Prime95 Question

Firegod · 8 Sep 2006 at 21:03

I've noticed that a few games have been hanging recently, the latest are the Warhammer Demo and CoH Demo.

I decided to run Prime95 anyway to see if anything failed and I ran the small FFT Test for an hour and had no problems. CPU temp was around 50.

I then went and did the in place large FFT test and it failed after about 10 minutes. Am I right in saying that it'll be the RAM causing the error? As nothing occured when running just the CPU test.

Look forward to a quick reply, I don't know much about Prime95!

Thanks!

Raze · 8 Sep 2006 at 21:19

Could be your memory...

Try this to be sure!

http://www.memtest86.com/

jpe20 · 8 Sep 2006 at 21:30

Sounds like your RAM. Memtest is not really the answer as in my experience Prime95 spots memory problems quicker than memtest. I have run memtest for 8hrs with 0 errors only for Prime95 to fail due to memory in less than an hour.

Looking at your sig. I see that you have a massive overclock in place @2.75GHz!!! This is obviously the source of your problem. I would solve it by 3 methods.

(1) Increase the RAM voltage (I do not know the limits of DDR1 RAM as I have DDR2) Look at the memory manufacturers website and set the memory voltage in the BIOS to the maximum quoted in the specifications. You can overvolt but take advice from people with similar memory/setups to you. (I have a DDR2 system)

(2) Lower your memory divider. This will turn down the RAM FSB speed making it more stable, but you do lose a little memory performance bandwidth.

(3) You can loosen the memory timings if you are running at a high memory frequency.

The final "stable" solution may be a combination of all three. First I would try method 1-3 in order.

Make a small change each time and run Prime95 (large ffts) and note how long it runs for without error. The longer it runs the more stable your system is becoming.

HTH
Jules

PS If you are a real tweaker then you could try as a final measure to fiddle around with the data read drive stengths of the memory.

jpe20 · 8 Sep 2006 at 21:35

Just as a first thing to try b4 the stuff I mention above. I would run your system at default speed (take off the overclock). Then test the memory (memtest & prime95). This will remove the possibility of the RAM being faulty in the first place.

Most likely is that the memory is unstable due to your overclocking the memory too much. Increasing the voltage will give you more stability at the same memory frequency. Lowering the memory frequency and/or looser memory timings will give more stability.

Firegod · 8 Sep 2006 at 22:07

Thanks for the help, and the link Raze!!

Just ran MemTest for an hour (did just over 2 passes) and no errors. :confused:

I'm thinking either voltage is an issue or the CPU memory controller. Thanks Jules, I'll try upping the RAM and perhaps bump the CPU up a tiny bit more. DDR is good for anything between 2.6 - 3.0v but G.Skill uses ic's that like low volts. But I'll see what happens.

At the minute the Data Strength and Skew, etc are set to Auto. May try fiddling with them.

Watch this space!

Firegod · 8 Sep 2006 at 23:05

Looks like it was a few things...

I've increased the CPU voltage from 1.582v -> 1.610v and that has worked well. But also slapped a 5:6 divider on the system which does the following:-

CPU - 275x10 - 2.75Ghz

MEM - 229 @8-4-4-3.0

So a few things I can do over the weekend. See if the increased voltage can allow for a few more Mhz on the CPU. (At load atm it is 52 degrees)

And also, probably more importantly, try and tighten up the timings as the RAM is running ~20Mhz slower from stock.

Thanks for the help!

jpe20 · 9 Sep 2006 at 11:42

Just a note of caution on the level of volts you are running on Vcore. 1.61V(+ any offset added by the mobo) is on the high side. You are running the risk of shortening the life of your CPU. The temps you report ... are they with both cores loaded???? (most games tend to be single core with a few exceptions). My personal limit was 1.6V and 62C load temp which is high due to my willingless to replace the CPU (only 100quid or so) if it went bang!!!!!

I would not be happy running over 1.6V Vcore 24/7 unless you had some serious cooling.

Not a helpful post I fear as I am sure you are chuffed to run 2.75GHz!!

Anyhow I am off to read about the Scythe Mine cooler ..looks a real possibility!!!

Jules

Nami · 9 Sep 2006 at 15:38

Tbh I'd spend the time to read the overclocking sticky in this forum and then start ** OC from scratch. It's highly likely that the instability is coming from the OC, but which component? Well you'll never know until you take the time to figure it out by testing each thing individually. Note ** current BIOS settings and start again, making sure to find the ceiling overclock for each component (mobo, CPU and RAM). Then try to combine those ceiling overclocks together with the best compromise, and stablity test.

Firegod · 9 Sep 2006 at 15:49

Cheers peeps.

All is well since the voltage bump so I think it's the mem controller on the CPU causing the problem. I played the CoH demo again today and it never crashed once.

Jules, I have been running this overclock speed for a good 2+ years, but I'm guessing the problem has only occured now as I upgrade recently to 2Gb of RAM. Previously I have not experienced any issues. So the CPU is still giving as good as it gets, and I'm hoping she'll be running like this for another year. But I defo won't be going higher on the VCore as I think the air cooling has reached it's limit at just over 50 degrees on load. Oh, and it's only a single core CPU mate.

Nami, if the problem re-occurs I think that's a sound plan to start again. Basically I have put in the 2Gb of RAM and just tweaked the old settings to get it working, I did not start from scratch to find the best configuration. So possibly may do that if things start to go wrong.

Thanks for the replies again!