You don't have to burn 24 hours ...10-12 will do fine.
Download the orthos progam.
Download the A64Info program.
Once you're done you should be able to run f@h or prime all day long. For right now orthos will simulate each one of whichever you chose. This way you won't lose any work.
What is your highest stable clock at 1.35?
If you begin your burnin at this voltage and you get errors on either the ram or the cpu then lower the frequecy about 5mhz.
A64Info will give me all of the answers to any questions I might have later on if you start getting errors so it would help much to get that ...plus the latest version has an awesome temp monitor. Works as well as core temp IMO.
Here is a pic of the two programs running together. Look at all of the info on A64Info.
I am working on to determine the highest clock at 1.35, I will let you know soon.
I tried the latest A64Info beta in my configuration; Win2K, Asrock 939Dual-VSTA, and Opty165. It is greeting me with an error message as follows "Error loading driver". I searched for a solution, and the recommendation was to make sure user has admin privileges. I have admin privileges, and still searching for a solution.
Hm ...maybe win2k wont support. I assumed you were using xp.
At any rate ...try again when you think you're stable maybe.
Tried to use version 0.5 of A64Info, and it started but the information I got was completely off, e.g. memory frequency 350, CAS 1.5, ... I will try to use systool
I am in business with Systool, it worked fine. I found that at 1.35 vcore, maximum stable frequency was 2790MHz@310 HTT. Silly me, I had arbitrarily reduced the voltage to 1.5 while backing off HTT from 333 to 311. In fact I could have gone all the way down to 1.35. Anyway, burn in is running using ORTHOS Gromacs core test. I will let you know in the morning.
Regarding your question, I use prime, since it appears to me that it is a more sensitive test for stability; temperatures are generally higher when running prime maximum-temperature test compared to f@h.
If systool says that your highest frequency stable max is 2790 I would start at 2750. Let me know if you got errors during the night at 2790.
To give you an example I took another screenshot of orthos this morning after bumping up my voltage from 1.35 to 1.375 and increasing my frequency to 255 I still got lower idle and load temps. It went from 21c & 31c idle to 18c & 28c idle. After burning in for a week all of a sudden my temps have dropped 3c. I think burning in actually tames the TIM so that must be why it takes so long.
I'll add a load temps once orthos has had a long time to heat things up and you can compare it to the previous temps above. Keep in mind that I actually raised the frequency and voltage.
If systool says that your highest frequency stable max is 2790 I would start at 2750. Let me know if you got errors during the night at 2790.
To give you an example I took another screenshot of orthos this morning after bumping up my voltage from 1.35 to 1.375 and increasing my frequency to 255 I still got lower idle and load temps. It went from 21c & 31c idle to 18c & 28c idle. After burning in for a week all of a sudden my temps have dropped 3c. I think burning in actually tames the TIM so that must be why it takes so long.
I'll add a load temps once orthos has had a long time to heat things up and you can compare it to the previous temps above. Keep in mind that I actually raised the frequency and voltage.
Here you go.
2790 GHz with 1.35 vcore ran over night with no errors using Large-in place FFT test in ORTHOS. Today, I tested 3 GHz with 1.5 vcore again, but failed after 2 hours. Gromacs core is not giving any errors but Large-inplace FFT test is failing. Now I am running at 2.9 GHz with 1.4v, no errors since one hour, and core temperature is fluctuating around 50C. Do you think 10-12 hours burn-in was not sufficient and I need to burn-in longer? Maybe I need burn in for a week like you did.
Tomorrow, my super-duper memory (Ballistix DDR500 2x1gb) is arriving. So I will raise memory frequency to 250 MHz and run a couple of benchmarks to compare with the other Athlon64-X2-3800+ at 2.6 GHz. Are you interested in seeing any specific benchmarks?
2790 GHz with 1.35 vcore ran over night with no errors using Large-in place FFT test in ORTHOS.
After this frequency runs fine at this vcore for 10-12 hours only raise it 1 mhz to 2791 and stress it again for 10-12 hours. Only raise one mhz at a time.
