To Prime or Not to Prime? OC Guru's reply...

Hi People

Ok, I got kind of bored and have been slowly dropping the 3.0 from its oced setting of 3.4 which was prime stable to 3.2 then 3.0. I thought nothing of it until something cracked inside of me and I had to try this OCing thing again. So I jumped to 3.5 knowing it will not be prime stable at 1.65v. Usually running prime would crash the pc. Instead of prime I tried 32mardk both version for 30 mins each then sandra 04 burn-in for 30 mins and the pc is still alive and well Awe-f(uc)king-some!er is it?
Prime is one of those givens right? U use prime to test stability but without using prime a setting I thought was unstable seems to stable enough at present and temps make for good reading with the reserator cooling the cpu from idle 33c to load of 40c. Thats ok right? Low vcore of 1.65 and everything. Actually I found something weird, at idle the vocre hovers at 1.66 to 1.68 but when using the pc it jumps down to 1.65! Is that normal?
During the whole benching I never saw the vcore go over 1.65 but as soon as I stopped it went up...the full setting I am using are below.

Fsb 235/66/33.NB strap:auto.agp/pci fixed.ratio 1:1, vcore 1.65v, vdimm 2.8v.Ram @ 3-8-4-4 with game setting set at a-n-a-d-d.

What I want to know is if failing prime is the end of the bencing cpu world? I seem to pass everything else with no probs. The thing is that if the way I am doing it now is OK then I think I might just be able to push it further to 3.6 to 3.65 and then rack my head around the choice to raise the vcore to 1.67 to pull out 3.75 to max the memory out. I am potentially talking about 250mhz. I have pc4000 ram so getting it running is my main goal as it would stop/save me money buying a new cpu.

Whats your opinion? I works...oh and the temps were taken at room temp of 26c. With the AC cooling the room to 18c as usual I am sure I could get the temps down even more, maybe knocking off good 4-6 degrees. The reserator responce to the enviroment its in is just amazing...the temp rises slowly but drop like in a second.

Last thing I want to say is that while priming large tfts would use the cpu at 50 ( one at load the other not-HT is enabled) the sandra testing was making the whole cpu at 100% so it was properly maxed without having to run 2 primes which is nice.

Guide me so I can help me to help the pc to help me...translated that means can you help me?

No. Its not the end of the world in benching. If you can run benches just fine, thats ok. But stability is all relative. Others like long term stability (24+ hours of prime) while others are more moderate (12 hours prime). Prime does look like the main indicator of stabilty. I mean, if you are gaming, would you want your game to crash once every 4 hours? Or once ever 24 hours? Not to mention, some say that unstable systems my lead to data corruption. My suggestion is that you should prime to a point where you feel comforatable. It'd be always nice to know you are running a system that you can count on instead of benching at high ghz levels which might crash every now and then.

i think testing 30 min of sandra and 3dmark is not close to enough... if you really hate prime, at least set the burn in and loop for a couple hours. i mean Prime is a pretty good stability tester. and yes its normal for the vcore to drop slightly under load

Well, I like benchmarks, but not to test stability. I like them to test my overall performance gains from overclocking. As for testing stability, the best way is to run the computer as you normally do for a few days. If it's unstable, you will find out! Of course, a honking big torture program or game can speed up the process for you. Remember that all hardware and software are unstable. If you are searching for perfect stability, you're wasting time. It does not exist. What you need to find is PRACTICAL stability. That means crashes are rare with normal desk top software, andvery infrequent with torture software and games. Another thing: 3D benchmarks and game benchmark modules don't tell you much about your overclocking gains. Rather, an all-around benchmark like Performance Test (Used to be called Passmarks) does a much better job for bragging rights and comparing overclock performance. It tests everything, not just 3D crunching ability. Oh yea, and if your system is overclocked too high, PT will crash your *ick in the dirt!

I think that I would rather have a stable system than have data corruption and crashing. I used to have my 1.8A at almost 3.2 but it was only stable at high vcore and it would still crash several times a day. If I have it at 3.0, I can have the vcore almost at stock (1.54V) and it's prime stable for 10+ hours and hasn't crashed in the month or so that I've had it on. What I'm getting at is speed is great, but if your programs stop working right b/c of data corruption and your computer crashes in the middle of gaming, is it really worth it?

bobad said:

Well, I like benchmarks, but not to test stability. I like them to test my overall performance gains from overclocking. As for testing stability, the best way is to run the computer as you normally do for a few days. If it's unstable, you will find out! Of course, a honking big torture program or game can speed up the process for you. Remember that all hardware and software are unstable. If you are searching for perfect stability, you're wasting time. It does not exist. What you need to find is PRACTICAL stability. That means crashes are rare with normal desk top software, andvery infrequent with torture software and games. Another thing: 3D benchmarks and game benchmark modules don't tell you much about your overclocking gains. Rather, an all-around benchmark like Performance Test (Used to be called Passmarks) does a much better job for bragging rights and comparing overclock performance. It tests everything, not just 3D crunching ability. Oh yea, and if your system is overclocked too high, PT will crash your *ick in the dirt!

