A Problem I'm having with my DFI Infinity NF2, and 3D stability at high FSB speeds

[felinusz]

A Problem I'm having with my DFI Infinity NF2, and 3D stability at high FSB speeds

I am having a seemingly sourceless problem with 3D stability, and high FSB speeds, with my DFI Ultra Infinity NF2.

My memory, processor, and motherboard are all fully 24 hours stable in both memtest86, and Prime95, yet I cannot get 3DMark2001, or any other 3D applications, to run for more than a few hours with a high (245+ MHz) FSB speed.

What happens, is 3DMark2001 will run fine for 10 minutes to two and a half hours, and then close without explanation. There is no error message or anything, it will just kick out to the desktop without explanation. After the machine has been on for a while, the time 3DMark will run often decreases to less than a minute; 3DMark most commonly closes mid way through the lobby tests, or the Dragothic High test in 3DMark2001. I can run 3DMark long enough to get a benchmark score, but it is flakey, and *not* 3D stable at high FSB speeds.



I have tried eliminating my processor as a source of error, buy running it at 250 X 9 (2250 MHz), with a 1.9V Vcore. This chip does 2700 MHz Prime95 stably at 1.9V.

I have tried overvolting the AGP bus voltage to 1.8V, and also tried running the AGP bus as high as 80 MHz - neither helps.

I have tried loosening the Alpha timings on my memory, tried turning CPC off, and tried loosening the other timings as well - no go.

I have tried every single Alpha Timing combination that I've run across in all my searching.

I have tried overvolting my board's chipset as high as 2.2V, no help. I have a Thermalright NB-1 on the northbridge chipset.



I would like to stress that the machine is 24 hours memtest86, and Prime95 stable at a 250 MHz FSB. It's only in 3D that I run across this random application closing, and apparant instability.

This is a big deal to me, because the entire point of my using this board, was in the hopes of getting up to a benchmarkable ~260 MHz FSB, for some mad 3DMark scores, and hopefully a world record AXP+9800, with luck. 250 MHz FSB is "Benchmarkable", but NOT STABLE. I'm big on stability, especially when I expect that this hardware should be capable of 250 MHz on the FSB for day-to-day use.

Oh, I should add that I had the EXACT same problem with my old NF7-S, except that it's limit seemed to be ~225 MHz FSB. Could my video card be the problem?



My detailed system specifications are below. I'm currently using Hellfire's 8KA REV 2 Modified 6/19 BIOS.

Mobile Barton 2600+ IQYHA 0348 SPMW
2 X 256 Megs Corsair XMS PC3500 BH-5 (does 250 MHz memtest86 stable 2-2-2-11 with 3.45V of VDIMM)
DFI Ultra Infinity A NF2
ATi Radeon 9800 Pro
2 X 36 Gig WD Raptors in SATA RAID0


Any advice and help is greatly, greatly appreciated. Thanks in advance, and a standing ovation for anyone who can help shed some light on my problem.

I'm desperate for help with this. I've tried everything I can think of, and am running out of options .

I'll be crushed if I can't figure this out . I have also posted this problem, in a similar thread, on the dfi-street forums.

 

[deception``]

Before I begin....

Allow me to just say that, believe it or not, this is indeed a very common issue and not just to DFI motherboards. There are many instances in which a system will memtest certain speeds stable, yet the 3D benchmarkes provide otherwise. So don't feel like you are the only one in this situation, as there are many others looking for 250+ glory at the moment. Now, let's delve deeper into this matter:

1. First and foremost, let me simply re-iterate that patience must be synonymous with these boards. I do feel your frustrations, but just know that your quest for fsb glory will not be realized so soon. One thing that most of us had to learn the hard way (e.g., without a plethora of information as you have available now) is that these boards require a lot of patience and diligence before true results can be had. So, above anything, stay patient with your setup. And remember, hard work will go a long way with any DFI offering.

2. Secondly, also know that certain setups respond differently to the numerous bioses available for these boards. Of course, I recommend the HKA lineup for their wonderful bios options and SATA support, but there are a number of other bioses you can try. Also, there are some old-school DFI enthusiasts that only use some of the older bioses, as they feel as if they provided the highest clocks and 3D Mark stability. So you may need to stray away from the timing-rich 6/19 bios variants to achieve your ultimate goal. However, be warned that, in dealing with the older beta bioses, there will most likely come a degradation in stability. Trust me when I say that you're patience has not been tested until you give some of those a shot.

3. Next, you should take the time to play with as many bios options as possible. As I mentioned in a pm earlier, I had to increase my AGP voltage to 1.8 simply to achieve 3D stability at fsb speeds past 248 or so. Of course, your results may vary here. So it may turn out that you'll have to play with settings as meager as the PCI latency. I cannot quite explain the sometimes uncanny combinations necessary, but such are the things that must be done to push a board to it's limits. Oh, and you should also disable any uncessary ports and/or settings that you do not use regularly, too.

