Overclock passes 8 hrs P95 @ priority 10, but fails 1min of LinX?[SOLVED]

[poohbear]

Hey all, I have what i would like to consider a stable overclock on my X6 1100T (3300mhz stock, specs in sig). I have the CPU overclocked to 3800mhz @ 1.40vcore & it passes 8 hours of Prime95 on priority 10 (the most intense priority). I leave it overnight and dont touch it and no programs are running. I used the Small FFts and Blended tested and it passes both for 8 hours. I DON'T have the CPU/NB overclocked, only the CPU.

However, 1 minute of LinX or OCCT & the computer reboots. At stock speeds it passes LinX or OCCT fine, but once i overclock it, even its a measly 3500mhz (200mhz increase), it reboots. I've tried pushing the vcore incrementally up to 1.525 and it still reboots. What gives?

I read if the mobo can't handle the sudden voltage and stress loads of LinX & OCCT it will reboot, because its like a jackhammer hitting the CPU on and off & on & off (i saw this on the cpu workload when LinX was testing it at stock speeds, all 6 cores would jump to 100%, drop, then jump again to 100%).

It runs everything fine, but i just have doubts that it might crash one day when i least expect it and when im doing something important! Any feedback? Thanks in advance.

 

[TakeNoPrisoners]

I do not like P95 for testing. I use intel burn test and it finds instability much quicker. That said I find gaming even better to use for stability testing because that introduces more heat into the case and will cause things to crash that would otherwise be stable without the GPU running.

 

[classy]

I do not like P95 for testing. I use intel burn test and it finds instability much quicker. That said I find gaming even better to use for stability testing because that introduces more heat into the case and will cause things to crash that would otherwise be stable without the GPU running.

I agree. In all the years, I have found 4-5 hours of serious straight gaming to be the best test for OC stability. So of those programs are good, but many don't tax the whole system.

 

[Schmide]

Do you have the latest bios for that MB?

 

[Big Lar]

It could be quite a few things including voltage for the NB, as well as maybe try a different version of OCCT. If memory serves, P95 needs a tad More voltage to run than OCCT, sometimes OCCT will act that way due to that slight bit Too much voltage that P95 needed to be stable.

In my testing Intel Burn Test isn't reliable enough and I do agree that Gaming seems to be a decent stress test. However were it me, I would first try a different version of OCCT before playing with the voltage any more. Just IMHO.

Larry

 

[eternalone]

OCCT destroys these so called stable overclocks some people post on here, thats why I laugh when I read some of these ridiculous overclocks and the people that think their rig is stable. Prime 95 is also known to kill motherboards, OCCT destroys only your sense of satisfaction when you realize you have to lower your clocks. LOL

 

[Suspicious-Teach8788]

OCCT is truly the best test out there. I use the OCCT Linpack test for 24 hr stability with the regular OCCT test too.

Prime 95 just isn't sufficient anymore. It's a 2005-esque test lol.

 

[richierich1212]

It's not your vcore that's causing instability. Your motherboard probably can't handle the extra juice required.

 

[pantsaregood]

I'd usually swear by OCCT/Linpack, but I actually had a bit of a reverse experience once. I ran Linpack for ~12 hours at 4.5GHz on 0.02v above stock with my 2500K without issue. It, however, failed Prime95 almost immediately. I had to bump it up to something like 0.06v above stock to Prime95 without errors.

 

[john3850]

Your ps specs are 12V1@18A,+12V2@18A,+12V3@18A with a {125W} OC Phenom II X6.
I know that power hungry 5870 has a seperate line but it still pulls power from your mb.
I ran off a {95watt 2500k} with the same spec ps with 12V1@17A,+12V2@17A any thing over 4200mhz with LinX I got restart.
I put my system in a case with larger single rail ps no more restart problems even at 4800mhz.

 

[poohbear]

Thanks for all the feedback guys! yes mobo has the latest bios.

Your ps specs are 12V1@18A,+12V2@18A,+12V3@18A with a {125W} OC Phenom II X6.
I know that power hungry 5870 has a seperate line but it still pulls power from your mb.
I ran off a {95watt 2500k} with the same spec ps with 12V1@17A,+12V2@17A any thing over 4200mhz with LinX I got restart.
I put my system in a case with larger single rail ps no more restart problems even at 4800mhz.

