Quick sanity check for noobish OCer - UPDATE: cpu/ram question at higher OCs

[praktik]

Hey all!

please see latest post, can even skip to it really..


-------------

JUst looking for more experienced OCers to chime in with a "ya thats happened to me before".

System:

P5e3 Premium. x48
Q9550 @ 3.4 ghz (1.3 volts, 8.5 x 400 FSB)
4 gigs Patriot DDR3 @ 1600 9-9-9-24 CR2 (1.92 volts)
FSB Strap: 333
I've bumped the NB and that FSB termination voltage just a tad too.

OK so here's what happened:
- I did keep getting Prime errors until I relaxed the ram timings. THis is strange because the kit is rated at 7-7-7-20 @ 1600 and 1.9 volts. But with CPU underclocked to 2/4 ghz and RAM at above relaxed timings 9-9-9-24 I ran prime using more memory and had no errors for like close to 20 hours...

ok it was the ram I figured... now we know thats cool... i can tweak later lets get the CPU up!

- so I put the multiplier back to 8.5 and this puts CPU at 3.4 ghz. Looking around though plenty of peeps at that speed with my CPU and less voltage. It does run fine though @1.32 volts but I peel it back and get a few degrees saved an no errors in prime for 10 hours @1.3 volts.

Then, I'm thinking well ok then things are going great! Maybe I can reduce the voltage a bit more. BUt I stop Prime running cause I have some things to do and figure I'll resume it in a bit. BSOD.

One the IRQ LESS OR NOT EQUAL, one talking about NTFS.sys, and one with no real sign of what went wrong all happening pretty quickly one after the other...

Now I downlocked back to 2.4 ghz by reducing the multipler - have been stable since then.

So here is why I need the sanity check:

- Prime ran under settings where BSODs occured for like 10 hours with no BSODs, rounding errors or anything...
- these BSODs happen with no change to settings that were good for all that Prime work...

I guess my inuition was saying: well if Prime ran for so long then why the hell these BSODs? Why no BSODS during the 10 hour prime run?

I have not reformatted this WIn 7 install for a long time.. lots of uninstalled devices and programs, recently took out two internal cards for a pro audio external interface and a consumer end USB card from creative labs for gaming. Was thinking maybe this could be a coincidence of windows deterioration - but then, I[m solid at lower CPU clock...

I dunno just confused. I see my cpu running at 3.4 with no voltage bump sometimes out there in forums. I am suspicious of the RAM: not running at rated timings even with low CPU clocks.

I am curious: could my RAM be causing these problems, and act in such a way that they only come out at higher cpu clocks? Or is this a sign my particular CPU really does need more voltage than others at the same speed? Have any of you passed long prime runs only to have windows instability after?

One last q: my mobo is used to providing 1.9 volts to the ram, but most kits out there now take less voltage, 1.65 cause of the i7s. Its no issue running a 1.65 kit in an LGA 775 board eh? My mobo is such that it doesn't turn up in many QVL lists cause of its age, I know my board was popular though if anyone has good suggestions for it lemme know..

 

[praktik]

UPDATE: ok bumped RAM to 1.98 volts, CPU is at 1.3 in windows 1.31 or something like that in BIOS...

20 hours of Prime no problem... stopped the tests this morning and hit another BSOD - considering a reformat this week...

Is my intuition correct in assuming that with 20 hours of Prime running, that I am just having coincidental OS instability and my OC is ok?

 

[CLite]

Asus website shows that you need the 0803 Bios to support that chip (2009/06/22 release). Is that your current BIOS?

http://support.asus.com/cpusupport/cpusupport.aspx?SLanguage=en-us

My build had a similar issue (I think?) I would get very random BSOD after successful long-term Prime95 runs. After I updated the BIOS I stopped getting BSOD. [note my build is not the same mobo/cpu, it just had the same kind of issue]

Just be careful when flashing.

 

[praktik]

oh ya! Actually its 808 I'm pretty sure, I'll be able to double check in a sec.. testing @ 6x multiplier in memtest 86+ right now, almost done no issues.

Will repeat @ 8.5x multiplier.

