Stability is key. Stability is key. Stability is key.

[hat]

As the title says, stability is key in running WCG (and F@H, and whatever else). If you're going to overclock and you don't want to be faced with problems such as BSOD, lockups, or the ever so infamous you sending work units in and the server raises the BS flag on them.

If your overclock isn't entirely stable, it will come out in the wash in your crunching results. First and foremost you don't want to be turning bad results that could booger the research, and secondly, if you turn in bad work units and they get recgonized as bad, you don't get points for them.

My advice: make sure you know everything about the frequencies you change when overclocking. If you're planning on achieveing some really high clocks, get appropriate cooling (water is almost a given on overclocked i7 rigs) and make sure you get a motherboard that can withstand the voltage. What program to use? OCCT Linpack is the most stressful, but other members like to run this program called Linx and loop it for some number of times. 20 seems to be a good number.

[erocker]

I kind of thought this was obvious. Overclocking is like a bar stool. If the stool isn't stable, it will be wobbly and not perform very well to keep my fat butt upright. If it is really unstable, I will crash to the ground.

[Paulieg]

I just posted in another thread on this very topic. A very stable 3.8 ghz will give you better results than a semi-stable 4.0. I've found that testing with LinX to be the quickest way to verify stability, with a 25x cycle.

One thing I disagree with is the need for water on an overclocked i7. Obviously, water will give better temps, but high end air is certainly appropriate until you start bumping up the vcore over 1.35v.

[hat]

Well erocker my friend, you and I are what some folk call oldschool. There's lots of people here that think OCCT is way too strict and if it does what they normally do, it's stable, even if it crashes when they try to run the stress testing programs.

[imperialreign]

As the title says, stability is key in running WCG (and F@H, and whatever else). If you're going to overclock and you don't want to be faced with problems such as BSOD, lockups, or the ever so infamous you sending work units in and the server raises the BS flag on them.

If your overclock isn't entirely stable, it will come out in the wash in your crunching results. First and foremost you don't want to be turning bad results that could booger the research, and secondly, if you turn in bad work units and they get recgonized as bad, you don't get points for them.

My advice: make sure you know everything about the frequencies you change when overclocking. If you're planning on achieveing some really high clocks, get appropriate cooling (water is almost a given on overclocked i7 rigs) and make sure you get a motherboard that can withstand the voltage. What program to use? OCCT Linpack is the most stressful, but other members like to run this program called Linx and loop it for some number of times. 20 seems to be a good number.



Yerp . . . and such apps are notorious for spotting OC errors really quick.

Big reason why I tend to use (and have recommended) F@H for stability testing. I've had numerous OC attempts that were stable enough to pass Prime, Orthos, 3m06, or whatever else without any problems for hours on end . . . but crash the entire system within 5min of starting F@H.


One thing I disagree with is the need for water on an overclocked i7. Obviously, water will give better temps, but high end air is certainly appropriate until you start bumping up the vcore over 1.35v.



Don't know about i7 from personal experience yet . . .

But, for the last two decades, I've always viewed cooling as "passive overclocking."

We all know that you can take two rigs that are identicle, but with different cooling setups - the rig that stays cooler will tend to perform just a bit faster than the warm running rig. As well, upper-end air cooling tends to have some low limits when it comes to dealing with heat output over time . . . plus, the warmer temperatures get, the less stable things can get, too . . . why add in un-needed stress?

The way I look at it - such measures are fine for day-to-day benching and testing . . . but if you're going to OC with such workloads in mind, you're going to put the cores through some stress to begin with . . . might as well make them comfortable, y'know?

[hat]

high end air is certainly appropriate until you start bumping up the vcore over 1.35v.

Well, I wouldn't mind welding a solid copper tube to my heater (electric radiator) to a copper block mounted on an i7 to see how well it heats up the room during the northern winter.

[Paulieg]

Yerp . . . and such apps are notorious for spotting OC errors really quick.

Big reason why I tend to use (and have recommended) F@H for stability testing. I've had numerous OC attempts that were stable enough to pass Prime, Orthos, 3m06, or whatever else without any problems for hours on end . . . but crash the entire system within 5min of starting F@H.

I've done some testing with WCG. I can run WCG for days on a semi stable overclock before I start getting error that LinX will pick up in 2 minutes.

Well, I wouldn't mind welding a solid copper tube to my heater (electric radiator) to a copper block mounted on an i7 to see how well it heats up the room during the northern winter.

I'm going on experience that I've had with 13 (yes, 13) i7 chips, both on water and high end air.

[hat]

I'm going on experience that I've had with 13 (yes, 13) i7 chips, both on water and high end air.

