BIG TROUBLE IN LITTLE CHINA: GA-EP45-UD35

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
Ok, so maybe I shouldn’t have attempted this after having been out of it for so long, but I’m committed (or MAY be if I can’t fix it).

First, the system:

Gigabyte GA-EP45-UD35
Thermaltake Pure power 500W ATX 2.0 power supply
EVGA GeForce 9500 1GB video
Intel Core 2 Duo E7400 Wolfdale 2.8Ghz
Seagate 7200.10 ST3250310AS 250GB SATA hard drive
Plextor DVD SATA burner PX-850SA
OCZ Platinum 4GB (2X2GB) DDR2 1066 (PC 8500) (OCZ2P10664GK)
MS Windows XP HOME SP3 for System Builders-OEM

The problem is that I cannot load Windows. Initially it went about ¾ through setup, then froze.

Repeated attempts to fix it only offered “cannot copy file” (dozens of them).

Tried with one stick of RAM, no difference. Tried putting the HD on the purple connector and it seemed to go further, but still failed.

Each time I set boot from CD and ‘load fail-safe defaults’ in the bios.

I’ve tried this maybe a dozen, or more times, no difference.

At this point, I have completely formatted the hard drive and need some help. I somehow feel the problem is RAM related as setup stops at different files and locks at different, random places. Haven’t fooled with RAM volts, don’t feel too comfortable doing that without astute direction. :whistle:

Anyone, anyone? Bueller? Bueller? :ange:
 

rockyjohn

Distinguished
You may be rigft about the memory. That OCZ memory is not on the mobo compatible memory list, probably because it is designed for 2.2v while mobo is set at 1.8v.

I suggest you look up the memory at the vendors site and see if it has alternate settings that let it run at other voltages, although this might not be likely since it appears mobo may not have found another setting to run.

What you can do is go into bios and change memory voltage to 2.2v (or +0.4v depending how the setting works) and hopefully this will work.

If you do get it working you might also try backing it off to 2.1 or 2.0 volts

I have a Gigabyte P35 board which also defaults at 1.8v. I purchased Crucial memory off the compatible list which I later learned was 2.2v. So I adjusted bios +0.4v to compensate. A Gigabyte tech later suggested I tune it back to 2.1v which I did and it has been running fine. It loaded Windows Ok but failed repeatedly on memtest. I had to make some other bios adjustments to get it to pass an overnight run of memtest.

 

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
Thanks so much Rockyjohn and zipzoomfly for the quick reply and the consensus.

I made some previous notes and see VCORE-1.268 and DDR18V-2.208. Suspect these are for the CPU and not the dram. I'll look through the manual.

Everything is brand new and can't believe it's the optical reader of the MB or a bad copy of windows, got to be the RAM.

I'll give it a try and report back. THANKS!
 

bilbat

Splendid
Your MOBO number is unknown; 'one wrong character' substitution shows three possibilities - this should work for any of them...

Get a copy a copy of MemTest86+:
http://home.att.net/~chip.programm [...] 1a.iso.zip
Unzip it to an .iso, and burn it to a CD - makes a bootable RAM tester...

Take out all the RAM, and unplug any USB devices other than keyboard; place one stick of RAM in slot 'DDRII1" (closest to the CPU)...
Power up, hit <DEL> to enter the BIOS, select "Load Optimized Defaults":


Reboot, hit <DEL> to enter the BIOS, select "Advanced BIOS Features" page;
Set "CPU Enhanced Halt (C1E)" and "CPU EIST Function" to "Disabled";
select "First Boot Device" and set to "CDROM";
exit back to the main BIOS page, select "Integrated Peripherals" page;
set "Legacy USB storage detect" to "Disabled"
load the CD we made above into the drive;
do an <F10> BIOS "Save & Exit";
on the reboot, MemTest will run and check your stick of memory...

Let it run at least one complete pass;
if it fails - you've found your problem;
if it passes, power down, remove the stick of RAM and replace it with the other one;
power up and run MemTest as above...

If they both test OK:
power up, again hit <DEL> to enter the BIOS again, select "MB Intelligent Tweaker(M.I.T.)" page;
set "Robust Graphics Booster to "Auto"
set "Performance Enhance" to "Standard"
set "System Voltage Control" to "Manual"
set "DR2 OverVoltage Control" to "+0.4"
set "(G)MCH OverVoltage Control" to "+0.1"
do an <F10> BIOS "Save & Exit";
power down;
re-install all your RAM;
power up;
(hopefully) should work at this point...
 

