Prime95 now failing right off the bat

[kronnyq]

e8400 w/ GA-P35DS3L mb, 4gb OCZ platinum @ 850mhz.

425x9 for 3.8Ghz, 1.4Vcore for 1.36 actual.

Some games have been crashing/ freezing computer lately so I decided to look into it.

When I first built this PC I tweaked it for about a week for overclocking, once I had it runnin good ran Prime95 for atleast 8-10 hours without a prob.

Using the old Prime95, it seems to work good but that's only because it spreads 100% load between the 2 cores. The newer Prime95, it puts 100% load on both cores, and the stress test immediately fails on the second core but keeps going on the first.

Granted the 250mm fan on the side of my cast is jammed full of dust and theres some dust in my CPU heatsink, I don't see how heat could be an issue because it's about 58F in this room, and CPU doesn't seem to top out more than 54C under max load according to speedfan.

I have noticed that the temp on the 2nd core is normally 3-5 degrees higher than Core1 temp. But still it's not even getting that hot, and even if speedfan is giving the wrong readings Coretemp tells me 56C max, which is still quite far from the danger zone for this chip.

Thought I might be having NB overheating issues again because thats what did it in the past, so I notched NB OC down from +.1v to normal, but no help there.

Have thermal throttling and all that junk turned off in Bios, ran memtest fine with no errors. Anyone got any ideas????

 

[BTRY B 529th FA BN]

What is the latest version of Prime?

 

[error8]

Originally posted by: kronnyq


Thought I might be having NB overheating issues again because thats what did it in the past, so I notched NB OC down from +.1v to normal, but no help there.

This is a mistake. You should up your NB voltage a step higher, instead of getting it lower. For 425 fsb you should be stable at around 1.4V V on your NB.

The temps are not to blame for your failing oc. You have one of the following: the cpu degraded and it's not stable at 3.8 ghz for that particular voltage and probably needs higher vcore to be kept stable or higher VTT. One thing that looks weird is that you have a very high voltage for 3,8 ghz on your E8400. Normally, 1.4 V in bios is used for 4 ghz or more, on these chips.
Also, your ram might be guilty and this is the first place I'll start looking. Do a memtest and see if you encounter errors. Maybe the ram is faulty or it needs a higher voltage to remain stable at 850 mhz.
Tell us exactly what voltages you set in bios.

 

[brencat]

Originally posted by: error8
The temps are not to blame for your failing oc. You have one of the following: the cpu degraded and it's not stable at 3.8 ghz for that particular voltage and probably needs higher vcore to be kept stable or higher VTT. One thing that looks weird is that you have a very high voltage for 3,8 ghz on your E8400. Normally, 1.4 V in bios is used for 4 ghz or more, on these chips.
Also, your ram might be also guilty and this is the first place I'll start looking. Do a memtest and see if you encounter errors. Maybe the ram is faulty or it needs a higher voltage to remain stable at 850 mhz.
Tell us exactly what voltages you set in bios.

I second this...and especially since you're using OCZ RAM too.

(Sorry, I'm just not a fan -- great PSUs, mediocre RAM broadly)

 

[Idontcare]

Originally posted by: kronnyq
Using the old Prime95, it seems to work good but that's only because it spreads 100% load between the 2 cores. The newer Prime95, it puts 100% load on both cores, and the stress test immediately fails on the second core but keeps going on the first.

I can't underscore enough just how wrong this approach was for you to take. The entire point of stress testing your rig is the part where it gets stressed.

Running a single thread (as you were doing) on an overclocked dual-core "so it spreads the load around" is the absolute opposite of stress testing.

The reason the new versions of Prime95 creates a thread for each core is intentional...given that the idea is to stress each core to its max.

Based on the info you provide here in regards to how you determined your system was stable in the past my conclusion is everything you thought was stable was probably not stable and it is really beginning to show.

Getting to stability (true stability when both cores are 100% loaded) is going to take more than a single tweak on the NB as you prolly were not really stable months ago to begin with.

Use the latest prime95, set you clocks back to stock, and start tweaking up your OC from the bottom-up just as if you purchased the system brand new today and you are OC'ing for the very first time.

If all your prior OC knowledge on this rig was generated by intentionally restricting Prime95 stress testing to just 1 thread on a dual-core rig then all that knowledge needs to be dumped and started over again.

 

[error8]

deleted

 

[kronnyq]

Originally posted by: error8


This is a mistake. You should up your NB voltage a step higher, instead of getting it lower. For 425 fsb you should be stable at around 1.4V V on your NB.

The temps are not to blame for your failing oc. You have one of the following: the cpu degraded and it's not stable at 3.8 ghz for that particular voltage and probably needs higher vcore to be kept stable or higher VTT. One thing that looks weird is that you have a very high voltage for 3,8 ghz on your E8400. Normally, 1.4 V in bios is used for 4 ghz or more, on these chips.
Also, your ram might be guilty and this is the first place I'll start looking. Do a memtest and see if you encounter errors. Maybe the ram is faulty or it needs a higher voltage to remain stable at 850 mhz.
Tell us exactly what voltages you set in bios.

The reason I had to go 1.4v is because I got one of the absolute worse batches of this chip out there, and after much testing I could only hit as high as 3.8ghz with 1.4v. Other voltages are as follows: Fsb +.2v, PCE-E normal, DDR +.3v (as said by OCZ), NB back to +.1v. When I had NB at +.2v it was overheating and causing freezups, I guess the cooling on this board for the NB really sucks.

I did run Memtest for about 15 minutes or so, not a full test but anytime I've had faulty ram it usually spits out errors within the first minute or 2. Ram is rated as 1066 and have it set at 5-5-5-18.

Now heres some interesting info....I put FSB down from 425 to 400 and now Prime has gone through Test 1 1024k already and half way thru test 2 with no errors. I'm starting to believe your theory about the CPU degrading...but why would this happen, is it because I have the voltage set so high? I've only had the PC for about 5-6 months now. I'd be scared to run it more than 1.4v even though I'm only peakin out at 52C atm.

It could be the Vdroop killin me as well...Cpu-Z reporting 1.328v on max load. I did see a Vdroop mod for this board but I think I have to rip my motherboard out for that and I'm not sure if I want to deal with it.

Probably won't see any real world difference running at 3.6ghz compared to 3.8ghz but being an OC junky it makes me :brokenheart:

 

[error8]

You should let memtest do its thing, at least one pass, to determine if the ram was indeed the culprit. You do have an amazing vdrop, the largest I've ever heard of, so probably 1,4 V is not that high, if you get 1.32 V under load,so maybe upping the vcore a bit won't kill anything. But before going into this, test the ram first, so we can rule that out.

You haven't mentioned your VTT voltage. E8XXX series are very sensitive when overclocking, at that voltage.

 

[kronnyq]

Not sure which is the VTT voltage here:

Speedfan

I will be runnin a full Memtest later on tonight