Question about starting pc after OC

[mosheg91]

yesterday i was on a computer lab to install my new gpu (HIS HD 4850 IceQ 4 512MB GDDR3) and psu (Tagan Piperock 500W TG500-BZ) and i just talk with him and i told him that i heard that oc needed to my cpu for decrease the bottleneck that my cpu is doing to the gpu ... in short, an a half hour later i back to the lab after he installed everything and he said that he oc'ed my cpu . of course i was happy and i asked him, if i'll format my pc the oc won't disapear ? he said no ... i came home, i was formating my pc and the settings realy saved ... but, after i installed service pack 2 the OC is disapeared ! i tried to ask for help in another forum but no one helped me . so i called him a day later and he told me how to oc back as he did . we successed to oc back . now what's the problem ? when i came home i have noticed that when you press on the power/off button the pc is starting and a second - second and half he restarts . i call to another computers technician (from the same stores line) because the computers technician (which oc'ed my pc) is went home and he told me that this process is normal (he explain to me that the pc is checking something i don't remember exactly what and then he starts the hard disk and everything else ...) and if i want to cancel that process i need to update my bios . i told him that i don't know and he said if you don't know, don't do it . and now when i'm starting my pc this process is dissapeared ... i wish to know why ?.. from a bit reading on the net about oc'ing i thought to my self that maybe the computers technician which oc'ed my pc is forgot to tell me (when we talked on the phone) to disable on something ... i was searching on the bios and i didn't find any option that releated to the oc manual which i found on the net ... someone can help me and explain me what is going on ?.. maybe i need to cancel or enable something for the process will back ?..

please read and pay attention to all what i have wrote ... do me a favor ...

p.s, i have Gigabyte GA-965P DS3
p.s 2, sorry for my bad english and grammer ...

regards,
moshe

 

[bilbat]

I'm not real familiar with GB's nVidia boards, but, my guess - it's pretty much the same as with the Intel MOBOs;

When changing the OC, booting twice is normal for it to accommodate the new clocking;
if it reboots a third time, it found something in the OC unacceptable, and is reseting itself to enable booting so you can try again...

Hope this helps - that was a little garbled???

 

[mosheg91]

but why now it doesn't reset itself ?.. only when i'm pulling out the power cable from the psu for a few minutes or hours (i didn't check it, we have lightnings and thunders in israel so every time i go to sleep i'm pulling out the power cable from the psu - don't worry, i'm turning off first the system and the power button on the psu and then pulling out the cable) ...

i checked my system stability with OCCTPT for 2 hours, is that program is good for testing ?.. or should i check with some other program ?..
i checked on the bios settings and i saw that eist and some other options are enabled (except C.I.A 2) and i read on the net that when you oc you need to disable those options ... do i need to disable them ?..

 

[bilbat]

The 'two resets' thing only happens the first time you change (or restore from a stored profile) the OC - and, even then, only sometimes: I haven't determined exactly when, but one thing I've noticed is that if it recalculates, or you manually change the tRD, it will always do the 'two resets'...

Perhaps this will explain a little: your CPU contains a set of registers that 'tell' the BIOS what FSB and multiplier it uses; the RAM contains a set of registers (called SPD) that also tell the BIOS how it should be set up for a given FSB - some (high performance) RAM also contains an EPP, which has settings for the higher speed the RAM is capable of (example: I have 1066 - it has an SPD to tell the BIOS how to set it up as standard 800 RAM, and an EPP to tell the BIOS what it wants to reach 1066...). What happens if the board resets twice (on a new overclock), and it finds it couldn't run at the new settings, is that it resets a third time, and reads the CPU and RAM, and sets itself to those safe 'stock' settings, and, in the process, wiping out your overclock settings...

Memtest86+ is probably the best program to run to qualify RAM settings - it's time-consuming (I reccomend letting it run overnight) but thorough:
http://www.memtest.org/download/2.11/memtest86+-2.11.iso.zip

Prime95 is a good CPU stress tester;
Win 32-bit version:
http://www.mersenne.org/ftp_root/gimps/p95v258.zip
Win 64-bit version:
http://www.mersenne.org/ftp_root/gimps/p64v258.zip
You'll want to visit here:
http://www.mersenne.org/freesoft/
for instructions on how to set it up...

 

[bilbat]

Sorry, I missed the last couple of questions (early here - half-asleep, I guess...)

Here's what you want for your OC testing:
Limit CPUID Max to 3: Disabled
No-Execute Memory protect: Enabled
CPU Enhanced Halt (C1E): Disabled
CPU Thermal Monitor (TM2): Enabled
CPU EIST Function: Disabled
Virtualization Technology: Disabled (Unless you are using MS' Virtual Machine s'ware...)
Full Screen Logo Show: (doesn't matter - if you leave it on, you seem to get some extra time to hit that <DEL> to enter the BIOS between boots; I leave it on while fiddling with the clocking, turn it off once I'm happy with the BIOS' settings for a bit quicker boot...)

