Finetuning the OC on my e8400 questions

andrei3333

Senior member
Jan 31, 2008
449
0
0
Specs in sig

There are 4 settings that i am curious about in hardware monitor
1)CPU temp
2)system Temp
3)Core0
4)Core1
Which setting means what ?

At idle:
1) 42C
2) 32C
3) 56C
4) 56C

During Orthos stress testing:

1) 43C
2) 59C
3) 72-76C
4) 72-76C

NO ERRRORS

IN BIOS :
FSB - 444MHz (x9=3996GHz)
Mem multiplier is 2.0 (1:1 ratio)
mem freq - 888
mem timings 5-5-5-15
DDR2 Voltage - +0.3 (which should result in 2.1V)
CPU Voltage 1.306V
MCH (NB) +0.1
FSB Voltage +0.1


IN WINDOWS
CPU Voltage 1.232-1.248V as per CPUID
CPU VOLTAGE 1.23-1.25V as per HW Monitor
DDR Voltage 2.10 - 2.14V as per HW Monitor

i guess my CPU is running around 74C at load while stress testing for i duno 30 minutes or so, is this too high ?

I do not want to go past 4GHz, so this is where i draw the line, so what should i tighten up?

I do not understand why in HW Monitor the ITE IT87 tab shows one CPU temperature and the Intel Core 2 Duo E8400 tab shows both core temperatures..... ;-(
 

andrei3333

Senior member
Jan 31, 2008
449
0
0
Update 1:

I tried the program called Core Temp and its showing the Core0, 1 temps 5C LOWER than HW Monitor is showing

Which program should i trust ?

(after a few steady minutes in Orthos its showing me 65-67C MAX and a VID of 1.2125 while HW Monitor is reporting 70-72C) i also lowered the voltage in BIOS to 1.275 and CPUZ is showing 1.216 - 1.232 so far everything is stable as i type this
 

Tempered81

Diamond Member
Jan 29, 2007
6,374
1
81
Open Coretemp, Realtemp, Orthos, and Cpu-Z simultaneously. Open cpu-z 4 times and display each tab of information. Start Orthos & after about 5-10 minutes take a screenshot. Post that screenie up here, and we can help you.

You want a delta to tjmax > 25C, IMO, you want stability, and actual vcore of < 1.400v

 

BonzaiDuck

Lifer
Jun 30, 2004
16,832
2,148
126
Originally posted by: andrei3333
Update 1:

I tried the program called Core Temp and its showing the Core0, 1 temps 5C LOWER than HW Monitor is showing

Which program should i trust ?

(after a few steady minutes in Orthos its showing me 65-67C MAX and a VID of 1.2125 while HW Monitor is reporting 70-72C) i also lowered the voltage in BIOS to 1.275 and CPUZ is showing 1.216 - 1.232 so far everything is stable as i type this

You say you over-clocked this puppy to just short of 4.0Ghz, but your sig says 3.6.

I also have an E8400, but couldn't get it past 3.8 with a voltage setting somewhat higher than yours. I'm wondering how many hours you are stressing the system to declare "stability." Maybe you're really lucky with what came as a prize in your Cracker-Jack retail-box, though. Or maybe it's just the P35 chipset: My E8400 was set up with an nVidia 680i board which is currently out in "RMA-land." Things have changed, though, because I'm running an E8600 and an nVidia 780i board, and . . . "it's happy, I'm happy."

Now, about the temperatures.

There's a good explanation of Conroe temperature sesnors at Tom's Hardware:

Core 2 Quad and Core 2 Duo Temperature Guide

I tell you to read it, because it provides insight to the basic facts about the TCASE sensor (measuring temp at the IHS-processor-cap) between the cores, and the Tj core temperature sensors -- embedded in each core respectively.

Then take a look at an Anandtech Article on the E8500 processor published in March of this year:

E8500: Best just got better

As good as it gets, this may not explain that many of the Wolfdales have been shipping with sensors that are "seen to be defective" when Intel is suddenly saying "the sensors are not meant to read idle temperature values." So people perceive that they are "stuck." They are "stuck" below a certain nominal value.

