Phenom II X4 940 Prime95 issues

[OVerLoRDI]

I recently upgraded to a Phenom II X4 940, on a Gigabyte mobo, with 4gb of GSkill DDR2 1066. Now I'm pretty certain I have not done anything stupid in my configuration of the hardware. My Ram is configured according to GSkills Timing and voltage config, the CPU is not overclocked, running at stock voltage and never gets above 55C under 100% load (using that massive Xigmatek cooler). I loaded up Prime95 64bit and chose the torture test of, "maximum heat, maximum power consumption" to test my cooler. Everytime I have tried it, within 10 minutes I get a rounding error. Now this worries me, stock voltage, stock speed, cpu temps very cool, and I'm getting rounding errors. I'm not sure what could be causing this, there are three things that appear to be likely the cause.

1. Bad CPU
2. Bad RAM
3. PSU not powerful enough

1. I do not think it is a Bad CPU, I'm not sure how often you can get bad CPUs, but it seems like I would see problems in other areas not just a stress test. The computer has not given me any blue screens, or any program instability. Which makes me thing the PSU might not be supplying enough juice (with all four cores at 100% using max power) and as a result the CPU undervolts and errors.

2. Again I have not experienced any other issues with the ram in user space, programs all fine, no blue screens or crashing.

3. My PSU is a 500watt coolmax thing I got at fry's in desperation when my 550watt antec died last summer right before I was about to move. The 500watt coolmax has worked perfectly day in and day out until this upgrade. My current computer configuration is as follows:

Phenom II X4 940
4gb Gskill DDR2 1066
Geforce 8800GTS 640mb
Geforce 7600GT, for 2nd and 3rd monitor so I can have the 8800 in 3d mode and the 7600 in 2d and be able to watch tv shows/movies while playing games
5 7200rpm hard drives of varing size and age
1 5400rpm laptop hard drive
2 DVD burners
and some fans, sound card, cpu cooler.
running windows xp 64bit.

So you see I have a couple of paths to go down. My current most likely theory is the PSU, when i crank the quad core to max it pushes the psu a bit too far and it can't quite supply enough juice. Sadly I don't have another powersupply with which to test this theory, so that makes troubleshooting that problem difficult. Do any of you guys have any ideas on how to track down the heart of my problem a bit more? Maybe ways to isolate and test individual components better? Thanks in advance guys.

OVerLoRDI

 

[richierich1212]

Which gigabyte mobo?

 

[OVerLoRDI]

GA-MA790X-UD4P, with latest bios

 

[richierich1212]

Ok bump up your NB voltage to 1.2v, because they undervolt the NB. And don't let the BIOS automatically set your RAM timings, you should put them in yourself. What is your G.Skill memory rated @ and what voltage?

 

[OVerLoRDI]

...weird, I hadn't considered the NB but that would explain things. I'll give it a shot when I get home.

 

[OVerLoRDI]

I set the memory timings myself according to GSkills specs. The ram is DDR21066 5-5-5-15 @ 2.0-2.1v Currently I have it at 2.1v at those settings.

 

[BTRY B 529th FA BN]

Make sure the vcore is within AMD's specs

 

[geokilla]

Take everything out except for the basics required for your computer to boot properly. Then boot into Windows and try stress testing again. If it still errors, then it might not be your PSU.

Did you check your RAM wit Memtest?

 

[OVerLoRDI]

I did run memtest via bootable CD but it seemed not to detect my ram properly and crashed after a while so I wasn't sure if the test held much water. I think I might have accidentally burned an old version of memtest, before it became memtest+

I'll double check the vcore, but like I said everything is stock so unless the bios is setting it wrong it shouldn't be an issue. 1.35v for phenom II correct?

 

[BTRY B 529th FA BN]

Usually test 5 will find the errors quicker, but when you check the ram via memtest x.xx run the ram at 1:1 at it's fastest rated speed: eg. DDR2 1000 ala 500fsb

 

[OVerLoRDI]

Yeah I burned memtest, not memtest+. I downloaded a burned memtest+, hopefully it will function better and I can do a full run.

 

[richierich1212]

Any Luck?

