PDA

View Full Version : E6600 and P5W DH Deluxe Overclocking Questions


ArrowMk84
10-10-06, 08:25 AM
Ok, I'm starting to overclock my system (in sig). I've only fiddled with it a couple of times, and right now my CPU is at 2.88 GHz, with the RAM at DDR2-800 (320 FSB with 4:5 DRAM divider). I did this this morning, and so far its stable in SP2004 with stock voltage (1.32v). I'd like to push this higher.

I know Windows will boot at 3.2 GHz and DDR2-710 (335 FSB 1:1 DRAM) with stock CPU voltage, but it gets a flakey at this point. How high can I safely push the CPU voltage? Is it safe to push my RAM over 800 Mhz (last time I tired to push my RAM just a few MHz over its rated speed, I ended up with two dead sticks).

Also, I've noticed that when my system is overclocked, any time I do a restarted (either from Windows or exiting the BIOS menu), my system actually shuts down for a couple of seconds and then powers itself back up. Is this normal?

DataMatrix
10-10-06, 09:09 AM
There should be some specifications provided on that by intel, although I THINK a rule of thumb is to never make any change over 5%. Although wait for other replies first.

Blacklash
10-10-06, 01:27 PM
There you go:

http://www.xtremesystems.org/forums/showthread.php?t=110193

For overclocking make sure you kill hyper path 3, ram throttling, and spread spectrum. They may gimp OCs. Lock your pci freq @ 33 and your PCI-e @ 100. You may want to try your luck with 1:1. My old P5W loved 1:1 and didn't care for dividers. If you have the latest BIOS that should be fixed though.

This is a good thread also. There is lots of good overclocking info @ XS:

http://www.xtremesystems.org/forums/showthread.php?t=115765

BTW the power off then on with OCs is an Asus "feature". I believe it has something to do with the board making adjustments if it doesn't accept your OC at first. P5B does it too.

For others, when you buy the P5W make sure its serial number begins with 68. Example, 68M0AG-. The old 66s, like mine, only had 1.65v MCH. The new revs go to 1.85v MCH which should enable a 450+ FSB on a 1067 strap. If you are expecting to do 450 x 9 I hope you have some serious cooling.

If you're planning on keeping that C2D a while I'd advise the best OC you can manage on up to 1.45 vcore with a solid aftermarket cooler on air. If you have high end water or better 1.52v. Phase 1.6v+. For me, on air, that's 3.6GHz 24/7.

Also use power management. With it on I am 2.4GHz idle and 3.6GHz load for 24/7 use.

ArrowMk84
10-10-06, 05:37 PM
Any idea where the spread spectrum option is in the BIOS? I can't find it under Jumperfree, Chipset or CPU.

Blacklash
10-10-06, 06:43 PM
Any idea where the spread spectrum option is in the BIOS? I can't find it under Jumperfree, Chipset or CPU.

It may not be there :p It may be the P5B Deluxe that has it. I haven't looked at the P5W BIOS screen in a while.

You may want to increase the ICH and FSB termination in addition to the MCH voltage too when swinging for the fences with this mobo. I was able to get a 439 FSB on the P5W and I was restricted to 1.65v on the MCH because I had an early revision. That was a 439 FSB boot from BIOS and not a clockgen up.

ArrowMk84
10-10-06, 06:45 PM
It may not be there :p It may be the P5B Deluxe that has it. I haven't looked at the P5W BIOS screen in a while.

You may want to increase the ICH and FSB termination in addition to the MCH voltage too when swinging for the fences with this mobo. I was able to get a 439 FSB on the P5W and I was restricted to a 1.65v MCH because I had an early revision.

Heh, I don't think I'm going to get that aggressive. I think I'm going to try that 3.2 GHz OC again with the vCore bumped up a little and see how it runs.

Blacklash
10-10-06, 06:59 PM
Heh, I don't think I'm going to get that aggressive. I think I'm going to try that 3.2 GHz OC again with the vCore bumped up a little and see how it runs.

All my C2Ds would do 3.2GHz on an "auto" voltage setting. That's an E6400, E6700 and this E6600. It should be doable.

Good luck. Just make sure you kill hyperpath3. My OCs were so gimped they pissed me off till I killed that feature and ram throttling on the P5W.

Now I am going to go watch Tito beat up on Ken again.

ArrowMk84
10-10-06, 09:12 PM
All my C2Ds would do 3.2GHz on an "auto" voltage setting. That's an E6400, E6700 and this E6600. It should be doable.

Good luck. Just make sure you kill hyperpath3. My OCs were so gimped they pissed me off till I killed that feature and ram throttling on the P5W.

Now I am going to go watch Tito beat up on Ken again.

3.2 was working fine, so I set the divider up from 1:1 to 4:5 (888 Mhz). That ran fine for about half an hour of running SP2004 on both cores while using FireFox. Went to load a page, and the system shut itself down. The timings at SPD (5-6-6, IIRC), and I had the voltage at 2.1 (which what's its spec'ed for at 800Mhz). Should I up the voltage a little more? Am I risking my RAM with this?

