Asus P5B-D & 3x E8400 @ 590fsb dual 32M superpi stable. One off freak result?

[davemelb]

Hi everyone,

I am David from www.overclockers.com.au. I am very much a novice overclocker with an Asus P5B Deluxe Rev 1.04G + pencil vmod. Now I seek your help and comments on the situation below:

Recently THREE different retail E8400's was put in my motherboard with, what I now realise astonishing results. Astonishing because they did dual 32M superpi stable @590fsb on air in the middle of the warm Australian summer.

The complicated part:
Two of these CPUs was sent to dino22 to test, however, dino22 couldn't reproduce anywhere near the FSB I was getting, even with the same motherboard as I have, using the same BIOS and BIOS settings. He also tried some of the best and latest motherboards in his posession, but no go.

The third CPU was sent to another person, which also failed to replicated my results. He informed me he uses an Asus P5K deluxe.

So here I am, scratching me head and confused. But I post the screen dumps and validations of those tests below for your viewing:


CPU1. Q744A.

590FSB dual superpi 32M. Begin of test.
http://img301.imageshack.us/my.php?i...sbstartaq4.jpg

590FSB dual superpi 32M. Middle of test.
http://img504.imageshack.us/my.php?i...0fsbmidep7.jpg

590FSB dual superpi 32M. End of test.
http://img168.imageshack.us/my.php?i...0fsbendop5.jpg

CPUz Validation:
http://valid.x86-secret.com/show_oc.php?id=296598

600FSB dual superpi 32M. Begin of test.
http://img150.imageshack.us/my.php?i...sbstartrn4.jpg

600FSB dual superpi 32M. Middle of of test. Shortly after this, Windows frozed at the 3/4 stage, test uncompleted.
http://img110.imageshack.us/my.php?i...0fsbmidgw2.jpg

CPUz Validation:
http://valid.x86-secret.com/show_oc.php?id=296599


CPU2. Q748A.

594FSB dual superpi 32M. Start of of test.
http://img340.imageshack.us/my.php?i...sbstartff1.jpg

594FSB dual superpi 32M. Middle of of test.
http://img237.imageshack.us/my.php?i...4fsbmidmf1.jpg

594FSB dual superpi 32M. End of of test.
http://img175.imageshack.us/my.php?i...4fsbenddd7.jpg

CPUz validation:
http://valid.x86-secret.com/show_oc.php?id=297368

Did no try 32M SuperPi test at 600FSB. But here's a validation at 600FSB on the second CPU.
http://valid.x86-secret.com/show_oc.php?id=298246

This is the result of the third CPU:
Single 32m superpi, 590fsb.
http://img443.imageshack.us/my.php?image=590pihz4.jpg

Dual 32M superpi, 590fsb. Half-way stage.
http://img413.imageshack.us/my.php?i...ay590pihe1.jpg

Dual 32M superpi, 590fsb. Completed stage.
http://img413.imageshack.us/my.php?i...te590piom8.jpg

Interesting to note: The three E8400's seem to max out at around 590 - 600 fsb.

My system:
P5B-D Wifi Rev 1.04G,BIOS 1226. Pencil vdroop mod.
2GB Crucial Ballistix PC8500 RAM
Ultra 120 extreme cooling + 120mm fan.

BIOS settings:

BIOS settings:
FSB = 400 - then SetFSB in Windows to around 590 - 600.
PCI-e = 100
33.33MHz
Sprectrum = Disabled
vcore = a high value to withstand the test, usually round 1.5v
vdimm = 2.25v
FSB termination = 1.45v
NB = 1.65v
SB = 1.8v
ICH = 1.2v

Diabled everything under "CPU Configuration", except multiplier.

RAM timings = loose, 5,5,5,15, 5, 35/42, 10,10,10,0
Static Read = Auto

I am happy to upload the CPUz validation files saved on the HDD for you people to analyse if they wish.