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
I'm workin' on it, Bill.....

But have a question. Finished one pass with one stick of ram but show 848 errors (!) Is this normal?

mem86.jpg



UPDATE: the second stick only had 26 errors in a full pass compared to 848 errors on stick #1.

What is acceptable? None??

Should I have raised the voltage to 2.2V?
 

bilbat

Splendid
Should be able, essentially, to run for days without an error - once we get 'er tuned properly; first thing I'd try is backing the voltage down a tenth at a time with the worst stick in (848 errors on stick #1 - eeek!), and see what it does. Meantime, first thing in the AM, I'll do some broad-spectrum research and see what I can come up with; I know the board and the chipset are up to it - we just have to figure out 'what it wants'! One thing that might help is a copy of CPU-Z:
http://www.cpuid.com/cpuz.php
I sometimes use it to read the EPP (Enhanced Performance Profile) stored on most 'fast' RAM, which contains quite a few more parameters than the usual 5-5-5-18 we're handed - sometimes will point out a mis-match in an odd spot... OCZ has great support, and lots of info accessible - don't fret - we will get it working!

Talk atcha tomorrow - I just ate six ibuprofen and an oxycodone - gots to get to sleep while I'm temporarily feeling no pain - have an ungodly toothache...

Bill
 

bilbat

Splendid
Rockyjohn said:
I have a Gigabyte P35 board which also defaults at 1.8v. I purchased Crucial memory off the compatible list which I later learned was 2.2v. So I adjusted bios +0.4v to compensate. A Gigabyte tech later suggested I tune it back to 2.1v which I did and it has been running fine. It loaded Windows Ok but failed repeatedly on memtest. I had to make some other bios adjustments to get it to pass an overnight run of memtest.
So I thought, in the temporary (toothache) absence of my ability to help with anything else (further research) it couldn't hurt to try...

I, too, am running 8Gb (4x2) 1066 G.Skill F2-8500CL5D @ 1080 at a tenth under spec (2.0 vs 2.1 specified) - runs demonstrably faster at 2.0 than at 2.1
 

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
Bill, hope the tooth is better today....there is nothing worse.....wiskey seems to help me a lot... :D

Not much time as family is visiting, but I raised the ram voltage to 1.84V and both sticks together are stable!!! :eek:

Windows loaded and I'm doing the drivers now...so far, so good.

I'll keep you posted, my friend.
 

bilbat

Splendid
Thank you for inquiring - it's better, but not fixed - I've got it a bit loose, and may try the yankeroo later - but my favorite anesthesia is 151 rum; I actually started (back in the days when I was drinking) on the stuff 'cause I went to a dental school or an emergency room to get something done about an impacted wisdom tooth that flared up (as Murphy's law would have it) on a Friday evening of a three day weekend, figured I'd die of it if I had to wait until my dentist appeared on Tuesday; the guy told me he wasn't allowed to prescribe anything for pain, but that I might try swishing some "151"around in my mouth... Took me a second to get it, and I asked "you mean, like, 151 rum??" He said yeah - it'd deaden things up good. Somehow, I forgot the instructions to 'rinse and spit', swallowed a couple times, and found - yup, I was getting pretty 'numb' - and a new bad habit was born! I remember getting so 'comfortably numb' a couple times that I laid on a pool table motionless for a couple hours...

Ah well, I digress! I have learned boat-loads about your RAM; one interesting thing is that especially with the P/Q4x northbridges, not only does it react differently with each board implementation, but sometimes on 'supposedly' identical boards - eeek! Voltages, and response to the MCH voltage variations, seem to have radically different behavior from board to board. One thing that seemed consistent, though, is that the RAM seems to 'like' having its tRFC (Refresh to ACT delay - in the lower RAM timing adjustments) somewhere between 60 and 72 - they actually have an SPD 'patcher' that fixes the SPD, so that boards whose normal tRFC adjustments don't go that high, can work by reading the modified SPD, and 'forcing the issue'...

When you have the time, here's some of the stuff I found:
Memory Voltage.....what do i set?

--------------------------------------------------------------------------------

The question of memory voltage comes up a great deal, i will try to explain the best way to find the optimum voltage you need and explain why you will see 2.1V DDR2 or 1.8V DDR3 when in some cases we will advise you to test and run lower than this.