Also, am I losing much using 1:1 vs 4:5 (or even 2:3)?

ArrowMk84
10-10-06, 10:13 PM
Just for the hell of it, I set my FSB to 385 and my DRAM to 1:1, with a vCore 1.45. Hard lock after 15 minutes. I was doing the same thing as above, SP2004 on blend and web surfing - the lock happened when I was typing a reply to the new R600 thread. That was actually my target OC of 1 GHz, too.

More vCore?

Blacklash
10-11-06, 02:26 PM
3.2 was working fine, so I set the divider up from 1:1 to 4:5 (888 Mhz). That ran fine for about half an hour of running SP2004 on both cores while using FireFox. Went to load a page, and the system shut itself down. The timings at SPD (5-6-6, IIRC), and I had the voltage at 2.1 (which what's its spec'ed for at 800Mhz). Should I up the voltage a little more? Am I risking my RAM with this?

Also, am I losing much using 1:1 vs 4:5 (or even 2:3)?

Hmm. It's hard to say without knowing your complete BIOS settings. A system can shut itself off when the CPU gets too hot. I would install core temp or Asus PC probe and check that. Perhaps try running your rig with the side off the case and see if it behaves the same.

Some C2Ds are not perfectly flat and if good contact isn't being made between the heatsink fan and the chip you may have temp problems. I am assuming you used arctic silver 5 or another high quality thermal paste between the heatsink and chip. Also my Ninja has a thermaltake fan capable of up to 2.6k RPMs on it I run @ 1-2.4K. Make sure your fan is clipped on so its sucking air from the front of the case and towards the rear exhaust.

I did not care for the fan that came with the Ninja.

My ram runs on low voltage and does better on 2.25v. I wouldn't be afraid to use a little more. I actually get my best timings and up to 1150 speed on 2.45v. Some people go higher than that and I wouldn't without great air flow or a small fan near the RAM, or on it.

Shutting down can be heat related, and yes it could be voltage related. I'd try the other things I suggest before upping vcore. FSB termination @ 1.40v may help and MCH @ 1.55v. I really don't think you should need it at this point. I suspect more voltage needed on the ram, or over heating. Try 2.25v if you want to stick with +800 DDR.

The differences between 1:1 and the dividers are small. My ram will do 3-3-3-10 on relatively low voltage below 800 DDR @ 1:1. You may want to try something akin to 4-3-3-12 1:1 and see it it holds. Then tighten the Tras a bit and perhaps the CAS to end at 3-3-3-10. If you want safe and stable manually set your ram to 2.25v 4-4-4-12 and be done with it. If you use dividers and head for DDR 1000 you may need to relax up to 5-5-5-15. Again voltage increase on the ram will counter that and do not exceed 2.45v.

Make sure you are using the latest BIOS, some of the old ones on that board were unstable with dividers.

You may want to fire up superPi 32m with Asus PC probe open and watch the temps.

Blacklash
10-11-06, 02:33 PM
Just for the hell of it, I set my FSB to 385 and my DRAM to 1:1, with a vCore 1.45. Hard lock after 15 minutes. I was doing the same thing as above, SP2004 on blend and web surfing - the lock happened when I was typing a reply to the new R600 thread. That was actually my target OC of 1 GHz, too.

More vCore?

Again, could be heat or volts. Monitor your temps with core temp and or Asus PC probe when SP2004 is going.

Try this:

CPU/Memory: 1:1
DRAM Timing: Manual 4-4-4-12 or by SPD if you prefer.
Hyperpath 3: Off
DRAM Throttling: Off
CPU Frequency : 356
DRAM Frequency : 712
PCI Express Frequency : 100
PCI Clock Synchronization Mode : 33.33MHz
Memory Voltage : 2.25V
CPU VCore Voltage : 1.45v (I'd urge trying for a lower vcore 1.35-1.40v ideally for 3.2GHz)
FSB Termination Voltage: 1.40v
MCH: 1.55v
ICH Chipset Voltage : MAX

Having to use more than 1.45v for 3.2 GHz isn't worth it IMO. I'd just do 3.0GHz if it truly does take more than 1.45v for you to hold 3.2GHz. I'd try booting and running 3.2 again with between 1.37-1.40 vcore. IMO 1.45v is a lot for 3.2GHz. If you did get a dud chip e-bay it and try again.

Get 3.2 stable first. Stick with 1:1 until you get things sorted out. When you can run SP2004 stable for 6-8hrs you are likely set. I really suspect heat because I was able to run three different C2Ds @ 3.2GHz in a P5W with vcore on auto and ram @ 1:1 2.25v, all else to auto voltage.

A freeze certainly can be voltage related and I'd only increase that after I was sure my temps weren't out of control.