 

[Ben90]

Could just be a bad part; i had a AMD 6000+ that failed prime @ 205x15 (a 2.5% overclock)

 

[OVerLoRDI]

I haven't had much luck. I ran memtest and it seems to always freeze (not error, I have seen memtest report bad ram) about 9-10 minutes in so I think my problem is likely my ram. I've tried various configs and voltages and no luck so far. The ram is DDR2 1066, but I'm dropping it down to 800 with everything on auto and seeing if that yields any luck.

Next I'll try the sticks individually or try swapping the ram out completely since I have an extra stick of DDR2 that I know is good.

The great surprise is that even though my comp freezes during memtest and fails prime95 stress tests it played Civ 4, for about 7 hours last night while bittorrenting and decoding high def without a single issue. I'm starting to wonder if perhaps the stress testing software is having trouble with the Phenom II memory controller or something...

 

[Arkaign]

Check to see what others are running your memory at. It might require 1.9v or 2.0v to be stable.

EDIT : 2.1V for some people to be stable at 1066, looking at forums and newegg reviews.

 

[OVerLoRDI]

With the ram set back at DDR2 800 at auto voltage 5-5-5-15 all voltages auto on NB and CPU it works perfectly. Memtest successfully completed all the way through and prime95 has been running for an hour heating my room with out a problem. So it is the ram.

Arkaign, I have tried 2.1v on the ram and no luck.

 

[OVerLoRDI]

I found a sticky on the Gskill forums showing the proper config for my motherboard and ram combo. Their config is almost identical to what I had before so hopefully the small difference will fix it. The one thing they do say is that the ram should be installed in slot 1 and 2.... Which I cannot do since my massive xigma heatsink blocks slot 1. In theory using slot 3 and 4 over slot 1 and 2 shouldn't matter right?

Edit: No luck, froze same spot, about 9-10 minutes in using the settings from the sticky in the GSkill forum. Looks like for now I'll be running DDR2 800 waiting for a bios update from Gigabyte since this ram and motherboard doesn't like to play together properly.

 

[Idontcare]

Another thing you could do to eliminate concerns over the possibility of a defective cpu is to significantly underclock it while keeping the volts up just to see if the rounding error goes away.

Based on your description of the test you selected in the OP I presume you are testing with what folks around here refer to as "small FFT", which indeed places the most stress on your CPU and its on-die caches.

If you are concerned with the possibility of your ram or NB being the culprit then you should really be testing with the "large FFT" torture test in which the matrix sizes become large enough that they no longer can reside fully within the on-die cache of your cpu.

 

[OVerLoRDI]

It isn't the CPU or the powersupply it is the ram and motherboard combination. The GSkill ram I have uses 5-5-5-15 timings on the standard 4 but there is one of the more obscure timings on the ram that needs to be at 48T where the motherboard can only go as high as 42T. So currently I am running the ram at 5-7-7-20 at 1066 and everything is rock solid. GSkill says in their forum that they are aware of the issue and have alerted gigabyte, hopefully I will see a bios update in the future, if not I'm not going to lose sleep over 5-5-5-15 vs 5-7-7-20.

Thanks for everyones help in the first place, always frustrating putting a new machine together and having to track something like this down.

 

[cusideabelincoln]

Yeh I've had troubles with G.Skill and Gigabyte, specifically the parts listed in my signature. When I first built the machine it would not post at all. I had to borrow a different set of RAM (G.Skill DDR2-800) and update the BIOS before I got them to play nice.

 

[Markfw]

Personally, I think that a 500 watt PSU is not enough for all of that hardware. Also, coolamx does not seem to be a highly rated PSU. A good 600-700 watt Seasonic or Fortron (OCZ now uses Fortron, and rebadges them) is my recomendation. If I am right, take out all but one hard drive, one video card, and one optical drive, and see if it makes any difference. If it does, then that is your problem.

 

[OVerLoRDI]

Well now that I have the issue locked down everything is working fine. No games have had any trouble or any Prime95, so I think the PSU is magically doing okay. But you are right, a good PSU can be had for not that much and it wouldn't be a bad idea to upgrade. I'll keep my eye out on newegg. How do you feel about Corsair PSUs?

 

[Markfw]

Not sure, I don't own one, but I think they are better than what you have. I have not heard anything bad about them. I LOVE My fortron 700 (>86% efficiency)