After this frequency runs fine at this vcore for 10-12 hours only raise it 1 mhz to 2791 and stress it again for 10-12 hours. Only raise one mhz at a time.
Csimon, you are turning me into a believer of burn-in. Here is a screeny for ya
and vcore is, guess what, 1.4. Yes it is 1.4 at 3Ghz and the core temperature is 48. Even if it crashes after two hours, (not crashed yet) this is a big improvement after your recommendations. I will let you know if it can make it over night.
Thanks
Edit: crashed .... going back to last stable frequency of 2.9Ghz and continuing burning in
Thanks for the PM mirage. I got it today ...CCBBE 0615EPMW! Finally on my third try with Opterons I got a good stepping. The 175 and 170's I ordered before were all poor steps from Monarch. Never will I shop there again for a processor ...feels like $750 down the drain but I know I got what I paid for. But now I feel like I'm getting my monies worth!
Thanks for the PM mirage. I got it today ...CCBBE 0615EPMW! Finally on my third try with Opterons I got a good stepping. The 175 and 170's I ordered before were all poor steps from Monarch. Never will I shop there again for a processor ...feels like $750 down the drain but I know I got what I paid for. But now I feel like I'm getting my monies worth!
That is one of the best steppings, I am pretty sure you will hit 3GHz easily. Please post the news and enjoy !!!
Hows your OC coming? I'm doing burn in and I'm up to 297X9@1.35v. Temps are average 40C & 50C underload but they cool down a few degrees after about 10 hours.
Hows your OC coming? I'm doing burn in and I'm up to 297X9@1.35v. Temps are average 40C & 50C underload but they cool down a few degrees after about 10 hours.
Congratulations, it seems you have upside potential. Currently, I am running at 311x9@1.35v 100% stable around 48-52C core temperature. This is the highest 100% stable clock at default voltage. Last time I tried one week ago, 333x9 still needs 1.55v but core temperature exceeds 60C at that voltage. What is your cooler ?
My cooler?
Well ...I have a v2000 case modded 3 (5.25) bay to 4 (3.5) hd up top. The whole front uses mesh 5.25 covers for max airflow. All hd bays at the bottom compartment have been removed for radiator and pump.
I use a swiftech apogee waterblock, a shrouded pa160 radiator w/ 80cfm evercool spider, a D5 pump, 1/2" tubing, and non-conductive coolant mct-5.
I got most parts from dangerden except for the apogee ...with a coupon code I think I payed $35 for it. It works well. This is a low restriction/high flow setup.
As soon as I have time to rip the motherboard out again I will remove the NB heatpipe system (since it does no good being inverted) and replace it with a passive zalman cooler I bought with the mobo. I'm debating on whether to yank the heatpipe out and use the mosfet sink on or not.
My cooler?
Well ...I have a v2000 case modded 3 (5.25) bay to 4 (3.5) hd up top. The whole front uses mesh 5.25 covers for max airflow. All hd bays at the bottom compartment have been removed for radiator and pump.
I use a swiftech apogee waterblock, a shrouded pa160 radiator w/ 80cfm evercool spider, a D5 pump, 1/2" tubing, and non-conductive coolant mct-5.
I got most parts from dangerden except for the apogee ...with a coupon code I think I payed $35 for it. It works well. This is a low restriction/high flow setup.
:wow2: You will not have much trouble cooling when the time comes to raise your core voltage. There is definitely potential to overclock further.
I have a question. Is there any performance penalty by using odd memory dividers? I have divider 11 right now (yields 254 DDR frequency at FSB=311), should I try 12? Actually BIOS do not allow me to choose 12, it is 11 by default when I set 166 as DDR frequency.
:wow2: You will not have much trouble cooling when the time comes to raise your core voltage. There is definitely potential to overclock further.
I have a question. Is there any performance penalty by using odd memory dividers? I have divider 11 right now (yields 254 DDR frequency at FSB=311), should I try 12? Actually BIOS do not allow me to choose 12, it is 11 by default when I set 166 as DDR frequency.
