Q6600 B3 on a P5B Dlx: delayed vdroop

[jonny_ftm]

Hi
As it is my first post, I'll first put my system with some images. It is my first WC setup, mounted with a Swiftech H2O-Apex Ultra 220 kit and a PA120.3


Side panel modded with a 120mm FAN blowing on the mb



The MCR-220 in the rear with the PA120.3 mounted on the modded lateral panel. Because of screws incompatibility, I had to put the center fan on the PA120.3 in a blow configuration between the radiator and the panel





The mcres in place
It was impossible to mount it as Swiftech claims in the manual because of the thik grommets



The mb mounted with HR-05 IFX and the 2 HR-09U type 2.
Note that the upside HR-09U is some displaced because of mb components interferring with it in length. However, the 8 mosfets are in tight contact.
I opted for a HR-05 IFX and 2x HR-09U type 2 (and yes, the P5B dlx could fit two of these mosfet coolers). I also removed the SB HS and replaced the thermal compound with some MX-1. I mounted the HR-05 IFX with a 80mm fan undervolted to 5v




And a full view with some work on the cabling. I was too impatient to make an effort on the SATA cables and the cables in the bottom rear part of the case. In any case, they are not on any air path

[jonny_ftm]

Well, now my problem

I have the system in my signature and with the photos above (Q6600 B3 @3.5Ghz, 1.50 vcore bios, 1.51/1.52 load)
All is working fine, and my temperatures are 60°C-63°C on the cores when priming
The P5B deluxe is pencil modded for the vdroop, and on load the vcore increases now (many passes with the pencil).

The problem is that after exactly 2h of priming or OCCT in RAM mode, the vcore drops slowly till it goes below 1.5v and system fails.

As the NB and mosfet heatsinks were burning to the touch, I mounted the mosfets with 2 HR-09U type 2 and the NB with a HR-05 IFX and now they are cool to the touch, but it didn't help.

It's strange that exactly after about 2h, the vcore begins to drop like if it was not modded. Increasing the vcore won't help either, it will progressively drop and the system fails at 1.495v

Any one having this board with a Q6600 on a high vcore and having same results?
Is it possible that the graphite of the mod drops its resistance with the heat that's why the vcore drops again?

I put a 120mm fan on the mb between the RAM slots and CPU waterblock that blows air directly on the graphite vcore mod: still vdroops after 2h

I tried to open the window so that I have cool ambian air, it vdrooped in 2h while the cores are about 59°C

With the 120mm fan on the graphite mod area, the system temperature is around 43°C

Please any one could help me with that vdroop delayed issue? Is it known with the P5B dlx and quads?

By the way, I can fold@home for over 14h and 3DMark06 in loops, cinebench, PCMark, memtest and SuperPI 32Mb on four cores with a stable system. The only hanging with an empty blue screen is when the vcore drops after 2h of priming

Many thanks

[higgins]

Nice rig man. Where did you buy the HR-09U type 2? Checked at performance-pcs and sidewinder but doesn't say if they're Type2. And whats the diff between type1 and 2 btw?

And sorry, got no clue why your vcore drops man. :-/ PSU maybe?

[Timbosan]

I experienced something similar, but i wasnt running a quad, or the deluxe board. I was running an e6300 @ 3.43 GHz on a P5B vanilla. I also did the vdroop mod and also my vcore increased under load. At first everything was ok, but after one month, the vcore began "dropping" about 0.2v from the stable load voltage, whereupon the chip would fail. I tried (stupidly) increasing the vcore, but this only made the problem worse. I removed the vdroop mod, and even though i have to put up with horrible (0.07v) droop, i have stable voltage under load now. Because you are running a quad, it would put more strain in the board than my chip so you may have different or worse symptoms.
What you have to realise is by doing the vdroop mod you are putting a lot of strain on the voltage filtering capacitors on the mobo. They are not able to cope with the load and effectively buffer the vcore, and the vcore becomes unstable.
My advice - remove the vdroop mod and see if the problem is corrected. I know vdroop sucks, but at least it may be stable.

EDIT: see this thread post #22:

http://www.xtremesystems.org/forums/...d.php?t=160592

[jonny_ftm]

Nice rig man. Where did you buy the HR-09U type 2? Checked at performance-pcs and sidewinder but doesn't say if they're Type2. And whats the diff between type1 and 2 btw?

