New to OC, AMD Athlon II X4 635 help

[trents]

By the way, its time for you to put your system info in your "Sig". You now have a star under your name so you have access to that feature. Include as much info as you know about your CPU, Ram, Motherboard, PSU (make and Wattage), Case, Video card, CPU cooler, and OS.

 

[Neutrolizer]

Oh I see that makes sense. That explains why everyone was reporting higher numbers when I was only seeing half of it. I did bump my CPUNB Voltage up to 1.24. VCore to 1.54. The increments was at .02, not .2 so I messed up reporting that earlier.

My core temps are definitely higher, think it was up to 45 degrees playing WoW. I also left WoW running overnight and woke up to discover my computer just recovered from a blue screen so I guess my comp isn't exactly stable right now.

Was reading that I should maybe increase my CPUNB Voltage some more? Bumping the memory voltage might help but you said before that the mobo doesn't like the ram having a high voltage so that's out of the question. Or should I try to increase my CPUNB maybe to 2600 by lowering my multiplier to 14x and then bumping my FSB up?

If I were to buy DDR3 1600 memory, can I combine them with my old patriot DDR3 1333 memory? I remember somewhere that having the same kind is better not but not really how different types is bad.

Thanks

 

[trents]

Oh I see that makes sense. That explains why everyone was reporting higher numbers when I was only seeing half of it. I did bump my CPUNB Voltage up to 1.24. VCore to 1.54. The increments was at .02, not .2 so I messed up reporting that earlier.

My core temps are definitely higher, think it was up to 45 degrees playing WoW. I also left WoW running overnight and woke up to discover my computer just recovered from a blue screen so I guess my comp isn't exactly stable right now.

Was reading that I should maybe increase my CPUNB Voltage some more? Try 1.3. Bumping the memory voltage might help but you said before that the mobo doesn't like the ram having a high voltage so that's out of the question. Or should I try to increase my CPUNB maybe to 2600 by lowering my multiplier to 14x and then bumping my FSB up?

Yes, I would try lowering the CPU multiplier so that you can work on getting your memory speed up. I suspect you haven't run out of FSB room but that the BSOD you were getting was due to exceeding the CPU capability. You have a C2 stepping core and those would max out at 3.5-3.6 ghz. Lowering the multiplier is certainly a way to test this theory as it would eliminate the processor as the cause of instability. Does that make sense?

If I were to buy DDR3 1600 memory, can I combine them with my old patriot DDR3 1333 memory? I remember somewhere that having the same kind is better not but not really how different types is bad.

No, don't mix memory speeds. You're asking for instability if you do. Besides, the bios will default to the speed of the slowest module anyway so there' not point in it. If you think you need more memory then go for a 2x4 dual channel kit rather than a 2x2. But beware, increasing the amount of memory or the number sticks, either one can have a negative effect on overclocking. It puts more strain on the already challenged AMD integrated memory controller.

Thanks

 

[Neutrolizer]

Oh I see, yeah that makes sense. So if I were to buy more memory, it'd be better just to replace my 4 gbs with the new one. But if I were to buy 8 gb of memory, it would mess up my overclock that I'm already working on and then I would have to start all over again basically? One more thing, do you think its worth it to get 8 gb memory or is 4gb sufficient? Thanks

 

[trents]

. . . One more thing, do you think its worth it to get 8 gb memory or is 4gb sufficient? Thanks

It depends on what you mostly use your computer for. If you do a lot of Photoshop or CAD or A/V editing you would benefit from more ram. If you just do stuff like surfing the net, word processing and occasional gaming - probably a waste of money. The only thing is, ram is so cheap right now and who knows how long it will stay that way. Hard drives are terribly expensive currently because of the flooding in SE Asia where most of them are produced but ram is dirt cheap.

 

[RGone]

From the outside looking in...

Only a few of things I am going to put forth.

1. I never did see a baseline stability test of Prime 95 for at least 2 hours in the very first post of this now 2 page thread. Nothing but push, push push to overclock.

2. The ram in use now is only DDR1333 if it is run in an XMP mode; which is Extreme Memory Profile mode. That does not at all sound like straight up DDR1333 memory to my mind.

3. Looking at the first 10 user reviews for the motherboard at NewEgg for the fall of 2011, the users with problems was 7 out of 10. Maybe there was a motherboard revision and things got worse.