I don't doubt you. Hell, you alone make up for 65% of all buying done on the B/S/T forum :laugh:

Perhaps I'm a bit early in ushering in the i7 heat jokes?

[Paulieg]

I don't doubt you. Hell, you alone make up for 65% of all buying done on the B/S/T forum :laugh:

Perhaps I'm a bit early in ushering in the i7 heat jokes?

LOL. You give me too much credit. Maybe 65% of all higher end, reasonably priced Intel processors and motherboards. :laugh:

[hat]

Really? My origional figure was 80% but I thought that was just shy of the actual figure. Still, I thought it simply impossible for one man alone to be able to buy all that so I dipped down to 65%.

[Velvet Wafer]

As the title says, stability is key in running WCG (and F@H, and whatever else). If you're going to overclock and you don't want to be faced with problems such as BSOD, lockups, or the ever so infamous you sending work units in and the server raises the BS flag on them.

If your overclock isn't entirely stable, it will come out in the wash in your crunching results. First and foremost you don't want to be turning bad results that could booger the research, and secondly, if you turn in bad work units and they get recgonized as bad, you don't get points for them.

My advice: make sure you know everything about the frequencies you change when overclocking. If you're planning on achieveing some really high clocks, get appropriate cooling (water is almost a given on overclocked i7 rigs) and make sure you get a motherboard that can withstand the voltage. What program to use? OCCT Linpack is the most stressful, but other members like to run this program called Linx and loop it for some number of times. 20 seems to be a good number.

under 150 runs with at least 1024,better 2048, it isnt stable. it can crash after 100 runs, even though its unlikely. at least on phenom, I7´s are another story;)

[Paulieg]

Really? My origional figure was 80% but I thought that was just shy of the actual figure. Still, I thought it simply impossible for one man alone to be able to buy all that so I dipped down to 65%.

Drama llama, anyone? ;) I know someone has it out there. Erocker, care to help with this?

[RAMMIE]

I know all those stability programs are crap and give a false sense of security.If your computer runs the apps you use without error then your computer is stable.You can pass all those prime,linpack,occt and furmark programs and still crash in games or whatnot.

[Velvet Wafer]

I know all those stability programs are crap and give a false sense of security.If your computer runs the apps you use without error then your computer is stable.You can pass all those prime,linpack,occt and furmark programs and still crash in games or whatnot.

it can, but thats really rare... me myself luckily never had it:D

[RAMMIE]

it can, but thats really rare... me myself luckily never had it:D

Well you can waste your time running hours of stability programs or do something productive with your machine.
I haven't run any of those apps in years.Do you think my machines are unstable or I don't get the most out of them?

[Velvet Wafer]

Well you can waste your time running hours of stability programs or do something productive with your machine.
I haven't run any of those apps in years.Do you think my machines are unstable or I don't get the most out of them?

ever considered, doing both? if your rig is stable, it will even work at the same time, it will just lower your output for a few hours, in which it stays on. !1! rig,thats not i7. 1 rig is nothing, and you know that:D
linx WILL error earlier than WCG,thats a simple fact. proof me im wrong, and i, of course will stop using stability tests....

[Paulieg]

I know all those stability programs are crap and give a false sense of security.If your computer runs the apps you use without error then your computer is stable.You can pass all those prime,linpack,occt and furmark programs and still crash in games or whatnot.

I don't believe they are "crap", as I've never had a system crash after completing a thorough stability test. I do know that a few months back when I was testing an i7 920, it ran WCG for a long time without error results, but it crashed in just over 2 minutes in LinX. Everyone is going to have an opinion on this, but I chose to make sure that my system is stable by using stability software, crunching and gaming.

[erocker]

Erocker, care to help with this?

http://i403.photobucket.com/albums/pp112/erocker414/drama_llama.jpg

:D

[RAMMIE]

ever considered, doing both? if your rig is stable, it will even work at the same time, it will just lower your output for a few hours, in which it stays on. !1! rig,thats not i7. 1 rig is nothing, and you know that:D
linx WILL error earlier than WCG,thats a simple fact. proof me im wrong, and i, of course will stop using stability tests....

Remember my challenge? You put any CPU against mine for a week and if you win you get my farm for a week.
http://forums.techpowerup.com/showpost.php?p=1627601&postcount=85

I'll play for a week,a month or a year.You wanna have a go?
Anybody can challenge.Let's see what you can do.
My machine will out do yours.I am confident.

[RAMMIE]

....
linx WILL error earlier than WCG,thats a simple fact. proof me im wrong, and i, of course will stop using stability tests....

I have no reason to run linx hours on end because it doesn't do any work that needs to be done.