Can anyone on here with an Asus P5B-D please test your board with a dualcore 45mm to see if you can replicate my results? If you can, would you please also test it on a P35 or x38 or x48?

[dinos22]

it will be good to see what we can figure out once your board arrives over at my place for testing. Extremely weird this occurrence.

Just to add to your post it's summer time in Australia so ambients were most likely 25C+ easily

in fact during our talks when i was making arrangements to buy these CPUs there was a heatwave in melbourne with average 40C temps so no sure what ambients Dave had

I've personally tested the two CPU with
Gigabyte X38-DQ6 F2ES, F3ES and F3 bioses (this motherboard is very reliable for me and i know i get really good FSB out of CPUs with it)
Gigabyte X38T-DQ6 F2 and latest bios
Asus P5B DLX as per Dave's instructions

his CPU that did 18 loops of dual 32M SuperPi at 600MHz was doing around 530Mhz single 32M SuperPi
Second CPU was better with single 32M runs at 545MHz FSB but that's still way short of the monster 595/600MHz runs (dual 32M too lol)

[zizo]

Was your board loading with 1X Pci express? I can clock my E6400 higher with 1X, but that could be the NB limitation not the cpu.

[dinos22]

have you tested how much higher your gain was
P5B Has a bug if you apply a lot of volts to NB the PCI Express goes down to 1x

but were those gains from extra NB or the fact it dropped to 1x

[zizo]

I have one of the first board revisions, so it doesn't clock very high. If I load at 400FSB, I can go as high as 460 with 1.55v NB and 16X, or 490 1.65v NB 1X pci express. Even with 401FSB I can't go higher than 530FSB with 1.65v.

I think you can test that with a pci vga.

[-n7-]

Wow.

P5B-D won't die!

[davemelb]

Was your board loading with 1X Pci express? I can clock my E6400 higher with 1X, but that could be the NB limitation not the cpu.

I would say yes, it was 1x PCI express. I maxed NB 1.65v and SB 1.8v.

I saw 1x in CPUz a few times, but I didn't know what causes it or howto fix it.

Maybe it doesn't really matter but, I think I installed either Intel chipset driver 8.3.1.1009 or 8.4.0.1016 with Windows XP SP3 build 3264 beta.
http://downloadcenter.intel.com/filt...4&submit=Go%21

http://majorgeeks.com/Intel_Chipset_...ity_d1667.html

[Kunaak]

too low of a NB voltage makes it go down to 1x.

raise it to 1.45 or higher to maintain 16x.

[davemelb]

hmm, dino said too much voltage causes 1x pci-e, and now you say too little also causes it. I'm abit more confused now

dino, I think the ambient temp was around high 20's celcius (or about low-mid 80's F) at the time.

[jonny_ftm]

Like Processors, NB chips are also a sort of CPU. They are still in the 90nm process though, while CPU are in 65-45nm process. The NB is one of the most limiting factors in OC on new processors.

I got a first P5B Deluxe 1.03G, impossible to get it at 400 FSB on my quadcore despite any NB voltage. Also, it had the 1xPCI-E bug when you load more than 1.45v in the NB.

I RMAed recently that board and got a new one, sadely same 1.03G revision (that is revision C1 of the NB instead of C2). Guess what: that new P5B deluxe clocked my Q6600 stable at 400 FSB (strap 266), with DDR2 at 1000MHz and 5-4-4-15-4 timings, and only 1.45v on the NB. This is very high performance, as even the P35 and X38 chips aren't able to get that FSB, DDR2 overclock and timings with a quadcore on such a low voltage for the NB and especially on strap 266.

Conclusion:

Like with processors, it is a pick and luck story. Some chips are best overclockers than others, be it a CPU or a NB (that is motherboard). If you read well on the P5B deluxe main thread, you'll have noticed that some motherboards are Roxors, while others s**k.

David got suerley a roxor motherboard, and by far, not all chips are so good performers

[jonny_ftm]

hmm, dino said too much voltage causes 1x pci-e, and now you say too little also causes it. I'm abit more confused now

dino, I think the ambient temp was around high 20's celcius (or about low-mid 80's F) at the time.

