G.Skill 8 GB (DDR2-1000) on GA-MA78GMS2H 4-Minutes to POST

[Carson]

Hello,

I built this system last November, and for 7-8 months now, I have been trying to solve this startup issue. NCIX told me the RAM was "99.9% sure" to work with this mobo. And it does—but first, on any cold start, it takes some 3 to 12 minutes JUST to POST. (That is BEFORE it even thinks about proceeding to the bootscreen and up into Windows, so it is a very long startup.)

The BEEP at the POST is always a single beep, BUT it "stumbles" or "hiccups", like this: bi-BEEP! (or any variation of that). But it is DEFINITELY A SINGLE BEEP, as it should be—just a distorted one, as though it trips.

Restarts from Windows always work properly. (So do their beeps.) Windows XP x64 and XP x32 work fine. The problem is only with the cold boots' very long delay before POST.
— The long pre-POST times are with the BIOS DRAM set to AUTO (default Trfc @ 127.5ns). No errors are detected in the CMOS. But if I manually set the Trfc ratings to 105ns, 127ns, or 327ns, I get a nice quick POST time but then a CMOS error is detected. I can easily reproduce this; never an error when the Trfc is set @ AUTO. (I keep charts of all cold boots on an available .PDF.)

The DDR2 voltage control in the BIOS is set @ +0.2V = 2.0V. No other voltages are modified. The system runs Windows XP x64 or x32 without bluescreens and without crashes.


Trying the system without the hard drives and with just one stick of RAM reduces the POST speed but does not correct it. In other words, it is evidently not a hardware problem.
—I have Thermaltake's Dr. Power, which says my Corsair HX520W PSU is fine. Both Corsair's calculator and NCIX agree this PSU is right for this load.
—Temperatures usually run @ 40º-50º C.

After endless experimenting, we think it might be a problem with the RAM's compatibility with this Gigabyte mobo. (Firmware on the mobo is up to date.)

I believe the tRFC is supposed to be the result of a calculation. I think two sub-timings added together should be less than the tRFC. However, this is the limit of my knowledge and at this point I quickly don't know what I'm talking about: can someone please help?

At this level which I don't understand, I have been told there is an SPD flashing tool and instructions that might correct this problem—or that all the memory timings can be entered manually, bypassing the SPD settings currently being used. (That would include the sub-timings too.)

But here I am sitting in the Restaurant at the End of the Universe, and I really need some guidance to proceed. I enjoy learning, but I don't want to make any rash mistakes. And I would LOVE to hear nice quick POST beeps, that really say BEEP!—the way normal folks do.

Thank you very much.

[BeepBeep2]

Tried any other BIOSes? I just skimmed through the post. It's also set to 2T also, right?

These CPU's don't exactly like 4 modules at the same time either.

[Carson]

@BeepBeep, it is set to 2T. Thanks for your reply.

I've flashed the BIOS from its original F4 to a new one last Christmas to its current F8. I've also recently removed the battery just to be sure their were no CMOS clutter issues.


These CPU's don't exactly like 4 modules at the same time either. . . . but if that were the problem, the issue should have vanished when I tried it with just a single stick of RAM.

I'm seeing some evidence the CMOS Error! popup might be displayed whenever the DRAM setting is on Manual rather than Auto—regardless of changes.

[Carson]

Here is a .JPG showing the DRAM settings possible in my BIOS.

http://uppit.com/v/VZKT12QV

I also have a .PDF available which details an agonizing 2 dozen or so startups, showing all the logic and all the failed attempts. It's about 22 pages now.

[BeepBeep2]

I'm stumped and it would be nice if G.Skill support would enter the thread.

See, I know hardly anything about intel memory timings, I'm only used to AMD/nvidia settings. Though I would suggest moving the trfc up to about 192ns

[Carson]

@BeepBeep, I appreciate your input.
My system IS an AMD, similar to yours in that my CPU is likewise an Athlon x2 5600+.

