New 1090t 4ghz funbox

QuietIce said:

With water cooling (I emphasize that for others reading this, not you, Peeved Kitten ) I'd say 1.525v (AMD safe + 0.05v) is what I'd call good - maybe even a hair more depending on your load core temp ...

Click to expand...

Ok, I'm going to play with it some more tonight. I will be on here while doing that to post pics & results (and to ask for help most likely!) I'd like to think that with my cooling I could aim for 4.4 or 4.5. With LLC on my idle volts are nice and low, since the BIOS update to the CHIVF it has been rock solid on voltage handling. I also upgraded to a 950W power supply which probably helps a great deal.

Can you link me to the beta bios? I can't see it on Dolk's page - not for the Crosshair IV...

Running a Phenom II at 1.5v 24/7 will damage the CPU. I'd recommend keeping it around 1.45v for 24/7 usage.

Dolk said:

Running a Phenom II at 1.5v 24/7 will damage the CPU. I'd recommend keeping it around 1.45v for 24/7 usage.

Click to expand...

I wish my system was 24/7 but I cant afford to run it that much, it is on probably 4 hours a day at max, I'm not too worried about it. right now my volts are around 1.47-1.48 I still have a bit of leeway to the 1.5v limit, although LLC occasionally shows 1.5 on CPUZ.

Is 1102 the latest bios for the board?

Please could you tell me what BIOS are you using for your clocks, and which settings?

gcwebbyuk said:

Is 1102 the latest bios for the board?

Please could you tell me what BIOS are you using for your clocks, and which settings?

Click to expand...

I'm using 1102, I'm having issues getting it to boot at 4.3 ghz

Here are some stats at 4.2 ghz: these are temps and stats after 30 mins of Prime95 and Furmark.

Peeved Kitten - what was your CPU temp, as that is what you need to keep under 55/62c, not the cores - as per quietice and my posts above.

How is your coretemp minimum 12c ? What is your ambient temp there? Seems way too low? Unless may be the PC is in a shed out in the freezing cold?

Core temps are inaccurate. You need to go off the cpu socket temp.

crunchie said:

Core temps are inaccurate. You need to go off the cpu socket temp.

Click to expand...

On a few systems, yes, the core temp is inaccurate - and when they are it's pretty obvious, like this one.

For most systems the core temp is a much better choice ...

From what I have read around various forums, the cpu socket temp seems to be the most accurate reading to go off.
If the CT is out, how far out is it?
No-one (that I have seen) has appeared to have any problems with inaccurate cpu socket temp readings though.
If core temps are to be used though, I have found it better to boot the pc and get a stable temp, then boot into the bios and check the temp there, then back into windows to see the difference.
If you use CoreTemp (software) you can then adjust the offset.

I was put off using core temps as they are lower than ambient. I was advised to use the CPU temp and keep it under 62c, or realistically high 50s for stability. Someone once linked me to an AMD article that explained this too.

Using the motherboard's CPU socket temp is all well and fine if you're running a horizontal heatsink or a tower heatsink with "normal" case airflow (and a not-cheap motherboard). Get outside of that margin and the CPU socket temp gets farther and farther away from the core temp. If you would like to test this take your stock set-up, add water cooling to the CPU (preferably with an external rad), turn off the case exhaust fan and THEN take a look at the CPU socket temp. That's a very extreme case used to point out a simple fact; the CPU's internal sensors are right there while the CPU socket temp can be influenced by outside factors like the NB, a video card (in the top slot), RAM, and MOSFET heat.

If you look at the technical data from AMD the internal sensors don't actually measure "temperature" per se. A "55°C core temp" is not necessarily 55°C - but for our purposes that doesn't really matter. For the CPU's internal thermal protection circuitry (not motherboard) the core temp is all that does matter because that's the sensor it monitors. For over-clocking the actual temp doesn't matter either because we've found from observation that somewhere around a 55°C load core temp the CPU tends to become unstable. In both cases it's the core temp, not the socket temp, that's being used as a reference.


Yes, there are some systems where the core temp doesn't seem to be relayed correctly by the motherboard (and it tends to be ASUS boards for some reason). With those systems the best thing to do is get a good ambient temp at idle (temp IN at whatever heatsink/rad you're using) and add ~3-10°C to it for the core temp (depending on how good your heatsink/WC loop is). You can also use the CPU temp but if you've got a cheap board don't expect that to be particularly accurate, either ...

Those numbers are not fake or wrong. My machine was outside in the garage, (it's where all my media toys are) Core temp is VERY accurate at least for my board/processor. Last night I actually had ambients in the 8-9c range and was able to get my temps about that low on idle. Going by the CPU socket sensor I run 44c at full load. I also managed to get 4.3 stable last night. I was unable to increase the multiplier any more, every attempt was a failure. I moved the FSB up and had no issue whatsoever, I didn't even have to touch the voltage.

I'd like to mention I did a slight motherboard mod. I ground down the standoffs on the CHIVF NB heatsink and re-set it in place with real TIM. My NB temp dropped 21c at idle just by doing this. Heat isn't an issue with my rig, last night at it's coldest I was running about 25c at full load. I use core temp hardware monitor and Asus Probe II to watch my temps/voltages.

