OC Assistance: E5200 & P35-DS3L

[theonedub]

I am trying to get this cpu to 3.0-3.6ghz stable for WCG. I read a lot of FAQs, stickies, and threads before attempting, but evidently I missed something.

@stock 200x12.5- Vcore BIOS- 1.2375, CPUID Idle 1.20, OCCT Small Data 1.18

@333x9- Vcore BIOS- 1.325, CPUID Idle 1.28, OCCT 1.26

At 333x9 it ran OCCT Small Data High Priority for 2 hours. No errors and a max temp of 56C with high ambient temps. Only other voltage increase was +.1 DDR2 w/ Auto Timings 5-5-5-15 I believe. It ran WCG for 8 hours with no errors. I figure it was stable enough for its purpose. I've read a lot of other WCG members running OCCT for a shorter amount of time with no issue, which is why I chose 2 hours for stability.

Today I was using the computer with WCG running and the computer froze twice and BSOD'd once. Max temp 55C. OCCT freezes the computer within 5 minutes. I upped the Vcore .0125, +.2 DDR2, and +.1 MCH with the same results. I think the voltage is already too high for a paltry 3ghz, no?

I can get it to post 3.33 w/ BIOS Vcore 1.3937 which has me thinking 3.6 is probably not going to happen. Right now I would be satisfied with 3.0 stable and elated for 3.33 stable w/o excessive volts.

The rest of the components are listed in my system specs (WCG0). BIOS is the latest F9 version. Any help is appreciated. Let me know if there is any other pertinent info I forgot to add. I am not an avid OCer so bear with me please, thanks!

[boomstik360]

You got a bad cpu man most e5200's hit 3ghz-3.33ghz stock vcore. Try lowering the ram divider

[theonedub]

Attempted 250x12 @ stock Vcore, +.1 DDR2 and MCH, 2 SPD Divider posted but no Windows. Bumped Vcore up to 1.3265, still no Windows. Did the same thing with 333 FSB with no luck either.

I also tried out 220x12.5, stock Vcore, +.1 DDR2 and MCH and it got to Windows and failed OCCT 4 seconds into the load test.

This is very disappointing. I can't imagine how hard it will be to sell an E5200 that cant clock. The member I bought it from thought it would do upper 3s, but in the end it is all luck of the draw.

Thanks for the tip, I am pretty much ready to call it. Anyone have any other tips before I try to get rid of this CPU?

Originally Posted by boomstik360

You got a bad cpu man most e5200's hit 3ghz-3.33ghz stock vcore. Try lowering the ram divider

[p_o_s_pc]

any possible way you can RMA it? Also you try upping the NB voltage and PLL maybe even the term.volt alittle

[a_ump]

how could he rma it? just cause it doesn't oc won't be a reason to get it through intel's doors. I guess he could try to explain some issue he has with it to the store he bought it from.

[Mussels]

use OCCT's linpack test, its more stressful than the regular test, so it will show up errors faster. 1-2 hours of linpack will show up any problems that may have been missed in other tests.

I'd be watching your ram, often people miss simple things like ram overheating, or NB overheating.

[theonedub]

How high is safe for the NB voltage? From what I have read, the P35-DS3L runs hot as is. Whats safe for the term volt? Just to verify that is the FSB Volt setting in my BIOS, right? I have read something about PLL before, but I think with my FSB being under 333 I shouldn't touch it.

I don't think RMA is an option. I bought it from a member here and shipping it back and forth to complete RMA in addition to getting the RMA approved is probably not worth the effort.

Sorry for all the questions. Thanks!

[Mussels]

for a mere 333 FSB, you should be able to use stock voltage on just about everything.

[p_o_s_pc]

Originally Posted by a_ump

how could he rma it? just cause it doesn't oc won't be a reason to get it through intel's doors. I guess he could try to explain some issue he has with it to the store he bought it from.

i have heard stories of people RMAing there chip because it didn't overclock good. Some even told intel that others made up lies

[Mussels]

333x9 = 3Ghz

273x11 = 3003Mhz (aka, 3GHz)

Why not try the far lower FSB, same CPU clocks, lower NB stress, possibly lower voltages needed?

[theonedub]

It doesn't pass OCCT small stress test with 220x12.5 stock Vcore, +.1DDR2 and MCH.

[Mussels]

try the stock FSB's (200, 266, 333) with lower multis.

If you cant pass 266x9 for example, you're missing something - ram set wrong (voltage/timings/MHz) or somethings weirdly incompatible (need a BIOS update, etc), or PSU is too weak

[theonedub]

266x9, 266x10, booted into windows all stock settings, did not run any stability tests.

266x11 no go @ stock volts, but with a slight bump +.0125 it can get to Windows to fail Linpack in 10 seconds.

It's late so I am going to quit on this for tonight.

Tomorrow I am going to attempt it again and keep a clear log of exactly what voltages were along with the multiplier settings. That way I can be very clear as to what I tried and the results (my piece of scratch paper is not cutting it).

Thanks for the tips, hopefully you will be here tomorrow as well.

edit-
Just remembered, I posted it @ 3.6 and it froze in the BIOS earlier today. Had to reset the computer 2x before it would enter the BIOS again. Should I reflash the BIOS to make sure its fine, or does that happen sometimes and I shouldn't worry about it?

[Mussels]

dont reflash the bios, thats not needed.