Well the WC setup is probably more than adequate but a good air setup will get the same results actually. The trick to a good WC is doing it right. You certainly don't want to remove all of that heat from the rad back into your case/sys or you're defeating the purpose. That's what I like about this case ...the bottom comparment is seperate from the system above and the holes that I use for tubing are already there. They are originally intended for psu and hd cables. The best benefit to WC is that it can be less noisy ...depending on choices you make.
Are you talking about the 11x-12x mem divider?
As for odd number dividers the answer is no. It is discouraged to use 1/2 mulitpliers like 11.5 and so on although I've used them before with no problem.
I have mine set at 333 which I'm assuming is equal to your 166 ...when I reach the top I only have the option to go down to 266 or lower but nothing in between. The Asus allowed other in between which was nice. Perhaps I can do it with A64Info but I'll cross that bridge when I get there. Probably around 305-310 ht or something.
I haven't tried 1.55v yet or 333x9 but I imagine mine would get to 60c under load as well.
Well the WC setup is probably more than adequate but a good air setup will get the same results actually. The trick to a good WC is doing it right. You certainly don't want to remove all of that heat from the rad back into your case/sys or you're defeating the purpose. That's what I like about this case ...the bottom comparment is seperate from the system above and the holes that I use for tubing are already there. They are originally intended for psu and hd cables. The best benefit to WC is that it can be less noisy ...depending on choices you make.
Are you talking about the 11x-12x mem divider?
As for odd number dividers the answer is no. It is discouraged to use 1/2 mulitpliers like 11.5 and so on although I've used them before with no problem.
I have mine set at 333 which I'm assuming is equal to your 166 ...when I reach the top I only have the option to go down to 266 or lower but nothing in between. The Asus allowed other in between which was nice. Perhaps I can do it with A64Info but I'll cross that bridge when I get there. Probably around 305-310 ht or something.
I haven't tried 1.55v yet or 333x9 but I imagine mine would get to 60c under load as well.
Yes, that is mem divider and thanks for the clarification. By the way, I checked your motherboard, I really liked the northbridge, power-circuit cooling design. Elimination of unnecessary ports to enable better air circulation for the power circuit is neat.
Yes, that is mem divider and thanks for the clarification. By the way, I checked your motherboard, I really liked the northbridge, power-circuit cooling design. Elimination of unnecessary ports to enable better air circulation for the power circuit is neat.
It's a really nice board and with the rebate it sells for only $100!
Comments
I am working on to determine the highest clock at 1.35, I will let you know soon.
I tried the latest A64Info beta in my configuration; Win2K, Asrock 939Dual-VSTA, and Opty165. It is greeting me with an error message as follows "Error loading driver". I searched for a solution, and the recommendation was to make sure user has admin privileges. I have admin privileges, and still searching for a solution.
At any rate ...try again when you think you're stable maybe.
Tried to use version 0.5 of A64Info, and it started but the information I got was completely off, e.g. memory frequency 350, CAS 1.5, ... I will try to use systool
I am in business with Systool, it worked fine. I found that at 1.35 vcore, maximum stable frequency was 2790MHz@310 HTT. Silly me, I had arbitrarily reduced the voltage to 1.5 while backing off HTT from 333 to 311. In fact I could have gone all the way down to 1.35. Anyway, burn in is running using ORTHOS Gromacs core test. I will let you know in the morning.
Regarding your question, I use prime, since it appears to me that it is a more sensitive test for stability; temperatures are generally higher when running prime maximum-temperature test compared to f@h.
Good night
To give you an example I took another screenshot of orthos this morning after bumping up my voltage from 1.35 to 1.375 and increasing my frequency to 255 I still got lower idle and load temps. It went from 21c & 31c idle to 18c & 28c idle. After burning in for a week all of a sudden my temps have dropped 3c. I think burning in actually tames the TIM so that must be why it takes so long.
I'll add a load temps once orthos has had a long time to heat things up and you can compare it to the previous temps above. Keep in mind that I actually raised the frequency and voltage.
Here you go.