At idle, and with on-the-edge-current BIOS update, my E8600 shows 56C/56C at idle. Those are the Tj core temperatures. The TCASE temperature is shown in BIOS, and seems to vary properly between idle and load. When my load temperatures for the cores exceed 56C, the temperatures start to move.

Because your monitored voltage values seem so low, I'm not all that happy with your load temperature values. If you were to run IntelBurnTest, I would guess they would climb to the low 80's Celsius. By itself, this is not a problem, but it is getting close to one.

The thermal spec for the Wolfdale -- the point where the CPU will shut down from over-heating -- is a TCASE value of around 75C. What seriously bother me here is that your show a CPU temperature that only moves between 42C and 43C idle to load, but a system temperature (motherboard) that moves between 32C and 59C. This latter variation seems more characteristic of a CPU TCASE variation idle-to-load, while the former -- for being 5C or so too high -- is acting like the motherboard sensor.

And for the voltages being reported in the monitor, your Tj core temperatures still seem too high (I already said that). So in addition to getting your monitoring software to read and report temperatures with the right labels, I'd try to focus attention on CPU cooling and case-airflow.

The Freezer 7 Pro should do a better job for you than it is.

I'm only over-clocking my E8600 to 4.0Ghz, and my load temperatures under INtelBurnTest, room-ambient around 78F and a Noctua-NH-U12P CPU cooler are peaking at 65C degrees. [Of course, I lapped my IHS and heatsink-base to bare copper and used synthetic-micronized-diamond thermal paste.] But my reported VCORE or CPU voltage is around 1.26V.

If you just didn't make a mistake of massive transpositional dyslexia in telling us your temperatures, see whether there isn't a newer BIOS for that motherboard. There are ALWAYS bugs in BIOS's, and some of the "Cadillac" boards have to go through five or ten BIOS revisions before all the little inconsistencies are cleared up. That's why I usually wait 6 months after a board is released before I bother to buy it.
 

andrei3333

Senior member
Jan 31, 2008
449
0
0
awesome response - but i have hit a snag - something is wrong in the bios - it shows all my latest OC values, but when the PC is in windows and monitoring software is running the numbers are way way down, ram is down to 667 and its ato stock 3.0GHZ

nothing i change in BIOS seems to affect anything in windows. i am going to
1) load safe values from bios
2) failing that i will take out the bio battery and short the pins
3) failing that i will update the bios with the same one ( since its the latest)
4) failing that i will break my PC

oh and my sig has not been updated since i have started finetuning to 4.0GHz in the last 2 days

i will respond to yous post in detail once everything resumes as normal ( i am praying no rma will be required)
 

andrei3333

Senior member
Jan 31, 2008
449
0
0
Ok option 1 worked out so i am back on track, something did go wrong though, the last time i was in windows before this happened i was stress testing and orthos crached on me twice completely closing itself...time to start from scratch

Update 2:
i am at 3.6GHz now and everything is at stock voltage and ram is 5-5-5-18 1.8V @800MHz obviously - running orthos now
 

BonzaiDuck

Lifer
Jun 30, 2004
16,832
2,148
126
Why haven't you stopped using Orthos and switched to the multi-core version of PRIME95 for stress-testing? It will create automatically as many threads as you have cores.

I use PRIME95 to find failure thresholds, and I may run it for 15 or 16 hours to verify stability at some choice of settings. Then I run IntelBurnTest for 2 hours.

I've also said this before. There are a set of exponential probability distributions in relationships between heat and voltage, failure over time, failure with increasing voltage, and other parameters such as speed.

I have an E8400 which I managed to over-clock to 3.6Ghz. I was trying to reach 3.8 before I encountered a series of problems. The motherboard -- with an nVidia 680i -- died. It is currently in "RMA" status.