Yes, I'd like very much to hear from G.Skill, for sure.

This afternoon I am trying 192ns. I'm also trying 6-6-6-16, but I am very uncomfortable with it, because honestly I don't know what I'm doing.
This is what I call "flailing" rather than informed tweaking. I don't like doing it at all.

I had my voltage set +0.2 = 2.0 total, and 6 months ago that seemed to be required to prevent blue screens. This afternoon I've returned that to AUTO, hoping the system knows what it wants more than I do.

I'm on the verge of wasting the $100 I spent on this GA-MA78GMS2H 1.1 (F8) mobo, and getting a new one. I can't afford to do that AT ALL, but things are looking desperate, unless I become very lucky very soon.
— Also I'm reading about BIOS, but I feel lost. Surely the RAM and the other components are designed by engineers who know what they are doing. I am definitely not qualified for this. I'm happy to give it a seriously good try—and I have been trying now since last November—but I believe I am way out of my territory.

Thanks again.

[bigdaddy25fb]

@BeepBeep, I appreciate your input.
My system IS an AMD, similar to yours in that my CPU is likewise an Athlon x2 5600+.

Yes, I'd like very much to hear from G.Skill, for sure.

This afternoon I am trying 192ns. I'm also trying 6-6-6-16, but I am very uncomfortable with it, because honestly I don't know what I'm doing.
This is what I call "flailing" rather than informed tweaking. I don't like doing it at all.

I had my voltage set +0.2 = 2.0 total, and 6 months ago that seemed to be required to prevent blue screens. This afternoon I've returned that to AUTO, hoping the system knows what it wants more than I do.

I'm on the verge of wasting the $100 I spent on this GA-MA78GMS2H 1.1 (F8) mobo, and getting a new one. I can't afford to do that AT ALL, but things are looking desperate, unless I become very lucky very soon.
— Also I'm reading about BIOS, but I feel lost. Surely the RAM and the other components are designed by engineers who know what they are doing. I am definitely not qualified for this. I'm happy to give it a seriously good try—and I have been trying now since last November—but I believe I am way out of my territory.

Thanks again.
Sounds like you should have RMA'd the board. See if Gigabyte will accept it for an RMA? Any post longer than 1 minute for a system set to QuickPOST is an indication of errors with the memory or the motherboard logic. If you can test the RAM in another board and if it posts fine, I'd say you need a new mobo.

[GSKILL TECH]

Have you increased NB Voltage to 1.4V? It definitely needs that to operate 8GB. Will the beeping occur with just 4GB in either 2 slots? tRFC is not a calculation, it should be set at 127ns or 195ns. Command Rate should be set at 2T. Let me know how that goes, sorry I had to rush this because I'm already suppose to off work but I thought you really needed a response. =P

Thank you
GSKILL SUPPORT

[Carson]

Hello, G.Skill Support

Thanks very much for your response here. No, in all my work on this system, I had never tried setting the Northbridge Voltage up to 1.4.

So I have done that now!

This morning I have made the following changes. I haven't properly tested the results yet, so I will add to this thread when I know more.
NB Voltage — I have now increased it the max available (+0.3V) to get 1.400V
tRFC — I have set all 4 to 195ns
Command Rate = 2T — Yes, it was already @ 2T

Other settings currently include:
DDR2 Voltage — increased by +0.2V to get 2.00V — most of my experimenting has been at this setting

CPU Voltage (for my Athlon x2 5600+) has always been left @ "Normal" but there is really a list of available manual settings from 1.3250V to 1.9000V, in small increments.

Thank you very much for your help. This is something of a breakthrough, and my optimism has flared accordingly!! :D
_____________

[Edit, later same day] I have tried several start-ups, but nothing conclusive so far. I'll know more tomorrow morning after a cold-start after a longer power-off time.