In Europe: Pc-cooling.de (where I got them)
In north america: directcanada.com

There's a size difference I think between type 1 and 2, if I see the thermaltake specs, but didn't find anything about it on the net. type 2 are the newer models and also I think compatible with more boards, so future proof

My advice - remove the vdroop mod and see if the problem is corrected. I know vdroop sucks, but at least it may be stable.
EDIT: see this thread post #22:
http://www.xtremesystems.org/forums/...d.php?t=160592

Many thanks for your answer. I understand better now. I'll try to find a compromise: redo my vcore mod so that I get a smaller vdroop, thus maybe having less stress on my PWM. I'll end up with a higher idle vcore but yeah, maybe more stable

I'll give it a try and let you know

Also, I'm finishing a test with the 120mm fan spotted on the HR-09U and they are rather cold now. I'll see if it gets better

[jonny_ftm]

Ok
It was a cooling issue. With a fan spotted on the HR09-U, the vcore is rock stable and system temperature reported by speedfan dropped by 10°C (32°C now)

Many thanks anyway

[jonny_ftm]

Ok,
Got my self a G0 and I can have a stable 3.6GHz at 1.43v real and 20°C lower than my B3 at 3.5GHz. Story ended, my time spared, and even money spared if I would get so crazy to buy a DOD to cool that su**ing CPU

[Timbosan]

Glad you sorted it out mate. So it was 'heat soak' ? after time the temp of the fets climbed to the critical point & the vcore began dropping? hmmmmm i may have to put a sink on mine - theyre naked. I have never touched them while pc has been on so dont know how hot they get. I did notice the pc was much less stable on hotter days, and the higher the vcore, the less stable it was.

awesome that you updated to the G0, 3.6 @ 1.43 must be nice. Whats the load temp on it?

2 qstns - do you need the vFSB set at 1.45v, and do you need the vNB set to 1.55v for 400 FSB? i would be surprised if the vNB needed to be over 1.30v for that. I found on my P5B, i needed exactly 1.39v vNB for 490 FSB with static read control set to "faster" (Performance Level = 10) running 4-4-4-12 primary ram timings. I only had to crank it up to 1.5v+ for 4:5 divider max FSB testing.

[jonny_ftm]

It's a known problem on the P5B with quad core. It is FSB limited with quad cores, I saw some reviews where they needed 1.65v for the NB at my FSB. At 389MHz, vNB 1.35 will reboot on windows start while 1.45v is ok. At 400MHz, same vcore, I need vNB 1.55 as it will reboot at 1.45v
With C2D, there's no problem

[Timbosan]

right i wonder why that is? good to hear its working now though

[jonny_ftm]

Ok, followed your advice. I removed the vdroop mod, but as expected, I should have to live with a 0.06v droop on load and 0.04 at idle
My Q6600 G0 finally needs 1.44v to remain stable in all circumstances. 1.432 failed

I reapplied the vdroop but lighter and testing after each appliaction of the pen. I now let it so that on 1.45 bios, I have idle 1.440v and on load it increases slightly to 1.448v. I see no more any variation in the vcore on load or idle

About the vFSB/vNB/vICH, I maxed them and put the memory timing to 5-5-5-18 so that they are out of the equation if Prime95 fails so that I can get the lowest vcore needed for my OC

Now, I put the memory timings to 4-4-4-12 and hey, my so bad XMS2 C5modules did it, I'm still stable

Next step: lowering the vFSB to 1.40 and vNB to 1.45 to see if still stable (below: instant reboot on windows load)

Again thank you for your help, I'll drop another post when finished tweaking the voltages. Hope this thread will help some of the P5B owners

[jonny_ftm]

A quick note.
Even 400 MHz was unstable on blend tests or OCCT in RAM mode. In small FFT it rocks.
400-405 MHz same thing (NB strap 1333). After 405 MHz, no way to boot windows. 390 MHz is the last Mhz stable on ram mode

So, I have to live with a 3.51GHz OC at only 1.38v (and from my previous tests, I should be stable even below 1.38v real on a modded mb)

It is same limit (exactly 390 MHz) as my previous Q6600 B3. Timings of 6-6-6-18 or every other voltage combination for NB/SB/ICH and DDR won't change it.

I just hit my P5B delx FSB wall with quad core chips


Sad

[dnottis]

A quick note.
Even 400 MHz was unstable on blend tests or OCCT in RAM mode. In small FFT it rocks.
400-405 MHz same thing (NB strap 1333). After 405 MHz, no way to boot windows. 390 MHz is the last Mhz stable on ram mode

So, I have to live with a 3.51GHz OC at only 1.38v (and from my previous tests, I should be stable even below 1.38v real on a modded mb)

It is same limit (exactly 390 MHz) as my previous Q6600 B3. Timings of 6-6-6-18 or every other voltage combination for NB/SB/ICH and DDR won't change it.

I just hit my P5B delx FSB wall with quad core chips


Sad

P965 = old. Try a p35 or x38.

[jonny_ftm]

I see your setup as a Q6600 on the P35 chip

Can you boot it above 400 MHz at idle? (no need to stress it as B3 would be too hot)

I didn't find a suitable P35 X38 board yet. P5K premium having the SATA2 issue and no RAID (bugs). The Maximus having some bugs with prime95 and quadcores at above 3.8GHz and 1.5v...