ArrowMk84
10-11-06, 06:55 PM
I've been playing with the voltages today (got the day off to do some other stuff), and right now, it looks like the RAM voltage ws just too low. I've got the CPU running at 3.2 GHz, with vCore set to 1.32, FSB/MCH/ICH all set their mid values (1.3 for FSB and the other two have slipped my mind, but I'll up them per your values) and the RAM at 2.2v with 5-6-6-18-21 on a 4:5 divider. After three hours of SP2004 and three hours memtest86, this should be stable (need to do a longer SP2004 run to verify, and tighten the ram timings), but overall, I like this overclock. It also at just about the same temp as stock.

I had it running 3.4 GHz stable with a 1:1 divider, but I had to use 1.5v to get there - I didn't like those temps one bit. The motherboard reported a CPU temp of 51c+, and Coretemp was reporting 75c, so that had to go.

The cooling on for the CPU is a bit different. My TJ-07 case sucks in air from two 92mm rear fans, and then exhausts from two 120mm fans on top. All four of these fans are pretty much right at the heatsink. I've replaced the stock fans with two panaflow-m's (92mm) and two tri-cools (120mm), so the Ninja's got a quite a bit of air moving through it. For some added airflow, I also installed a Kama drive bay fan and moded the side panel for two 120s over the PCI slots (which does a nice job off keeping my GX2 under 75c).

I seriously appreciate your help with this. This is the first OC I've done that hasn't turned out to be a failure or blown up a component.

Blacklash
10-11-06, 08:54 PM
Glad to hear it.

3.2GHz is a safe and rational OC 24/7 on air in a case. I wouldn't try the higher OC with temps you are indicating.

If you want to keep temps down use the power management features that reduce your CPU speed to 6 x on idle.

Don't increase any other voltages unless you experience further instability. Perhaps the ram boost was enough. I didn't need higher FSB termination, MCH and ICH for 3.2 GHz on my P5W. Hardware does vary. I did need them increased for 3.4-3.6GHz.

I have all my fans on a sunbeam rheobus and vary them according to workload.

Roadhog
10-11-06, 10:19 PM
Glad to hear it.

3.2GHz is a safe and rational OC 24/7 on air in a case. I wouldn't try the higher OC with temps you are indicating.

If you want to keep temps down use the power management features that reduce your CPU speed to 6 x on idle.

Don't increase any other voltages unless you experience further instability. Perhaps the ram boost was enough. I didn't need higher FSB termination, MCH and ICH for 3.2 GHz on my P5W. Hardware does vary. I did need them increased for 3.4-3.6GHz.

I have all my fans on a sunbeam rheobus and vary them according to workload.

bump the fsb termination and southbridge, chipset, and the MCH to max voltage... dont worry, it wont do damage lol.

Blacklash
10-12-06, 09:10 AM
bump the fsb termination and southbridge, chipset, and the MCH to max voltage... dont worry, it wont do damage lol.

My advice is to never use more voltage than needed for a task. I only find some of those increases needed @ 3.4 GHz and up. So if this fellow stays @ 3.2 GHz I see no reason for the increase save further instability. YMMV.

@ArrowMk84

You may find this thread interesting and I'd recommend it.

http://www.xtremesystems.org/forums/showthread.php?t=112776

HYBRID
10-12-06, 09:35 AM
I posted my results here using 1.35 as my vcore maybe this will help you a little.
http://www.nvnews.net/vbulletin/showthread.php?t=77462

good luck

ArrowMk84
10-12-06, 10:05 AM
Nice post HYBRID. After reading that, I'm thinking of using a 1:1 divider a go with the 4-4-4-12 timings for my memory. I honestly can't see there being too much difference between 1:1 and 4:5, except for a couple of benchmarks (my 4:5 is currently using 5-6-6-18 SPD timings). Plus I could lower my vDimm a bit. Any opinions on which way is the best way to go?

I had to up vCore a tad to 1.35. SP2004 had an error this morning on AUTO (and PC Probe was reporting 1.30v for vCore).

ArrowMk84
10-12-06, 06:47 PM
Sweet Results! (http://home.comcast.net/~arrowmk84/Overclock.JPG)

Thanks for all the help guys!

Now it's time to enjoy this beast... then I'll try OCing the GX2! :beer:

Edit: spoke a wee bit too soon. 3DMark05 dumped during the CPU test, and one core kept on trying to execute code, so I had to bump up the voltage to the next increment. Then it passed with no problems, clearing 15k points, and 3DMark06 gets ~8.5k points.

lduguay
10-13-06, 07:07 AM
Any idea where the spread spectrum option is in the BIOS? I can't find it under Jumperfree, Chipset or CPU.
At one time it was an AGP option used to reduce EMI and had no effect on performance. Could be something else now.

Blacklash
10-13-06, 11:57 AM
At one time it was an AGP option used to reduce EMI and had no effect on performance. Could be something else now.

Right and wrong.

SS is indeed for emi reduction.(That's electromagnetic interference for those unaware.)

System stability and performance may be slightly compromised with SS enabled. Especially with timing critical devices.

For overclocking even a 0.25% jitter can cause a temporary boost in clockspeed which may cause an overclocked processor to lock up.

If you aren't having EMI problems keep it off.