chemojay Posted October 10, 2008 Posted October 10, 2008 hi, I just completed my first so far stable overclock on a computer I have built for my friend. Specs AMD Phenom 9950BE Asus M3A78-T Asus 3870x2 (1 card) OCZ Reaper 1066MHz PC Power & Cooling 750 Zalman 9700 copper heatsink Lian-Li PCB 65plus case with window So far I have it up to a 3007MHz clock speed. I have ran Prime95 1hr preliminary test. A 2hr OCCT stable, a 9hr OCCT stable and I have just started a 12hr run of Orthos times 2 to test all four cores. So far 2hrs in and no errors. I may let it run for 24hrs. So far in the 9hr OCCT test the highest CPU temp reported was 45C. And all I had to do was change the CPU multiplier to 14 and tested. Then I changed it to 15, and manually set the DRAM timings. I left the FSB at 200 and my ram is running 534Mhz according to CPUID. This is with a voltage of 2.0V to the ram and 1.300V stock to the CPU. I am pleased with the overclock if it turns out to be stable which I am hoping it will. I also tried a multiplier of 14 with a FSB speed of 215. It clocked the CPU but dropped my memory back down to 400MHz or ddr2 800. So I stuck with the multiplier setting 15. I know I have read it somewhere, but why does increasing the FSB drop the memory to ddr2 800? Anyone else overclocking these? This has improved the 3dmark06 score from 13000ish to 14300ish. Which to me, seems to be a bit of a difference in benches. In the favor of the overclock. Just wondering if this CPU could see 3100-3400MHz. So far in the 9hr OCCT test the highest CPU temp reported was 45C. So it may have a little room to go. I'll post some screens at some point when its not 3am. I need to read the OCDB guidelines and see if this would qualify for the AMD 3GHz club. Share this post Link to post Share on other sites More sharing options...
dr_bowtie Posted October 10, 2008 Posted October 10, 2008 the memory speeds are calculated off the CPUs multiplier and thats what makes the difference... with the architecture they arent like normal CPUs...same way Brisbanes are in AM2 dual cores.... nice job none the less.... just make sure you actively cool the CPUs MOSFETs...so the board doesnt die....big problem with most boards and 140w TDP CPUs... Share this post Link to post Share on other sites More sharing options...
chemojay Posted October 10, 2008 Posted October 10, 2008 (edited) Well I hit 1 error code in orthos about 6 hrs in on core #3. No explanation. Weird that I ran over 9hrs of OCCT stable. With the same settings. I am going to try again. I was really hopeful that the first 12hr orthos I ran would be stable. Any tips? I guess I can probably google the bristbaine arch. and see what you are talking about. I know that the AMD is not like a regular cpu but are they a different arch from the windsor? I need to read more and more. I am starting another orthos run to see how far it goes. Don't these motherboards cover the mosfets with a heatsink? there are 2 or 3 heatsinks on the MB itself. I will have to look again when I get home tonight. Do you think that could have caused the error? Heat. I am supposed to hook my buddy up with this computer on saturday. I guess I am pushing it kinda close. If I can't get a stable run I will just take it back to stock clock and let him have it I guess. IDK at this point. Work that pays my bills calls so catch you later. Thanks Dr_Bowtie Edited October 10, 2008 by chemojay Share this post Link to post Share on other sites More sharing options...
cchalogamer Posted October 10, 2008 Posted October 10, 2008 That's the joy of stability testing, you may run it for years and never see a crash, yet fire up something new that does something different and BOOM, error after 5 mins. Best you can do is tweak settings until it goes away or live with the possibility that it's not going to be rock solid (something I tend to pass on for OTHER ppl's PCs I OC) Share this post Link to post Share on other sites More sharing options...
chemojay Posted October 10, 2008 Posted October 10, 2008 BTW how long does Orthos run? or does it just run untill you stop it? Share this post Link to post Share on other sites More sharing options...
chemojay Posted October 11, 2008 Posted October 11, 2008 (edited) well time to try something different because I tried again and same result. about 6hrs in instance 2 of Orthos give 1 error on core #3. only core #3. the instance 1 will run for 12 hrs stable on core#0 and core #1. I just don't understand why only 1 core. I guess I don't have to understand it though for it to give me the error. lol I have been playing around with some settings and found that on this motherboard you have to manually set the ram timings on both sticks from what I can understand. I have these both set now and am trying again. I also found a article with this MB and processor exactly here. http://www.neoseeker.com/Articles/Hardware...gx_phenom_9950/ just incase anyone else may find it useful. I didn't up the CPU voltage quite that much because I am not going for that high of a stable clock. I also didn't see any reference in the article about stablity test. just benches and read write test. Edited October 11, 2008 by chemojay Share this post Link to post Share on other sites More sharing options...
dr_bowtie Posted October 11, 2008 Posted October 11, 2008 yes you will get errors from heat...especially on the MOSFETs...those regulate the power and when they get hot the power starts to drop... check temps of the northbridge and if you have a thermal probe check the MOSFETs and the PWMICs you'll want to keep them as cool as you can on high wattage CPUs during stability runs... the more you overclock the more wattage it's going to pull...you could be reaching over 200watts right now...alot of times if the board as a heat pipe over the MOSFETs with a thermal pad i will pull the pad and either use direct contact (providing nothing touches anything else) or use something else for pulling the heat away....I use Zalman TIM exclusively in all my builds... the rest is pretty easy.... you may try setting TRFC looser...for all slots occupied...like 105-112.5 75 is way to tight TRC of 26 is good as well Share this post Link to post Share on other sites More sharing options...