And if I pass after that, I will consider pulling the trigger on a reformat to be honest I've needed to for a while - removed a RAID array and would like to reinstall under AHCI.. and well, just to redo a clean image I can use in the future with some differences I'd like to put in to make it more "agnostic" (ie, an image of a fresh install with no vendor specific video drivers and such)

 

[praktik]

yep 808... memtest passed at lower multi - 8.5 running now...

 

[praktik]

ok blame the multitasking.. yes its 803 dunno why I was typing wrong #..

memtest passed 1 pass at both cpu multipliers

 

[praktik]

OK so update: reformatted and it was a breeze.

Snappy new updated install, fresh image ready if I need it..

So I return to 3.4ghz/1600 9-9-9-24 and Prime goes for 11 hours no issues.

Pause it, start a video - pause the video for a phone call - and boom BSOD.

Ok - so my OC was causing a teency bit of instability. Here's my question: I see my CPU with less volts and more frequency out there all the time, am I right in assuming that as the CPU freq gets higher, that it can impact how my RAM responds even if the ram speed/timings remain unchanged? My guess is that the RAM is fine at its settings only if CPU is below a certain threshold... does this happen?

Or do I just have a chip that can't work as well as others of the same type? MIne is E0 stepping btw...

Just took .5 off the multiplier and took some off of all my voltages except for Vdimm, seeing if 3.2 ghz/400fsb/16000 ram does better with no BSODs... have a new vid card on the way this week want things solid for my first PC gaming weekend in some months! (the gf is away for a bachelor party!)

 

[Spikesoldier]

watch your vdimm, 1.98v is a lot for ddr3 dimms.

 

[Ben90]

Oh jeeze, I haven't touched a 775 BIOS for a long time, but those have Load Line Calibration correct?

Basically what LLC does is compensate for the voltage drop under load (vdroop). For example lets say you have your VCore set to 1.3v in the BIOS. When idling your CPU will be receiving a near 1.3v, but once you start loading it up without LLC, the voltage will start to drop. When you enable LLC, the motherboard tries to compensate for the increased/decreased load.

There are benefits and drawbacks for using LLC. One of the benefits is you are able to use a lower VCore which can help in heat bound situations. My 4ghz overclock on the 920's stock cooler wouldn't be as comfortable without LLC for instance.

A huge drawback to LLC however is that during light-heavy load transitions your voltage still drops as your motherboard is not able to compensate in real time. This can create sporadic crashes that you can't really check for that easily. The only real solution to this is to up your VCore slightly higher than what would be considered stable at a constant load or turn off LLC which will also require more VCore.


This of course is assuming that LLC is the problem, which very well might not be since your RAM isn't stable at its rated frequencies and there is always a plethora of things that go wrong when overclocking.

 

[Ben90]

Also I would try running something like LinX instead of Prime95 for quick load stability checks. While no single program will stress every part of the CPU perfectly, Linpack normally finds general instabilities much much faster than Prime. Just be sure to watch your temps as well, because it also creates a lot more heat than Prime.

I would say 10-20 passes of max RAM to be considered stable. Random nonload crashes aside.

 

[praktik]

Hmm I think I will give that a shot Ben90!

So far, no errors since 11 oclock last night running continuously, no issues yet... starting to tire of this round of OCing a bit I think I will play around some other time and give that LinX a shot... also considering getting RAM but haven't been able to committ myself to it - might be easier when bonus comes late feb....

As for the ram voltage ya 1.98 is a lot nowadays but when I got the kit was kinda normal. Had an OCZ pair before this one that stated to run it at 2.1 (mine's rated 1.9)! I just bumped it up a tad cause I was suspicious of the ram...

I'll just reiterate the question I had above though to tap the minds of you experienced OCers:

- All RAM settings staying the same, can a higher frequency CPU setting "bring out" problems in RAM that you wouldnt see at lower frequencies with same settings?

 

[Ben90]

All RAM settings staying the same, can a higher frequency CPU setting "bring out" problems in RAM that you wouldnt see at lower frequencies with same settings?

I believe it can; although, I've never really double checked my quick experience with your question.

My Uncore (memory controller/L3$) can run up to ~3GHz and my ultra budget RAM is good up to around 1500Mhz. However when I run both of them at that speed together I crash pretty quick. Again though, I'm too lazy to really run stress tests when adjusting my settings so it could have been a fluke.