Jump to content
Corsair Community

BIOS and CPU-Z don't match in OC attempt


Nevets

Recommended Posts

I am a bit inexperienced in OC'ing (OK I'm just a newbie in dual core OC'ing who has OC'ed a single core CPU) and I am trying to OC the FSB on the EPoX MF570SLI AM2 board.

 

I am using the latest BIOS and the only significant changes that I made on the memory page was to change the 5-5-5-18 timings to 5-5-5-12 as the stick should have been read. I enabled the PCIex Spread Spectrum, CPU Spread to center. I have the memory set at 800 instead of auto, interleaved, and I have set the VDimm to 1.9 instead of auto. I also have the CPU Clock set to 11, and I have stock FSB (In this BIOS it is called 'CPU Overclocking') which I beleive to be the FSB.

 

When I increase the CPU Overclocking to 10 (it ranges from 1 to 250) and reboot the POST indicates that I am running at 2310 (210X11) with DDR2 800@385Mhz. However when windows starts and I check with CPU-Z, it shows the board in the default clock mode of 2211.8; 11 X 201.1. The memory change that I made to the BIOS memory page going from 18T to 12T sshows up in CPU-Z.

 

The same results are given when I am at 11 +25 which the POST reads as 2475 (225X11) DDR2 800@412Mhz which is what I expected. But CPU-Z and CBID both indicated that I am still at 11 and 201.1.

 

At 2970 (270X11) POST with DDR2 800 @495Mhz 5 5 5 12 2T. No change in CPU-Z.

 

I played this game last night and went all the way up to 3025 (275X11) POST but no change in CPU-Z.

 

Now this board has a nice Thunder Heat Probe that I placed on the side of the CPU Heatsink and there has been no increase in the temperature of the CPU. This is confirmed by Ghost Monitor that shows the CPU at 38C throughout the experience.

 

This leads me to believe that either the POST is wrong and the CPU-Z is right and that I have not done anything at all to the FSB or I actually clocked the CPU at 275X11 for 3025. The only other obvious thing is that I am off in never-never land and I really don't have a clue. :D:

 

Also, if this memory is PC6400 800 why is the POST defaulting to 386Mhz instead of 400Mhz (DDR 200X2=400 X 2 = 800 for PC6400 800)?

 

How do I resolve this so that I can get to OC'ing.

 

TIA

Link to comment
Share on other sites

Try these settings.

 

Power BIOS Frequency

 

CPU Frequency = 210

PCIE Clock = 100Mhz

CPU Clock Ratio = 11

AMD K8 Cool 'n' Quiet = Disable (When overclocking you must set to disable)

CPU Voltage = 1.40V

Chipset Voltage = 1.5V

HT Voltage = 1.2V

Memory Clock = DDR2 800

 

Keep in mind that as you increase your FSB, you will very likely have to drop this DDR2 800 to DDR2 667. Cool 'n' Quiet if enabled when overclocking will disable your overclock so I surmise this was the issue. :)

 

Neither was wrong. Your BIOS was correct and your settings were saved. Cool 'n' Quite recognized that the settings were out of spec and it must have spec in order to power save so it disabled your settings and reordered back to default the specific settings that would not allow it to work. CPU-z was correct for it recognized your Cool 'n' Quiet defaulted spec. settings.

 

Note:

 

When overclocking, disable all spread spectrum's as they have a propensity towards instability. Enable when at stock and utilizing Cool 'n' Quiet.

Link to comment
Share on other sites

Try these settings.

 

Power BIOS Frequency

 

CPU Frequency = 210

PCIE Clock = 100Mhz

CPU Clock Ratio = 11

AMD K8 Cool 'n' Quiet = Disable (When overclocking you must set to disable)

CPU Voltage = 1.40V

Chipset Voltage = 1.5V

HT Voltage = 1.2V

Memory Clock = DDR2 800

 

Keep in mind that as you increase your FSB, you will very likely have to drop this DDR2 800 to DDR2 667. Cool 'n' Quiet if enabled when overclocking will disable your overclock so I surmise this was the issue. :)

 

Neither was wrong. Your BIOS was correct and your settings were saved. Cool 'n' Quite recognized that the settings were out of spec and it must have spec in order to power save so it disabled your settings and reordered back to default the specific settings that would not allow it to work. CPU-z was correct for it recognized your Cool 'n' Quiet defaulted spec. settings.

 

Note:

 

When overclocking, disable all spread spectrum's as they have a propensity towards instability. Enable when at stock and utilizing Cool 'n' Quiet.

