OC Input with on water cooled IC7-MAX3 and P4 3.2EE

[StangMan]

OC Input on water cooled IC7-MAX3 and P4 3.2EE

Hi all... Although I also built another water cooled system about six months ago with a different motherboard (GA-SINXP1394 & P4 3.06 OC @ 3.55), I'm not the most experienced around with overclocking yet. However, I just completed a newly built system (both utilize Koolance water cooling rigs). I'd love any advice those more experienced with similar configurations can offer, as I am trying to tweak the most out of it while also remaining stable and not setting myself up for S.N.D.S. Here are the specs:

Koolance PC2-650BK watercooling system
ABit IC7-MAX3 Motherboard
P4 3.2 Ghz Extreme Edition 2MB L3 HT CPU
Cosair 2 x TWINX512-PC4400 DDR Memory modules
Radeon 9800 Pro AIW Video Card
4 x Maxtor 160 GB 7200 RPM SATA Drives (RAID 10)

The CPU, MB Chipset, Video card, and all four hard drives are water blocked. The DDR modules also have ThermalTake active coolers. There are three 80mm case fans, BGA DDR Ramsinks on the Radeon, and a twin 80mm side mounted card cooler. The system runs 30 C under a full load per the Koolance display... although, the BIOS shows 42 to 46 C under load/tests. But, I've read on here that this motherboard shows about 10 C hotter than it really is. Correct? In addition, I've looked at the board and it is a newer revision that has the MC64 cap removed already.

So far, the best stable (i.e. I can run all benchmark suites and SETI for quite some time with no crashes) overclock I've been able to get the system to run is at 3.7 Ghz (3712 Mhz) using the following settings:

FSB: 232 (3712 Mhz)
vCore: 1.7v
N/B Strap: PSB800
DRAM Ratio: 1:1
AGP Ratio: Fixed (66/33)
DDR Voltage: 2.8v
AGP Voltage: 1.65v (Radeon OC @ 432 Core / 378 Memory)

DRAM Timing
- CAS : 3
- Act to Precharge: 8
- RSA to CAS : 4
- RAS Precharge : 4

PAT (Game Accel): Auto
- Refresh Cycle: Auto
- Read Delay : Auto
- tRDA : Disabled
- CPC : Disabled

MPS: 1.4
IDE Bus Master: Disabled
USB: Set to OS, not BIOS
BIOS Cacheable: Disabled
Video Cacheable: Disabled
AGP: 8X
Fast Writes: Enabled

I recently read some references to the type of DDR chips used with this board (BH5 vs CH5 I think). I do not know if I have CH5 DDR chips on those Cosair modules or not, so I am wondering if that is my problem. In addition, I am also wondering if I am too high running my vCore at 1.7, although it seems that I have to in order to run stable at the 232 FSB. I am going to try again backing off to 1.675 though. It also seems from what I've read that you just don't want to go over 1.75 vCore with water... and, you want to stay at or below 1.7 with air (from what I've read thus far). I want to make sure I don't kill this thing two or three months from now seeing as this CPU was a hefty $850 bucks.

Any ways, a lot of these settings I figured out from doing some back reading on the forums here. Unfortunately, I seem to now be at a brick wall. I was hoping I would be able to FSB this thing at around 240 to 250 (near 4 Ghz) due to the water cooling and PC4400/550 Mhz RAM I'm running. Can anyone offer any suggestions at this point? Oh yes... and, should I or shouldn't I do a burn-in process at stock settings before running OC'ed? How big of a deal is that any ways?

My current benchmark scores at this OC on the CPU and GPU, with just a standard non-tweaked/streamlined install of XP, are:

CPU Arithmetic Benchmark
-- Dhrystone: 11360
-- Whetstone: 4752 FPU / 8307 iSSE2

CPU Multimedia Benchmark
-- Integer: 28866
-- Floating-Point: 40974

File System Benchmark
-- Drive-Index: 100 MB/s

Memory Bandwidth Benchmark
-- Integer: 5652
-- Float : 5645

Cache & Memory Benchmark
-- Index: 12311 MB/s

3DMark2001SE Score: 33607

3DMark03 Score: 6743

PCMark2002 Score
-- CPU Score: 9079
-- Membory Score: 16009
-- HDD Score: 1267

PCMark04 Score: 5850

Here's a photo I took while assembling the system and after (the only thing missing is the side card cooler on the Radeon). Gee... it was nice and neat until you run a gillian power and drive cables all over to hook it all up.

 

[batboy]

Impressive rig.

Abit mobos are known to read a little higher than other mobos. Is this really inaccurate or do other mobos read low? I've got two regular IC7's and they only seem to read about 4 degrees high. Some people claim up to 10 degrees higher, but I'm not convinced that they all do.

My recommendation is that 1.7v is max for long term even with watercooling. The less vcore the better. Personally, I don't run my vcore above 1.65v for long term. I'm talking actual measured voltage, not what it's set at in the BIOS.

My guess is that your RAM probably has Hynix chips.