And yes, for the P5B deluxe, too much voltage on the NB causes the bug. Usually 1.55 and 1.56v settings. 1.45v is not bugged. It is maybe a bug in intel NB, as they released the C2 revision and new C1 ones without the bug I think. Or a bug in ASUS PCB layout corrected on new batches (as My new C1 revision is not buggy while my previous same C1 had that bug)

[dinos22]

as far as i know and in my experience with the boards too much NB causes the 1x bug in most cases

jonny we are talking about some monster differences in FSB though
i can and have seen how a better board clocks RAM or FSB better but 50-60MHz is a bit crazy :s unless some sort of freak board

[Zeus]

Dinos, did you test the cpu's on the same board you got your golden E6700 to 600+ FSB?

If you did, it can't be your board, right?

[dinos22]

same board

i also tested that E6750 on SS 600+MHz FSB too

[vanovich]

well ,its sometimes wierd, i tried a 6850 in a giga p35 ds4 and giga x38 .
the same 6850 did 540fsb on p35 and only 500 on x38 .though i could op the multipler on the x38 ending getting same oc .
dinos your giga x38 is also just one of the few ,that ive seen with such high fsb.wonder how the new 8### serie cpues runs best combo ? p35 or x38 ?

[metro.cl]

Interesting i guess i'll have to try the 1x speed to see if my FSB increases

[dinos22]

well ,its sometimes wierd, i tried a 6850 in a giga p35 ds4 and giga x38 .
the same 6850 did 540fsb on p35 and only 500 on x38 .though i could op the multipler on the x38 ending getting same oc .
dinos your giga x38 is also just one of the few ,that ive seen with such high fsb.wonder how the new 8### serie cpues runs best combo ? p35 or x38 ?

it's just bioses that's all

if you have them still try F3 bios for DDr2 board

[davemelb]

And yes, for the P5B deluxe, too much voltage on the NB causes the bug. Usually 1.55 and 1.56v settings. 1.45v is not bugged. It is maybe a bug in intel NB, as they released the C2 revision and new C1 ones without the bug I think. Or a bug in ASUS PCB layout corrected on new batches (as My new C1 revision is not buggy while my previous same C1 had that bug)

Now that you mention it, when the NB voltage was too high, I always saw 1x in CPUz. I have the C2 version. However, when it's 1x I tend to get better CPU overclock.

Interesting i guess i'll have to try the 1x speed to see if my FSB increases

Give it a try.

Also like to update that I have sent the motherboard away for testing. I am not sure what will happen, but we'll find out abit more about this soon

[mrcape]

wonder how the new 8### serie cpues runs best combo ? p35 or x38 ?

This is the big question I'm wondering about too. I have an e8400 arriving tomorrow. I have the DFI P35 LP UT T2R (long name to remember and type!) set up for the chip, but I also have brand new unopened P5E and P5K-Premium sitting on the table.

So again the question, what would you guess to be the best fit for the e8400? x38 0r P35? I really hope it's the DFI coming through with the fine tuning we need for this baby.

If there's no clear answer I guess I can always play around with them all and see.

[luihed]

Dave, maybe you should go to Dinos place and teach the dude how to OC lol.... Dont know what you said when he 1st PMed you about his results but it should go something like " well dude itll do 580fsb but you gotta know what youre doing lol" .... (j/k) ... sorry bro, couldnt resist

On a serious note, could it be a strap thing? also since dave is just starting out as an overclocker, maybe his p5B is in better shape than Dinos who beats the hell out of his for sure ( like most of us do )....