chemojay Posted October 11, 2008 Posted October 11, 2008 (edited) yes you will get errors from heat...especially on the MOSFETs...those regulate the power and when they get hot the power starts to drop... check temps of the northbridge and if you have a thermal probe check the MOSFETs and the PWMICs you'll want to keep them as cool as you can on high wattage CPUs during stability runs... the more you overclock the more wattage it's going to pull...you could be reaching over 200watts right now...alot of times if the board as a heat pipe over the MOSFETs with a thermal pad i will pull the pad and either use direct contact (providing nothing touches anything else) or use something else for pulling the heat away....I use Zalman TIM exclusively in all my builds... the rest is pretty easy.... you may try setting TRFC looser...for all slots occupied...like 105-112.5 75 is way to tight TRC of 26 is good as well I have the TRFC set to auto because there is a tRFC0, tRFC1 ect ect. is that refering to each slot? I only have 2 sticks in. 0 and 1 slots are occupied. Updated 12am. well I ran 5hrs 24 min earlier today and it froze. I made a little more voltage adjustment up to 1.450V. that's a 0.150 increase from stock setting of 1.300V. And I set the TRC set to 26. I still have the TRFC set at auto because I don't know if I need to change all or just one. In the bios it looks like the following below in the dram settings DRAM Timing mode [Auto] [DCT 0] [DCT1] [both] CAS TCWL TRCD TRP TRAS tRW tRFC0 tRFC1 tRFC2 tRFC3 TRC TRRD tWRT tRTP then below it repeats all of those again for what I believe to be the second DIMM being used. So what is confusing me is the tRFC0 ect ect. I found this on google or something tRFC Timing: Row Refresh Cycle Timing. This determines the amount of cycles to refresh a row on a memory bank. If this is set too short it can cause corruption of data and if it is too high, it will cause a loss in performance, but increased stability so does that mean each DIMM has 4 banks of chips? so would you loosend them up to 127.5ns or 195ns. Those are the two closest higher than what you recommended. In this case higher is looser correct? Should I just set all the tRFC timings the same? Thanks for the ideas thus far. I am going to pull the MOSFET heatsink and use some artic silver or thermalright TIM. I am out of the Zalman you recommended. If I can talk him into a 40mm fan I will try to figure out a way to get it on the MOSFET. You have made custom heatsinks before haven't you? Any chance you could make one with a way to mount a fan if I send you a pic of the MB with dimensions of the heatsink? On the heat issue I noticed around hour 5 that it was hot in the house and that the cpu temp was creeping up and down from 43C to 46C at highest. So I went and turned the air conditioner on and moved the case intake towards the vent. The CPU temp has dropped to around a steady 38C creeping up and down to 37C and 39C. I am at 7hrs so far. and I have 2 sets of DRAM settings for each DIMM slot in use, I think. In this build here it would be 2 slots occupied. Slot 0 and Slot 1. Edited October 12, 2008 by chemojay Share this post Link to post Share on other sites More sharing options...
chemojay Posted October 12, 2008 Posted October 12, 2008 Got it Share this post Link to post Share on other sites More sharing options...
dr_bowtie Posted October 12, 2008 Posted October 12, 2008 TRFC is essence is the memory slot frequency....the lower the number the tighter/higher the frequency = more work load on the memory controller.... the higher the number the looser/lower the frequency = easier it is on the memory controller...it's sorta like the Northbridge Strap control on Intel board but for each ram slot... Most all AMD boards default at either 75/75 or 105/75 for dual channel....75 is way to tight... I always set it to 105...for all slots you do lose performance if you set it too loose but it will almost always help if the ram is the question... And I have said it 10000000000000000 times...in AM2/AM2+ even if Memtest says the ram is fine...IT's NOT....memtest can pass for days and orthos/prime will fail in seconds and all the voltage in the world doesnt help much but as soon as you start adjusting the secondaries it starts passing... the key to getting over 200fsb with a Phenom is all in adjusting the secondaries... Share this post Link to post Share on other sites More sharing options...
chemojay Posted October 13, 2008 Posted October 13, 2008 TRFC is essence is the memory slot frequency....the lower the number the tighter/higher the frequency = more work load on the memory controller.... the higher the number the looser/lower the frequency = easier it is on the memory controller...it's sorta like the Northbridge Strap control on Intel board but for each ram slot... Most all AMD boards default at either 75/75 or 105/75 for dual channel....75 is way to tight... I always set it to 105...for all slots you do lose performance if you set it too loose but it will almost always help if the ram is the question... And I have said it 10000000000000000 times...in AM2/AM2+ even if Memtest says the ram is fine...IT's NOT....memtest can pass for days and orthos/prime will fail in seconds and all the voltage in the world doesnt help much but as soon as you start adjusting the secondaries it starts passing... the key to getting over 200fsb with a Phenom is all in adjusting the secondaries... I am actually running a 200fsb. all I adjusted was secondaries. I may go over to his house and set those timings later. Also made him order a Asus mosfet fan and when that comes in we will pull the heatsink off and thermal paste it with some better TIM. Share this post Link to post Share on other sites More sharing options...
Recommended Posts
Please sign in to comment
You will be able to leave a comment after signing in
Sign In Now