4. I did a little sratching around looking for 'reviews' of the GA-870A-UD3 and what I found for that relatively inexpensive motherboard may be one reason you have issues once you push-push the board to overclock. No MOSFET / VRM heatsinks.

NOTE: I look at many review sites but trust very few.

5. Knowing #4 as above then this may well explain why overclocks with high Vcore can fail. With the Hydro H60 CPU cooler as your cpu cooler you have a situation where there is absoutely no air cooled cpu cooler blowing any air in the MOSFET circuit area. I see you already talking about putting 1.5+ Vcore to the CPU. From what I read there were heat issues with CPU VRM circuit way before reaching 1.5+ Vcore.


LINK >>>
While the 870A-UD3 does have many good features, there were some shortcomings that need to be addressed. Firstly, there are problems with the current fan header layout as there is no dedicated header for a rear exhaust fan. Furthermore, placing the clear CMOS header underneath the PCI-E x16 slot makes it cumbersome to access when a video card is installed. Last but not least, and perhaps the Achilles heel of the board is the lack of a MOSFET heatsink. Through our testing at a CPU voltage of 1.45v, we recorded MOSFET temperatures which were quite high considering the relatively low ambient temperatures. Trying to get the motherboard to stabilize at higher voltages proved to be impossible without placing a high speed fan over the area to keep the MOSFETs cooled.

 

[Neutrolizer]

Thank you for looking into this RGone. I never did do a baseline stability test as I didn't know I should of done so. Since I don't have any air blowing into the MOSFET circuit area, would it be possible to reverse my rear fan to have it blowing in instead of out?

 

[RGone]

Thank you for looking into this RGone. I never did do a baseline stability test as I didn't know I should of done so. Since I don't have any air blowing into the MOSFET circuit area, would it be possible to reverse my rear fan to have it blowing in instead of out?

Since about 1995 or thereabouts, I have never overclocked a system without establishing a baseline first. Not doing so leaves me guessing if the thing is able to even run the standard speeds and do so with some form of expected stability. I am speaking for me and my years of testing stability and I came to my conclusion that my system stock or overclocked should pass at least 8 hours of Prime85 running on all cores. Once I established this pattern for myself, my systems had failure rates so few that over those years, I cannot even remember any failures when the system was put into use after passing P85 for my subjected 8 hour runs.

I personally would not reverse the rear fan. Most cases are designed with a fan drawing air into the case down low in the front and the air moving upwards from heat and exitting thru the rear fan drawing air out.

 

[Neutrolizer]

Ok then I will run Prime85 tonight. How would I show it being run for 8 hours? Any settings I should change for Prime85? Do I just take a screenshot of HWMonitor to show core temps after said 8 hours?

Is there a easy way to reset my bios settings to default in the bios menu? Its true about what the review said about the mobo, the clear cmos is badly situated under my graphics card

My case does have a lower fan in the front. So without cooling to the MOSFET, does that mean it is dangerous to overclock my CPU as that link you showed me shows how critical it is to keep the area cooled if and when overclocking. Is it possible to get a side fan to blow in? But then it seems cumbersome if the fan is attached to the side of the case and when detaching the case, it may seem impossible without pulling the wires unless they have one long enough to reach the 3 pins.

 

[trents]

Does the case have more places for fans than you have fans currently installed in? If so, fans are cheap. Install one in every available hole, orienting them to move air from front low to back high. Side panel fans should blow in toward the motherboard.

 

[trents]

Neutrolizer, if you go back and read post #20 I advised you to run a Prime95 blend test for at least 2 hr. to confirm stability. I guess you didn't do that, huh! RGone advises 8 hr. and so do many (some advise much more than 8 hr. to confirm stability). Two hrs. has served me just fine but regardless of what number you decide on the point is you need to make the system work at virtually 100% load for an extended period to confirm stability. Casual computing and even gaming does not stress test adequately.

 

[Neutrolizer]

Haha kind of missed that trents! I did one earlier before we bumped it to 1.55v, but I didn't afterwards. But I think it isn't stable as it couldn't run with WoW on overnight.

I thought Rgone was advising to run Prime95 at system defaults. Should I be doing that then?