OCZ qualify memory on many boards, for the sake of this guide i will talk about 8000 Reaper X 4GB, take this though as all our memory with the references to voltage etc covering all our product.
8000 Reaper X http://www.ocztechnology.com/product...pc_4gb_edition has specifications that say 2.1V operating voltage. The reason for this is as follows.

OCZ test and qualify the memory on many different boards, not all boards need 2.1V for the memory to work well, but some do. Because of this we have to state the highest needed voltage that was needed for the modules to work as advertised, often though this is NOT needed. So board A needed 1.85V but board X needed 2.1V, to make sure A and X work with everything in between we state the voltage for board X.

Now this also covers OCZ a little as yield at any given voltage can go up and down. This means over the life of the part the quality of the IC's may get better and require less voltage to run at 1000MHZ or may get worse and require a little more voltage to run 1000MHZ.

So why do OCZ forum staff ask you set ONLY the voltage the memory needs and not just set this max voltage as stated on the website?

We do this knowing in many cases there may be no need to run the voltage that high. Many boards (high end) have awesome memory overclocking and usually operate at much higher frequencies at lower voltages...hence we advise you test and ONLY set the voltage that is needed to be stable.

So why do some boards work better than others?

Many of the top enthusiast boards overclock memory very well, the reasons for this are the PCB's and components used on the motherboard have been tuned to keep memory signaling as clean as it can be, this in turn means it is able to run at higher frequencies. This is why top flight enthusiast boards are more expensive than mid range channel boards ( which usually focus on feature set over overclockability)
These enthusiast boards also feature settings which allow much finer tuning of the memory voltages, timings and clock signals. This makes it possible to fine tune memory at higher clocks using less voltage.

I thought just adding more voltage helped memory overclock?

It does, but there are 2 ways to get the memory doing what you want.

1 Fine tune the clocks and timings using the options in bios.

2 Force more voltage to the clocks and memory and use brute force to make it work.

Now obviously option 1 is what we want end users to adopt, the reasons for this are mainly it helps to ensure a long a productive life for the memory, and it produces less heat which can actually mean it runs better.


Overall you can either just follow the specs that came on the packet or off the website and set that voltage; or you can fine tune using that voltage as a base to start and set the voltage the memory actually needs to run and no more. Be aware though many boards actually overvolt by default, if you set 2.1V in bios you may be actually supplying 2.2V or higher to the memory. Also the AUTO setting does not usually set 1.8V DDR2 or 1.5V DDR3, it normally set quite a bit higher than this. On some boards I have seen as high as 2.0V for DDR2 and 1.78V for DDR3 all hidden behind the AUTO setting.





If you are running ANY type of Intel or 680/780i chipset DDR2 motherboard with 2x2GB for 4GB or 4x2GB for 8GB of ram, tRFC will have a major impact on how your system will run.

you may see option for tRFC called "Refresh to activate delay" so please don't be fooled. If your bios only goes as high at 42 you are going to have issues with speeds much over 900MHZ, in some cases some boards even have issues at 800MHZ.

We have changed SPD on many of our 2GB modules to set tRFC to 54 leaving the bios option at Auto, you do NOT need this SPD if your motherboard has this option in bios and it sets higher than 42, only ask for this update if you have a board that sets 42 or lower AND you can supply the SPD file to us off the modules you have showing a tRFC values of lower than 54 clocks.(we will supply the tool that allows you to send SPD to us)

tRFC has a massive impact on stability with 4 and 8GB kits due mainly to the massive increase in density and the way the memory has to be refreshed, most boards were originally setup for 1GB modules, this is the reason why some set tRFC to low for 2GB modules to work well.

Please if you have bought a 4GB kit and you are having issues look for tRFC in bios and set to at least 54 Clocks, in many cases you will see that you can lower voltage to the memory quite dramatically....





I've been experiencing similar difficulties with 4x2gb of OCZ2RPR10664GK modules.

Things I've learned (so far):

1. The relationship between DRAM Voltage, MCH Core, and tRD is very touchy. You just need to play with combinations of values @ stock frequencies; and eventually you'll get there. Once you establish a solid baseline, you can try to overclock your system from there.

2. You'll probably need to set your DRAM voltage near the upper limit of the rated voltage for your modules to run 4x2gb. My max rated v is 2.30, and I've determined my "Sweet Spot" is at 2.24v-2.26v.

3. Add to MCH core slowly. I've found that immediately jumping MCH up to 1.2v is not a good idea. Try 1.6 and then 1.8 first. From trial-and-error experience, I've learned that too much MCH is just as bad as too little.