[Carson]

Unfortunately, this morning's startup time was no different from before. This is with the suggested NB voltage of 1.4:

0:00 Power ON
4:19 POST (= d-BEEP! "stumbling BEEP") — i.e. 4 minutes and 19 seconds to the POST
5:00 Boot Screen
5:40 Welcome Screen
6:22 Windows XP x64 Desktop
7:46 Windows XP x64 totally ready to work, with all applications fully operational

[Carson]

Update:

Yesterday I received this email from G.Skill:

Dear Customer

That is a very odd issue, although it can run properly without any issues, it takes over 4 minutes to POST. It simply seems like a motherboard issue, the memory would not cause such issues. Is that the latest BIOS for your motherboard? If so, all I can further suggest is resetting the board to see if it would clear the issue, otherwise you would need to RMA it for an exchange.

Keep me updated on the progress.

Thank you
GSKILL SUPPORT
___________________________________________

Ever since I built the system last November, the mobo has been the main suspect. As bigdaddy25fb says, I would have done well to RMA the board some time ago. However, this mobo was a replacement—unfortunately, an exact replacement—for another motherboard which showed the same problems right at the beginning, when I trusted myself less and trusted Gigabyte's brand new motherboards more. I wanted to give Gigabyte its fair run, before concluding the mobo was performance-challenged. So, because the system functioned properly once it got beyond its 4-to-12 minute delays before every single POST, I have continued testing and trying every imaginable variable all these months.

This G.Skill letter was welcome because here we have an expert on RAM advising that RAM would apparently not cause a recognition problem like this and then work just fine throughout a normal computer day. I had already tested my HX520W Corsair PSU, so am now looking directly to the Gigabyte GA-MA78GM-S2H ver 1.1 motherboard, which, yes, has been flashed to the latest F8 BIOS firmware. (But the problem has remained identical through 3 different BIOS designations on this board.)

Well, I'll certainly be happy to trade UP to a better motherboard if either Gigabyte or NCIX will meet me halfway. I'll post back again here with my eventual results.

[BeepBeep2]

Take out two sticks and run them in same colored slots.

Keep the trfc at 195ns, tell me how long it takes to post.


5600+ 65nm or 90nm?

These things really hate 4 sticks at one time. And anything more than 4 GB.

Gimme a list of all the voltages you can tweak.

[Carson]

This post has 2 parts:
— NCIX response
— BeepBeep's help and voltages

==================================

1.
NCIX

Carson to NCIX, May 4, 2008:


Have I made a mistake in my RAM selection? I have just noticed that Gigabyte does not include DDR2-1000 in its RAM list at all. It does include DDR2-667, DDR2-800, DDR2-1066, DDR2-1100, and DDR2-1200.

This worries me, because DDR2-1000 should come right in the middle.
http://www.giga-byte.com/FileList/MemorySupport/motherboard_memory_ga-ma78gm-s2h.pdf

I still have not begun to install anything. I won't be putting anything together until I have everything I need.

Thanks.
_________________________

NCIX to Carson, May 5, 2008:


Hi,

As I believe, the ram will actually work fine. They don't include that simply because ddr2-1000 is not that common. You can try to install them and let me know if there is any problem.

Thanks
_________________________

Carson to NCIX, August 4, 2009:


Greetings,

Here is a response to email we exchanged last year. As you may recall, after I built my computer system last November 2008, I returned the motherboard to NCIX for a replacement, because the mobo would not work properly in the system. Your technician. . .agreed that the motherboard was not performing right. The system includes 8GM of RAM.

The replacement motherboard, an identical Gigabyte GA-MA78GMS2H ver 1.1, has always required between 4 and 12 minutes to POST. I have tested it extensively all these months, as well as receiving input from G.Skill and other technicians. I have purchased a test device for my PSU, and found the PSU works perfectly.

I am attaching a .PDF which explains the problem in a recent G.Skill thread. There have been many other similar threads. I have documented tests for months now.