Any card to advice me please? I see in ASUS Maximus thread rare people getting above 3.6GHz on a quad qith that card too.

[hstuehmeyer2000]

I see your setup as a Q6600 on the P35 chip

Can you boot it above 400 MHz at idle? (no need to stress it as B3 would be too hot)

I didn't find a suitable P35 X38 board yet. P5K premium having the SATA2 issue and no RAID (bugs). The Maximus having some bugs with prime95 and quadcores at above 3.8GHz and 1.5v...

Any card to advice me please? I see in ASUS Maximus thread rare people getting above 3.6GHz on a quad qith that card too.

why only asus?

[Comminadian]

Any card to advice me please?

Abit IP-35 Pro? Maybe an IX38 Quad Gt? Pro is here and with good results, IX38 is on the way and if it is anything like Abit's P35's it will be a winner.

[jonny_ftm]

Why ASUS? I don't know in fact. Most reviews put the ASUS on the top of the line. But, all my ASUS cards had bugs: bios issues, RAID, Sound, SATA2 incompatibility, vcore PLL... (these are all broken advertised functions on the P5K premium and some of them on the Maximus)

Many of them are still unresolved. I really don't understand how such cards can go retail sometimes when most of the advertised functions are just broken.

The problem, is that Gigabyte is lazy on bios updates and my old MSI and Gigabyte cards had bugs and bluescreens with XP 2-3 years ago, so I switched to ASUS

Also, the P35-DQ6 and X38 from gigabyte got worst reviews than the P5K premium and Maximus... So, it is really hard to decide which crappy card is best (as it seems with their bioses they are all crappy actually)

I saw the reviews on the Gigabyte DQ-6 series, and I'm not convinced. ASUS is crap, but at leaste after a year they fix their bioses

Abit, I may take a look at it, I really never watched at reviews with the Abit P35 and a quad core. You know any of them?

Many thanks again for your help

[jonny_ftm]

Ok, looked at the Abit IP35 Pro: it has some drwabacks sadely

No way to do a vdroop mod using a pencil: it badly vdroops for a quadcore, so overclocking will be really limited above 1.5v

It won't fit my HR-09U and HR-05 IFX

I think I'll be soon very limited with my Q6600 G0 if I needed 1.5v-1.6v to reach >3.8GHz

[Timbosan]

Yeah board brand selection is really hard. ASUS is medium quality, medium price, with regularly updated support/bios etc, so you may run into problems but you may get a solution. Harware failure happens but rarely in normal and air/water OC situations. DFI make high quality but their support is lacking although improving. Gigabyte are pretty low quality it seems ive had nothing but crappy boards from them right back to the P4 days.
Ive read good things about the Asus Maximus, DFI UT-LP and the bloodiron, also P5K-Plus and Premium.

[jonny_ftm]

Yeah board brand selection is really hard. ASUS is medium quality, medium price, with regularly updated support/bios etc, so you may run into problems but you may get a solution. Harware failure happens but rarely in normal and air/water OC situations. DFI make high quality but their support is lacking although improving. Gigabyte are pretty low quality it seems ive had nothing but crappy boards from them right back to the P4 days.
Ive read good things about the Asus Maximus, DFI UT-LP and the bloodiron, also P5K-Plus and Premium.

Yep, but the maximus seems still to be some buggy in high quad OC, P5K premium is great for OC, but has a serious SATA2 and RAID problems

I think the maximus will be the way to go despite its prise. I like it's built in NB WB. Just waiting to see if the bios gets mature or if ASUS will release a new revision because of hardware faults

[Timbosan]

i also like the maximus, however i'd try to get the non watercooled NB as the WC one has very small (~8mm) diameter connectors, which would be a massive restriction on your WC loop. Also air should be fine to cool the NB (mount a 40 or 50mm fan on it), i saw temps of ~50 - 60*C with the WC block not connected to a loop (passive).

[jonny_ftm]

I see, yes, 8mm is really too small for a LPDC circuit
The problem is that in any new MB I'd prefere using thermalright solutions because they are much more effective on air. If I don't use the NB WB of the maximus, I'll have to buy a new HR-05 IFX to cool both SB and NB or a WB for the NB which is rather expensive

Anyway, for now, I'll stick with my current settings unless the maximus issues are fixed. There are so many that I feel a new hardware revision will come soon

[jonny_ftm]

Here we go, my new P5B deluxe came back from RMA. Got same 1.03G revision :-(
But hey, this one rocks and I finally get my 400FSB stable on my quadcore. Anything above 400MHz needs more volts and impacts on RAM performance because of the NB strap change at 401MHz
I'm happy with this one