Just took a look at the screenshots and Im pretty sure he wasnt running 590x7... not saying he cheated but theres something wrong with his mobo... that time is too slow, even for somebody that doesnt know how to bench... dont know whats wrong with daves setup but I dont think you can run that slow even if you tried.... Never ahad a wolfie but from my experience with the QX9650, My guess on a fully tweked run at 590x7 will yield a PP in the low 25xxx, maybe even 24xxx( ~10m ) ... Sad to say but it looks like some weird aussie bug.... $.02

Dinos, when you get his board and you still have your 600+ c2d, you can have alot of fun if my theory is correct....

[adamsleath]

what would you guess to be the best fit for the e8400? x38 0r P35?

or 965?

[davemelb]

luihed;2719369 Dave, maybe you should go to Dinos place and teach the dude how to OC lol.... Dont know what you said when he 1st PMed you about his results but it should go something like " well dude itll do 580fsb but you gotta know what youre doing lol" .... (j/k) ... sorry bro, couldnt resist


Well believe it or not, dino sent one of his EE8500's down. We chucked it in my P5B-D, the first thing I did was raise the FSB to 580 - I was shocked thinking to myself 'the hell, dino would never let go of a chip this good. It did 1M superpi easy (I didn't capture a screen dump). Then my curiosity went got the better of me, I SetFSB it to 596 and it was easy too. I didn't bench superpi at this fsb, but luckily I saved a validation file, below:
http://valid.x86-secret.com/show_oc.php?id=298776

Immediately, I raised SetFSB to well over 600 - then suddenly Windows crashed, every frozed. Forced reboot. Unfortunately I haven't been able to SetFSB anything over 500 ever since I hope the mobo is not damaged.

Just took a look at the screenshots and Im pretty sure he wasnt running 590x7... not saying he cheated but theres something wrong with his mobo... that time is too slow, even for somebody that doesnt know how to bench... dont know whats wrong with daves setup but I dont think you can run that slow even if you tried.... Sad to say but it looks like some weird aussie bug.... $.02

O.K, educate me further, what sort of time should it have been? I am not familiar with benchmarking. The RAM timings was very loose in BIOS.

Maybe the mobo loosened memory bandwidth (and possibly it slowed the CPU clock) at those FSBs?
typo:

[luihed]

Performance product ( PP ) is simply mhz x time ( in sec ) ... so your PP is 4135 x 843 = 3485805.... Im not sure what an average begginer gets but to give you an idea, I did 10m59s running 4180mhz ( PP=2758800) on my QX9650 at 440fsb.... In 45nm its all about the fsb, take a look here and youll see my time is at par with guys running ddr3 and Im not in the same league as "before" in terms of spi skill, but since my fsb is higher I made up the diff using only ddr2.... see the guy with 501fsb ( #1), look how far ahead he is, now imagine what kind of damage one can do with 590fsb.....

Edit: Just noticed somebody ran 578fsb on a wolfie and his PP is 2632500, using his PP, your run will be 10m36s... thats a long way from 14m03s and since youre running 1:1 already not really much you can do to really slow down your time... unless you were running orthos while benching lol....

Edit again: ran this last summer when I had no clue whatsoever in spi32M... PP=3368220... tight timings but at only 486fsb, no tweaks and all 4 cpuz and speedfan running while benching.... Im not trying to be pessimistic (sp?) but from what ive read here it really does look like a bug of some sort.... Unless you found a tweak, then you can kindly PM me about it and not broadcast it to the rest of the world lol....

[davemelb]

I appreciate your explanation on PP I have never heard of PP before.

No other programs was running in the background during those.

So it doesn't matter that two superpi was running at the same time? I would have thought this makes it slower than one superpi.

The benchmarks you showed me ran single superpi and very tweaked.

I wish there was a trick to get that high FSB, unfortunately I can't take credit for it. A NB bug?

Oh, I only got this motherboard too, about three months old as a run out model.

[luihed]

Dint noticed you had two running lol... I guess that will make a difference ... Not sure by how much but forget my theory for now, you could be into something good here....

Now that you pointed out you had 2 spi running, it only means one thing... Dinos really needs to learn how to OC ..Sorry agan Dinos, I just cant resist haha...