What I would like to do, if NCIX is agreeable, would be to trade this motherboard UP to a better AM2+ motherboard that you might wish to suggest. This one is an mATX, but I would be happy with an ATX, and maybe an ATX would be preferable. (It must be AM2+.) Perhaps we could negotiate an agreeable price. I am obviously unhappy with my purchase, and I have spent many, many hours of my time confirming it is indeed a motherboard limitation.

I am open to suggestion. Thank you.

Copy of INVOICE: November 27, 2008
Gigabyte mainboard would not POST. Exchanged at NCIX for over-the-counter replacement (also ver 1.1)

_________________________

NCIX to Carson, August 4, 2009:


Hi,

I have checked on the purchase, but the motherboard is purchased probably more than 1 year and I can't see anything on the invoice anymore. And if the motherboard is defective, you may want to contact manufacture for the warranty. However, if the motherboard is purchased like 1 year ago, then we would not be able to exchange for that anymore since we can't even return to our supplier or resell it. We can try to help you to check and see which part is defective and try to fix that for you, but we won't be able to exchange or replace the parts for you any more.

Thanks


================================================== ==================

2.
BEEPBEEP

— My CPU is an 89 watt 90nm AMD Athlon 64 x2 5600+ 2.8GHz AM2 1000FSB Windsor.

VOLTAGES:

trfc: 75, 105, 127.5 (= AUTO), 195, 327.5

DDR2:
Normal
+0.1V = 1.900V
+0.2V = 2.000V
+0.3V = 2.100V

Northbridge:
Normal
+0.1V = 1.200V
+0.2V = 1.300V
+0.3V = 1.400V

CPU
Normal CPU Vcore = 1.3500V
Increments:
1.9000V, 1.8500V, 1.8000V, 1.7500V, 1.7000V, 1.6500V, 1.6000V
1.5500V, 1.5250V, 1.5000V
1.4750V, 1.4500V, 1.4250V, 1.4000V
1.3750V, 1.3500V, 1.3250V, 1.3000V
and similarly down to
0.800V

RAM

CAS# Latency:
RAS to CAS R/W Delay:
Row Precharge Time:
— all = 5T (of available 3T, 4T, 5T, 6T)
Minimum RAS Active Time = 15T (of available 5T to 18T)

Command Timing = 2T
Tw Tr Command Delay = 3T (of available 1T, 2T, 3T)
Write Recovery Time = 6T (of available 3T, 4T, 5T, 6T)
Precharge Time = 3T (of 2T or 3T)
Row Cycle Time = 25T (of 11T to 26T)
RAS to RAS Delay = 3T (of 2T, 3T, 4T, 5T)

I have now removed 4GB of RAM, and tomorrow morning I shall post the results of a cold boot after the system has been off all night, using 2 banks of RAM (= 4GB; not 8GB of RAM) and with the trfcs @ 195ns. (My system seems very good at holding onto a bit of capacitor energy, so I don't trust "contrived" cold boots.)

— I already know that any trfc setting apart from the AUTO 127.5 default will cause a CMOS ERROR popup.
— USUALLY when there is a CMOS Error, the POST BEEP will sound within just a few seconds, as it should—except in these cases with the CMOS claiming there is an error.
— Tonight I am setting the BIOS to HALT ON NO ERRORS. This won't get rid of the error (whatever the error is), but it will allow the fastest possible boot.

Thanks.

[Carson]

@BeepBeep, please also see last night's post (#13).

August 5-09, 6:45 am, cold start after 7 hours with all power totally off

RAM = 2 sticks only (= 4GB; not 8GB)
Trfc = 195ns
DDR2 @ +0.2V = 2.000V
NB @ +0.1V = 1.200V
Halt on NO ERRORS
Adjusted settings were saved via F10 to the CMOS, but the Profile was NOT modified via F11

0:00 Power ON
10:43 POST (= d-BEEP! "stumbling BEEP") — i.e. 10 minutes and 43 seconds to the POST
11:11 Boot Screen
11:51 Welcome Screen
12:22 Windows XP x64 Desktop
13:45 Windows XP x64 totally ready to work, with all applications fully operational

[BeepBeep2]

You can really only adjust the nothbridge voltage by .1v at a time? That must really suck man...
I would say try +.2 on the northbridge (memory controller) (I run mine at 1.31v to get 2x2GB (4GB) stable on my motherboard.)

