3.4Ghz Prescott failing Prime95 1st Test

joetechman

Distinguished
Dec 27, 2009
13
0
18,510
Hello everyone,

I just switched my CPU in an old machine of mine. I put in a 3.4Ghz Prescott to replace a 2.4Ghz Northwood. This machine is running 32bit Windows 7 Ultimate. The Mobo is an Asus P4P800 Deluxe, and the ram is 3GB Kingston DDR400 in dual channel mode. I ran Prime95 while having HWMonitor running to see temperatures and Prime95 failed in under two minutes with 1 error. The error says rounding 0.5, expected less than 0.4. I tried running Prime95 a second time and the same thing happened. I did the blend test and selected 1 test thread. The CPU idles b/w 50C and 54C and the highest temp it hit during the testing was 69C. The BIOS settings are all set to Auto, and the RAM timings shown in CPU-Z are 3,3,3,8 at 2.5V. Cpu core voltage is approx 1.392. I used to have 1GB of ram (2x512MB) in this machine and about 2 months ago I added in the other 2GB (2x1GB, this was while the CPU was the northwood 2.4ghz. I ran memtest86+ back then when I added the ram and it passed it fine with no errors.

I hope somebody can help me out here. Thanks everyone.

Edit: Heres a screenshot of CPU-Z, HWMonitor, & Prime95.
prime.jpg
 
Solution
Well if your not seeing any errors after a few hours then i say your good to go. There nothing really you can go if it not showing now. Just leave it as is. If it comes up again in the future just let us know.

werxen

Distinguished
Sep 26, 2008
1,331
0
19,310
You need to recheck the ram - that would be my first guess. If it passes - next step is to see if the CPU itself is forked. Try lowering the frequency or multiplier of the chip (i am not too sure how to OC P4s personally) but lower it to around 3.0 gighz. If it fails, try lowering it some more while upping the voltage. If it still fails yet your ram is fine then I am out of ideas. Make sure you memtest for 24+ hours.
 

gelde3001

Distinguished
Aug 5, 2007
47
0
18,540
Right then its almost certainly your cpu not being cooled properly - Id reseat the heatsink and use a small ammount of fresh paste, preferably arctic silver or better on that cpu, its a 120w+ processsor !.

As you have already checked the ram its unlikley to be the problem unless its not seated quite right - try unplugging all and reseating maybe, but 2 passes on memtest is enough to make sure the ram is ok - 24hrs is more for a brand new test but to re-check 2 passes is more than enough, one to heat up the ram and the second will then find any probs usually.

If the cpu isnt at about 35c idle then Id suggest a new heatsink or simply put the northwood back in as clock for clock its faster than the presshot and in all honesty wont make that much performance difference really, unless one has ht and the other doesnt of course as that can help esp on win7 / vista. Of course you could overclock that northwood to at least 2.8ghz and then it would likely be closer to the 3.4 and still run cooler.
 

joetechman

Distinguished
Dec 27, 2009
13
0
18,510
From what I've read 50-55C idle is actually pretty normal for the 3.4ghz prescott since it is a very hot running CPU. And I already put on a small amount of fresh arctic silver 5 when I seated the heatsink today.
 

sabot00

Distinguished
May 4, 2008
2,387
0
19,860
Look's like we've got a PresHot problem on our hands.
You can't change the multiplier unless the chip is a Pentium 4 EE.

You OC just like any other non-multiplier unlocked chip.
Change the base clock.

Check your temps.
 

joetechman

Distinguished
Dec 27, 2009
13
0
18,510
My BIOS can only set DRAM Frequency to 266, 320, 400, or Auto. It can only set DDR Reference Voltage to Auto, 2.55, 2.65, 2.75, 2.85. I have AMI BIOS v.2.51 on a P4P800 Deluxe Motherboard.

@apache_lives / everyone: Should I try 320, 2.65? or 2.55?

@sabot00 / everyone: Is the "CPU External Frequency (Mhz)" the equivalent of "base clock"? Because right now in my BIOS, CPU External Freq is set at 200mhz and it wont allow me to change it. When I highlight it and press enter, nothing happens.