4. Try "loosening" your DRAM timings a bit. My rated timings are 5-5-5-15, but I'm not stable (so far...) unless I set them to 5-5-5-18.

5. Use MemTest86+ as your first stress-test evaluation of a settings scenario. If it passes that, move on to Intel Burn (set @ maximum to make sure most of your 8gb is utilized...) If that passes, only then move on to prime95 testing. The concept here is to gain confidence that your DRAM is stable before committing to a 12hr. p95 test.
 

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
Wow, Bill...151 proof! We have that on Barbados, but they tell me that you cannot bring it home on a plane due to fire restrictions (!). I usually just stick with the normal Mt Gay rum and it seems fine. I remember my father going out in the garage and pulling his own teeth, but that's not me. :eek:

Thanks for all the research while you're under the weather! WOW! I've got the family here so will have to put that on my 2am agenda, but it seems unbelievable, yes? All is now solid at 1.84, so, we'll see.

I'll keep you posted and we'll improve on my setup when you get fixed.

Take care and thanks, my friend.
 

rockyjohn

Distinguished

Unless I missed it, I still don't see OldAnal (sorry just had to say it) report the correct number for his mobo. I assumed, based on just on what was close - that UD35 had to be UD3R - easiest to mistake an "R" for a "5". I did not specifically check but I think all the EP45-UD3X boards must have the same voltage. Still it would be nice to know specifically.

Old - are you sure you are running at 1.84v? I did not think board allowed that fine of adjustment. Are you sure you are not running at 1.8 + 0.4 = 2.2?

Whatever its at it is good its running. Now the goal is to get that overnight test with no errors. If you get errors, and your voltage really is only 1.84, I would jump it right up to 2.2v and test that. The process I recommend is to get it running stable at the manufacturers suggest voltage, then after everything is rock solid, try reducing the voltage to 2.1 and maybe 2.0 and see if it still is stable.

If you cannot get it stable at 2.2, I recommend leaving it there - letting us know - and perhaps we can suggest some other bios changes to make it so. Then when stable, try moving the voltage down.

Are we in sync on this bilbat?

Also Old - while bilbat is very knowledgeable about these things - there is another resource you consider - especially if bilbat is wounded and MIA - tweaktown.com is actually has a official Gigabyte forum monitored and supported by Gigabyte techs to help with problems like this. I got excellent help from Lsdmeasap and strongly recommend him. I posted my problem on Thanksgiving day no less and had a response from him within an hour. He walked me through making several bios changes and tests over that weekend - and I always received prompt responses from him. You might want to give it a try. If you do - please keep us posted as well.

http://forums.tweaktown.com/f69/
 

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
John!

It's a EP45-UD3R, revision 1.1, bios F5.

Voltage is truely 1.84VDC. Selection starts at 1.8 and goes up to 1.82, 1.84, 1.86 all the way to 2.3 if I remember correctly.

It's solid. Runs 3Dmark6 at 7500, going to try prime 95 soon and see if I should go higher. The blurb that Bill send indicates that OCZ ram may be happy anywhere, depending on the board/chipset.

I'm sure learn a lot from you guys and I greatly appreciate it.
 

rockyjohn

Distinguished
Glad to here its stable. How long did you run it on memtest? You should run it over night. When I was working through my problems I sometimes had memtest run through 7 or 10 times correctly, then start making occasional errors on later tests. Need to run overnight - or about 50 cycles - to really prove it is stable. If you have not, I suggest doing this before moving on to running prime95.

I suggest you still go to the tweaktown site just to become familiar with it. Also near the top of the forum pages is a stickie about memory timings with suggested timings including for the P45 boards. It might be more info than you want right now if everything is stable - but it might also give you some ideas about things you might want to tweak later. You can find a lot of information about overclocking too from the techs and other forum users with similar boards - just search on P45 and you should find a lot of interesting and useful info.
 

bilbat

Splendid
So why do OCZ forum staff ask you set ONLY the voltage the memory needs and not just set this max voltage as stated on the website?

We do this knowing in many cases there may be no need to run the voltage that high. Many boards (high end) have awesome memory overclocking and usually operate at much higher frequencies at lower voltages...hence we advise you test and ONLY set the voltage that is needed to be stable.

So why do some boards work better than others?

