FSB Wall - Q6600 and P5K Dlx

[r1ch]

Hi Guys, need to rack your brains ahead of a little bench session I have planned.

I've a P5K Deluxe and probably 3 Q6600's to test with.

1. VID of 1.25v, and hipro method = "00"
Result - 4Ghz@1.46v, 4.25Ghz wall at voltage up to 1.65v.
2. VID of 1.2125v and hipro method = "23"
Result - 3.6Ghz@1.37v (P5K Vanilla board so huge vdrop and vdroop, Bios was 1.50 or something?)
3. VID of 1.20v and hipro method = "20"
Result - 3.6Ghz@1.35v

#2 is in my mates PC at the moment (hence P5K Vanilla and a crappy PSU) but I aim to get it back from him for the testing. All results were Quad Prime stable for 10 minutes, expect #3 which i think was 4 hours stable. All run fine, but are quite hot and running up to 60C loaded with my GTX in the loop.

My question is mainly about where I got to with #1. I couldn't get anywhere past 427 FSB (4.25Ghz) on the P5K Deluxe no matter what I tried.

Any suggestions? Am I looking at mods to get higher?

[blindfitter]

I would think you need to look at your cooling, your on water, which rad,

I could only get 4100mhz on my Q6600 GO under water and 4365mhz under my Single Stage cooler these were benchable stable. that was with P5K3 and a voltmodded P5B.
Do your benching outdoors or in the shed, it will help the cooling, Tis no good voltmodding a board if you can't keep it cool,

Good luck with your tests, looking forward to the results

[4Qman]

FSB is really decided by the CPU itself. However there are some things you can do to help to get some more :)

PLL voltage plays a part in the FSB wall of a CPU. Stick under 2v for benching with extreme cooling as this is the biggest killer to the CPU's life span. Some have killed there CPU in a few hrs with 2.2v. Try in increments however i believe the deluxe has a PLL limit of 1.8v :huh: Volt Mods?

Cooling itself will help with high FSB,. I normally expect a increase of around 30FSB from Air to cascade with increased voltages however this is again if the CPU will actual take the extra speed itself.

North bridge is a important part in FSB over 500. Start around 1.75v as this is touching the sweet spot for most 65nmn CPU's. I use 1.8v with my P5K-E ;)

Good Luck

[r1ch]

Sorry, forgot to include my system specs in my sig. They're there now.

PA120.3 with 3x Antec TriCool fans (from my P180) on high.

The 472 FSB was with the PC next to the window and Rad practically hanging outside just after Xmas with 1-2C ambients, load temp max I saw was 40C across all cores so temp wasn't the issue I don't think.

Was a pretty firm wall, 1.46v@4Ghz Quad Prime stable had me hoping for a lot more...Guess I might just be expecting a bit too much on water.

[Raja]

What happens when you try above 427 fsb on the 7x or 8x multipliers?

If you can't go over 427FSb on a lower multiplier, there are still a few variables (chipset (trd etc)/memory), but as this is a Quad core CPU the key would first be VTT and GTL reference voltages. For explanation's sake (something you may have tried already), is to use a VTT of around 1.54-1.6V and a CPU GTL of 63X (or 67X if it's available), set the NBGTL to 67X too.

Unfortunately ASUS's implementation of GTL is very limited and basic (something Kris and I have asked them to change).

If you can go above 427FSB on a lower multiplier the suggestions by the guys above sum up most of the problem.

regards
Raja

[r1ch]

472 FSB Raja, but thank you very much for the advice. I'll do some testing now and get back to you with the results. :)

If you don't mind, could you explain exactly what the GTL is and how the 63X and 67X affect the voltages or overclocking? I'd be very grateful :)

[maverik-sg1]

Try a x6 multi and getting max FSB - I read that PLL helps climb the ladder but shortens the life of the CPU - 2.4v for burst benching, 2.2v for a couple of hours and 2v for a full days benching - but for 24/7 1.8v is as high as you should go safely.

Also, 1.9v into the NB helps (vMCH).

You may need (if you dont already have) volt mods to get these voltages.

Can you link us to the hipro methods please.

[r1ch]

http://www.hipro-tech.com/images/hipro5/QX9650_RETAIL/QX9650_RETAIL_2.jpg

........the "secret" of the Stepping/wafer for those who know.....
Once again I was lucky.....

Thread: http://www.xtremesystems.org/forums/showthread.php?t=167336

Looks like I can boot up to 477 (randomly) or a more stable 475 (pretty much everytime) with the 6x multi, and whenever it boots, it Quad Prime's nicely.