What you need to do is test CPU clocks, and memory clocks seperately - so you know CPU needs 1.30v for 3.2GHz, ram needs 1.95v for 960Mhz, NB needs 1.4v for 450 FSB, that kinda thing


You keep track, so that you know what you need to do when you want a certain combination.
my CPU needs 1.18v for 3.8Ghz (i know, awesome chip), so i can do 400x9.5 or 450x8.5 - i know that both will need the same CPU voltage, but that the mobo will need more NB voltage (since i tested at 450x6 until it was stable) and that the ram needs 1.90v (since i tested 200 FSB at 800Mhz ram, and kept raising and testing in memtest, til it wasnt stable)

[theonedub]

333x9 1.3Vcore idle, 1.28Vcore OCCT Linpack, DDR2 @ 2.096v = pass OCCT Linpack 1hr 90% test.

Safe to leave my RAM V that high? I have G.Skill F2-6400CL5D-1GBNQ. The heatspreaders on them are very cool to the touch (my case has excellent air flow).

Going to let it run WCG for the next 8 hours again. When I wake up I will try browsing, play movies, and Crunching to see if I can duplicate the freezing and BSOD, hopefully it will be fine. Thanks for all your help, I am learning a lot about this process.

[Mussels]

DDR2 ram is safe to about 2.2v, as long as its cool. if you cant hold your finger on it, its too hot.

[phanbuey]

I would RMA both board and chip... "it doesn't work - BSOD at stock settings" (lie). That being said it sounds like a bad board, not a bad chip.

[a_ump]

it does pass linpack or OCCT's stress test at stock right? if it doesn't like mussels said something else in your config is messing up. Mussels has you covered here for advice but i just wanted to throw in my opinion

Also, i personally never use linpack, i've always used OCCT and my pc never crashes on me if i pass it, linpack is just a rediculous stress test imo. no useful or entertaining program will EVER come close to stressing you CPU like linpack does.

[theonedub]

Originally Posted by phanbuey

I would RMA both board and chip... "it doesn't work - BSOD at stock settings" (lie). That being said it sounds like a bad board, not a bad chip.

I don't think RMA is an option for a variety of reasons. What exactly makes it seem like a bad board? I do not have another 775 board (that supports 45nm) around to test the chip in unfortunately.

Originally Posted by a_ump

it does pass linpack or OCCT's stress test at stock right? if it doesn't like mussels said something else in your config is messing up. Mussels has you covered here for advice but i just wanted to throw in my opinion

Also, i personally never use linpack, i've always used OCCT and my pc never crashes on me if i pass it, linpack is just a rediculous stress test imo. no useful or entertaining program will EVER come close to stressing you CPU like linpack does.

It passes OCCT and OCCT's Linpack at stock, and it did pass OCCT Linpack for 1 hour max stress at 333x9 with the voltages mentioned in my last post. I don't understand how the original 333x9 settings (in my first post) pass OCCT Small Data Sets for 2 hours then BSODs with a combination of Windows Media Player, FireFox, and BOINC.

It did crunch 6 hours straight no issue so far. I had a movie open in WMP, opened 25 tabs in FF, and started seeking around to random points in the movie with BOINC running and it did not BSOD. Im going to try it again when it gets hotter.

[JrRacinFan]

Just out of curiosity;

Is your chip an R0 or an M0? Can you post a CPUz screenshot for me.

[theonedub]

Its an M0, here is a CPUz capture (currently running WCG in background @ 100% load). Let me know if there are other tabs you would like to see.

[phanbuey]

What makes it seem like a bad board is the probability of not being able to hit 3.0Ghz with even high volts. Even a dud e5200 is likely to get past 3Ghz;

Thing is, when you start reaching the limits of the chip/board you really can't tell what is failing the tests, because they have nearly identical symptoms. That is when you have to step back and think... Hmm the average OC for this chip is 3.6+Ghz with ease, the board, however, is a bit more temperamental, replete with massive v-droop and FSB holes (i used to have that board).

My problem was always NB heat on that board, as the puny heatsink they use on DS3L will get melting hot (touch it - its fun). Oh btw i found the F7 bios to OC the best (got my e4300 to 3.4Ghz), but am unsure if it works with the 45nm's.

Edit: One good way to find out if it is the board is to run a 400Mhz FSB, 800Mhz ram and a 7 multi - if it posts and is stable in windows, then it is the chip that is holding you back. If it doesn't post even with a low multi then it is the board.

Don't forget to touch the nb heatsink . Hope that helps, good luck on your oc

[JrRacinFan]

Thanks onedub

Try out 10.5x333 1.4v in bios; Im guessing 1.4v in bios would give about 1.38v in Windows. Trust me on this that M0 revisions of these chips dont mind high volts, you have to clock them like they are a 65nm chip.

Also, what do you have your CPU GTL's set at?

[phanbuey]

Originally Posted by JrRacinFan

Thanks onedub

Try out 10.5x333 1.4v in bios; Im guessing 1.4v in bios would give about 1.38v in Windows. Trust me on this that M0 revisions of these chips dont mind high volts, you have to clock them like they are a 65nm chip.

Also, what do you have your CPU GTL's set at?

Last I remember, that board does not have GTL ref options.

[yaji]

I have got the same CPU and board. I cant go above 3Ghz too. I ended with 266x11.5@1.325v. I think that DS3L dont "like" e5200...