[KieranD]

Remember my challenge? You put any CPU against mine for a week and if you win you get my farm for a week.
http://forums.techpowerup.com/showpost.php?p=1627601&postcount=85

I'll play for a week,a month or a year.You wanna have a go?
Anybody can challenge.Let's see what you can do.
My machine will out do yours.I am confident.

Fill in your specs.

[RAMMIE]

Fill in your specs.

I have 40 machines.Quite the chore.My specs are my contribution to TPU WCG

[KieranD]

I have 40 machines.Quite the chore.My specs are my contribution to TPU WCG

oh right so you dont have a main machine okay

40 machines would really outdo most people never mind techpowerup

[RAMMIE]

oh right so you dont have a main machine okay

40 machines would really outdo most people never mind techpowerup

No I don't have a main machine.My top producers are 14 q6600's and 1 I920
All you have to do is put one CPU up and if I have one of the same the challenge is on.
The one I'm typing on is an E6300

[Paulieg]

The argument for and against stability testing will go on forever. To each their own.

[RAMMIE]

The argument for and against stability testing will go on forever. To each their own.


You're absolutely right.

[mrhuggles]

do you have an e2140?

[bogmali]

Mike (RAMMIE), which rig and what specs? I might take you up on your offer. I will match it give it a go for a week:D

[mike047]

I have never used linpac/linx or whatever the current utility is. When initial build is done, 1 hour OCCT and I put the box to crunching.

The key as Mike has found out is moderate overclocks. Long term it will beat any cutting edge box.

I ran my farm this way long before it was "cool" to run long term stability tests. It is an old school "proven" method.

My boxes do not BSOD/stall/lockup/shut down, they just reliably crunch. The only time my boxes don't produce is either power outage or project issues.

YMMV but it works.

FAH and gpu's are the same...moderation and NO problems.

[Velvet Wafer]

I have never used linpac/linx or whatever the current utility is. When initial build is done, 1 hour OCCT and I put the box to crunching.

The key as Mike has found out is moderate overclocks. Long term it will beat any cutting edge box.

I ran my farm this way long before it was "cool" to run long term stability tests. It is an old school "proven" method.

My boxes do not BSOD/stall/lockup/shut down, they just reliably crunch. The only time my boxes don't produce is either power outage or project issues.

YMMV but it works.

FAH and gpu's are the same...moderation and NO problems.

OCCT linpack? that would be the same than linx or ordinary linpack,am i right?
1 hour occt is relatively equal to testing 75 runs linpack,i believe to know;)

we all know that average overclock,that will crunch forever. but some of us, are tempted, to bring higher clocks to crunch, because they lag rigs. for this people, a stability tester has its use. otherwise your right, why loosing crunching time?

[MetalRacer]

Mike (RAMMIE), which rig and what specs? I might take you up on your offer. I will match it give it a go for a week:D

How about a three way i7 920 matchup? :D

[bogmali]

How about a three way i7 920 matchup? :D


The more the merrier:D

[mike047]

OCCT linpack? that would be the same than linx or ordinary linpack,am i right?
1 hour occt is relatively equal to testing 75 runs linpack,i believe to know;)

we all know that average overclock,that will crunch forever. but some of us, are tempted, to bring higher clocks to crunch, because they lag rigs. for this people, a stability tester has its use. otherwise your right, why loosing crunching time?

Plain old "old school" OCCT. NO linpac or whatever.

But, marginal overclocks that yield bad or poor work is not acceptable and defeats the purposes of crunching:D

[Velvet Wafer]

Plain old "old school" OCCT. NO linpac or whatever.

But, marginal overclocks that yield bad or poor work is not acceptable and defeats the purposes of crunching:D

should be prime95 or something similar.at least its heat output is comparable for me...:confused:;)

[Chicken Patty]

I just posted in another thread on this very topic. A very stable 3.8 ghz will give you better results than a semi-stable 4.0. I've found that testing with LinX to be the quickest way to verify stability, with a 25x cycle.

One thing I disagree with is the need for water on an overclocked i7. Obviously, water will give better temps, but high end air is certainly appropriate until you start bumping up the vcore over 1.35v.

I agree Paul, with either a Mega or the new Noctua, water is just not worth it. My mega by far does better than my water at high clocks/voltages. I had a pretty good water setup with a 120.4 rad, the mega beats it by over 10 degrees at high clocks.



Stability is key, i ran 100 passes of intel burn test on my rig :laugh:

[niko084]

Well erocker my friend, you and I are what some folk call oldschool. There's lots of people here that think OCCT is way too strict and if it does what they normally do, it's stable, even if it crashes when they try to run the stress testing programs.

Haha, with you there.. I'm a 72 hour stress guy myself.

I did punch up a video card a little too hot and killed my PPD for a night :(