Ocing P4 2.8

[imported_goku]

Ok I have a pretty decient fan on my P4 2.8C with thermal control and I have been wondering why it has been freezing on Prime 95 when oced to 3.5GHZ despite the fact the fan wasn't even running at full speed when it finally froze. I first tried at 3GHZ then 3.2 then 3.3 and finally 3.4 all reliable now but before when I was using cheaper ram with a 5:4 clock ratio I had difficulty getting to 3.4 with out stablity issues.

Now that I have Corsiar XMS TwinX PC4000 1GB ram (512MB Dual channel) I am able to have a 1:1 ratio and stable overclocking at 3.4. Problem is like I said when I get to 3.5 GHZ I noticed that after a while when my CPU gets to 51+C*/ run a program TO get the cpu to high temps, it freezes. These seem like the same symptoms I had before when I had the Cheaper corsair "Value ram" at a 5:4 ratio at 3.4 it made it seem like the cpu was overheating but was only like 40C*. So now I'm wondering what could be causing my system to freeze even though the fan hasn't been able to reach it's full speed. Should my CPU be crashing at 51c*+ or should it be allowed to go higher. Also remember I have MBM5 compensate like
-14*C (Meaning Subtract from what the bios says) because I hear that the ABIT bios over states temps. To add a little not, my case temp is at a cool 18*C.

Oh forgot to list specs:
Abit IC7-G (Rev 2. I think) Pentium 4 2.8C PC4000 1GB ram (Corsair XMS Twinx) Thermal take POLO 735 fan (Valcono 12 combo) ATI Radeon 9800Pro and so on. Ram is 2.8V (2.75 actually) CPU 1.7V (1.65/68 actually). Thanks very much!

 

[Navid]

To confirm that it is the CPU that is limiting, I would run memtest.
Memstest stresses both the CPU and the memory. But, it stresses the memory more!

Your VCore is high!
After making sure that it is the CPU that is limiting, I would run prime95 and gradually change VCore, and see how high I can go at each VCore setting. You will find out that there is an optimum VCore that gives the highest overclock. In other words, it is NOT true that the higher the VCOre the faster the CPU!

 

[Navid]

Originally posted by: goku2100
Ok I have a pretty decient fan on my P4 2.8C with thermal control and I have been wondering why it has been freezing on Prime 95 when oced to 3.5GHZ despite the fact the fan wasn't even running at full speed

How is "thermal control" achieved?
It may be that the threshold for increased RPM is set to a temperature higher than 51.
You can force the fan RPM to maximum and see if you can overclock more.

 

[oldfart]

Is thermal throttling disabled in the IC7 BIOS? 3.5 GHz @ 1.7 Vcore is A LOT of heat. I wouldn't run that. You are asking for trouble.

 

[imported_goku]

CPU 1.7V (1.65/68 actually) BTW, meaning that the bios setting is 1.7V but is acutally outputting 1.65/1.68. Yea I guess I should run memtest but the problem is I dont have a FDD and not sure how to make a bootable ISO of it. Yea I determined that I need 1.65 actual voltage to overclock at all! I tried 3ghz and I had stability issues when it was on default voltage. Navid your absolutely correct when you say that the thermal control limit/fan speed limit is above "freezing temps" but the problem is that the cpu won't get that high. I'm just wondering why the cpu wont get the high if the fan would support it, one time when I had the fan postioned incorrectly the bios said I was hitting 68*C idle and 75*C load and thought it was a little hot, fixed it and went down to 36*C. When I say thermal control, I mean the heat sensing diode you tape to the HSF. Oh one more thing, I have also determined that using thermal throttle is the best way to OC because when I use manual control, I some times set it too low or higher than needed. Like I said, I'm just confused why the CPU wont let it's self get hot enough TO enable the max speed/Higher OCing. I'd rather the CPU overheat then "protect" it's self.

 

[imported_goku]

Oh I Just realized that Memtest wouldn't help in the situation because of the fact I had the Ram at a 5:4 ratio assuring the ram isn't overclocked. I've already proved that it oc's well, infact I'm running the system at 3.5GHZ the one I'm typing on but when I run prime it just gets to 53*C and then "freezes" Oh the IRONY! Systems freezing because they are overheating!

 

[PhoenixOrion]

Sometimes its not the heat that causes instability.

Maybe its just that the 2.8C has cap of 240fsb and thats it.

 

[JohnAn2112]

Originally posted by: PhoenixOrion
Sometimes its not the heat that causes instability.

Maybe its just that the 2.8C has cap of 240fsb and thats it.

Right. Just because someone gets DDR500 RAM does not guarantee that they will hit 250 FSB on their system.

 

[imported_goku]

Well like I said, Im running at 250FSB right noW! Infact I just played GTA III for 4 hours at 100% CPU utilization and it would seem the system is fine. But what I am wondering is why does prime 95 crash after like 20 or 30 minutes?

 

[Navid]

How are you measuring the CPU temperature?
I hope that you are not talking about the values measured by the diode that you put on the heat sink.
You should use Motherboard Monitor 5, which uses a diode on the CPU die and a sensor chip on the motherboard to report the CPU temperature.



Why the system freezes when you run prime95.
Because prime95 stresses the CPU and the ram.

You should have a system that is stable with prime95 running for hours preferably overnight. Then, you can be sure that it will be stable under all conditions.

Your room temperature is not always the same. You run different programs on your computer.
If you only have the system stable running windows, you may not have a margin of safety.
Then, one day you turn off the air conditioning and your system freezes! You don't want that to happen when you are falshing your bios!

You may hear some people say that prim95 is an overkill and you do not have to worry about it even if your system fails under prime95. Don't listen to that. You need a safety margin of stability.

 

[imported_goku]

Ok here is how I have been determining my CPU TEMP, I have done it two ways: First I use MBM5 and look at the temps its giving me for the CPU, then (After hearing ABIT boards over exaggerate temps) I have then hooked up the thermal sensor to the HSF but not knowing the temp its detecting, so I look in MBM5 for the RPM the fan is spinning at and check the box to see at what temp would the RPM be at so it goes like: 2000RPM 20*C 5500RPM 55*C then see in MBM 5 that the fan is spinning at 4500RPM so I have determined the HSF is 45*C so I do a compensation and make the CPU temp match the RPM of the CPU Fan.

So the CPU says the temp is 59*C so I compensate -14*C and get 45*C and thats what the CPU's actual temp is. It might be a little bit off but is much closer than the MB heat sensor. Please let me know what you think of this method and also please let me know what I should do to be able to get it stable at 250MHZ FSB, 60-65*C should be fine (According to bios) or 52*C with my method.

 

[imported_goku]

Interesting, I just retried the whole Core Voltage and CPU FSB max thing and now I am running the system at 3.5GHZ with default voltage!!! I think the reason why I had to increase the voltage to 1.65 before ie because I had a combination of crappy ram and PSU. Is there anyone on the Anandtech forums who has their CPU OCed with Default voltage and how much is it OCed? Thanks!