Then, when you have a good, stable, tested overclock, you can turn on C1E and EIST for energy savings, and see if it's OK with your intended usage pattern, and if it 'returns' from sleep mode OK; what I actually have is one stored profile that's 'balls to the wall' wide open, for extreme work where I've got AutoCAD 64 electrical, a logic designer, and Visual Studio all open & working hard at the same time; another 'milder' overclock for day to day work, with quite a bit less fan (13 fans in the box...) noise but no energy saving features turned on; and a 'stock' (CPU and RAM set to no overclock at all) with all the energy savers ON, for, say, a weekend, when I'm using it mostly as a TV, and for e-mail and web-browsing...

If you want to learn more about the actual overclocking, there's a GREAT Gigabyte forum over at:
http://forums.tweaktown.com/f69
You'll find a pair of step-by-step 'stickies', one on RAM testing/tweaking, and one on GB overclocking.

Hope this helps; good luck!!!

Bill

 

[mosheg91]

you help me very much !!
but i have some questions ...

so if i understand right, the reset thing have to be only if it's the first oc with this mb ?.. and do i need to check memory even though i didn't oc them ?..

this is my oc settings:
Robust Graphics Booster --------- [Auto]
CPU Clock Ratio --------------------- [11x]
CPU Host Clock Control ------------ [Enabled]
CPU Host Frequency (MHz) ------- [266]
PCI Express Frequency (Mhz) ---- [100]
C.I.A.2 -------------------------------- [Disabled]
System Memory Multiplier (SPD) - [3.00]
Memory Frequency (Mhz) ---------- 800 800 (not changeable)
High Speed DRAM DLL Settings -- [Option 1]
System Voltage Control ----------- [Manual]
DDR2 Overvoltage Control -------- [+0.3V]
PCI-E OverVoltage Control -------- [Normal]
FSB OverVoltage Control ---------- [Normal]
(G)MCH OverVoltage Control ----- [Normal]
CPU Voltage Control --------------- [1.40000V]
Normal CPU Vcore 1.32500V ----- (not changeable)

and i forget to tell you my chipset is intel's not nvidia's ...

this is my specs:
Intel Core 2 Duo E4500 2.2GHz 2Mb @ 2.93GHz
Gigabyte GA-965P-DS3 Intel P965 LGA 775
HIS HD 4850 IceQ 4 512MB GDDR3 PCIe
OCZ 1GB DDR2 800 PC2-6400 Platinum Dual Channel CL 4-5-4-15
GEIL 1GB DDR2 800 Dual Channel CL 4-5-4-15
Western Digital 160GB 3.5 7200Rpm 8Mb Buffer SATA II WD1600JS
Tagan Piperock 500W TG500-BZ
LG GSA-H62 SATA Multi DVD-Recorder Bulk
Compucase 6A19 Black Midi Tower No PSU

p.s, the technician told me that the oc he did is stable ...
p.s 2, is OCCTPT not good enough for system stability checking ?..

 

[bilbat]

First question: 'double reset' can occur after saving ANY changes to your OC - as I said, it doesn't ALWAYS do the double reset, but it can, any time you change the OC, not just the first OC for the board...

Your overclock settings look good, and should be stable; I don't know the meaning of the "High Speed DRAM DLL Settings -- [Option 1]" - I'll dl the manual and try to give it a look by tomorrow; your DDR2 overvoltage setting gives you ~2.1 V RAM setting - is any of your RAM rated that high? If not, you might be able to lower it a bit; also, to my taste, 1.4V Vcore is kind of high - I try to never go above 1.375 EVER - it's not heat that kills CPUs, it's voltage...

I am not familiar with OCCTPT, but, from their website, it looks thorough enough to trust...

 

[mosheg91]

what do you mean when you ask if any of my RAM rated that high ?..
i called again to the technician and asked about the cpu voltage and he said that the voltage 1.40000V is fine ... it's not that high to kill the cpu ...

 

[bilbat]

What I mean is the manufacturer's specification for the RAM voltage; JEDEC specifies 1.8V for 'standard' DDR2-800 RAM, but each manufacturer will have their own voltage rating - the OCZ platinum I found shows 1.9V - 2.1V, but I was unable to find that model GeIL, and their RAM (at least where I searched) could be anything from 1.8V to 2.4V (awfully high...). It needs to be set to the highest of the two maker's minimum spec:
Say you have RAM 'A' that says it's 1.9, and RAM 'B' that's 2.0-2.2 - you would want to start testing at the 2.0 voltage - the minimum of the highest spec...
Once more: RAM 'C' that's rated at 2.1, and RAM 'D' that's 1.8 to 2.0 - you'd pick the 2.1...