@everyone: I will run memtest86+ after attempting the previous two suggestions. I've also added a screenshot to the first post.
 
As for temp, there fine. Thats common for a cpu under netburst architecture. (i know i still have a prescott.) Unless you underclock and undervolt. you'll wont get under 50c on stock cooling.

My question is, are all of these ram cas latency the same? Maybe a dumb question but it could be what causing the problem.

 

joetechman

Distinguished
Dec 27, 2009
13
0
18,510


Okay I just opened up my box and pulled out my RAM sticks to double check. I made one error and in which I said the all my sticks are 1GB sticks totalling in 4GB. I actually have 3GB of ram in total, two 1gb sticks and two 512mb sticks. Two of my ram sticks (the original two) are Kingston KVR400X64C3A/512 and the two new ones are KVR400X64C3AK2/2G. So two of them are 512MB sticks plugged into the black ram slots and the other two are 1GB sticks plugged into the blue ram slots (all running in dual channel mode).

@warmon6: so yes the cas latency for all 4 sticks is the same. They are all rated CL3 (3-3-3) and 2.6V.
 

whitewhale1

Distinguished
Apr 7, 2009
20
0
18,520
i have had exactly the same problem with my old system

while i havent solved it yet i think this comes when all 4 ram slots are occupied

its not a temp thing those temps are normal

try just using 2gb of ram and see if it still happens
 

werxen

Distinguished
Sep 26, 2008
1,331
0
19,310
Obviously its a RAM or CPU issue or else the bloody thing would not even boot. The thing is it has to be either 1) his CPU frequency is not stable anymore and his cpu is slowly dieing (very very very very very rare) or 2) his ram is forked. Simple problem really...
 

joetechman

Distinguished
Dec 27, 2009
13
0
18,510
Okay... so Ive been running memtest86+ v.2.11 for almost 12 hours now, and it is now on the 10th pass. So far, there are 0 errors.

*Update*: It is has now been running for almost 20 hours. Pass 17 and still 0 errors. I am now stopping the test and will try some of the other suggestions.
 

joetechman

Distinguished
Dec 27, 2009
13
0
18,510


Back with more.... I tried setting the DRAM to 320MHz and the voltage to 2.65v. I then ran memtest86+ for one full pass with 0 errors. I booted to Windows 7, and ran Prime95. CPU temps had no change from the previous temps, and this time I got Prime 95 through the first four 1024k tests and the first three 8k tests before manually stopping it. Before when it was failing it would fail one the first or second test, so I figured the CPU/RAM is probably okay at 320,2.65.

So then I rebooted and went back to the BIOS settings. This time I changed the settings back to the way they were originally, which is everything on auto. So 400Mhz DRAM and voltage of either 2.5v or 2.6v (theres no way to actually check what its at since it is on auto). I tried Prime95 again, and this time it is no longer failing on the first few tests. So far, I have passed the four 1024k tests and the four 8k tests without errors and am now on the second 10k test.

I'm not quite sure why it hasn't failed the Prime95 tests yet, because originally it was failing every time on the first few tests and I had tried like 5 or 6 times.

Edit: On second thought...When I looked inside my machine to check the RAM model #'s, I pulled out 1 512mb stick and one 1gb stick to read the stickers on them. Could it be that when i re-seated them, that did the trick??
 

joetechman

Distinguished
Dec 27, 2009
13
0
18,510
*Update*

I left Prime95 running all night and it still has not had a single error. The highest temperature the Prescott has hit is 70C. Could it be that when I re-seated the two of the four RAM sticks the other day(to check the sticker), this fixed whatever issue was causing Prime95 to fail? Should I still be weary of a possible CPU and/or RAM issue? Should I try something else for testing?
 

joetechman

Distinguished
Dec 27, 2009
13
0
18,510


 
Well if your not seeing any errors after a few hours then i say your good to go. There nothing really you can go if it not showing now. Just leave it as is. If it comes up again in the future just let us know.

 
Solution