So far I have two open slots available on my side panel, Do fans made for the side panels have their wires long enough to ensure I don't break em when taking my side panels off when they're attached? Would this help the MOSFET part?

 

[RGone]

1. You never advised us of completion of P95 nor at what speed, voltage etc, the pass might have occurred at.

2. For fan leads n0t long enough for me to get side of case off and completely out of the way, I put connectors in the wiring so I can open case side slightly and unplug the connectors I put in the wires themselves and move case cover out of the way. I have a NOTE on the side of the case that says "Hey stupid, remember the fan wires!!!", that way I don't tend to forget and snatch my side-cover off. I never enter the case except to clean dog hair out every 4 months or so and the n0te reminds me.

3. Are side fans going to help cool the MOSFETs? Don't know. Personally I doubt it, but only testing will say for sure. I do not care for side fans just blowing in my case disrupting other air-flow. Keyword = " I " don't.


I use something like this most of the time so I can get the AIR more nearly where I wish the air to be. I am sure there maybe other types out there. I am saying what " I " use. YMMV. Adding a NOTE: This bracket linked comes with a fan. Cheap with fan as without, BUT it can be had without fan at some locations. Once I have the bracket in hand...well then I can put smaller fan on the bracket if I want to and have 92mm fan for other uses. Suit one's ownself or whatever.
ZALMAN FB123 Video card fan



OCCT Perestroïka v4.1.0 Available !

This should go to the main English page. Buttons across the top for Home, News, Download, Forum (Closed), FAQ, and PRO (none available)

" I " use this because it has a number of tests. These tests are one's we already know, BUT the front-end for reach the tests IS oh so MUCH better than the individual tests by themselves.

1. The first test is Prime95 in a very easy to get to GUI. You can choose all cores. You can choose a specific TIME for the test to run and IF the test fails you can esaily see at what time the test fails, that is unless the test crash is so hard it reboots the computer.

2. The second test is another front-end for LinX/LipPack/IBT and again time for testing can be set as well as knowing when the crash happens unless the computer reboots.

There are other tests as well of which I have no knowledge since I have not used them. Like Video and Power Supply. I am n0t going there.

NOTE: I seldom say much about what to test with. Too many people want to come and put their dog or 2 cents in the race. I will only say this one thing. I have used this since version 0.99a way back there. I have used about 10 of the betas as they have come over the years. I still have the last beta on my hard drive and n0t yet the new one since I had not gone to the site in a while. I am n0t doing any testing right now.

NOTE2: I will be downloading the latest 4.1.0 version for "MY" use now that I know it is out of beta.

NOTE3: I did not see any pictures. Would have gotten some if I had seen them and am not going to other sites chasing pictures. The program setup is n0t rocket sciences.

RGone...ster.

 

[trents]

If you get fans that have a four pin molex connector that are powered right from the PSU they should work. I know for a fact that Yate Loon fans will meet your need in that regard. They have long wires. They are pretty good quality fans that don't cost much. You can usually get them for under $5 a piece. Look for deals on Ebay.

Yes, I think RGone was advising 8 hrs. of Prime95 at stock settings. That was my impression as well. I was recommending 2 hrs.of Prime to test overclock settings. I have found that sufficient to demonstrate stability while not putting undue stress and strain on the system.

 

[RGone]

Yes, I think RGone was advising 8 hrs. of Prime95 at stock settings. That was my impression as well. I was recommending 2 hrs.of Prime to test overclock settings. I have found that sufficient to demonstrate stability while not putting undue stress and strain on the system.

I am easy...gimme whatever. Hehehe.

 

[Neutrolizer]

That's a good note to have lol. anyways I ran OCCT and within a minute of testing one of the core crashes.

I might just take up one of the fans. thanks for the recommendations on that!

Now I'm not really sure if I should go back down on the FSB or lower the CPUNB down a multiplier. Unless this is one of those situations with the GA-870A-UD3, that no matter the voltage, it won't rally stabilize?

 

[RGone]

Now I'm not really sure if I should go back down on the FSB or lower the CPUNB down a multiplier. = Really who knows? As far as I can tell going all the way back up to the very top of page #2 here; there is no indication make that no real indication of what the setup is. Don't see a CPUz of CPU, Memory or for ease a capture also of SPD tab. Keeps us from having to go many pages back to find SPD.