 

[StangMan]

Well, I have done a lot of playing and testing with the system over the past few days. In addition, I also ran a day long burn-in process on the system. I did the burn-in going by the rule of running low clock speed, maximum voltage, and minimum temperature. In other words, I set the bus (Mhz) to stock (200), I set the vCore to 1.75v, and I had the water cooling fans set to max... and, I even opened the window to the cold winter air in the room. I then ran Sandra's burn-in wizard set on low utilization, with only the CPU and memory tests, to run continuously... and, I also ran SETI@home. This maxed the CPU at 100% for ~24 hours. Also, I ran Motherboard Monitor and had its dashboard up so that I could somewhat monitor vCore during this process when I wanted to watch things run.

Some observations I made are that regardless of what you set the vCore to, it actually runs about .02 less with no load. So, if you have it set to 1.75, it sits at the desktop running around 1.73v. If you set it to 1.7v in BIOS, it sits at the desktop around 1.68v and so forth. Now, with a full load, as described above, the vCore drops quite a bit. Running 1.75v (or 1.73v no load per BIOS), it would drop all the way down to 1.68v... sometimes fluctuating to 1.66v even. The same for running at 1.7v (1.68v with no load). It will drop to 1.63v under full load, and flux to 1.62 at times and back to 1.63v.

So, from what I've observed there is a 5 to 7v swing in voltage based on CPU load of the system. The trick seems to be setting the vCore high enough for whatever Mhz you want to run so that even with max load & flux, it will never drop lower than the minimum voltage level required for that overclock to successfully operate it seems. It just seems to be lots and lots of experimenting to find that point and what Mhz level you can OC to while also keeping your vCore at what will be a safe level for long term use... Does all this sound logical or reasonable?

Regardless, the burn-in did seem to help the system run with slightly more stability. Currently, I have found that the maximum FSB I can now run with ~98% stability is 230, which gives me 3680 Mhz (3.68 Ghz) with a 1.7 vCore. I CAN run it as high as 3.88 Ghz with a 1.75 vCore, but it will not run stable with SETI any real length of time. In fact, some times bench marking programs (3DMark, PCMark, Sandra) will run ok and then sometimes the system will just lockup or crash trying them at OC's over 3.68 Ghz. It seems to get touchier the higher I clock it beyond 230 FSB (at least not going over 1.75v). Perhaps upping the vCore again to 1.8v would solve that... but, I don't want to push it beyond the 1.7v mark for permanent usage any ways, as you suggested.

Here were my benchmarks clocked at 238 FSB (3.80 Ghz) level though:

CPU Arithmetic Benchmark
-- Dhrystone: 11160
-- Whetstone: 4824 FPU / 8313 iSSE2

CPU Multimedia Benchmark
-- Integer: 29308
-- Floating-Point: 41183

Memory Bandwidth Benchmark
-- Integer: 5747
-- Float : 5737

Cache & Memory Benchmark
-- Index: 12494 MB/s

PCMark2002 Score
-- CPU Score: 9334
-- Membory Score: 16323
-- HDD Score: 1272

PCMark04 Score: 6120

3DMark2001SE Score: 37463

3DMark03 Score: 6920

*The Radeon AIW 9800 Pro was OC'ed at 432 Core / 378 Memory.

 

[batboy]

That 3Dmark2001 score must be a typo. That would probably be a world record if it was correct.

 

[Digital_Cowboy]

Perhaps it's his scores when running on lowest settings?

 

[Digital_Cowboy]

Hey Stangman, do you boot off those RAID Drives. I have had some trouble with setting my RAID Raptors to the main boot drives.

 

[StangMan]

Yes, I did get that score (honest). I got it when I had everything at the highest overclock I could reach. I also had Rage3D set to the 3DMark2001SE profile, which I believe sets all the video card settings to max performance. However, I have since tried to duplicate the score/result... I've rerun the thing with the same overclock settings several times and I can't get that high again. Go figure! The best I've been able to duplicate is around 33 to 34K. Something else odd I've notice is that if I run 3DMark and get a score... and, then run it again (repeat that process a few times), the score will drop lower and lower each time... WTF? That seems weird to me.

Any ways, Digital_Cowboy, yes I have a SATA Raid 10 array configured (four Maxtor SATA 160GB drives) and I am booting XP Pro off them. I also have a single 120 GB IDE drive as the Master of IDE1. What you have to do is turn OFF the standard IDE channels while you install XP on the Raid array (because if you don't it will write MBR sectors to the damn thing even though you aren't installing to the standard IDE drive). Also, you need to configure in BIOS to have SATA Raid bootable and change the boot order (once you add the standard IDE channels back in), so that the add-in bootable device(s) is listed before your regular drives. Make sense?

