Socket 939 AMD X2 4400+ - Page 24 - Overclock.net

blitz6804 · 07-19-08

Rig in my sig updated. Rig in the post to follow once I get the OCs hammered out.

blitz6804 · 07-19-08

Quick update (READ: Wall-of-text warning):

While I had a nice 3DMark06 score (9217), I would occasionally crash the video driver. It does not recover as nice with this video card as with the AGP-variant, and pictures would be distorted until I restarted the PC.

I took a 7 MHz drop on the GPU core (about 50 3DMarks) but now am Real-Time High Dynamic Range Image-Based Lighting stable for 1/2 hour. Unlike the AGP card, which could not do much more than 1300x1100 before it would cause the program to crash (over 100% load) this card can run 1920x1200x32 at 16x Anti-aliasing and Bump maps turned on. Granted, it only does it at 3-5 FPS, it does it all the same. (At 8x AA, you get a "playable" picture, that is, 60+ FPS at that resolution. Nice. Crysis time soon methinks.)

In other news, this motherboard has an "annoying" feature: unplug it, and when you boot next time, most things set to default. Granted, resetting the external clock to 200 MHz is nice if you went to far, resetting to stock without fiddling with the CMOS jumper, but there is another side effect: all the fan throttling, monitoring, and system alarms return to default. PCI-E and RAM settings are unaffected, but it is still annoying.

I discovered this this morning. I needed to fiddle around with some wires to try to improve airflow. When I restarted, I also figured I would drop my GPUC clock as I said above. I dropped it 20 MHz, and upon running 3DMark, found I lost over 2000 3DMarks! *Gasp!* This made me about 1800 3Dmarks BELOW AGP. I started going crazy and then thought to check CPU-Z. Sure enough, I had dropped back to 2200 MHz.

Long story short, I fixed it as above, and now am around 9170. I do not remember, and I did not submit it.

Before I fixed it though, I did some stock of my temperatures both before and after 3DMark. All readings are done at 24º C ambient with the PCI-slot cooler installed between the secondary PCI-E-slot and the Socket:

STOCK (2205 MHz)
Case: 32º C idle, 38º C loaded
Proc1: 27º C idle, 44º C loaded
Proc2: 34º C idle, 47º C loaded

OVERCLOCKED (2801 MHz)
Case: 34º C idle, 38º C loaded
Proc1: 30º C idle, 46º C loaded
Proc2: 36º C idle, 48º C loaded

I would take those temperatures to mean I have some overclocking headway now. Comparing cold-start temperatures, before the GPU has a chance to mess up case temperatures, I would estimate that my lap job has improved idle temperatures by as much as 12º C, and load temperatures as much as 8º C. Pre-lap, on the AGP board, my cooler core ran at 40º C idle. Post-lap, the cooler core ran at 30º C in a warmer case temperature. Pre-lap, 3DMark made CPU temps rise to the mid-upper 50s; now they remain sub-50.

Only thing left to do is OCCT I suppose. Oh, that, and wring some more work out of the processor.

It should also be noted that I am seriously acknowledging now that the GA-K8NSC-939 was holding me back. When set to 1.35 V in the BIOS, it would provide 1.328 ± 0.028 V (quantum = 0.014 V). Cool-n-quiet dropped the multiplier from 11x to 5x and dropped voltage to 1.080 ± 0.008 V (quantum = 0.004 V). Cool-n-quiet worked in XP and Linux, not Vista.

On this board, when set to 1.35 V in the BIOS, it provides 1.380 ± 0.010 V (quantum = 0.010 V). Cool-n-quiet has two settings. The multiplier at 9x drops the voltage to 1.260 ± 0.010 V (quantum = 0.010 V) and the 5x drops the voltage to 1.120 ± 0.010 V (quantum = 0.010 V). In other words, I have a much more stable (and higher) voltage to work with. For those of you keeping score at home, my average load VCore is about 3.91% higher than before, meaning I should be able to keep a higher, S&M-stable, 24/7 configuration. So far, I am running 2801 (versus 2750 previously) with more expected to come.

</WALL_O_TEXT>

I seriously need to edit the first post...

blitz6804 · 07-25-08

According to Everest, PCI-E voltage is 1.2 V. I am trying 1.3 V now to see if that fixes anything.

And I have good news and bad news for the club. The bad:

Try as I might, my old silver lady will not get me into the 1 GHz OC Club. With Thlnk3r egging me on, I even tried an uncharacteristic 1.800 V VCore, a 3.000 V DDRV, and a 2.000 V NBV. Couple this with 2 GB of RAM running at 3-5-5-10 running at DDR 386, and you still cannot make it POST.

The good news, I finally broke 30s in SuperPi. I was fiddling for the highest OC I could get that would run SuperPi. Any OC higher than this either crashes or runs OC super-30s.

As you can see below, I am running an uncharacteristic voltage for me of nearly 1.6 V:

The CPU-Z validation link

Needless to say I was flabbergasted at me finally breaking (significantly) over 3.0 GHz, as my Gigabyte could just barely squeak by.

This got me curious. Could I replicate the feat with four DIMMs of RAM instead of two? As Joe always says, "Testing, testing, testing!" So, I dropped those DIMMs back in and fired her back up. What would you know?

The CPU-Z validation link

Needless to say I am elated to see that she could do it. Temps idle were 45/50º C. Running SuperPi, about 51/54º C.

Right now, I have dropped my VCore back to 1.380 V, and am resuming experimentation to get my new S&M/24-7 stable settings going. Since I know that the RAM will never run at DDR 500 with four DIMMs (but apparently DDR 444 is not unheard of) I am going for tighter timings instead. I am running sub-SPD timings of 2-3-2-6-2T-11-17 @ DDR 401, 2.82 V to see what will become of it.

(This post is cross-posted in the Socket 939 Appreciation Club and Knowledgebase.)

**thlnk3r** · 07-25-08

Blitz, great job

What was the highest overclock you were able to achieve at 1.380volts?

Good luck

StormX2 · 07-25-08

thats not bad man

blitz6804 · 07-25-08

So far, I am running 2801 MHz at 1.380 V. I want to see if it is S&M stable before I move up any further. I know it is OCCT2 stable, max temperature in the mid-50s.

blitz6804 · 07-27-08

The Xaser TASER is once again.

.

Max temperature obtained was 60º C on the high core. It only briefly appeared, more often fluttering between 58 and 59º C.

blitz6804 · 08-21-08

Update:

I added a second 3850. Graphics are stunning until the PC goes into hard lockup.

A bigger PSU is on order.

blitz6804 · 09-02-08

Another, and in fact, nearly the final update.

The PSU was fine. The second 3850 on the other hand was DoA.

A tragedy befell this morning which struck the end of Xaser TASER. Some idiot *points at self* decided he would try to delid the Toledo. I failed. At the same time, the motherboard decided to commit seppuku. My San Diego works in my Gigabyte, but not in the Abit. My Toledo's memory controller is gone on both motherboards.

Whence my AM2 parts get here, I will post a link to the new build thread.

Moral of the story: do not attempt to delid a processor after being awake in excess of 36 hours. Do not attempt it if you have bad eyesight. Do not attempt it if you have unsteady hands. Do not attempt it if your name is Blitz6804 and all of the above apply to you. Oops; too late.

blitz6804 · 09-06-08

Since the Xaser TASER is dead, I decided to retire the thread.

Visit Xaser Brixfire's thread instead.

For the record, TASER topped 3DMark06 at 9005.

		Overclock.net - Overclocking.net > AMD > AMD CPUs
Socket 939 AMD X2 4400+

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)