Click to expand...

Don't listen to this guy....Use Prime95......
If you don't you have a chance of corrupting data and major problems that cannot be seen through "PRACTICAL" stability as this joker calls it.

I believe he was investigating stability not benchmarking

Which 3.0 do you have Jenbell? the Northwood or Prescott 3.0C or E?

It's been mentioned before, stability is in the eyes of the beholder... or something like that anyway. Prime is the highest standard that is universally accepted. If you are doing important things like Folding or SETI or crunching scientific research data, or mission critical NASA calculations... then yes, it better be Prime stable and then some.

However, if your activites mainly include playing games, downloading naughty photos, and surfing the forum, then as long as you don't experience crashes or lock-ups... then you are probably ok. I've seen plenty of times that when I was testing a new O/C setting, that my system would "seem" to benchmark Sandra and 3Dmark2001 ok, but then it would crash on PCmark2004. If it won't pass PCmark2004... then it sure ain't stable in my book.

aaroncat said:

Don't listen to this guy....Use Prime95......
If you don't you have a chance of corrupting data and major problems that cannot be seen through "PRACTICAL" stability as this joker calls it.

Click to expand...

don't flame him man come on, we can all have our opinions. i agree with you, but you can def put it in nicer terms.

stan03 said:

don't flame him man come on, we can all have our opinions. i agree with you, but you can def put it in nicer terms.

Click to expand...

agreed

i dont like prime all that much myself, it does work tho.
mainley its a ram tester in my opinion.ive found better programs for stability.
cpu stability test is better imo,so is cpuburn4

i also prefer memtest test to test ram.

everything has its uses and purposes and prime is just an decent well rounded test for ram and cpu,but i like others better.

are there any technical "theorems" on stability? Like a program or something that computer companies use or something. because i know everyone has things they like to use, but I would like a more definitive answer. I guessing by a search there isn't going to be one, so why doesn't someone create one?

My experience tells me this:

IF prime can run 12h+ without crashes, its totally stable.

It always work for me.

I like Prime95 and other stability programs cause I can stress the cpu when im not using it while Im at work. I cant be near the computer all the time expecially 7 hours on just 1 high cpu intensive program to see if it will crash which is the case with me at 233Mhz FSB compared to my 100% stable 231Mhz Fsb.

I run 2 instances of Prime95 both on Large FFT. 1 set on Real Time priority and the other on High. I close all applications running in the back plus disable the screen saver. Doing this doesnt give the stress testing any time to breath inbetween and forces all its time to concentrate on both Prime95's. At 233mhz FSB I can do it for 7 hours untill it fails. When I keep the priority at default setting and allow a screen saver to run plus have stuff running in the back then Im 100% stable at 233mhz which I clearly am not.

So its best to only allow prime95 run and have the cpu concentrate itself on it alone and at highest priority possible without any interuptions on Large FFT.

You should always run F@H under the user name sautegod to get the best results when testing stablety

batboy said:

It's been mentioned before, stability is in the eyes of the beholder... or something like that anyway. Prime is the highest standard that is universally accepted. If you are doing important things like Folding or SETI or crunching scientific research data, or mission critical NASA calculations... then yes, it better be Prime stable and then some.

However, if your activites mainly include playing games, downloading naughty photos, and surfing the forum, then as long as you don't experience crashes or lock-ups... then you are probably ok. I've seen plenty of times that when I was testing a new O/C setting, that my system would "seem" to benchmark Sandra and 3Dmark2001 ok, but then it would crash on PCmark2004. If it won't pass PCmark2004... then it sure ain't stable in my book.

Click to expand...

LOL, sir!

I concur. P95 is necessary if you intend to do heavy scientific calculations or mission-critical work.

You should always run F@H under the user name sautegod to get the best results when testing stablety

Click to expand...

Yah huh.

Idd, i never crash but if i followed prime to 12h+ i wouldnt get much over 230fsb -- im sorry but i'd rather have a decent overclock and save the chip from higher voltages. If my PC crashes, THEN i will increase the vcore. Im a gamer, not a scientist. If i was a folder, sure i'd make sure it was prime stable for a long time.

~t0m

making your rig prime stable feels good even if you don't fold. it means you got all those extra MHz for NO COST, as opposed to it costing you even the slightest stability. its good to know i'm as solid as a stock 2.4c.

I agree, I fold like a mofo, but even before I did, I wouldn't want to even have the possibility of the proc missing calcs. I have a buddy that oc'ed when he didn't know what he was doing, and had an extremely unstable comp, and his windows registry was so messed up after about 2 months he had to reformat. I know some of the guys on these boards are so into comps they reformat once a month, but I don't see that as being a TRU OCer. Because after all, if you have to sacrifice stability for speed, you are defeating the purpose IMHO.