GA-EP45-UD3P - Won't boot with 450FSB

[D-an-W]

I ruled the 4 extremes out and are currently working my way down the middle of my chart. I am also keeping a note of how many lines Prime gets to before it fails, I am right in assuming that lasting longer before a fail is better or is it just a random coincidence when it fails (I know what I mean )?

[BoombasticSloth]

I make note of the test used, the FFt size of failure, duration till failure, max temperature, and voltage after vdroop. When analysing the spreadsheet with this information along with the bios settings, it really helps to narrow down the options and to see where the strengths and weaknesses are in the approach.

[D-an-W]

blimy, I just drew some lines on a sheet of paper and are scrubbing out the ones that fail

[BoombasticSloth]

hahaha, you are such a kidder.

But seriously, it is not enough to just knowing whether a particular bios setting change passes or fails. The more important thing is why it passed or failed and how it helped or made things worse. While the first part is hard to determine, the second part can be determined. Further, it can help you identify how combinations of bios settings work in tandem with each other, for better or for worse. Hence, the myriad of options can be reduced when trying to achieve ones oc goal. That is why I feel oc'ing is as much an art as it is a science. Lay all the information in your speadsheet and really look at it, you might just have an epiphany.

[devoh]

Originally Posted by devoh

Wow I've never really played with the CPU PLL. I set it to 1.57 and have left it there.

I am trying the CPU term of 1.32 and now I think I have 4.2Ghz stable.. will know in another 10 hours (or so )

-devoh

24 hours stable.. 12 hours prime small in place FFT, 12 hours large in place FFT, and just to show D-and-W I can go over 24 hours I'm going another 12 hours with prime blend.. oh not to mention the UT3 playing and IBT runs.

-devoh

[D-an-W]

Don't do it devoh! profJim will be challenging you to a 72hr Prime off!!!

[edge929]

I've been reading this entire thread over the past couple of days (at work of course) and wanted to chime in on my findings with the love/hate relationship I have with my UD3P. I apologize ahead of time for the long post.

I've been building computers for about 9 years now and overclocking for all 9. In that time I've built well over 200 computers and overclocked about 15% of those. Everything from a Pentium II to a Phenom and Core i7. I've used stock coolers all the way to custom, high-end, chilled water-cooling (damn you Prometia for being so expensive). I've done soft-mods, hard-mods, soldering and more BIOS flashes (mobo and GPU) than you can shake a stick at.

Coming with that territory, I'd like to tell you I've seen it all but I haven't and never will. Overclocking from rig to rig is different. I've prime'd 36 hours stable only to launch Crysis and BSOD within 30 seconds. I've run Orthos for 24 hours straight only to have SuperPi blow up on 32M. I've run SuperPi 32M 10 times in a row only to have the 11th time blow up. I've seen a workstation run 24/7 for 1 year mildly overclocked but memtest86 v3.4 threw errors within 10 seconds. As you can imagine, none of this promotes confidence in any 1 testing method. I'm sure many of you have learned this already.

My personal "rock solid stable" award only goes to rigs that can run Prime95 (or Orthos) stable 24 hours, SuperPi 32M for 5 runs, can unRaR a 1GB+ rar file made up of 100Kb or smaller files (try it sometime, unleashes hell on your L2 cache) and most importantly, can endure my 6-8 hour WoW marathon sessions (full graphics during raids 1920x1200).

Here is my current setup (back to high-end air)
CPU: Core 2 Duo E8400 C0
Cooler: Prolimatech Megahalems 120mm fan hooked up to fan controller (set at low)
Mobo: Gigabyte EP45-UD3P BIOS F7
RAM: 2x 2GB GSkill 1066 5-5-5-15 2.0-2.1
HDD: Intel 80GB X25-m
GPU: Radeon 4870 Space heater
Case: HSPC Topdeck station (google it)
PS: Corsair HX750w (no not the TX)
1x 120mm fan on NB/Memory
1x 120mm fan on mosfets/CPU area

I should note that this rig isn't an entirely new build. The CPU, GPU and RAM were salvaged from my old 650i SLi rig where my E8400 was on water running 4Ghz 24/7 for over a year.

Stable: 500x8
Current testing: 533x7 is Orthos stable 15 hours
Goal: 533x8, RAM 1:1 5-5-5-15, 9 or 10 tRD

My goal is for 24/7 use. When I first got the board, I couldn't get 480 stable and I was pulling my hair out reading that Anandtech and countless others are sitting pretty with 550 and using THEIR settings (yeah I know better). After playing with the countless voltages I finally found a balance of MCH core, VTT and references to get 500 stable. I already knew my CPU could run 4Ghz at 1.35Vcore all day (even being a C0) and my RAM had passed 50+ hours of memtest v3.4 from my old rig, so I knew my problem lay with the MCH core. I've been literally lol at some of your guys comments about where the MCH core should be and where it should NOT be. With that said, I'm sitting pretty at 1.56 MCH Core with a 120mm fan on it (low setting). I did the highly sophisticated finger-touch and it's barely warm. I can leave my finger on it all day. Dropping this below 1.5 will immediately give me instability, I can POST but Windows freezes on boot-up. I have not taken off the NB heatsink to apply my own thermal paste but I may still do this if I can't get 533x8 stable. Has anyone taken off the cover to get more airflow to the fins? Bad Gigabyte.

