Fortron 350w PSU and DDR RAM problem

[xoham]

Hello,

Once again I am in need of the collective wisdom of this forum.

I've just read through the stickies and didn't really see the answer.

I have a problem running two sticks of DDR 3200 RAM in my system. Of course I checked them both out by themselves and they work fine. But with both, I cannot even post. I updated every driver including the processor(!).

The one remaining idea I have is that maybe I have insufficient power?

I have a Fortron 350w with the fan speed turned all the way down. It is powering the following
AOPEN 89Max mobo
AMD 64 3200+
Radeon 9800 pro
2 Seagate 120 SATAs
1 CD/DVD r/w combo
1 floppy
Integrated audio and ethernet
1 120mm fan
1 Zalman AlCu psu heatsink fan
1 Kingston 512 3-3-3 2.5v DDR 3200

If I add another identical stick, no POST. I don't think this is a heat problem because it happens immediately from power up and I suspect nothing is very hot yet. The AOPEN guys are trying to help me but so far we haven't figured it out. I followed on of the links I found here to a web page that estimates power usage based on components you select and it estimated 348w. This is where I started to get suspicious.

The problem is, I don't really know which devices are using which voltage (3.3, 5, 12). If I did, it would be very easy to see if I am beyond the limits of the PSU. I could use some help identifying which voltage rail is used for each part. Note that the microprocessor alone is near 90w.

Also, if anyone has solved such a problem by turning up voltage on the RAM, I would like to know how high they went up. The Kingstons are specified to require 2.5 +/- .2v so I turned it up to 2.7 and no luck. I will not hold anyone responsible if I burn them up, just looking to see what people have tried here.

Thanks in advance for your sound advice (pun intended).

[luminous]

To me it sounds like a mobo issue. Given that you have been in touch with Aopen I suspect that you have the latest BIOS. If not, do this.

As this is a fault finding mission, disable anything that you do not need so that the machine can POST. With the one stick of memory take a trip into the BIOS and disable any fancy RAID controllers etc.

I would suggest that after doing that you unplug hardware components until only the following items remain: CPU,HSF, Video card, ONE stick of memory, keyboard and mouse. After this the machine should power up nicely as only the one stick is present. Then add the second stick and see what happens.

With only those components present your PSU (as long as its not defective) will not have any problems powering up your machine. The video card you have is powerful, but will not draw any where near its maximum draw when attempting to POST. With that config you will not be close to even 200W.

If you cannot POST with both sticks of ram, and there is no BIOS update. You will probably have to change out either the mobo or the RAM. With luck Aopen will work with you to solve this problem.

I hope you get it fixed soon, post back here with the results so that we can try and help again (if needed)

[MikeC]

That's a board, CPU & HSF I am using, and I had an odd problem that stumped me for a week or more. It had to do with the system not posting at all, suddenly, after it had been running fine for a several days.

I tore my hair out -- and every part in the system -- for days, maybe weeks. It finally came down to this: The back plate for the HSF that sits under the motherboard had to be changed for the Zalman Z7000. It turned out that the insulating material on the plate was just thin enough that a sharp portion of the underside trace on the PCB made contact with the metal of the plate, which of course, caused some kind of short back to the CPU.

I pulled the back plate off, added a thin piece rubber sheet for a better insulation layer, reassembled, and the system worked perfectly well after that.

Maybe you have a similar problem, but somehow a different part of the motherboard PCB is making contact and it has to do with the memory portion of the circuit??

I'd say it's worth taking a look & adding another insulating layer between motherboard and HSF back plate. See if it fixes the problem. Not much to lose...

>edited for terrible typos!<

[NeilBlanchard]

Hello:

I just built a similar machine:

AOpen AK89 Max
Athlon 64 2800+ (@1.35v!)
Alpha PAL-8150 (which uses a backplate) w/ 80mm Enermax
Mushkin PC-3200 512MB 2.5-3-3 - 2 sticks -- this is just $85 each @ NewEgg!!
Samsung SATA 80GB - 2 drives in RAID 1 array
Samsung CD-RW
Sony floppy
Sapphire ATi Radeon 9600SE 128MB
3Com V.92 Data/FAX modem
CleverPower 300watt PS
120mm Enermax
Evercase 4252

I would definitely try boosting the RAM voltage at least to 2.6 volts -- this is the standard for most RAM. And if that doesn't work, go up to 2.65v or 2.7 and see what happens. Also, you might have to try using different slots -- there may be issues with timing, and who knows how this might change things?

