Dual Core Opteron Thread #2

[scoobie dave]

watching prime like a moron go right ahead

I tend to leave it running when I go to work/sleep at night. I have better things to do than watch prime also.

If you only prime for 30min then your making false stability claims.

[alpha0ne]

If you only prime for 30min then your making false stability claims.

Agreed, the very bare minimum should be 2 hours at least and lest we go overboard no more than 12 hours should be needed IMHO

[itznfb]

If you only prime for 30min then your making false stability claims.

i use 30 min primes for benching and numbers, for complete stability i use 2hr, anything beyond, as i've already explained is a complete waste of time and resources. it may prime for another 100hours, and then on hour 101, then temperature in your room may change 1/10 of a degree and prime fails. when stability testing you're testing to make sure the chip is stable at the current voltage under full load. an OCed chip will ALWAYS fail at some point if not kept in a contained thermal environment. if i could somehow keep this room at 20c all day everyday, nothing could change that, then i could run 335x9 prime for years, until the chip dies in fact.

false stability claims. i don't think so. why don't you not make false accusations. i know what i'm talking about.

[itznfb]

on another note: if you guys really want to argue with me about this, i'm going to work, so send me a PM, i'll argue about all day, and i'll eventually convince you... but i'd rather not see this brought up again in the thread. there is a lot of good information in here and i don't want to be a part of it getting so far off track that it gets shut down. everyones opinion has been strongly stated. so let it be, or take it to PM.

[ChkDsk]

Hmmz... Sorry I ever asked

But I didn't ask for a definition of prime stable, just for how long you have to prime to get prime stable in the database... What's prime stable or not will allways be a discussion I think

[alpha0ne]

Hmmz... Sorry I ever asked

But I didn't ask for a definition of prime stable, just for how long you have to prime to get prime stable in the database... What's prime stable or not will allways be a discussion I think

Unless a minimum is stated

I am in no way claiming that some ppl included in the database are dishonest or that their OC's are not stable but not having a specified acceptable minimum will always lead to this same discussion

[ekjr]

Is this prime stable???

[ChkDsk]

Is this prime stable???

My suggestion is to let SP2004 run 1 round.. So about 3,5hrs... But this discussion will probably never end... Maybe another thread with a poll will do the trick for the DB. This thread is to show results (prime or not primestable) and ask questions, ... (thats my opinion, no offence)

Has anyone removed the IHS from his CCBBE 0610DPMW? What temps are you getting? My IHS is still on and I think the temps are pretty good... And a 3100MHz 6hr prime isnt bad either.... Still, I'm wondering. My only DC with the IHS on and the best clocker... Can't believe it

[scoobie dave]

i know what i'm talking about.

I don't think you do.

[Getttosmurf]

going a step further in the prime stable discussion , is it considered stable if it can pass 24h dual prime95 at winter but creashes instantly during summer due to higher ambient temps ? my ambient temps wary quit a bit since im in northern sweden where temps wary from 30c to -35c not so much inside though but still primeing does give me no clue wheter my machine is stable or not , unless i run it 24/7 365 and that kind of defeets the purpose of having a fast dc when prime uses 100% cpu :P

Regarding databse entrys i think anyone reading them thinking they can just by the same stepping and get the sam oc is a bit off , isnt it smarter to take the average oc of a known stepping and ecpect somwhere around that speed ?

perhaps wee should start a thread where we can discuss what is stable and not further ?

Regards
Smurf

[ChkDsk]

going a step further in the prime stable discussion , is it considered stable if it can pass 24h dual prime95 at winter but creashes instantly during summer due to higher ambient temps ? my ambient temps wary quit a bit since im in northern sweden where temps wary from 30c to -35c not so much inside though but still primeing does give me no clue wheter my machine is stable or not , unless i run it 24/7 365 and that kind of defeets the purpose of having a fast dc when prime uses 100% cpu :P

Regarding databse entrys i think anyone reading them thinking they can just by the same stepping and get the sam oc is a bit off , isnt it smarter to take the average oc of a known stepping and ecpect somwhere around that speed ?

perhaps wee should start a thread where we can discuss what is stable and not further ?

Regards
Smurf

What he said...

[ekjr]

What he said...

Agreed

[babalouj]

Well it seems like my burn in session was a complete success! I couldnt boot at anything lower than 1.525V @ 2.9Ghz last night and prime instantly failed at that. Now after a 10 hour burnin with toast I can boot and prime for 14 minutes at 1.475 (turned prime off after my shower to turn toast back on, it didnt fail after 14 minutes it was stopped).