4. Before I conclude this post, consider this: Unlike your NF7-S, the DFI draws the vdimm from the +5 volt. I say this because there is a slim chance that your Antec 550W simply is not up to the task of supplying the necessary voltage on your memory. Personally, I don't even think that your Corsair bh-5 should need 3.45-3.5 volts to run 250 stable, but that's just me. Anyhow, know that earlier Antec models are castrated due to the fact that they were not built according to more future proof (namely A64) specifications. So my older Antec TP 550W only had about 24A on the +12 volt, whereas my newer 550W came with 30A on the +12 volt. Of course, this is simply an example, but the fact of the matter is that I do not feel that the Antec is adequate to get the job done.

Once again, these are merely things to consider. Please post back if you have any questions.

deception``

 

[felinusz]

Thanks for the detailed and informative post deception.

I really appreciate your words dude .

2. Secondly, also know that certain setups respond differently to the numerous bioses available for these boards. Of course, I recommend the HKA lineup for their wonderful bios options and SATA support, but there are a number of other bioses you can try. Also, there are some old-school DFI enthusiasts that only use some of the older bioses, as they feel as if they provided the highest clocks and 3D Mark stability. So you may need to stray away from the timing-rich 6/19 bios variants to achieve your ultimate goal. However, be warned that, in dealing with the older beta bioses, there will most likely come a degradation in stability. Trust me when I say that you're patience has not been tested until you give some of those a shot.

Time for me to start downloading BIOSes. Do you know of a good resource that has a wide selection of different DFI Infinity/LP BIOSes for download, both official, and modified?

. Next, you should take the time to play with as many bios options as possible. As I mentioned in a pm earlier, I had to increase my AGP voltage to 1.8 simply to achieve 3D stability at fsb speeds past 248 or so. Of course, your results may vary here. So it may turn out that you'll have to play with settings as meager as the PCI latency. I cannot quite explain the sometimes uncanny combinations necessary, but such are the things that must be done to push a board to it's limits. Oh, and you should also disable any uncessary ports and/or settings that you do not use regularly, too.

I'll start in on the miscellanious BIOS settings tommorrow night, and see if I can find the golden configuration for my purposes. I hadn't thought of PCI latencies, but indeed, I really should be trying everything at all possible before 'losing hope' on this setup.


You're entirely right, I've only had the board up for a few days now, and it's far, far, far too early to call it a night.

So my older Antec TP 550W only had about 24A on the +12 volt, whereas my newer 550W came with 30A on the +12 volt.

The one I'm using is the same, it is the older (oldest?) model that only has 24A on the 12V line. Unfortunately, I can't afford a new PSU at the moment (I guess it's finally time to sell that "unnessacery" half of my liver...), due to a semi-stupid, and altogether unnessacery impulse purchase (see my 'location').

My rails have testing out to be very close to where they should be, checked with a multimeter from the PSU molex, although I have no idea what kind of voltages the board is actually seeing, and is subsequently providing to the various components.

MBM 5, and the Winbond Hardware Doctor both report different voltages.

My VDIMM voltage checks out as well, but I have no real way of checking to see if the 5V rail is truly being stressed. My 3.3V rail is also overvolted signifigantly, something I might try removing.

Personally, I don't even think that your Corsair bh-5 should need 3.45-3.5 volts to run 250 stable, but that's just me.

This struck me as strange as well. However, these sticks are responding really well to the memory burn-in sessions I've put them through (the first time I've ever actually gotten any results or gains from a burn-in on any hardware...), leading me to believe that they're relatively "fresh" sticks, that haven't seen much use at high voltages (I bought them used).

 

[Dukemurmur]

I had the same problems with my board old setup in sig i freid the board and killed the 9800pro....any way i had the same problem above 248 i could get in at around 255 and run widows and basic apps but AS SOON as i ran anything like sandra or 01 or 03 it would run then crash and so on i tried EVERY bios under the sun to no avail...so i eventally settled with 248 "stable" ran all games fine for HOURS but would only run 01 for about 4-5 hours....why idk but this issue i have heard mostly only with DFI boards

 

[felinusz]

I've been through every single official DFI BIOS, and all of Hellfire's modified BIOSes, no go . Tried fooling with pretty much every setting in the BIOS, from CPU Thermal Throttling, to AGP latencies, no go .

10 hours of trying BIOSes, and messing with BIOS settings later, I still can't figure out how to get past this wall.

It looks like the board will simply not run stable in 3D, over a 245 MHz FSB. At 245, and upwards, it just closes 3D applications without explanation, no matter what I try.

It's benchable at a 250 MHz FSB, just not stable for day to day use.


I've accepted the fact that this board of mine simply has a slightly bunk chipset/AGP controller; not a lemon, but slightly below average.

What really gets me, is that there's nothing I can do about it either, except buy a new board (don't worry , I'm not going to waste *more* of my meagre savings) .

 

[RJARRRPCGP]

Maybe you're required to raise the Vcore setting even a little bit more.
Not likely the processor, despite programs closing with no message is likely the processor core. Maybe, (likely) the AGP bus is being overclocked. Make sure it's locked at 66 mhz!!!

 

[Dukemurmur]

it ALWAYS is.....i have never seen a case with this mobo that the AGP bus sint locked...