But there is a chance you can kill your now $40 CPU. :P

Also, locate the BIOS chip and push it a little to make sure it is locked completely in place.

[Carson]

I'm not ignoring any advice you guys have offered. Nothing new to report yet, but I am experimenting every day.
@BeepBeep, tonight when I shut it down I'll try a touch of pressure on the BIOS chip, just to be sure. Your advice had me go looking to see where the BIOS chip is on my mobo. I was surprised when I found it—it is TINY, just about 0.5 cm x 0.5 cm. If that is what has caused me so much grief for all these months, well. . . . :shocked:

Now I have these settings always:

DDR2 +0.200V = 2.000V (I could go up one more notch, but the next setting is in red lettering in the BIOS, and I think it is extreme. I want every-day settings.)

Northbridge +0.3V = 1.400V, which does indeed seem to improve things a lot.

I have also been experimenting with Delay to HDD (from 0 to 15 seconds).

I'm concerned about my DRAM Trfc settings. A fellow who eventually replaced his PSU said he had the same weirdness, where ANY setting except the default 127.5ns would popup a CMOS ERROR warning. Even so, I have tried running it @ AUTO and the computer can take as long as 20 MINUTES to get its POST! So most of my experiments now are @195 or 327.5ns and I set HALT ON NO ERRORS. But that seems wrong, too.

Cold starts require at least 3 hours. I have tried to release the capacitor memory by powering off the system at the power bar and then holding in the case power button for as long as 100 seconds (timed). Apart from wearing out the spring in the button—I'm afraid it might—that doesn't seem to avoid warm starts posing as cold starts. I have got some really good starts after quick off-and-on power-downs that way, but later, when I've tried a "genuine" cold start after at least 3 hours without power, I have not had such good results.

I am an inch away from RMAing my motherboard to Gigabyte. The problem is, people can point to my mobo, my RAM (DDR2-1000 is fairly rare), my PSU, or my CPU—and people have. I have been told to RMA just about everything I have. My reaction is to believe I might really be able to tune the system if only I can get everything just exactly, perfectly RIGHT.
:rolleyes:

[GSKILL TECH]

Your processor is holding you back. It has a 1000 FSB so you would need to overclock to even have a shot. In fact, AMD states AM2 processors will not operate above DDR2-1066, so that is why you are having these issues.

Edit: I meant AM2 processors will not operate above DDR2-800.

Thank you
GSKILL SUPPORT

[Carson]

@Mr. G.Skill Tech,

Thanks a lot for your help. Earlier, when you told me I should have my NB voltage @ 1.4, I made that change and I have never looked back. You were absolutely right, and now I'm doing ALL my testing with that setting.

As well, my DDR2 is set at +0.200V, although I could go one notch higher into (red-lettered in the BIOS, with a warning) +0.300V.

A bit odd is another apparent requirement that the delay for HDD must absolutely be set at the highest level @ 15 seconds. 12 seconds won't do; the system recycles or "goes around" to a second POST.

So these have become "absolutes" on my system. I have had one crash @ 195ns tRFC, and so far the difference between 195 and 327.5 hasn't been discernible, so I'm testing those two against each other at the moment.

I haven't touched the CPU settings. I'm taking your comment as good news, actually. Lately I have come as close as some fairly reproducible startups at 20 seconds to POST and a total of about 1 minute to the bootscreen. That is very fast for my system! :p:

So if I can OC the processor and get the RAM to function as DDR-1000, instead of at its present DDR2-800 (i.e. the stock limitation of the processor) and if you say that there is a bit of a chance there, well—I am VERY interested in trying that.