Many of the top enthusiast boards overclock memory very well, the reasons for this are the PCB's and components used on the motherboard have been tuned to keep memory signaling as clean as it can be, this in turn means it is able to run at higher frequencies. This is why top flight enthusiast boards are more expensive than mid range channel boards ( which usually focus on feature set over overclockability)
These enthusiast boards also feature settings which allow much finer tuning of the memory voltages, timings and clock signals. This makes it possible to fine tune memory at higher clocks using less voltage.

I thought just adding more voltage helped memory overclock?

It does, but there are 2 ways to get the memory doing what you want.

1 Fine tune the clocks and timings using the options in bios.

2 Force more voltage to the clocks and memory and use brute force to make it work.

Now obviously option 1 is what we want end users to adopt, the reasons for this are mainly it helps to ensure a long a productive life for the memory, and it produces less heat which can actually mean it runs better.
 

m4rmite

Distinguished
Jul 19, 2009
16
0
18,510


i have the ep45-ud3lr motherboard and have ocz 2x2gig ram pc1066 2.2v @ 533mhz 1.8v @266/333/400mhz
mine runs fine
goto the bios and select auto on the ram voltage and it will run fine
mine is set at 450mhz and it uses 2.18v
my e8500 = 9.5 multiplier x 450mhz ram = 4275mhz runs stable no issues at all
 

rockyjohn

Distinguished


So OldGuy - it you are running your memory at 1.84v, what speed is it? Are you running at the full 1066MHz or is it throttled down?

What timings are avail for the RAM and what are they currently operating at? I know on my setup (P35 described in earlier post) on reason I went to the higher voltage was to enable the faster timings.

Obviously the most important thing is to run stable - but I don't think you have tried the higher voltages so you don't know if they would be stable. Please correct me if I am wrong.
 

m4rmite

Distinguished
Jul 19, 2009
16
0
18,510
what i stated was the ram runs @ 1.8 at speeds of 400mhz or less
and 2.2 @ 533mhz

also the last line i posted says mines running at 450mhz @ 2.18v
i use the 2.00 ram multiplier so it runs @1800mhz as its quad pumped
this is the cpu @ 4500mhz with the ram @ 500mhz /2000mhz fsb
and you can see the super pi 1M result as well
http://i204.photobucket.com/albums/bb37/m4rmite/295.gif

this is the normal setting i run
http://i204.photobucket.com/albums/bb37/m4rmite/427.gif

this rig runs folding 24/7 ( i give it a rest ocassionally) and has no issues at all
 

rockyjohn

Distinguished
Yes - you said that before. I was asking what OldGuy's - the thread originator - was running at after showing that the reason I was asking is your comment about the options. He might be running it at a substantially slower - less than optimal - speed without knowing it.
 

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
Help me out here, guys.....I'm just a novice at this.

Had a few minutes so took a couple of pix of the complete 1st bios page.

Let me know how screwed-up I am so far...LOL

bios1.jpg

bios2.jpg

bios3.jpg
 

rockyjohn

Distinguished
Your last report prior to the one immediately above stated that you were running stable. Are you still getting errors? Am I reading the DRAM voltage correctly as 1.84v? If yes to both I can only repeat what I said two days ago - the very first thing you need to do is increase it to 2.2v.

And if still not stable after that, the next thing I recommend doing is what I suggested last night (see above) - use the forum at tweaktown. It has been two days since you first posted and it appears you are little further along than when you started - this should not be that difficult to fix - take some trial and error testing - but I think youi would really benefit from the official Gigabyte techs that specialize in this mobo.
 

m4rmite

Distinguished
Jul 19, 2009
16
0
18,510



lol whoops just thought u aws calling me an old guy after i saw the quote :sol:
 

OldAnalogGuy

Distinguished
Jul 8, 2009
17
0
18,510
Good morning Bill!

Hope your tooth is better today, or you’re going to see the dentist.

Was able to read most of the thread you posted. I have CPU-Z and Prime95 and, of course, memtest86+.

System had been running fine at 1.84VDC and passing many passes of Memtest, however I would get a lock-up here and there, so I raised the voltage to 2.0 and it has been solid since last night.

One thing that bothers me, though……

According to 3DMARK6 this new system is a slug. 4463 for the new rig and 13,209 for my old rig which is a Q6700 (2.66Ghz) and a 8800GT video card. I know that’s quad Vs dual, but could it be that much different in performance? The old one one is 50-60 FPS while this new one is 15-25!

I’ll be in and out today, but ready to get started. My shooting buddy’s birthday is this Thursday and I’d like to give it to him then. But only if it’s stable, yes?

Thanks again for all your help, Bill.