Im pretty sure its something mobo related, but u think its PSU related? The Corsair HX520 has 40A total & is a pretty badass PSU, and the CPU & Video card are supposed to be on separate rails... but u might very well have a point cause i did some reading and got this from a review site:

"On the back, we can see the traces and how the two different 12V leads go to the two different PCI-e connectors and the different peripheral connectors.

The "official" word from Corsair was the following:

12V1: ATX 20+4, 4-pin +12V and 8-pin +12V
12V2: PCI-e 1 and first two peripheral connectors
12V3: PCI-e 2 and last three peripheral connectors

It gets better...

Looking at the main PCB, we only see two points where 12V leads are soldered down. One labeled 12V1 and another labeled 12V2. For fun, I put a 30A load on the 8-pin EPS+12V connector. This should have tripped the PSU. It did not, which tells me that there is no OCP limiter set for this rail. I did the same with the ATX main connector and again the PSU did not trip....there is no OCP (over current protection) on each rail. Outside of a few traces zig zagging across PCB, I couldn't find how even 12V1 and 12V2 are separate, but I'm going to give Seasonic (the OEM for the Corsair units) the benefit of the doubt and say that we seem to have two 12V rails here, neither with any kind of "limit" on them."

So, does that mean they're not on two separate rails? The HX520 is a solid PSU that was the most expensive in its class for its specs. LinX does pass all the tests at stock, but once overclocking the CPU alone (CPU/NB at stock), it restarts, even @ a measly 200mhz overclock regardless of how much vcore i feed the cpu or how high i push the fans.

I'm more inclined to think the CPU circuitry on my ancient AM+ mobo simply can't handle the stresses of an overclocked 1100T (it has 3 revisions, wherein they're supposed to provide better overclocking support... i have the original 1.0 version!). Also, the jackhammer style test of OCCT & Linx (basically the same style test) might also be too much for the CPU circuitry on my mobo. Wouldn't that make more sense?

 

[john3850]

Corsair had problems and set it up as a single 12v which is common.
Glad your ps is cool.
On my i7x58 @4200 I can run 7 threads of LinX no problems.
When I try 8 threads on LinX instant restarts after a minute.
Since no game etc will ever pull 100% on 8 threads I left it.
My cheap 12 phase mb suks but has never restarted in normal use.

 

[bunnyfubbles]

I do not like P95 for testing. I use intel burn test and it finds instability much quicker. That said I find gaming even better to use for stability testing because that introduces more heat into the case and will cause things to crash that would otherwise be stable without the GPU running.

I agree. In all the years, I have found 4-5 hours of serious straight gaming to be the best test for OC stability. So of those programs are good, but many don't tax the whole system.

I don't agree, at all

In all my years of gaming I have hardly ever taxed a CPU more more than two cores (or even two threads going back to the days of P4) near 100% for the entirety of my multi-hour gaming sessions

Even now my most CPU intensive games will hardly push my quad core CPUs beyond 60% usage.

Maybe for people still gaming on a dualcore gaming might be an acceptable stability indicator, but it hasn't really been one for me for years and years. If anything the people I see making such a claim are those who will fail a stress test like P95 or LINX but will be "stable enough" for games and thus they'll use that as an excuse to maintain their precious epeen status instead of running a truly stable rig.

OCCT is truly the best test out there. I use the OCCT Linpack test for 24 hr stability with the regular OCCT test too.

Prime 95 just isn't sufficient anymore. It's a 2005-esque test lol.

IME Linpack isn't necessarily any better of a stability indicator (vs. say OCCT's own stress test) as much as it is an extreme heat indicator, its kind of like the Furmark of CPU stress tests in that its simply way over the top of any normal 100% CPU uses and is really only a tool I'll use for rigs I build that have to be "110%" stable.

That being said, I wish OCCT would get updated, it really was the perfect all-in-one tool but I'm pretty sure it hasn't been updated since X58 came out and thus is largely broken for when monitoring newer hardware, although I suppose the stress tests are still functional. It used to be my "one-stop-shop" of overclocking tools, with SandyBridge I now need things like CPU-Z and realtemp or coretemp along with the stressing app like P95 or OCCT.