 

[Jimbob7]

hmm submit you 3dmark score something is wrong the best is 31k i think, thats with vmods etc, theres noway you could have that score :O hehe

Very nice job on your rig, now just enjoy it

 

[Digital_Cowboy]

Any ways, Digital_Cowboy, yes I have a SATA Raid 10 array configured (four Maxtor SATA 160GB drives) and I am booting XP Pro off them. I also have a single 120 GB IDE drive as the Master of IDE1. What you have to do is turn OFF the standard IDE channels while you install XP on the Raid array (because if you don't it will write MBR sectors to the damn thing even though you aren't installing to the standard IDE drive). Also, you need to configure in BIOS to have SATA Raid bootable and change the boot order (once you add the standard IDE channels back in), so that the add-in bootable device(s) is listed before your regular drives. Make sense?

Thanks, StangMan. My problem was with the IDE drive taking some of the MBR like you said. I just unplugged the dang thing and the install worked great. I also noticed that I had to have the newest version 14 BIOS for it to work. Works great now. NOw I can enjoy those superfast 7.8ms seek times on my system HDD's.

 

[StangMan]

hmm submit you 3dmark score something is wrong the best is 31k i think, thats with vmods etc, theres noway you could have that score :O hehe

Very nice job on your rig, now just enjoy it

I had to run the thing at 3.88 Ghz with the 432/378 OC on the Radeon to get 30K+. It's not stable at that set-up and randomly crashes while running benchmarks/SETI. So, it's a moot point any ways. I've settled with a 230 FSB setting (3.68 Ghz). I get around 29K scores at that speed with the Radeon at 425 / 378.

Thanks for the compliments on the rig.

 

[StangMan]

Thanks, StangMan. My problem was with the IDE drive taking some of the MBR like you said. I just unplugged the dang thing and the install worked great. I also noticed that I had to have the newest version 14 BIOS for it to work. Works great now. NOw I can enjoy those superfast 7.8ms seek times on my system HDD's.

Awesome. Yeah, I ran into that the first time I tried to install the system. I have the four drive SATA set-up on the four Silicon Image SATA ports and then I have just a single HD on the primary IDE1 master spot. I wanted to run the RAID array as drive C and the single as drive D. With them all hooked up, I told windows to install on the RAID array, after loading the 3114 RAID drive during boot. It acted like it installed fine, but then when I tried to reboot it was trying to load off the single. Also, if I unplugged the single it still wouldn't load and gave me a no NTLDR error. I finally figured out that it was writing the MBR boot records to the single, even though it installed XP on the RAID array... just screwy.

Any ways, I used a Partition Manager to inactivate the single drives boot record. I then did as I directed you to do with turning off the IDE port(s) during windows installation, which keep it from doing that. Then, after install turn them back on and order the boot devices correctly and you're all set.

The system runs fine from the RAID array, but the issue I'm now noticing with it is that if windows locks up and I have to shut it off to do a reboot without "properly" shutting down and restarting... then, after I reboot it causes the RAID controller to "sync" and start a rebuild of my channel 2 drive. That is one of the drives running as one of the mirrors it appears. My other system running IDE ATA133 drives in a RAID 10 array NEVER does that. It's very annoying, because it takes an hour and a half to do the rebuild. And, if you shut down before it completes it will DROP the drive from the array and then you are SCREWED. You have to delete the array and recreate (reloda the whole system) to correct it... WTF?

 

[ShoNuff]

Stangman, that is an awesome set up you have. Looking at my sig we have very similar systems and you are getting great results from the Koolance system you have. I've heard nothing but good things from the new Koolance water block that comes with the Koolance system you have. Viperlair did a review on the water block and the results were impressive. The Koolance H20 block beat the Swiftech by more than 3c. Again, very impressive.

At 1.7v, my load temps are 55-57c with the Max3. On the same loop I am cooling the NB and GPU.

Looking at your post you have some good memory. If you have BH5 modules they can take a lot of juice. 3.2v is not out of the ordinary and may be required to run tighter timings. I've heard the Max3 vDimm is stable at 2.8 and 3.2v if you venture to give it a shot in order to run tighter timings. At 3.2v your memory you should be able to do much better than 3-4-4-8.

Your bios settings look good. The only thing I have done differently is listed below. My 3.2 runs up to 250 fsb at 1.75v stable. 245 fsb is rock solid at 1.70v with the following GAT settings:

PAT (Game Accel): Auto
- Refresh Cycle: Normal
- Read Delay : Auto
- tRDA : Enabled -Try this first and if this isn't stable change to disabled
- CPC : Disabled

These settings may get you a little bit more out of your CPU.

Speaking of CPU, what is your FPO/Batch# and pack date? I have a 3344A570 with a pack date of 01/09/04. I think they are all SL7AA's at this point.

 

[Quicksilverta]

Nice setup stangman. I've got almost the same system as you but not the 3.2EE. I've only got a 3.2 northwood. Im running 3.685 with a 1.575 vcore. RAN deal prime95 for 15 hours no errors. Temps are at 39.5 at idle and 50.5 at a load with air SP-94. Like ShoNuff said i would change you GAT settings to A-N-A-D-D. Your v-core seems kinda high to me for what your running. Just ordered my koolance setup. Can't wait to see what i can do with it. Also have you ever ran memtest98? I would get memtest and see if your memory holding you up.