One of the stranger aspects of the GB MOBOs is that they allow ridiculously high Vcores; your 1.4 is not THAT high, but read here:
http://www.dwpg.com/content.php?contid=2&artid=68
Again, the problem is not heat (of course, that's a qualified 'not' - enough heat WILL let the 'magic smoke' out) (all electronic devices run on 'magic smoke' - once you let the smoke out, they quit working!) but voltage causes electromigration, which causes degradation over time...

What you are trying to do in an overclock is achieve a multidimensional balancing act:

MoreFanNoise<HigherHeatLoad<BetterPerformance<HigherFrequencies<->LowerFrequencies>SlowerResponse>LowerHeatLoad>QuieterFans

LessReliability<Higher HeatLoad&Electromigration<HigherFrequencies<HigherVoltages<->LowerVoltages>LowerFrequencies>LowerHeatLoad&Electromigration>MoreReliability

LessStability<MoreDemandingTimings<HigherFrequencies<->LowerFrequencies>LessDemandingTimings>MoreStability

&on&on&on...

Another thing that must be considered in this balancing task is the Pareto principle:
squeaking out the last (top) 20% of the heat load will cost 80% of the cooling budget...
getting the last (top) 20% of the possible frequency from a system will take 80% more capable (costlier) components...
the last 20% of the high frequency/high voltage range will produce 80% of the heat load...

&on&on&on...

So, you see, there is no set rule for configuring your system - you pays your money, you picks your targets, and you takes your chances!

 

[mosheg91]

so how much voltage to put at CPU Voltage Control to prevent this electromigration thing ?.. and would it handle the oc ?..

how can i check those GeIL sticks voltage ?..

 

[bilbat]

There should be a little sticker on your RAM stick(s) that will have a model number - may even say right on it what the preffered voltage is...

Only way to find voltage vs stability 'sweet spot' is to try it, a bit at a time...

 

[mosheg91]

so you think he put too much volt on the ram sticks ?.. and do you think i should always to decrease the cpu voltage until i get stability problems ?.. like resets and etc ...

i have another question too ... when the technician did the first oc (when i was on the store) i saw that the ram sticks were on 1067 at Memory Frequency (Mhz) and at the second oc (when we did through the phone) he told me to choose 3.0 on System Memory Multiplier (spd) and it go back to 800 (the original clock from the factory) . it means that he did oc to the ram sticks when i was on the store ?.. the sticks are working now like the factory original settings ?.. is it work better\worst ?..

you help me alot, i really appreciate the time you're spend to help me ... very very thanks

 

[bilbat]

You're very welcome - I answer questions here because I learn a lot, and I rely tremendously on information from these sorts of forums in my own work (I'm a systems engineer - I do hardware and software to run machinery, factories, process plants, and the like...)

Ok - I rooted around on the web, and it would seem that your E4500 is quite tolerant of high Vcores - actually, could go up to a volt and a half, with good air cooling, over that on liquid... One thing you haven't told me is what kind of CPU cooler you have - the following steps (and how far we take it) will depend on this...

For the time being, we'll leave most of your voltages where they are, and play a little...

The plan is to lower your CPU multiplier to 8x, and go for a FSB over 400, with the RAM multiplier at 2.0 - which will give you RAM at rated DDR2-800, so each time we 'bump up' the FSB, your RAM speed will go up with the processor speed... You mention that at one point the tech had the RAM at 1066, but I'd advise against that for two reasons: firat, you're using RAM from two different manufacturers, which can be a problem at higher, tighter timings; and second, the 2.0 (1:1) multiplier is the most efficient 'hook-up' between the MCH (Memory Control Hub - or NorthBridge) and the CPU - at other multipliers, the MCH has to 'wait' to do certain transactions with the RAM until both busses are synchronized (I'm running 2.4, which is 6:5 - means that, at times, my MCH has to wait 5 bus cycles to commit a transaction...) We'll start at 400, and then, once we've got it stable, we'll take it up 5 mHz at a time, aiming for 425 - which, at an 8 multiplier, will give you CPU @ 3.425, w/RAM @ 850...

 

[mosheg91]

i have Big Typhoon

i would like to keep on 2.93 and wouldn't continue to higher rate ... and the same on memory ... as you said these two from other manufacturer so that could be a problem because i'm not an expert on oc and you're not from my place (israel) so you can't come and oc my pc

so in general the oc settings seems to be good ?..