My other findings on this board:
As others have said, upping a voltage may just cause more instability despite logic. It took me awhile to figure this out. My entire overclocking experience says just the opposite, more Volts = more stability. I was almost mad when I figured this out. I dropped some voltages and was immediately stable.

Pushing my CPU to 500x8.5 needs much more Vcore, on the order of 1.45+. My computer is located in my basement and ambient temps are rarely above 72F. My temps across the board are well under the radar (yes, even with 1.56 on the MCH).

I'll admit, RAM timings were never my strong suit. I've usually left most settings on default/auto after the usual 5-5-5-15 2T settings. I wasn't aware the tRD was as important to performance as LSD's posts showed.

If anyone is interested in my settings/voltages/love life, let me know and I'll post back, can't remember all of them off the top of my head.

Sometimes I wonder if all the time I spend tweaking/reading about tweaking is worth it, because after all, I should be using that time to enjoy the computer experience, not work on the "engine". Will I really benefit from another 200mhz? Maybe if I encoded video all day, but I don't. In the end, knowing my rig will decimate all but 1% of computers out there makes me sleep better at night

[Psycho101]

You'd be welcome, I'm sure, to post your current settings. This thread is probably the definitive E8xxx and EP45-UD3(x) resource. The more settings and experiences the better.

My personal "rock solid stable" award only goes to rigs that can run Prime95 (or Orthos) stable 24 hours, SuperPi 32M for 5 runs, can unRaR a 1GB+ rar file made up of 100Kb or smaller files (try it sometime, unleashes hell on your L2 cache) and most importantly, can endure my 6-8 hour WoW marathon sessions (full graphics during raids 1920x1200).

You play WoW AND realise the importance of MCH/CPU stability whilst using compression software. I like you! You can stay. Many a time have I gotten 24 hour Prime Blend stability and 12 hour small FFT only to un-rar an ISO of 6GB+ or some family photo's (re-prints for an anniversary etc) and get corrupt file messages. Large FFT would also fall flat on it's ass after 10 hours with the unstable settings.. MCH Ref sorts it for me almost always. That or a Vcore bump.

WoW is an extremely CPU intensive game, and thus a great test of stability. I don't stability test in raids or instances though, I'd get my ear chewed off by people. Have you noticed, no matter your CPU and RAM speed and the power of your GPU, dynamic world shaddows (2 notches from minimum and up) still pop in and you can also see higher res shadow-maps being drawn as you get closer.

I'm sure you've tested already, just asking to make sure but... that's a sh1t load of MCH Core indeed! Have you tried altering MCH ref down some at 1.2VTT? If not give it a go and MCH Core may come down sharply. VDIMM is worth pushing up to 0.08V out of spec if it gets that monster voltage down. However, if it's barely warm to the tough, it doesn't really matter. I'm sure you'll move on to another board before it ups and dies on you.

Post that templare, it would be interesting to see!

[creamsoda]

With that said, I'm sitting pretty at 1.56 MCH Core with a 120mm fan on it (low setting).

Sorry to say this but, 1.56 mch is way too high for 533fsb, even on an ep45-ud3p. On my ep45-ud3p board, I ran my E8500 at 526x8 (4.2GHz), and I only needed 1.34v for MCH Core to get it stable in Prime Blend & OCCT Large Data Set for 10hrs+.

Many a time have I gotten 24 hour Prime Blend stability and 12 hour small FFT only to un-rar an ISO of 6GB+ or some family photo's (re-prints for an anniversary etc) and get corrupt file messages. Large FFT would also fall flat on it's ass after 10 hours with the unstable settings.. MCH Ref sorts it for me almost always. That or a Vcore bump.

Hey Psycho101, I agree with you on the winrar part. It's a very good way to test the stability of your northbridge. And just like you, I fix the problem by adjusting my MCH reference (GTLref) voltage. Sadly not many ppl know about this... I'm glad to see someone with the same knowledge of Winrar & MCH reference as me finally.

[Psycho101]

Hey Psycho101, I agree with you on the winrar part. It's a very good way to test the stability of your northbridge. And just like you, I fix the problem by adjusting my MCH reference (GTLref) voltage. Sadly not many ppl know about this... I'm glad to see someone with the same knowledge of Winrar & MCH reference as me finally.

It's the first test I do before going for a Prime Large FFT starter. Waiting 4-5 mins for a big .rar file to decompress sure beats 4-8 hours of waiting for prime to fall over.

As far as edge929's MCH core goes, I reserve judgement until I see those BIOS settings. My initial reaction was "Woah! DUDE! To much MCH veee cooooore!" but when someone who obviously isn't new to overclocking states they need that much, I need more info before I make a judgement. If edge929 is willing to experiment a little it is likely possible to reduce it. However it may be that he got a P45 chip that was close to being put in the reject bin, and it really does need that many volts to prop it up.