 

[poohbear]

My cheap 12 phase mb suks but has never restarted in normal use.

yes mine never restarts in normal use or games or even Prime95 @ priority 10 for 8+hours, only in LinX & OCCT when overclocked. How about yours, does it restart when stress testing?

 

[Minerva]

That being said, I wish OCCT would get updated, it really was the perfect all-in-one tool but I'm pretty sure it hasn't been updated since X58 came out and thus is largely broken for when monitoring newer hardware, although I suppose the stress tests are still functional. It used to be my "one-stop-shop" of overclocking tools, with SandyBridge I now need things like CPU-Z and realtemp or coretemp along with the stressing app like P95 or OCCT.

Try the 4.x beta...

http://majorgeeks.com/OCCT_d5612.html

 

[richierich1212]

I think a better test for everyday stability is using BOINC or WCG

 

[JimmiG]

OCCT destroys these so called stable overclocks some people post on here, thats why I laugh when I read some of these ridiculous overclocks and the people that think their rig is stable. Prime 95 is also known to kill motherboards, OCCT destroys only your sense of satisfaction when you realize you have to lower your clocks. LOL

To some, "stable" means it can finish one loop of 3DMark without crashing

Prime95 doesn't "destroy" motherboards, though. It contains no code to do this. It just tells the CPU to do what it was designed for - calculations. Sure, it tells the CPU to do a *lot* calculations, but again, that's the purpose of the CPU. if that causes something in your system to break, then it just exposed some weakness that was already there.

 

[john3850]

LinX on my 930
I can only run LinX at 3800-3900 with 200BCLK @1.29v with no problems.

The higher the oc the more vcore at load is needed which in turn increases droop = restart.
Cpu set for 4000hmz linx
1.376vcore idle
1.280v- load with LinX vcore is on low side
0.096v = vdroop at 3800 never mind 4200

What I learned is a oc x58 pulls a heavy load and more then a cheap 12 phase vrm is needed to maintain a healthy voltage.

 

[poohbear]

argh guys im still trying to get this overclock 100% stable.... my CPU/NB passes 8 hours of stability @ 2800mhz @ 1.375v, but once i combine it w/ my 8hr prime95 stable CPU overclock of 3800mhz, it fails P95 in 3 hours. I've tried upping the vcore on the cpu all the way up to 1.55, but no luck. does the cpu/nb need more voltage when i combine it w/ a CPU overclock? mind you the overclock works perfectly fine in all my apps and games (Crysis, Shogun 2, DoW2), but none of them utilize 6 cores so im guessing that's why im getting away with it?

what gives? i read that AM2+ mobos can't handle 2800mhz+ on the CPU/NB, i need an AM3 one for that, is that true?

 

[Thor86]

Always use "custom" settings when stress testing overclocks using Prime95.

Select the right number of threads in relation to number of cores, and setup the test FFT size to 4096 minimum, and setup memory at least to a number rounding to 66-75 percent of your total physical memory installed.

This has always worked for me since the P4 days.

 

[poohbear]

Always use "custom" settings when stress testing overclocks using Prime95.

Select the right number of threads in relation to number of cores, and setup the test FFT size to 4096 minimum, and setup memory at least to a number rounding to 66-75 percent of your total physical memory installed.

This has always worked for me since the P4 days.

but its automatic now. It does 6 threads for my 6 cores without me telling it to.

like i said testing the NB/CPU & CPU individually is not a problem, but testing em together and comp reboots in P95.

 

[poohbear]

the crazy thing is the cpu seems to get MORE UNstable (reboots in less time in P95) the more vcore i give it, not more stable. In all my other cpus if i give it a little more vcore it usually takes longer to reboot as i approach a stable vcore. But w/ this one each time i give it a little more vcore it reboots faster in P95? This is even with the 2 fans on my massive Frio heatsink blasting away at 2500rpm each which sounds like a helicopter, normally the pull is @ 1200rpm and push is @ 1400rpm, and that works fine for a 3.8ghz overclock at a mere 1.425. This is all with the CPU/NB @ default. So a mere push up 200mhz (for 4ghz) and more voltage should be no problem, but no matter how much vcore/cooling i give it, it just reboots @ 4ghz. which leads me to believe it might indeed be a psu issue as john3850 pointed out.