As for the stability discussion. Throughout all my experiences with the A64 series (I have had about 9 chips, not alot compared to a bunch of people around here) I have found that 30 minutes of prime (large fft) + windows memtest at the same time for single cores and dual prime instances (large fft) + windows memtest for 30 minutes 90% of the time can tell stability. After that I run dual super pi 32m's and then loop 3dmark 06 for 30 minutes. Once I got all my chips to pass that, I never had problems. Most of the time if you fail prime after 2 hours or so its a memory issue in my experiences even when running large ffts. If people want to run 24 hours of prime to be sure they are stable then thats fine with me. Usually when I get to a stopping point when overclocking I will run prime at night until I get home from work the next day. Its nice to see that it can pass over 12 hours.

[itznfb]

ghetttosmurf and chkdsk made some good points. regarding the DB, it is meant to be a GUIDE, not a "if you buy 'this' you will get 'these' results". if you guys really want a concrete answer on how to get your OC into the Prime Stable category in the DB, then i think chkdsk had the best idea, let prime run for 1 round. if its 1 round prime stable, then its stable enough for the database. unless someone is really against this decision, this is the way it will be. from here on out, i will enter Unknown in the Prime Stable Category if you don't provide ss of 1 round completed in prime. if it isn't prime stable, but benchable, i'll enter Benchable in the category.

Another comment on the database. I don't feel this is MY database, so I am more than willing to listen to everyone's suggestions. I've already made tons of changes based on other peoples suggestions. Hell I re-wrote the entire thing last week because of a single suggestion. I would like to get this method of entry to be a standard way of stepping results analytics. and i'd like to get it up on the web so anyone would be able to enter their results, no matter what chip, intel, amd, so on and so forth.

[itznfb]

last run before i switch over to the 0609FPMW

[burningrave101]

i use 30 min primes for benching and numbers, for complete stability i use 2hr, anything beyond, as i've already explained is a complete waste of time and resources. it may prime for another 100hours, and then on hour 101, then temperature in your room may change 1/10 of a degree and prime fails. when stability testing you're testing to make sure the chip is stable at the current voltage under full load. an OCed chip will ALWAYS fail at some point if not kept in a contained thermal environment. if i could somehow keep this room at 20c all day everyday, nothing could change that, then i could run 335x9 prime for years, until the chip dies in fact.

false stability claims. i don't think so. why don't you not make false accusations. i know what i'm talking about.

On what research though is it a complete waste of time and resources? Do you even have a clue whether or not the data being processed from your overclocked CPU, that is not Prime stable, is even accurate? If you get rounding errors in Prime then that means your processor cannot calculate numbers accurately when under full load and at different points it most certainly is falsifying the data it is outputting. If all it takes is a 1/10 of a degree in variation to crash Prime for you then your system isn't stable and needs ither better cooling, more voltage, or you've pushed your chip too far and need to back down a few MHz. A stable machine can be left running for hours and hours and without failing Prime. Its when you push the chip too far and its already reaching temps too high or you have voltage fluctuation caused by your power supply or power regulator on your board that slight variations cause Prime to fail. The author of the original Prime95 program suggest leaving it running for at least 6 to 24 hours and even goes on to say that it can take days to find an issue with a system. Now likely you would never find that issue in everyday usage but then again you may fire up a program one day that does. Point is though that a measly 2 hours should not be considered Prime stable. You can say that it passed Prime for 2 hours but to say that its actually "Prime stable" is false in all actuality if your not even running it the recommended time. You can argue all you want but the simple fact is i dont think you have any qualified understanding of the inner workings of a processor to be able to tell someone how long is long enough to equate to a stable processor outside of what you personally feel to be right. No crashes or lockups does not mean no errors. Thats actually that farthest thing from the truth. You can think of it the same way as you do your RAM. If you overclock it too far and its spitting out constant errors, you can still opperate the system just fine. It will still boot into Windows and you can still run your programs and they probably wont crash. But most likely down the road your going to start getting blue screens and finally corrupt the data on your hard drive.

IMO the reason some of you choose 30 mins or 2 hours is not because you have better things to do, but because you can't get the processor Prime stable at those speeds you're trying to run for any long period of time and if you had to run it for 6-8 hours you might have to back your clock back down a little and it wouldn't look as good in your sig or in the database. I run Prime while i'm browseing the internet and carrying on with my business and other things online. You dont have to completely leave the system to run Prime. You shouldn't run a whole lot of programs though because that steals CPU cycles away from Prime but just a web browser and simple applications are normally fine. When i'm going for a long run though i usually leave it running when i leave to go do something.

