ABIT IC7 G system freezing occasionally

[txyz]

I apologize for the length of this posting, but the system details are essential to understand the problem. It is a new build system:

ABIT IC7 G motherboard with retail P4 3.0 Ghz Northwood CPU (SL6WU). Flashed to latest BIOS 28. 1 Gig of OCZ Platinum PC3200 Series II RAM -- (purchased as a matched pair). CPU voltage currently is set to 1.675v, DDR SDRAM is set to 2.8v and AGP is set to 1.65v in BIOS. Overclocked to 3.54 GHz -- FSB 236MHz. ThermalTake XP90 cooler with Artic Silver 5 thermal compound -- running at 44C idle. Running Windows XP Pro. ATI Radeon 9600XT graphics card. C drive is a Western Digital 200Gig drive alone on the primary IDE connector. The power supply is a new ThermalTake 480w PurePower. In a well cooled full tower case. Almost no software or hardware cards in the system yet. CPU/Mem set to 1:1. The PCI/AGP are locked at 33/66. Latency is is set in BIOS to 2.5, 3, 3, 7. The Northbridge strap is set to 800.

CPU-Z reports slightly different values. FSB measures at 236.2MHz so CPU is operating at between 3543.6 and 3543.9 (varies in that range). CPU voltage varies between extremes of 1.584 and 1.648. 1.600 and 1.632 are the most frequently reported values -- which are consistent with ABIT's undervolting of the CPU.

The problem that is driving me crazy is that the system freezes -- requiring a power down and reboot to regain control -- infrequently and randomly. No errors in overnight runs of OCZ Memtest, but system eventually freezes when running Prime95. Have run Prime 95 for as long as 24 hours with no errors but generally about every 8 hours or so the system freezes requiring power down to regain control. Does not show an error has occurred even though the system is frozen. I am an experienced overclocker and have spent several weeks varying voltages, latency setting etc. with no effect. The freezes are so intermittant that it is an all day job to check whether a change in a setting has made any difference or not. The temperature rises to around 50C when running Prime95. Given the ABIT bias to report CPU temperatures about 7C high, I interpret that to mean the CPU isn't going over 43 to 44C when running Prime95. So temperature is not the problem.

I am hoping that someone with experience on a similar system may have a suggestion. I have run out of things to try.

Thanks for your help.

PS
The system is a wonderful performer -- if only it didn't freeze. SiSoft Sandra reports 10,841MIPS and 4497/7811 MFLOPS in the arithmetic benchmark and RAM bandwidths of 5642 MB/sec Int and 5613 MB/s FP.

 

[batboy]

Back off the O/C a little and see what happens. I suspect if you have to use that much vcore (which is a little higher than I'd run), then you are pushing the limits of the CPU. It's possible the RAM is maxed out too, 236 FSB at 1:1 is a lot for PC3200. You could try 5:4 with all the other settings the same just to see if it's a RAM bottleneck. Those IC7 series don't like tight RAS to CAS delay, maybe try 4?

 

[txyz]

Batboy,
Thanks for the reply. I started with CPU/Mem set at 5:4 but the OCZ folks urged me to set the ratio to 1:1 saying this memory could easily run at a FSB of 250 if I set the timings to 2.5, 3, 3, 7 instead of the SPD settings of 2.0, 2, 2, 5. I have run the system at 240MHz FSB with no greater frequency of freezes than at 236. I backed down trying to find some combination that would work. I have run the CPU at every Vcore setting from the dfault of 1.525 to 1.7, again with no real change. For todays tests I had backed the Vcore down one notch (from 1.7v) since I was trying OCZ's suggestion to move the Vdimm up to 2.8. The system just hung after running nearly five hours. I will try your suggestion of increasing the RAS to CAS delay and start Prime95 again. I have only run timings of the SPD 2.0, 2, 2, 5 and the suggested 2.5, 3, 3, 7 so perhaps your suggested change will help.

Thanks. I will let you know the outcome

 

[batboy]

So, you still got the lock ups when you tried the 5:4 ratio? Is that OCZ RAM you're using have TCCD chips? If so, then I agree, it should handle 240 FSB or more at 1:1 if you loosen the timings. I think the vcore is still a tad high, I would not go above 1.65v myself. What do you have the game acceleration set at? Disable the last two settings, if you haven't done so already.

 

[hawtrawkr]

out of morbid curiosity what psu are you using in this system?

like batboy mentioned the game accelerator settings can make a diffrence ive found auto/auto/auto/disabled/disabled works about the best but ive seen alot of people use auto/normal/auto/disabled/disabled as well.

 

[deathman20]

out of morbid curiosity what psu are you using in this system?

He said ThermalTake 480w PurePower

I know my mem settings are auto/auto/auto/disable/disable

As you can see my ram settings it can run at faster speeds, but its more if the ram can handle that speed.

Now whats the lowest voltage you can run at @ your current speed? Maybe try backing it down to 230 or 225 FSB and try and see if its working better then.

 

[batboy]

According to the first post, power supply is a new ThermalTake 480w PurePower.

 

[hawtrawkr]

bah i see it now. oddly i read the first post then skimmed over it looking for psu and managed to miss it twice.

 

[txyz]

Thanks for the replies and the suggestions.