 

CPU Freq = 210 does not work. I suspect that you beleive that the range is 200 to 450 (original BIOS) where as now it is 0 to 250.

 

Cool and Quiet has always been disabled. I also disabled any spectrum changes as they refer to EMI which I am not concerned about.

 

The best I can get at DDR800 with increased voltages is 245 X 11 = 2695 at 5-5-5-12 then I kind of hit a dead end. Increasing the voltages any further does not seem to help.

 

I could loosen the timings but I would rather not at this stage (maybe later) so I guess my next step would be to drop the DDR2 800 to DDR2 667 and see if I can get any further. Would this be the correct way to proceed or should I think about DDR@ 400? Should I worry about the HT value or leave it in 'Auto' for the time being?

Link to comment
Share on other sites

The best I can get at DDR800 with increased voltages is 245 X 11 = 2695 at 5-5-5-12 then I kind of hit a dead end. Increasing the voltages any further does not seem to help.

 

I could loosen the timings but I would rather not at this stage (maybe later) so I guess my next step would be to drop the DDR2 800 to DDR2 667 and see if I can get any further. Would this be the correct way to proceed or should I think about DDR@ 400? Should I worry about the HT value or leave it in 'Auto' for the time being?

 

Set your HT at 4X and your FSB to 250. Leave your DRAM at DDR2 800 to see if it was the HT holding you back. If it wasn't then if the maximum FSB you can get is 250 at DDR2 800, then I would stay there rather than move to 250 at DDR667. You would suffer far more losses than gains in that move.

 

That's too bad that you can only effect a 250FSB. HT should not be raised passed 1000Mhz and 245 X 5 = 1225 so that is not good. Drop your HT to 4 X and then you are at a 980MHz HT. Keep in mind that the HTT does not get saturated even when at 800MHz so you want to be lower than 1000 and there is no gain with a higher HT than 1000, only a stability loss.

 

I assume that you now have your overclock showing in CPU-z?

I suspect that you beleive that the range is 200 to 450 (original BIOS) where as now it is 0 to 250.

Why do you suspect this? What have I said about your FSB range that would lead you to think that I believe your FSB range is 200 to 450?

Link to comment
Share on other sites

[quote name=...

 

Why do you suspect this? What have I said about your FSB range that would lead you to think that I believe your FSB range is 200 to 450?[/quote]

 

 

Thanks for the advice. As far as the FSB, every review that I have seen on this board (about 3 so far) shows in the writeup that that FSB is adjustable in 1Mhx increments from 200 to 450. My BIOS shows the values at 0 to 250. I checked on the BIOS page at EPoX and there is no mention of any change to the CPU Overclocking architecture. This has me a bit baffled though. This is from a review:

 

"BIOS and functions of other major components will be introduced in more detail over next page

BIOS overview

 

The main overclocking screen

 

http://img209.imageshack.us/img209/8049/mnb10us.jpp

 

The HTT (FSB) is ranging from 200 to 450MHz.

The lowest multiplier is x4; the highest one will depend on the type of the CPU used.

 

http://img209.imageshack.us/img209/139/mnb21wb.jpg "

 

All the other reviews same that the FSB is adjustable from 200 to 450 but I only have the option of 0 to 250.

 

Complete review is here:

 

http://forums.extremeoverclocking.com/showthread.php?t=223012

 

and here:

 

http://hardwarelogic.com/news/129/ARTICLE/1244/6/2006-08-02.html

 

Another one here:

 

http://www.madshrimps.be/?action=getarticle&number=2&artpage=1842&articID=441

 

Other than the 200 to 450 vs the 0 to 250, I don't see any significant changes in the BIOS.

 

I re-tried the 210 X 11 like you suggested but I could not even get the BIOS to POST.

 

And yes, the CPU-Z now shows the OC results as accurate. Thanks for the heads-up on the cool and quiet architecture.

 

Found another review here:

 

http://www.ocworkbench.com/2006/epox/MF570%20SLI/g1.htm

 

This also shows the 200 to 450 value.

Link to comment
Share on other sites

Contact your mainboard manufacturer and request information as to what the situation is as you should have a far better ability to upclock than 250Mhz.

 

 

I just downloaded the old BIOS (Jul 13, 2006) and that gives me the option to start at 200 and go to 450. Testing the FSB at low multiplier I am able to get up to 285 ~ 290 which gives me an effective FSB of ~342 at HT2X. At HT5X I again top out at 285 - 290 for an effective FSB of ~342.

 

Unless I loosen up the timings (not yet) I do not think that I will get any higher on the FSB with this memory unless you can suggest something else. I am going to try HTAuto next to see how far I can get on the FSB. After that I am going to hit the HT side of things.