Hmmz... Sorry I ever asked

But I didn't ask for a definition of prime stable, just for how long you have to prime to get prime stable in the database... What's prime stable or not will allways be a discussion I think

If a CPU is remotely stable it should be able to handle at least 6-8 hours of dual Prime Small FFTs. On hardforum the time is 6 hours to be entered into the database. On dfi-street it is 8 hrs large FFTs but it should really be small FFTs if its CPU overclocks. I think 6 hrs should be the bare minimum for an overclock to be entered into the database.

It should also be pointed out that CPU-Z normally reports low voltages on the DFI boards at least and that what CPU-Z reports isn't the actual voltage being used. You should only report the voltage actually set in the BIOS because normally that near exact what the voltage is if you measured it with a DMM. For example on my Expert board i'm testing at 1.42v (BIOS) and CPU-Z is reporting 1.392v. Thats only .028v differential but i've seen it as much as .044v+.

On that note i'm going to just drop out of the conversation though because i think we've all went off on this subject for a long enough period and should get back to discussing actual overclocks instead. I got a new CCB1E 0609FPMW in yesterday so i need to test it out today .

[Pinnacle]

On what research though is it a complete waste of time and resources? Do you even have a clue whether or not the data being processed from your overclocked CPU, that is not Prime stable, is even accurate? If you get rounding errors in Prime then that means your processor cannot calculate numbers accurately when under full load and at different points it most certainly is falsifying the data it is outputting. If all it takes is a 1/10 of a degree in variation to crash Prime for you then your system isn't stable and needs ither better cooling, more voltage, or you've pushed your chip too far and need to back down a few MHz. A stable machine can be left running for hours and hours and without failing Prime. Its when you push the chip too far and its already reaching temps too high or you have voltage fluctuation caused by your power supply or power regulator on your board that slight variations cause Prime to fail. The author of the original Prime95 program suggest leaving it running for at least 6 to 24 hours and even goes on to say that it can take days to find an issue with a system. Now likely you would never find that issue in everyday usage but then again you may fire up a program one day that does. Point is though that a measly 2 hours should not be considered Prime stable. You can say that it passed Prime for 2 hours but to say that its actually "Prime stable" is false in all actuality if your not even running it the recommended time. You can argue all you want but the simple fact is i dont think you have any qualified understanding of the inner workings of a processor to be able to tell someone how long is long enough to equate to a stable processor outside of what you personally feel to be right. No crashes or lockups does not mean no errors. Thats actually that farthest thing from the truth. You can think of it the same way as you do your RAM. If you overclock it too far and its spitting out constant errors, you can still opperate the system just fine. It will still boot into Windows and you can still run your programs and they probably wont crash. But most likely down the road your going to start getting blue screens and finally corrupt the data on your hard drive.

Well put

[itznfb]

you have got to be kidding me

if you really want to waste your time that much, and want to thread crap this bad, then take it somewhere else, and submit your info elsewhere

[Pinnacle]

you have got to be kidding me

if you really want to waste your time that much, and want to thread crap this bad, then take it somewhere else, and submit your info elsewhere

Who are you referring to?

[Reinvented]

*sigh* why the hell are we still on about prime stability? i thought we were over it already? Please, go thread-crap elsewhere burninggrave101. Your not wanted here. It was clean, until you opened your mouth.

Now, itznfb, more pictures please?

[itznfb]

anyone who wants to argue the point this much, i've already said do not waste the space in this thread with mindless factless arguments. if you really want to aruge about it, send me a PM.

[itznfb]

*sigh* why the hell are we still on about prime stability? i thought we were over it already? Please, go thread-crap elsewhere burninggrave101. Your not wanted here. It was clean, until you opened your mouth.

Now, itznfb, more pictures please?

i'm at work

i wish i could remotely play with my BIOS. all i can do is remote in a look at it basically.

Additional Comment:

this is my last comment on the prime stability DB issue, and should be the last comment period

as i've said before, the database is a guide for what you may be able to expect out of a certain stepping. not a who can prime the longest competition. and just so you know, just about every 3ghz OC in the DB provided a 12hr+ prime SS.

[Primoz78]

@itznfb off topic...look at your PM...

Regards,
Primoz

[The Nemesis]

can you guys point to me which online stores are shipping out the good steppings?

Flickerdown has retail 165 CCB1E 0609 FPMW's for $355 or so. Says the 246 3 Ghz cpus he sold had the exact same steppings. I have about the same steppings on my 170 so I orderd one.

[itznfb]

i wonder why the older chips are surfacing again. for the past couple weeks we've been seeing 0610's, now we are seeing 0609's, and 0606's. i guess its better than seeing 0551's :p