I doubt very much that the PS is at fault for the RAM not working, but try the RAM voltage increase -- at least to 2.6v -- you can always turn it back down again...

[xoham]

Hello:

I would definitely try boosting the RAM voltage at least to 2.6[/] volts -- this is the standard for most RAM. And if that doesn't work, go up to 2.65v or 2.7 and see what happens. Also, you might have to try using different slots -- there may be issues with timing, and who knows how this might change things?

I doubt very much that the PS is at fault for the RAM not working, but try the RAM voltage increase -- at least to 2.6v -- you can always turn it back down again...

As I mentioned, I have already tried 2.7 volts and still no post.

I will likely try disconnecting some components that I don't need to post. Aside from MikeC's idea, I can't see what else I might have missed other than power. Let you know what I figure out.

[MikeC]

I think checking for a possible short -- and adding more insulation if necessary -- pn the bottom support plate is worth doing. It is so unlikely to be a power supply problem that I would almost rule it out. And it sounds like you have checked everything else.

[lenny]

Maybe you can try increasing the RAM latency? I had a problem once where it would post but after a while it would hard freeze. I was using SPD and the information in there was incorrect. Maybe you can install one stick, set timing to manual and something conservative, then try again?

[SpyderCat]

Hi Xoham,

You're describing a very common problem with AMD64 CPUs/chipsets.
Try running the RAM in slots 1 and 3, not in 1 and 2.
And pay a visit to the MSI forum. It's amazing how many people have problems getting their machines up and running. Then again, the users that don't have a problem (like me) don't post.

EDIT: And try a different graphics card.

Have Fun!

[NeilBlanchard]

Hello:

I didn't see the part about you trying 2.7 volts. It could also be a RAM compatibility problem -- the PC4000 Mushkin I have tried only works at the higher latency (3-4-4) at PC3200 speeds and the super low latency PC3200 Mushkin (2-2-2) only worked on one motherboard that I tried it on (so far): the Asus. It would not even POST at 2.65 volts on the ECS mobo! So far, I am very impressed with the cheapest Mushkin PC3200, which runs *better* than it is rated for just using the SPD -- it runs at 2.5-3-3, and only costs $85 for a 512MB stick. I have only tried it on the AOpen mobo.

This is as good as the much more expensive Level one Mushkin!

http://www.newegg.com/app/ViewProductDe ... 219&depa=1

I also have had good luck with Corsair RAM, though I have not used it yet on a Athlon 64.

http://www.newegg.com/app/ViewProductDe ... 219&depa=1

There are apparently more issues with double sided RAM than with single sided -- does anybody know of a single sided PC3200 512MB, hopefully with good (low) latencies?

[luminous]

My own personal thought is to contact your mobo maker about ram compatibility. There are a few companies out there that will recommend modules that have known good compatibility. Some manufacturers will even say which ones won't work with their mobos. Thats got to be worth a shot rather than buying and hoping they work.

[xoham]

My own personal thought is to contact your mobo maker about ram compatibility. There are a few companies out there that will recommend modules that have known good compatibility. Some manufacturers will even say which ones won't work with their mobos. Thats got to be worth a shot rather than buying and hoping they work.

I have been in contact with tech support at AOPEN and they have said they have no trouble using the identical RAM. No luck there.

[luminous]

Well at least you know that you are not banging your head against a brick wall. Since they are meant to work in your mobo and they don't, is there any chance that you can RMA the memory?

I'm struggling to think of anything else that can cause you the hassle. It was suggested in this thread earlier to try and use different slots in the mobo - I take it you've done that and its not helped?

If you have another system lying around, you could try using a different graphics card and PSU. Computers are strange things sometimes. Logically swapping out components may just point to something we have missed.

[xoham]

It seems like a pretty subtle problem if both sticks of RAM work fine by themselves but not when both are plugged in. That leads me to believe that they are good and it is some other problem. I'm pretty sure I have also tried slots 1 and 3 but I can try them again. On my last system when I had a RAM problem, the new stick would not work at all, by itself or not.

This weekend I should have some more data as I will have time to test these ideas out.