I've been told that OCing the 5600+ won't have any effect on a slow time to POST, but my uneducated mind questions that. Without much knowledge, but just making guesses logically, I thought that a reasonably OCd system might be faster from the moment of power ON and might reach POST more quickly.

Thanks again for your input. It is really highly appreciated. :)

[BeepBeep2]

Hey Carson, forgot about this thread...let me read up a bit if you are still watching... :)

My CPU does 1100 just fine after tweaking drive strengths...and there is no reason this RAM should take 4 minutes to POST unless it's not compatible with your motherboard.

Also, a BIOS chip should be more like 2cm x 1.5cm if I remember right...
http://ts3.images.live.com/images/thumbnail.aspx?q=1120322388494&id=59b3e8db39e826af90c03f28d16d247e
Or
http://ts4.images.live.com/images/thumbnail.aspx?q=863206119855&id=87fccfc67c1dbdd84d356003375364e3&url=http%3a%2f%2fwww.nvcc.edu%2fhome%2fbholt%2fist 100dw%2fimages%2fbios.jpg

A NB voltage for a 5600+ seems a bit high to me, because that's the voltage for the memory controller as well.

And, GSkill_Tech is right about the 1000HT thing, I've found you need more HT bandwidth than memory bandwidth to run stable.
So, keep the HT multi at 5x at all costs. If you want to run the RAM at 1000 You will have to set the multiplier to either 11 or 12 and overclock the CPU, up to 3.05 Ghz for the 11x / 12x multi's.

[GSKILL TECH]

With your setup, the best settings would probably be to simply operate at DDR2-800 8GB and tighten timings as much as possible. Any more memory frequency above that will be minimal and simply cause instability, while at 800 it will be more stable and tighter timings would probably compensate for the loss of speed. OC your HT and CPU frequency as high as possible while being stable, and call it a day. You are very near hardware limits, so any more time used would simply be wasted.

Enjoy!

GSKILL SUPPORT



@Mr. G.Skill Tech,

Thanks a lot for your help. Earlier, when you told me I should have my NB voltage @ 1.4, I made that change and I have never looked back. You were absolutely right, and now I'm doing ALL my testing with that setting.

As well, my DDR2 is set at +0.200V, although I could go one notch higher into (red-lettered in the BIOS, with a warning) +0.300V.

A bit odd is another apparent requirement that the delay for HDD must absolutely be set at the highest level @ 15 seconds. 12 seconds won't do; the system recycles or "goes around" to a second POST.

So these have become "absolutes" on my system. I have had one crash @ 195ns tRFC, and so far the difference between 195 and 327.5 hasn't been discernible, so I'm testing those two against each other at the moment.

I haven't touched the CPU settings. I'm taking your comment as good news, actually. Lately I have come as close as some fairly reproducible startups at 20 seconds to POST and a total of about 1 minute to the bootscreen. That is very fast for my system! :p:

So if I can OC the processor and get the RAM to function as DDR-1000, instead of at its present DDR2-800 (i.e. the stock limitation of the processor) and if you say that there is a bit of a chance there, well—I am VERY interested in trying that.

I've been told that OCing the 5600+ won't have any effect on a slow time to POST, but my uneducated mind questions that. Without much knowledge, but just making guesses logically, I thought that a reasonably OCd system might be faster from the moment of power ON and might reach POST more quickly.

Thanks again for your input. It is really highly appreciated. :)

[BeepBeep2]

He's already runnin DDR2-800 5-5-5-15, posting after like 5 minutes :P

[Carson]

My motherboard is on its way to California now. I got my RMA from Gigabyte and sent it on its way—by truck, because air cost as much as the mobo.

I included a ton of documentation.

I'll let you all know when I have something to report, maybe around September 20th or so. Thanks for all your help.

My own best guess at the moment is that maybe the BIOS chip itself was faulty.

[BeepBeep2]

I hope the new one comes back working, I wish you the best.