Help please! Download corruption/failed hash while 100% CPU usage. OK when idling.

[alexhuzar]

Help please! Download corruption/failed hash while 100% CPU usage. OK when idling.

When I am testing for stability using orthos or prime (or anything else that makes my computer use 100% CPU) I get data corruption with download programs like eMule, Azureus and uTorrent. When the actual download finishes at 100% the file is perfectly fine (I guess the hash checking does it's job really well). It is only DURING the actual download that I get massive corruption leading to downloads taking 3 times longer than usual due to the amount of corrupted data that needs to be re-downloaded. The programs actually ban IP addresses because of this.

I only start getting corruption if I go above ~710 FSB. At stock setting everything is perfectly fine. I get this even when using settings that are orthos-tested for over 15 hours. No BSOD, no erratic crashes. Everything works great.

I use a ATA-100 western digital 120 GB drive for my eMule temporary files storage and even with that program saving to that drive I still get corruption (but nowhere nearly as much as with bittorent programs). I'm getting another motherboard in a few days so I'll see if it's the mobo's fault.

I would greatly appreciate any input!!!



My setup:
Asus P5N32-SLI Deluxe
Pentium D 805 (overclocked to 730*20 @ 1.525 V)
Corsair Twin2x 2048-6400 (5-5-5-18-22-2T @ 1.95 V)
2x eVGA 7900 GT KO Superclocked (factory settings)
RAID 0 - 2x "Seagate Barracuda 320 GB 7200.10" model: "ST3320620AS" (Note: the RAID drive is my primary boot/windows drive and uses SATA channels 1 and 2)

Cooling:
Corsair H20 220 Apex Ultra
2x Danger Den Koosah (one for each video card)

Temperatures under full load:
CPU: 42-45 celcius (45 on a 30 degree day with high humidity)
Northbridge: 34-35
GPU:
44/43 not running graphics
51/49 running "real-time HDR demo"

 

[Evilsizer]

depending on the ram divider you might need to add more volts to it or you need to add more to the cpu. as well data corruption would beside ram would be harddrive related.

 

[alexhuzar]

I changed my motherboard and CPU. The new board can actually run a command rate of 1T without ANY problems at all. People have reported my board to run at over 1.3 GHz on the RAM. So the memory was not the issue. I think the nForce 4 chipset was absolute garbage. Unfortunately it appeared to be the best board for it's time. I sold it along with the CPU to my neighbour for $200 . Good riddens.

WARNING: STAY AWAY FROM nFORCE 4!!! Learn from my mistakes.

 

[Automata]

People have reported my board to run at over 1.3 GHz on the RAM. So the memory was not the issue.

That doesn't make sense. Your RAM speed is limited by the RAM not the motherboard.

 

[alexhuzar]

While I agree that the 680i chipset won't pull a rabbit out of a hat, many people say that they get MUCH better results with memory than they do with even a Commando. Most notably the command rate of 1T is easier to maintain above 1 GHz RAM speed. If you don't believe me then check out this 160 page thread aboout my mobo:

http://www.overclock.net/intel-motherboards/167489-asus-p5n32-e-overclocking-guide.html

Here's another one on a site with the highest reputation one can find:
http://www.xtremesystems.org/forums/showthread.php?t=123227

The new BIOS revisions let you take the Vcore to 1.9 Volts so you can go crazy, and also the Vdroop is minimal (Vdroop = u set the bios to, for example, 1.5v but you only get 1.4v).

The pros on Xtremesystems.org also seem to be consistently hitting over 500 FSB with a few hitting 525 (the ******** won't reveal BIOS version or voltages thou ).

From personal experience I can tell you that tight memory timings on this chipset are a walk in the park. I can probably crack the 500 FSB if I get better cooling for the northbridge because right now it's friggin hot even with a thermaltake Spirit II cooler (yes I did put it in right...I even re-did the install just to make sure lol) - temps for NBcore at 1.45 volts are 34 idle and 36 load.

*ALSO*

I've heard reports that from factory this mobo (680i) tends to over-volt certain memory modules and fries them because it sets the volts to max (3.5v) right off the bat (most notably eVGA and Striker). I can tell you that this was not the issue at all with my board even thou it had a veeeery early 0308 BIOS which I immediately updated to 1103. My ram runs lukewarm to the touch.

Back to the nForce 4 memory thing: all I know is that no matter what setting I put on that board, even if it fails Orthos, or I get a BSOD, I get absolutely no corruption at all. Zip zilch zero nada ni'et. When I lowered the FSB to under 710 (177.5x4) the corruption doesen't happen at all.

 

[Evilsizer]

well first you said it was a NF4, now your talking about NV680I. on top of that how do you expect value ram to hit ddr2-1300 (670mhz). esp when the timings are 5-5-5-15. the memory was more of a problem then you think it was, its not the chipsets fualt if the ram can run that fast.

 

[alexhuzar]

On my new board I can run DDR-800 4-4-4-8 with 1T command rate and I get no corruption and it's also orthos stable for as long as I can run it (9+ hours). With old board at DDR-800 5-5-5-18 2T it would pass orthos 15+ hours but I'd get massive corruption while downloading. When I lowered the FSB the corruption went away completely.