Printable View
Thanx ..DankeQuote:
Originally Posted by Amurtigress
its a nice adventure
Q9550 E0 on an Asus P5 B (very first generation) :ROTF: :eek:
Prime 95 (custom incl.) for hours with 1,1 V Vcore
fails in seconds in IBT
---------------
now ....NB @ an insane 1,65 V (thanx uncleweb !) brr.... IBT stable (50 runs @512MB)
failure with 1024 MB RAM
Will for sure grill my P5 B .......
Best
M
There's a bit more to OCing a quad.
1st check the how to guides. There're several around the forums.
Then, you haven't mentioned or Vdimm, Vtt (or as another mentioned -- SB).
There's also fine memory timing and GTL ref.
It all depends on what your MB gives you control over.
You also didn't mentioned whether it was a failure in linpak, a windows lock, or a spontaneous reboot.
For now (as a quick test) -- I'd give Vtt a bump.
AND, (side note) I hope that you're running a few cycles of memtest #5 before booting windows. It'll minimize the risk of having to reinstall or restore from backup when you corrupt the OS...
1.65v is not insane on the NB.
Yes, I learned that the hard way with my AMD machine.Quote:
Originally Posted by davidk21770
Any plans to change error detection and the interface?
Regarding your question about temps,
I haven't done any Linpack testing lately, no time for tuning anymore :(.
I've been stable for two weeks with WCG 8 hours a day, 1.32v good enough for me.
well , got some strange readouts there
(qx9650 @ 3333 (x10) / CellShock MSC PC14400 PC3-1800(2gb kit black) /Rampage Extrem bios 0501 )
screen shot (large)
http://img396.imageshack.us/img396/5...39crdq1.th.pnghttp://img396.imageshack.us/images/thpix.gif
Add /Edit
FYI i was using XP pro SP2 32bit .. ( if it make any sense .. )
Cropped attached image :
Wow, nice bug :D
Hi,Quote:
Originally Posted by mine
actually it won't. I've had a P5B, P5B Deluxe, and a Maximus Formula. All 90 nm chipsets of intel require about 1.51-1.61V for a NB clock of around 500 MHz. Add 0.04V when using four RAM modules. and another bit when running a Quadcore. Too little voltage gives you some creepy problems, like using several PCI Express cards causing system stuttering. Which goes away magically at some more voltage.
45 nm CPUs tendentially require some less voltage on the NB, and make sure to keep it at +- 0.03V of the ideal settings. More or less will cause latent issues.
This is a clear hint that the NB is feeling better with more, so to say. But make sure you add a fan to your heatsink on the P5B. And probably replace the standard compound under the sink with Arctic Silver 5.
Oh, and be careful to not underestimate the Northbridge clock. The NB clock is NOT always identical to the FSB clock. It is only when the CPU multiplier is set to it's max. value!
Oh, and make sure to not use more than 1.4V as FSB Termantion voltage, for 45 nm processors everything beyond causes severe degradation and a quick death. Examples of my setups:
Intel X38, E6600 CPU, 4x1 GB DDR2-800 RAm:
FSB term voltage: 1.45V, NB Core at 490: 1.61V-1.63V
Intel X38, Q9300 C1 CPU, 4x1 GB DDR2-800
FSB Term: 1.36V, NB Core volt at 436: 1.55V
Intel P45, Q9300 C1 CPU, 4x1 GB DDR3-1600
FSB Term: 1.36V, NB Core at 460: 1.28v (The P45 is a 65 nm chipset!)
Just in the same position with my current OC bump (in fact, I was running prime to try to figure out what to bump to fix the linpak failure -- and it didn't fail :shrug: -- so I went back to the forums and did some research).Quote:
Originally Posted by mine
What it took FOR ME (YMMV)...
A Vcore/VID bump AND changing GTL from 0.67 to 0.635
(GTL changes can be easy or hard depending on the MB manu).
Check here for more info: http://www.xtremesystems.org/forums/...d.php?t=202292
thanx guys
you are very helpful
now I ended up with 1, 55 V NB
Primestable for 8 hours (small-large-blend -custom all)@ 3 GHz.. Vcore 1,16 V (CPU-Z)
IBT stable 50 runs @ 512 MB RAM
@ 1024 MB 50 runs ...5 failures (RAM = 2 x 2 GB Apogee 1150)
failures are only in IBT (no reboot -no bsd -no freezes)
Do you think my PSU has some influence ?