Link to comment
Share on other sites

I don't understand why you are running at over 1000Mhz HT. You are not going to saturate the HT Transport at 1000Mhz so you are sacrificing overclock and system longevity/stability for no reason that I can think of but hey, it's your machine to do with as you please.

 

The HT is a pipeline, no more, no less. There is no use that will advance a saturation of this transport even close to 1000MHz let alone higher.

 

My advice? Keep your HT close to 1000 and concentrate on your overclock. Once you achieve your highest FSB<-->DRAM throughput then you can test your HT and you will not find a difference between an HT of 900MHz vs an HT of 1100Mhz. You might very well shorten the life of your hardware though but as I said before, it's your machine.

Link to comment
Share on other sites

I don't understand why you are running at over 1000Mhz HT. You are not going to saturate the HT Transport at 1000Mhz so you are sacrificing overclock and system longevity/stability for no reason that I can think of but hey, it's your machine to do with as you please.

 

The HT is a pipeline, no more, no less. There is no use that will advance a saturation of this transport even close to 1000MHz let alone higher.

 

My advice? Keep your HT close to 1000 and concentrate on your overclock. Once you achieve your highest FSB<-->DRAM throughput then you can test your HT and you will not find a difference between an HT of 900MHz vs an HT of 1100Mhz. You might very well shorten the life of your hardware though but as I said before, it's your machine.

 

As I indicated, I'm new to this 2X Core overclocking. I do realize that anything over 1000MHz on the HT is a waste. So I am obviously doing this wrong.

 

I am just trying to find the max theoretical FSB. I figure that with this info and the max multiplier I will be able to find the happy medium that will give me the best FSB with HT~1000. Is this not the way to proceed?

 

Right now the HT is auto and I am seeing how far I can go. Am I wasting my time here or what? I used these 'runs' to tweak the max amount of voltage increase that I am going to do right now.

 

CPU = +0.1V

Memory = DDR800

HT = Auto @ 1.3V

VDimm = I won't say, but it is less much less than the board will allow.

Memory Timings are 5-5-5-12

Chipset = 1.6V

 

I won't at this time go any higher on the voltages until I get better air flow (when I run super-pi, I'm at 50%cpu at max temp of 44C).

 

How would you approach an OC solution to this set-up? I know this is a hell of a question to ask, but I want to learn the process that deals with the 2X Core.

 

By the way, thanks for you help....

Link to comment
Share on other sites

I have reset to:

 

CPU Frequency = 210

PCIE Clock = 100Mhz

CPU Clock Ratio = 11

AMD K8 Cool 'n' Quiet = Disable (When overclocking you must set to disable)

CPU Voltage = 1.40V

Chipset Voltage = 1.5V

HT Voltage = 1.2V

Memory Clock = DDR2 800

 

Do I now start creeping the FSB up or was this just to cofirm that CPU-Z would now track the results?

Link to comment
Share on other sites

Ok,

 

Here's a breakdown. You want to find the highest FSB of your system. The way to do this is to isolate all other variables in the overclock and then begin to raise the FSB. Once you can not raise it without Prime errors, then you incrementally raise the CPU core voltage and attempt to raise the FSB again. After all attempts have yielded a maximum FSB, then you begin to test your maximum DRAM speeds which are bandwidth and timings. Then you begin to work them both together once you have the CPU Core maximum, DDR Bandwidth and Timings maximums.

 

CPU Frequency = 210

PCIE Clock = 100Mhz

CPU Clock Ratio = 11

AMD K8 Cool 'n' Quiet = Disable (When overclocking you must set to disable)

CPU Voltage = 1.40V

Chipset Voltage = 1.5V

HT Voltage = 1.2V

Memory Clock = DDR2 800

 

This was to set to test if you were gaining an overclock. Now Set to:

 

CPU Frequency = 225

PCIE Clock = 100Mhz

CPU Clock Ratio = 7

AMD K8 Cool 'n' Quiet = Disable

CPU Voltage = 1.40V

Chipset Voltage = 1.5V

HT Voltage = 1.2V

Memory Clock = DDR2 533

 

This will isolate your dram from the overclock since it can not produce system errors at that low speed. Now you raise your FSB incrementally 225(1.575GHz), 250(1.75Ghz), 255( 1.785Ghz) until you find Prime errors using Orthos Stress Prime. When you find errors, raise your CPU voltage from 1.4 to 1.425. If an error still occurs, raise your Chipset voltage one increment and retest. Do not raise your chipset voltage past 1.65V or your cpu core voltage past 1.55V (these are my personal limits, although some do go higher).

