deneb's and that is why I think there is a small rash of High overclock threads that got everyone's hope up. There where a few reports of being able to unlock the L3 catch and that is what gives away that they where indeed deneb dies and not propus like the ones we are tinkering with.

Hear is one of those reviews.

And another

Maybe these truly where Propus 640's but if they where I would bet they where binned and the best of the best of the chip.
Edited by cook - 1/21/11 at 5:55pm
BullDriver
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX-8350 Vishera 4.0GHz Asus 990FX Sabertooth MSI N465GTX 8gb Gskill Sniper 1600/9.9.9.25 
Hard DriveOptical DriveOSMonitor
64Gb Crucial SSD(os) +1tb Baracuda SataIII Asus OEM SATA DVD RW x64 Windows 7 Home Premium 25" Hanns G 
KeyboardPowerCaseMouse
Microsoft Flatty Tethered RocketFish 550w Modular Corsair 600t Logitech MX Revolution 
  hide details  
Reply
BullDriver
(13 items)
 
  
CPUMotherboardGraphicsRAM
FX-8350 Vishera 4.0GHz Asus 990FX Sabertooth MSI N465GTX 8gb Gskill Sniper 1600/9.9.9.25 
Hard DriveOptical DriveOSMonitor
64Gb Crucial SSD(os) +1tb Baracuda SataIII Asus OEM SATA DVD RW x64 Windows 7 Home Premium 25" Hanns G 
KeyboardPowerCaseMouse
Microsoft Flatty Tethered RocketFish 550w Modular Corsair 600t Logitech MX Revolution 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › Problems Overclocking Athlon II X4 640