Its my brave 420 W Be Quiet ( 9550 -2 SataHD - 1 3850 - 3 fans )
Never had any problems with it with the E 6600...
@Amurtigress.....
thanx a lot for your settings ...rather similar here (fsb term @ 1,3 V)
@ david.....
P5B-DLX : ... sorry no such things as GTL... :D
Best
mine
mine
Try to pass Win MemTest 3.7. If you get errors, then you need encrease Vdimm.
I have 8GB ram, Vista 64 and Q6600 at 3.6GHz...
Using max stress, IntelBurnTest 1.7 went for 7 hours fine and then froze...
My temps max out at 90-93c, I know this is hot but AgentGOD has 95c.
My eVGA 750i (with pencil mod) voltages:
Vcore: 1.5375v
SPP: 1.5v
NF200: 1.4v
GTLREF: 105, 105, 100, 100
So... obviously it isn't "stable", but people are doing nowhere near as long as I am and calling it stable...
I increased Vcore to 1.54v and it went about 6 hours maxing out at ~94c then froze, so I don't think Vcore is the issue, it has to be something else.
My memory is G.skill PC2-8000 running at 900MHz 2.1v linked and synched, and I get no memtest errors etc, so no problem there. My FSB is at 1800.
Do you think this behavior just means that my CPU cant handle 3.6GHz regardless?
See this thread: http://www.xtremesystems.org/forums/...59#post3367859Quote:
Originally Posted by XxDeadlyxX
It'll take a while to sift through, but it's well worth it. A lot of really good tuning and setting info there.
And, yes, some people's definition of "stable" means that it doesn't crash with what they do -- not that it will never crash. I'm with you -- that's not stable, it's just lucky.
And don't forget good old prime95. When I'm testing a new OC, I've been lintest stable, but still had prime crash. Sometimes as quickly at 15 minutes, but more usually around 6-7 hours. If I can break 8 hours prime95, I'm usually good to go as long as I feel like letting it run.
And 95's too hot. I'll bet you're into throttling and you're actually running slower than at a lower FSB. Have you done any benchmarking?
Well, if you guys are running this for over 7 hours, you're going like 900 hour prime stability here :P
Woohooo! I just passed 100 runs of IntelBurnTest (13 hours) with these settings:
Vcore: 1.531v
SPP: 1.5v
CPU FSB: 1.45v (I hadn't had this up to 1.45v before, it must have needed to be)
NF200: 1.4v
GTLREF: 105, 105, 100, 100
That equates to... 1500 Prime95 hours, so I think I'm stable :banana:
Congratulations mate!
Whats the general consensus on this one?
10 iterations sufficient? only 5?
or more like 40-50?
I haven't had any trouble with rigs that will run 5 iterations.
But i have had OC's that will pass 2 iterations, but not all 5
It's great that you've gotten to that point, but don't rely on just this program. It's great to find errors quickly, but it's just one piece of the puzzle. My system was 60+ cycles of linpak stable and it still failed prime at 5:30.Quote:
Originally Posted by XxDeadlyxX
I have a question
If I run 15 loops 3 times at minimum ram and it passes, whats going on when I run 10 loops max ram and it locks the rig up?
I have up'd NB, FSBT, RAM, PLL and it still locks up
No I have not tested the ram yet was about to run memtest, it just passed today a nice 9 hour run of prime small fft's
Ah ok, Small FFT or Blend?Quote:
Originally Posted by davidk21770
I guess Blend, as small FFT involves mostly the cpu that would be stable after IBT.Quote:
Originally Posted by XxDeadlyxX
In my experience IBT is very fast to find pure cpu instability but very poor to find NB/memory one.
Chances are that's because your ram is buggered.Quote:
Originally Posted by davidk21770
If you run a bunch of linpack, and pass
and you run a couple memtest and pass
Then running blend seems silly if all your doing is stressing some of each, when you know 100% of each are good to go... no?
just to chime in
exchanged board from Asus P5b-dlx (very first gen. , no rev.)
which gave me so much problems with IBT runs
now : Giga EP-45 -DS 4 and
besides this board acts (memorywise) like a German Brewery horse
no problems at all with 50 IBT runs with full memory @1160 DDR2 .
thanx guys for your help
m
Not at all. Stability of individual components is no guarantee of system stability. No different then being 2d stable bit crashing in games. A variety of tools are necessary for an indication of overall stability.Quote:
Originally Posted by ThaLoveDocta
Will there ever be a fix for testing with this program under 32bit OS and maximum ram test ? I kindda like it, but some things still need some fixes