 

Once you can not raise your CPU FSB higher without errors and raising the voltage does not help, or you have raised your voltages to the maximals that either I have set or that you decide through research to set, then you have your Maximum FSB. Use Orthos Stress Prime to test.

 

http://www.techpowerup.com/downloads/385/mirrors.php

 

Now that you have your maximum FSB, you can now begin to test your maximum FSB and Multiple. Raise your CPU multiplier to 8 and your FSB to 250. See how high you can get with the same testing procedure. Test with 9 and then 10, then finally 11. Be sure to document these procedures as what you will be left with is a maximum FSB and multiplier. You want the highest FSB first, with the highest multiplier next. Once you have found that maximum you move to the dram.

 

Let us say, for theories sake, that you have found (stable for 8 hours of Orthos stress prime):

 

290 X 10 = 2.9GHz

265 X 11 = ~2.9GHz

 

You can be sure that the 290 X 10 is a far superior overclock than the 265 X 11 even though both equal 2.9GHz. This is because the CPU processing on the 290MHz FSB system is being done far faster on the data transfer lines than the CPU processing done on the 265MHz FSB system.

 

Now when you have found this CPU/Multiplier sweet spot, you must find the DRAM variable which has been set low. Since you know that 290 X 10 issues no errors on Large FFT's of Orthos Stress Prime for 8 hours you know that these characteristics in your overclock are stable. Set your DRAM to 667 and test with Memtest. Keep in mind that the DRAM will raise on a ratio to your CPU and multiplier. You can keep track of this in your CPU-z Memory Tab under FSB:DRAM. You want the highest Memtested stable Dram speed:

 

http://www.memtest.org

Link to comment
Share on other sites

I've hit a dead end as far as the FSB is concerned at 7 X 279.3 (for a freq of 325.9 on the memory). My voltages are:

 

VCore: 1.544

Chipset: 1.6

HT: 1.3

 

Do you recommend additional voltage to any of the above?

 

I am going to move on to 8X now.

Link to comment
Share on other sites

I've hit a dead end as far as the FSB is concerned at 7 X 279.3 (for a freq of 325.9 on the memory). My voltages are:

 

VCore: 1.544

Chipset: 1.6

HT: 1.3

 

Do you recommend additional voltage to any of the above?

 

I am going to move on to 8X now.

 

 

I only raise voltages when the overclock will not work and do so as a test to see if a bit more voltage will allow a greater raise of the FSB. Your voltages are high enough already as far as I am concerned,yet they are not at the damage levels so you are find there. Yes, onward to the higher multipliers. :D:

Link to comment
Share on other sites

OK. I found my sweet spot at 9 X 293 and have run the stress test ORTHO for more than 13 hours without any hic-ups. I can probably go higher with better cooling which will be here tomorrow. However, 2636 @ 9 X 292.9 on air using the default heatsink and fan seems OK.

 

Now when I try to change my DRAM2 speed to 667 instead of the 533, my system will not boot. I checked the memory using memtest and ran it for 12 hours with no errors (using the floppy method). So I am kind of stuck right now and do not know what to do next.

 

Should I loosen the timings, increase the VDIMM voltage, neither, both, interleave (on or off)?

 

I did not quite understand what you meant by keeping track in CPU-Z (Memory tab) when using MEMTEST86+. The version that I have only works in DOS off the floppy so I am lost on this one.

 

Thanks again.

Link to comment
Share on other sites

Now when I try to change my DRAM2 speed to 667 instead of the 533, my system will not boot. I checked the memory using memtest and ran it for 12 hours with no errors (using the floppy method). So I am kind of stuck right now and do not know what to do next.

 

When you set your dram to 667 you are also raising that 667 by the same percentage as your raise of the FSB. They are tied to each other. In order to keep stability you divide the FSB:RAM percentage (or ratio) down.

 

Should I loosen the timings, increase the VDIMM voltage, neither, both, interleave (on or off)?

 

This is up to you. Test the settings. Do not increase the Vdimm unless the tested setting does not work. You can also just set your Vdimm to 2.1v and test your timings that you set with memtest. If it doesn't work with 2.1v, then I advise you loosen the timings until it does work.

 

I did not quite understand what you meant by keeping track in CPU-Z (Memory tab) when using MEMTEST86+. The version that I have only works in DOS off the floppy so I am lost on this one.

 

If you look to your memory tab in CPU-z, you will find your CPU:RAM ratio and that will give you your bandwidth. Then you can document it, boot to your Memtest and test for stability.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...