The game settings are auto, auto, auto, disable, disable and yes -- the memory is the OCZ PC3200 platinum series 2 with the TCCD chips. I have run OCZ's version of Memtest overnight several times with no errors at all. That was why I was reluctant to follow the advice of the OCZ folks to raise the Vdimm to 2.8 volts since it appeared to me I would be stressing the memory for no good reason. They assured me that the lifetime warranty was good even if the memory was run at a Vdimm of 2.8v so I risked it for the tests I began this morning. If it doesn't make a difference I will go back to 2.7v for the next round of testing. The system hasn't hung up since I changed the RAS to CAS delay to 4 as suggested by batboy -- but it may still freeze in an overnight run of Prime95. I have monitored the PS voltages with a DMM, but do not have the measurements at the moment. The 5V and 12v rails were both slightly high, but in spec. I mainly was concerned with whether I saw any dips on any of the rails as the load changed. I did not.

 

[NuBie]

Similar Situation, HELP!!!

The goods: IC7-Max 2, Bios V.27, 3.4E Prescott, 2x 512MB HyperX UL 2225
BFG Ultra, XP90 w/ Tornado on Nexus, TT Butterfly 480W (550Peak) w/o PFCs Doh!

Been trying to get the chips to clock at their tightest timings and they fail memtest86 repeadetly. Even at stock speed of 200Mhz, they just fail. Tried 2,3,2,6 as Kingston said the 865 and 875 chipsets won't dig the tight latencies. I didn't buy it, as I've seen reports that these things can pull some small numbers even at 2225 on an IC7G (212 highest I've heard). I can't seem to get it stable even at stock speeds unless I run it at 2,3,3,6, and that sort of defeats the whole purpose for having bought this combonation in the first place. I was trying for a mild overclock using lower voltage ram and still be able to hold tight timings; say 236 x 17 @ 2336 since I've heard other peeps up to 240 or more, and at 2.5,3,3,6 they should run at 250, but I haven't the want nor the ability to go that high at 1:1. I haven't tried every combo under the sun yet, but I've had it up to 220 X 17 @ 2.5,3,3,6 for 3774Mhz stable. They won't do 2225, 2325, 2326 @ 2.6V, 2.7 or 2.8V @ 200Mhz at stock clock, w/ All PAT (or GAT) set to AUTO in the bios. Loosen to 2,3,3,6, and they seem peechy (ran for hours on end w/o mem error 1) I haven't tried to push them at all at these latencies, since I discovered they won't pull the tight timings I wanted. Should I not even worry that they don't pull the tightest timings and just focus on what I can get out of them when they are clocked higher with decent latencies? Or should I try to find a replacement??

 

[fabulouscoops]

I have a similar rig (see sig) One thing you could try is relax the mem timings to 2.5-3-3-8. Rule of thumb is the middle two numbers +2 = the high number. I get memtest86 errors with the T-ras at 7

 

[txyz]

The system has just frozen again -- after several hours of running Prime95. I will try upping the last timing number to 8 -- but it was OCZ techs (OCZ made the memory) that suggested the 2.5, 3, 3, 7 setting. I have built many systems (in fact I never purchased a built up computer in my life) and this one is the most frustrating I have ever encountered. Nothing I can do seems to make any difference in the random and infrequent freezes. I am beginning to think the behavior is not setting related -- but don't know where the hardware failure could be located. You can't simply swap out $300 worth of memory, $200 worth of CPU or $150 worth of motherboard on the chance it might be the culprit I will set the Act to Precharge value to 8 as you suggest fabulousoops and try to run overnight. The OCS SPD timings do not fit your rule: 2.0, 2, 2, 5. But I will try anything. If it hangs then, I will reset the default values with a 200FSB to see if it freezes then as well. If it does that will pretty conclusively show I have a hardware fault -- even if I can't figure out where it is. Thanks to all of you who have offered help and suggestions.

 

[Phrenetical]

dude 1. it could be your ram, even if ocz says it will do 250 doesnt mean it will, some sticks r just better than others

secondly, the 3.0c, 3.5 seems to be a common number with them for an oc, check my sig.

thats my max oc for this chip, but iv noticed this summer after a good 10-12 hrs of being on my pc will freeze like yours, its annoying, but its the price you pay with overclocking i guess. your better off backing down your oc a bit and lower that voltage of your cpu man, or its not gonna last.

i was running 1.6 through mine seemed more stable, but after 6 months it started to lock up every 2-3 hrs, i took it back and got chip switched and it s back at 3.55,

hopefully i will get my 3.2e soon and set up a new rig, with more headroom for oc'in.

 

[Sentential]

Check your RAM slots. I had a similar issue with my AS8. Slots 1+3 were 100% dead. When I moved my RAM to slots 2+4 for DC my errors went away completely. I think your situation is close to what happened to me

 

[Sentential]

Also to the forum!

 

[deathman20]

Check your RAM slots. I had a similar issue with my AS8. Slots 1+3 were 100% dead. When I moved my RAM to slots 2+4 for DC my errors went away completely. I think your situation is close to what happened to me

Oh yes very good call on that one. I know my older ABit board did a similar thing where only 2 of the 4 slots where really good to OC in. Yes the other slots worked and I could get a good OC but with the other 2 slots I even got a better OC with less voltage on the CPU also.

 

[Drop-Top]

You may have an unstable MB. I know I have had two identical asus p4p800-x boards as I screwed one up, and I was able to get almost 150 more mhz out of the chip on the second one in comparison to the first. IMHO, the most variable thing in a motherboard is how it regulates power, and unfortunately, that really is important in an OC. Even though the gauges may be saying one thing, they really haven't proven to be 100% accurate yet. Just a thought.


Drop-Top