This is with all the chipset BIOS voltages maxed out (cept CPU and Mem which are running well within limits), just so nothing gets in the way -

Bios 0404 - worth updating?
Memset - 5-5-5-5-15-5-42-12-14-11-3-8-3-5-2130T-2T
475x6=2850
FSB Strap - 333Mhz
PCIE - 100
DRAM Freq. - 950Mhz
DRAM Cmd. - 2T
DRAM Timings - 5-5-5-15-(All Auto)
Transaction Booster - Disabled
Relax Level - 0
Clock Overcharging Mode - Auto

CPU Spread Spectrum- Disabled
PCIE Spread Spectrum - Disabled

CPU V.- 1.4v
CPU V. Ref - 0.63v
CPU V. Damper - Enabled
CPU PLL V. - 1.7v (1.8v Max)
DRAM V. - 2.2v
FSB Term. V. - 1.4v (1.5v Max)
NB V. - 1.55v (1.7v Max)
NB V. Ref. - 0.67x
SB V. - 1.05v (1.20v Max)

Just dropped the CPU PLL v. the FSB Term v. and the NB v. down a notch and was still able to Quad Prime at 475x6 but still wont go higher.

[Raja]

Excellent reading info of GTL is here..


http://www.thetechrepository.com/showthread.php?t=87


regards
Raja

[r1ch]

Thanks for the GTL reading Raja, I'll sit down and have a serious read through it later, it looks a bit deep for a skim read ;)

Ok, PCIE - 101 seems to make it easier to boot 477, but not stable, even with max voltages. 475 is bootable, but fails Quad Prime in minutes with with SmallFFT or Blend Tests.

Temps are 59,59,54,54 with the Quad Core Nozzle in the FuZion read from Core Temp 0.96 and ambient of about 18-20C.

Just FYI guys, with Transaction Booster = Disabled you get the following...

BIOS - Relax Level - 0
Memset - Performance level - 12

BIOS - Relax Level - 1
Memset - Performance level - 13

BIOS - Relax Level - 2
Memset - Performance level - 14

BIOS - Relax Level - 3
Memset - Performance level - 15

[ncc-1709]

:D my wafer number is 03

[r1ch]

Nice! I've kind given up on the max FSB - it's somewhere between 470 and 475 depending on the multi and ambient temps I think. Will be needing those vmods to get any higher.

I've moved onto low vcore beching...my 1.20v VID "20" is doing pretty good now I'm spending some time with it - 3.6Ghz@1.28v Quad Prime'd for 30 mins before locking up a minute ago - trying with 1.296v now :D

Damn this stuff is addictive...I'm considering a second water loop for the GTX as I think it's dumping a lot of unnecessary heat into the loop and affecting the CPU temps...somebody stop me! :D

[MonsterMunch]

...somebody stop me! :D

Soz, can't mate... tis too late already by the looks of things :lol:

[jabski]

Soz, can't mate... tis too late already by the looks of things :lol:

lol :lol:

Yes and a second loop will sure help your temps :)

[r1ch]

Soz, can't mate... tis too late already by the looks of things :lol:

No kidding...

I just leapfrogged you on the benchtec hwbot team positions as well ;)

Options are, sticking to water and getting a second loop to help temps...or sticking with the current water for 24/7 day-to-day and looking at a pot and some DI for benching and to break my sub-zero virginity :biggrin:

I think I know the direction I'm headed :lol:

[r1ch]

So...3.6Ghz@1.3v ;)

Gonna take the fuzion off tonight and check everything is ok with the contact, maybe even lap it if I can find my kit. Might also take the GTX out of the loop and see just how responsible it is for those temps.

EDIT: Changed the thermal paste from Ceramique to some random stuff my mate gave me (looks like a Shin Etsu syringe - no markings) and dropped 1C off 2 cores and 2C off the other two - 58,58,52,52. Really need to take the GTX out of the loop...

http://www.digitalreport.net/images/stories/rich/forum/3600@1304.jpg

[Stocky]

Nice for 1.3V :)

Are you planning on giving the CPU a dedicated loop, then upping the VCore?

[Pjalchemist]

GTX will be dumping a lot of heat in there. My GTX falls infront of the SB and SB with no change from stock Vcore is at a toast 43+c!!

PA120.1 is more than enough for a GTX even voltmodded.

[r1ch]

Nice for 1.3V :)

Are you planning on giving the CPU a dedicated loop, then upping the VCore?

I may well do. I've got to say that I am a little worried about the temps being so high with so little vcore...

I think what I'll do is a little investigation:

- Intel Reference
- Tuniq Tower
- FuZion+PA120.3 loop
- 6mm nozzle
- 4mm nozzle
- Quad nozzle
- Fuzion+MCW60+PA120.3 loop

And then, I'll do the thermal pastes I have as well

- Arctic Silver 5
- Arctic Silver Ceramique
- Coollaboratory Liquid Pro
- Coollaboratory Liquid MetalPad
- Random mate's TIM (looks like Shin-Etsu)

Hopefully the best combination of all of these should result in sub 50C temps and allow some vcore upping for 24/7 use.

No doubt I'll be outside and benching in the freezing cold to see exactly how high the chip will go ;)

[Stocky]

If you're trying all those pastes it may be worth trying the MX-2, as that is becoming a favourite in the enthusiast world. Also, as Pete says the GTX will be dumping a lot of heat into the loop, so ideally a separate loop would be nice - but that is more cost!