 

[MMclachlan]

Hmmm. 1.4v Vcore does seem a little high. I had a E4300 that would do 2.8Ghz on 1.325v (350Mhz FSB).... I would suggest dropping it down to 1.35v and testing stability there, then if it crashes go up one step at a time until it is stable, if it is stable at 1.35v go down until it crashes then back up again a little after that. No point running excess voltage through your chip if you don't have to, you'll only cook it quicker!
As bilbat says the ideal situation is to have your FSB and RAM 1:1 at 400Mhz, although as you are running a slightly older 965 chipset board you may need to up the n/b voltage a fair amount when you get near this.

 

[mosheg91]

but the i have some option enabled which drops the voltage down if it's not needed ... in cpu-z i see that it uses 1.376v now ... so far i didn't see it higer ...

about the 1:1 + change the n/b voltage, i don't know how to do that and i don't know the meaning of n/b voltage ...

 

[mosheg91]

someone ?..

 

[bilbat]

Moshe -

I'll get back to you tomorrow AM - am busy knocking holes in walls, pulling wires & coax, and cursing - in equal doses...

 

[mosheg91]

ok thank you

 

[bilbat]

OK - second question first:

diagram is from your manual; note that the 965 is the CPU's 'connection' to the memory, i.e., the MCH, or Memory Control Hub; also note that the 965 is shown above the ICH8, which would be 'to the north' on a map (and you'll always see them represented this way) - thus the 965 is the 'NorthBridge' and the ICH8 is the 'SouthBridge'. System performance is intimately related to the transaction speed and coordination between the MCH and the CPU - so much so, that the AMD processors incorporate the MCH directly on the CPU die. The whole business of '1:1' memory speeds being faster is related to this: there are certain memory transactions that can only occur (and thus, are deferred, holding up the whole works) when the memory bus and the CPU are 'synchronized. For example, I am using the 2.4 memory multiplier, which is actually 6:5, so, occasionally, my memory transactions can be held up for 5 CPU cycles until bus synch - which means, onaverage, I'm losing 2 1/2 cycles for these transactions. At 1:1 (2.0 multiplier, the bus is always available on the next cycle... When you start 'pushing' things (say, if we go ahead and push the FSB up to 400mHz) you often need a little extra drive voltage on the MCH itself, which, much like higher DDR2 voltage, helps the performance at high frequencies/loadings.

Now, about the voltages:
First, you must realize that there is no calibrated 'reference' available on the system to measure voltages against - it's not like time, where we can, say, get access to the Bureau of Standard's absolutely accurate atomic clock. Every BIOS and program uses the method that they happen to believe in most strongly, so, often, different programs which purport to 'read' your system voltages will come up with (slightly, or sometimes widely) differing numbers. Second, there is a phenomenon in nearly every power control circuit that causes the voltage to drop slightly when the power loading peaks. OCing MOBOs often have a system which attempts to compensate for this (using methods which both vary by the manufacturer/BIOS writers, and in actual success at improving stability) called LoadLine Calibration, which, when enabled, may also cause variations in your voltage readings...

Back to the business of OCing your board - we CAN do it via forum; the main thing I'd like to know is whether you've been able to 'save' your BIOS settings to CMOS, as I'm not sure all GB MOBOs support this, and GB has not seen fit to put this function in their manuals, so I can't tell whether a specific MOBO will do this by looking at my (huge) collection of MOBO manuals. If you CAN save your settings, then, if we have a mishap overclocking, you can just go back to your 'last known good' settings, while we exchange info here for the 'next attempt'...

 

[mosheg91]

yes my mobo is able to save settings to cmos ... i even did that ...

 

[mosheg91]

i want to add, is saving the cmos settings is saving all other settings ?.. i mean like the simplest option like booting (first floppy, second dvd-rom and third hard disk), bios time and date and etc ?..

 

[bilbat]

I'm pretty sure it does, but I can't tell you for absolute certain, as, because of my boot manage, none of my stored configuratios has any changes to any of these settings... East enough to find out though - change a couple, sve 'em, change a couple more - save 'em, then restore, and see if they're still set correspondingly...

 

[mosheg91]

i understand ... look i tried to change the cpu to 3.2 by changing cpu host frequency (mhz) to 400, system memory multiplier (spd) to 2 (to make 1:1), cpu clock ratio to 8, saved it by hitting f10 and the pc didn't boot and the mobo reset the oc settings to the same i had (2.93 MHz) ... i think i need to up the DDR2 overvoltage control to +0.4 ... don't you think ?..

i have a question ... i saw that people changing the RAM sticks clocks ... why do they do that ?..

 

[mosheg91]

some fix to what i have wrote ...

to the same i had (2.93 MHz) before **