2790 GHz with 1.35 vcore ran over night with no errors using Large-in place FFT test in ORTHOS. Today, I tested 3 GHz with 1.5 vcore again, but failed after 2 hours. Gromacs core is not giving any errors but Large-inplace FFT test is failing. Now I am running at 2.9 GHz with 1.4v, no errors since one hour, and core temperature is fluctuating around 50C. Do you think 10-12 hours burn-in was not sufficient and I need to burn-in longer? Maybe I need burn in for a week like you did.
Tomorrow, my super-duper memory (Ballistix DDR500 2x1gb) is arriving. So I will raise memory frequency to 250 MHz and run a couple of benchmarks to compare with the other Athlon64-X2-3800+ at 2.6 GHz. Are you interested in seeing any specific benchmarks?
Csimon, you are turning me into a believer of burn-in. Here is a screeny for ya
and vcore is, guess what, 1.4. Yes it is 1.4 at 3Ghz and the core temperature is 48. Even if it crashes after two hours, (not crashed yet) this is a big improvement after your recommendations. I will let you know if it can make it over night.
Thanks
Edit: crashed .... going back to last stable frequency of 2.9Ghz and continuing burning in
Good luck, I will send you a PM if I see it available anywhere.
That is one of the best steppings, I am pretty sure you will hit 3GHz easily. Please post the news and enjoy !!!
Congratulations, it seems you have upside potential. Currently, I am running at 311x9@1.35v 100% stable around 48-52C core temperature. This is the highest 100% stable clock at default voltage. Last time I tried one week ago, 333x9 still needs 1.55v but core temperature exceeds 60C at that voltage. What is your cooler ?
Well ...I have a v2000 case modded 3 (5.25) bay to 4 (3.5) hd up top. The whole front uses mesh 5.25 covers for max airflow. All hd bays at the bottom compartment have been removed for radiator and pump.
I use a swiftech apogee waterblock, a shrouded pa160 radiator w/ 80cfm evercool spider, a D5 pump, 1/2" tubing, and non-conductive coolant mct-5.
I got most parts from dangerden except for the apogee ...with a coupon code I think I payed $35 for it. It works well. This is a low restriction/high flow setup.
As soon as I have time to rip the motherboard out again I will remove the NB heatpipe system (since it does no good being inverted) and replace it with a passive zalman cooler I bought with the mobo. I'm debating on whether to yank the heatpipe out and use the mosfet sink on or not.
:wow2: You will not have much trouble cooling when the time comes to raise your core voltage. There is definitely potential to overclock further.
I have a question. Is there any performance penalty by using odd memory dividers? I have divider 11 right now (yields 254 DDR frequency at FSB=311), should I try 12? Actually BIOS do not allow me to choose 12, it is 11 by default when I set 166 as DDR frequency.
Well the WC setup is probably more than adequate but a good air setup will get the same results actually. The trick to a good WC is doing it right. You certainly don't want to remove all of that heat from the rad back into your case/sys or you're defeating the purpose. That's what I like about this case ...the bottom comparment is seperate from the system above and the holes that I use for tubing are already there. They are originally intended for psu and hd cables. The best benefit to WC is that it can be less noisy ...depending on choices you make.
Are you talking about the 11x-12x mem divider?
As for odd number dividers the answer is no. It is discouraged to use 1/2 mulitpliers like 11.5 and so on although I've used them before with no problem.
I have mine set at 333 which I'm assuming is equal to your 166 ...when I reach the top I only have the option to go down to 266 or lower but nothing in between. The Asus allowed other in between which was nice. Perhaps I can do it with A64Info but I'll cross that bridge when I get there. Probably around 305-310 ht or something.
I haven't tried 1.55v yet or 333x9 but I imagine mine would get to 60c under load as well.
Yes, that is mem divider and thanks for the clarification. By the way, I checked your motherboard, I really liked the northbridge, power-circuit cooling design. Elimination of unnecessary ports to enable better air circulation for the power circuit is neat.
It's a really nice board and with the rebate it sells for only $100!