I'm going to try unplugging some stuff, drives to start, and seeing if that helps. Then I'll look at the back HSF mount for the problem MikeC saw. Again that seems like a unlikely set of circumstances since one stick works and the memory controller is in the processor. I'll also see check the reported voltages in BIOS and see if they are all sagging significantly.

[luminous]

Good luck. There is nothing worse than a computer that does not work

Hopefully in a few days you will be folding with the rest of us

[xoham]

Seems that Neil had the best answer. Ultimately, the memory voltage needed to be increased to 2.8v. I may experiment and dial this down to 2.75 and see how it goes because this is .1 volts over Kingston's specification.

This is all the steps I took:
-Upgraded AOPEN BIOS driver to 1.07a, a beta version
-Upgrade AMD CPU driver, first time I've ever upgraded a CPU driver.
-Tried all slot permutations
-Unplugged everything except what is needed to see POST. This did not help and basically proved that the PSU was not the problem as others have guessed.
-Read on AOPEN's forum about a related problem and followed the suggested solution of resetting the BIOS via jumper (after upgrading BIOS). I then loaded BIOS defaults and saved them. Now I can POST but still lockup on the windows desktop.
-Dialed the memory frequency down to 333. This stopped the crashing but is an unacceptable 25% decrease in memory performance.
-Reset frequency to 400 and turned up voltage beyond the 2.7 I had already experimented with. This stopped the freeze ups and allowed me to type this message in.

The answer may have simply been the memory voltage but I'll never know. Thanks for all the suggestions and I hope this solution helps others out. Note that I don't suggest monkeying with voltages as first solution as things may get fried. Tune at your own risk.

[luminous]

Yeeeeeeeeeeaaaaaaaaahhhhhhhhhhhh Result !

I do not understand why you have had to increase your mem voltage, they should have worked at their suggested settings. The good news is that you have done enough testing for you to know its not your PSU. At least now everytime you have a problem you will not be suspecting its your PSU

I'm far from being an expert in A64 stuff, but I thought it had an inbuilt memory controller. Afaik this meant that memory speed did not really matter. Maybe you will not see a performance hit even if you ran them at only 333? If you are uncomfortable at having to overvolt you could try this. Either that, or stay at 400 but relax the timings.

Glad to hear that you have got it all sorted out One day I may move into the 64 bit future. Something tells me that I will have to wait a while..........

[xoham]

Bad news. The system is not stable. I can play games for a few minutes before a crash-reboot or a lockup. Even if I go all the way to 2.85, it doesn't seem to matter. Turning down the memory clock seems to help, which of course is bad.

So, the problem is not really fully solved. I'm concerned the memory is simply not capable of running at DDR 400. Either that or the clock signal on the motherboard is not clean enough for the memory to function. I think in the future I will fork over more for either registered, and/or CL2.5 stuff.

Yes the A64 has the memory controller intergrated. But, I don't think that means the speed of the memory bus is no longer relevant. The speed of the memory bus will continue to determine performance.

As far as I can tell, this is not a PSU problem still. I think it is a clock signal intergrity problem or a memory problem.

[NeilBlanchard]

Hello:

I'd run the memory test that others have mentioned and if it fails (at 2.6volts) then RMA the RAM...

[lenny]

Turning down the memory clock seems to help, which of course is bad.

So, the problem is not really fully solved. I'm concerned the memory is simply not capable of running at DDR 400. Either that or the clock signal on the motherboard is not clean enough for the memory to function. I think in the future I will fork over more for either registered, and/or CL2.5 stuff.

What timing are you using on your RAM?

You may be right that your DDR 400 is not really DDR 400. A friend bought some no-name (at least, no one has heard of that name) RAM before, and when I tested his Athlon XP @ DDR 333 speeds with MemTest86, it flunked. It passed at DDR 266 speeds. And this is with 3-3-3-8 timing.

As for my own experience, I was using DDR 333 as DDR 266 (it was cheaper at that time). The RAM was 2.5-3-3-6 but SPD apparently told the system it is 2-3-3-6 @ 266 MHz. One stick worked fine. Two sticks lead to hard freeze after a short while. After changing memory timing to "manual" and forcing it to run at 2.5-3-3-6, it never froze. Memtest86 and Prime95 are all happy.

BTW I believe you won't be able to use registered memory on your system. Besides, they're typically slower.