Unless this is one of those situations with the GA-870A-UD3, that no matter the voltage, it won't rally stabilize? = Here again who can really say. I have not seen any evidence that the system is stable at any speed. And if it had been stable at a certain speed then what was shown of that particular setup to know how much to change what.

ALL this is why I said I never overclock a system without a known baseline. IF you had followed Dolks' tutorial, you would have found out what FSB your ram could have withstood and where to set the ram so it was not causing a crash just from increased FSB raising the ram frequecy.

I see two choices and it depends on your pocket book and intentions with your pocket book.

1. If your pocket book is flush instead of "flushed" like mine and you are going to attend to the project of getting more air to the VRM circuit and you are going to get some good 1.5V ram, then back it down to stock and use it until you get the projects done. No sense in us doing this 2 or 3 times with different parts and pieces unless you have learned how to do it yourself.

Give a man a fish and he is not hungry now...Teach a man to fish and he should never go hungry.

2. If you are not going to attend to air to VRMs and new ram any time soon, then Clear the CmOs and go back into bios and load optimized defaults or Setup Defaults or whatever your board calls board defaults and start OCCT and run for 2 hours and IF it completes we need the 3 captures of CPUz and OCCT when thru or when crashed. AS a simple safeguard to begin wth a capture of HWMonitor with those other 4 captures might be a good idea to make d*mn sure OCCT and HWMon are on the same page. After that probably only 3 Cpuz and OCCT will suffice.

After that first run at SetupDefaults and its' failure or completion, we will have some screen captures to know what in h*ll is going where you are.

There really is no such thing as a short cut unless someone like 'trents' or I were right there doing the key entering so we could know exactly what we keyed in. That keying in based on our years of doing it. Otherwise it is walk before running.


RGone...ster.

 

[trents]

Neutrolizer, if I were you I'd start from scratch with stock frequencies and voltages and start by seeing how far you can overclock the CPU without changing anything else. Start your ram at 400/800 to give it some headroom. Then raise the fsb by 5 mhz at a time. After each increase, run a 20 minute Prime95 blend test. If you pass, then increase the bus speed by another 5 mhz.

When you first fail the test, add a small increment of core voltage to make it stable again. Then retest as before until you fail the 20 minute Prime test and then add some more vcore. Of course, always have HWMonitor open on the desktop to monitor core temps and don't exceed 55 C.

After the second failure and vcore addition to re-stabilize, switch to two hour long Prime95 tests. Add a little vcore if necessary and CPUNB voltage to pass for two hours. Then we'll work on tweaking the CPUNB and ram frequency.

I'm thinking you should be able to get about 3.5 ghz on about 1.425 vcore. I had a Ahtlon II 630 for a couple of years and that's what it took for me to get 3.5 ghz stable with a fsb of 250.

 

[Neutrolizer]

Yeah that sounds good about starting from scratch. I was looking at the specs for my MOBO and it said this for Memory Standards: "1866(OC)/1333/1066" so does that mean I cant use DDR3 1600 or is the 1600 OC'd to 1866? Also what does VRM mean?

trents, what do you mean by "

When you first fail the test, add a small increment of core voltage to make it stable again. Then retest as before until you fail the 20 minute Prime test and then add some more vcore. Of course, always have HWMonitor open on the desktop to monitor core temps and don't exceed 55 C.

Did you mean that after I bump up the vcore and run Prime95 that I should increase my FSB by another 5 mhz and keep adding more vcore but stop before I reach 55 C?

I would love to upgrade my memory and if I do buy some, it will probably take a week or two. I would also be very thankful if you guys continue to teach me the fine art of tuning my computer as I would very much love to learn how. Now the question is should I continue to experiment with my current memory as a learning experience or wait a week or two? It is very clear that I am going to continue depending on you guys for your expertise!

 

[Neuromancer]

That little extra can help with stability but too much ram voltage has the opposite affect.

I assume this is a localized low clock function? Cuz I have never seen "too much vdimm causes issues" feature. On any architecture in recent knowledge. Then again I always BOOST vdimm to get higher bus clocks when they are not stable to increase stability. Side effect of having a IMC for the last decade (It is a known overclockers trick to lower vcore)

what?