I'm not familiar with the P35 chipset, although I would hope it would give you more flexibility with the E8400 processor than I experienced with the old 680i board.
 

andrei3333

Senior member
Jan 31, 2008
449
0
0
^ yeah i have realtemp 2.70 and it makes the numbers i like LOL

no it has the most reasonable ( and lowest ) values, strangely coretemp is exactly 5C higher than Realtemp and HW monitor is exactly 5C higher than coretemp, so its really weird
 

disports

Golden Member
Jul 18, 2008
1,176
0
0
you might want to use the latest versions: Real Temp 2.75 and Core Temp 0.99.3
 

Tempered81

Diamond Member
Jan 29, 2007
6,374
1
81

Nice. 4.0Ghz @ 1.248v Load for a C0 stepping is not bad. You either have 31C or 29C Distance to Tjmax. I wouldn't let this number get smaller than 24-25C. This is only about 4 to 6C degrees of play until your at max temperatures. You still have about 0.15 volts of play until you reach 1.400v. Overall great O/C! If you want more than 4000mhz it will take more vcore, but keep the temps reasonable, or look for something better than your AC F7P.

I can't see the speed of your Ram, could you show the Memory/ SPD tabs instead of the Cache tab? HM shows 2.0 Vdimm.

Originally posted by: andrei3333

FSB - 444MHz (x9=3996GHz)
Mem multiplier is 2.0 (1:1 ratio)
mem freq - 888
mem timings 5-5-5-15
DDR2 Voltage - +0.3 (which should result in 2.1V)
CPU Voltage 1.306V
MCH (NB) +0.1
FSB Voltage +0.1
-
Corsair XMS2 PC6400-DDR2 C4 2GB (specs up in the air)

How fast can the xms2 run at 5 5 5 15 2.1v?
 

andrei3333

Senior member
Jan 31, 2008
449
0
0
jaredpace

Thanks, BTW
these earlier specs you quoted me as posting are no longer correct as after the tyny meltdown i loaded the bios with default safe specs and started all over again before running those benchmarks

as of 09/25/08 it looks like this

445x9 @ 1.325V (in bios) which is 1.264 in CPUZ not 1.284 as you may have seen due to the crappy low res printscreen
mem freq = 890MHz (1:1)
5-5-5-18
DDR2 = ~2.0V (+0.2)
FSB and MCH voltage is stock

i have not yet optimized my ram to try anything faster, it is rated for 800MHz 2.1V 4-4-4-12 but i want more bandwidth rather than tight timings so i will try to increase it to 2.1V and run around 1000MHZ

HERE is a pic after running PRIME 95 as someone here has requested 2 hours and 20 minutes while fully computing right now after work...same exact results as fas as max temp goes as orthos
http://i150.photobucket.com/al...119/andrei3333/oc4.jpg

 

BonzaiDuck

Lifer
Jun 30, 2004
16,832
2,148
126
Looks pretty good to me. May be the limits I had with that CPU were due to the 680i motherboard. Actually, the more I think of it, I'd forgot to bump up the FSB/VTT voltage. Yup.

 

andrei3333

Senior member
Jan 31, 2008
449
0
0
Well from what i have read here and on Toms, MB does play a big role in OC's, and yes sometimes that means bumping more voltages all around like FSB and MCH as well as DDR2 and CPU, so it may be

i consider myself lucky to have done this, but it did take A LOT of rebooting and fine tuning and a meltdown and a half to get here, and i am still not done

one more thing,. my cooling and case neatness is pretty good, this weekend i will try to take some shots from the inside of the case and ask for the community's opinion...its an OLD case that did not cost me a lot, ( i do not even know the brand name) but it has served me very well so far...
 

andrei3333

Senior member
Jan 31, 2008
449
0
0
That is very ambitious, but why is CPUZ showing 4500MHz while Orthos and coretemp only show 4185 MHz ????

and whats the program on the bottom left used for ? and are such relaxed timings really that necessary ?