What should I check when prime runs for about 25 mins and then gives a memory management error? Is that timings or something else?

Peeved Kitten said:

Those numbers are not fake or wrong. My machine was outside in the garage, (it's where all my media toys are) Core temp is VERY accurate at least for my board/processor. Last night I actually had ambients in the 8-9c range and was able to get my temps about that low on idle. Going by the CPU socket sensor I run 44c at full load. I also managed to get 4.3 stable last night. I was unable to increase the multiplier any more, every attempt was a failure. I moved the FSB up and had no issue whatsoever, I didn't even have to touch the voltage.

Click to expand...

That's good to hear - thanks for clarifying!

I thought ASUS had fixed most of those reporting errors by now.

Peeved Kitten said:

I'd like to mention I did a slight motherboard mod. I ground down the standoffs on the CHIVF NB heatsink and re-set it in place with real TIM. My NB temp dropped 21c at idle just by doing this. Heat isn't an issue with my rig, last night at it's coldest I was running about 25c at full load. I use core temp hardware monitor and Asus Probe II to watch my temps/voltages.

Click to expand...

That was a good idea - might be useful for other boards as well with or without NB heatsink problems.

Peeved Kitten said:

What should I check when prime runs for about 25 mins and then gives a memory management error? Is that timings or something else?

Click to expand...

I've never heard of that error. It could be either RAM or cpuNB getting in the way ...

QuietIce said:

That's good to hear - thanks for clarifying!

I thought ASUS had fixed most of those reporting errors by now.

That was a good idea - might be useful for other boards as well with or without NB heatsink problems.

I've never heard of that error. It could be either RAM or cpuNB getting in the way ...

Click to expand...

If I can get it to do it again I'll grab a pic. It bluescreens and the message that pops up is Memory_Management Error. I think my cpu-nb freq is at about 3k with my ht at about 2600. I ran for right at 30 mins stable at 4.3 and the BSod'd with that error.

As for being asked for BIOS settings here is my view on that:
I can give them to you but unless you have a truly identical chip (remember every piece of silicon is different) They may not work. Couple that with the cooling etc. and it could be unbootable for you. I dont turn my cpu on until it's below 20 ambient in my room or the garage. I operate at the coolest temps I can get. I intend to use the BIOS feature to save my profiles. All of my heavy duty OC's are saved once stable. I may not run 24/7 at 4.3, but I can reboot activate it and overclock happily.

When I get home I'll try to post a pic or 2 of the bios, just take it all with a grain of salt.

4.3 seems to be getting stable!

Ok! I have finally gotten to the 30 minute mark of p95 stability testing at 4.3 ghz.

I'll post up more details in a bit but here's a screeny!

Peeved Kitten said:

Ok! I have finally gotten to the 30 minute mark of p95 stability testing at 4.3 ghz.

I'll post up more details in a bit but here's a screeny!

Click to expand...

Nice going . Any reason to have the HT link frequency so high? Have you noticed any performance increase at that speed, or is it for stability?
Everywhere I have read says that increasing above 2100Mhz is of no benefit. What nb are you running?
There seems to be a huge difference there between the cpu socket temp and the core temps. Mine has a delta of about 7C under load, with the cores reading lower.
Have a look here; http://www.overclock.net/11018198-post2078.html and here; http://www.alcpu.com/forums/viewtopic.php?f=63&t=892

Why is the temperature of my Phenom based processor lower than the ambient temperature?
Starting with the Phenoms, AMD's digital sensor no longer reports an absolute temperature value anymore, but a reading with a certain offset, which is unknown. It is estimated that this offset is between 10 - 15c.

Click to expand...

At the vcore you are running and the cpu frequency, it is extremely doubtful that your max load temps are 33C.
What are your ambients?
I have just run Prime for about 25 minutes @ 4050Mhz with 1.375vcore and am getting 44C at the cpu socket and 37C core temp in HWMonitor with about 20C ambients.
I am running a car heater core hooked up with an EK-CoolStream RAD XTC (140)

crunchie said:

There seems to be a huge difference there between the cpu socket temp and the core temps. Mine has a delta of about 7C under load, with the cores reading lower. Have a look here; http://www.overclock.net/11018198-post2078.html and here; http://www.alcpu.com/forums/viewtopic.php?f=63&t=892

Click to expand...

Unless you have very good case airflow the CPU temp will likely read higher with water cooling. There used to be a fan right above the CPU to mix the air up in that region of the case - what's there now?

My load core temp and CPU temp were very close to the same when I first set up my 940BE using only 1.36 vCore. I used a TRUE with a Panaflo medium and an extra S-Flex F near the base to keep the NB, MOSFETs, and RAM cooled down. (With no case I needed the extra fan.) If we assume (and, IMO, it's a BIG assumption) that "everybody" is right and the core temps are reading 10-15°C lower then they're all saying a TRUE and Panaflo is junk cooling since the cores were running 10-15°C hotter (according to "them") than the thermal sensor sitting right under the CPU.

With all due respect to The Coolest, I don't buy it.

What are your ambients?

Click to expand...

He posted above that they were 8-9°C on one run. The computer is in an out-building/garage ...