 

[jvroig]

Given what you say, I would suspect it is a motherboard power-delivery issue. From what you just described, my first thought is that your motherboard simply cannot deliver the power your CPU demands in its OC state as evidenced by additional vcore resulting in just reboots.

EDIT: Of course, it could simply be that your chip does not OC as far as you would want it to be. You know how it is - no OC is guaranteed. It could be just that, and the rest of your system (mobo + PSU in particular) is just fine.

EDIT 2: By the way, I have a 1090T. It can run stably at 4GHz core (and 2.8GHz CPU-NB). However, it needs 1.525Vcore. My old Phenom II X4 965 (C3) was a poor clocker as well, needing 1.55V to get to 4GHz stably. The reality is that no matter how many people go around saying "most Phenom II / Thuban can do X GHz easily", real life is not so forgiving. Moral of the story is not buying any chip on the expectation that it will OC to a specific target in mind, because it can most certainly fail to do so.

 

[poohbear]

EDIT 2: By the way, I have a 1090T. It can run stably at 4GHz core (and 2.8GHz CPU-NB). However, it needs 1.525Vcore. My old Phenom II X4 965 (C3) was a poor clocker as well, needing 1.55V to get to 4GHz stably. The reality is that no matter how many people go around saying "most Phenom II / Thuban can do X GHz easily", real life is not so forgiving. Moral of the story is not buying any chip on the expectation that it will OC to a specific target in mind, because it can most certainly fail to do so.

i understand & agree about no overclock guaranteed, but like i described, i have it running stable (stable is defined as 8 hours priority 10 Prime95 blended & Small FFTs test) @ 3.8ghz @ only 1.425v on my Frio with only @ 1200rpm pull fan & 1400rpm push fan (CPU only , not CPU/NB). I should definitely be able to up it a mere 200mhz to run @ 4ghz especially with both fans on this massive heatsink running @ ridiculously loud 2500rpm each. No matter what vcore i use @ those fan speeds it will just keep rebooting, and the higher up i go on vcore the faster it reboots! I went all the way up to 1.575 (in .25 increments), and it just kept rebooting in Prime95. LinX will ALWAYS reboot in 5 seconds regardless of the overclock or the vcore, anything above 3500mhz & it reboots.

Therefore it must be either a mobo or PSU issue. Im guessing mobo cause its a revision 1 (gigabyte made 2 more revisions of it) & AM2+ which were'nt really designed for overclocking these 6 core beasts. They can run em fine @ stock, but any decent overclock and it just chokes it seems. Is this a fair hypothesis?

BTW what voltage are you giving your CPU/NB @ 2800mhz? thanks!

 

[jvroig]

i have it running stable [...] @ 3.8ghz @ only 1.425v on my Frio with only @ 1200rpm pull fan & 1400rpm push fan (CPU only , not CPU/NB). I should definitely be able to up it a mere 200mhz to run @ 4ghz especially with both fans on this massive heatsink running @ ridiculously loud 2500rpm each

There is no guarantee that yo can go 200Mhz more. That's the whole point of overclocking being not guaranteed. Your chip could just simply not be capable of your 4GHz target.

My motherboard is also an AM2+ Gigabyte, MA785GM US2H. It can handle the power delivery just fine. Through an accident, I was able to determine that at 4.2GHz and 1.6Vcore, my thuban is stable.

I am not ruling out any other possible problems (for all we know, it could certainly be just a poor / faulty motherboard or PSU), but you can't say you "believe that OC is not guaranteed" and then come to a conclusion that just because you can reach X GHz easily, you must definitely be able to reach X+200MHz also.

Unless you are willing to buy a new motherboard and/or PSU for a mere 200MHz bump, the possibilities are:
1.) CPU limit
2.) Motherboard can't handle the required power for your target OC
3.) PSU can't supply enough power (I doubt this, especially since your GPU isn't active while stress-testing the CPU OC)

It's just 200MHz more. I don't think it is wise to make a gamble by buying a new board (or PSU) on the chance that your CPU might be able to do more.