armygreen11

Yeah, my SMBus dumps were all zero's except for one line (southbridge I'm guessing) nothing else was showing up. This is also all from memory. It's been a while since I did all this.

If you figure it out, definately post your solution up here, and if I use that board again, I'll definately give it a shot. I didn't think that the SMBus was completely dead, just one of the components on it. I think it's an ATXP1 or something... It's over by the CPU socket. I put a scope on it, and there was definately data on the SMBus pin, but it wasn't overriding the VID and FID from the CPU. Could be that that command just never got sent down from the bios. Although, I never got anything from CPU-Z's SMBus dump either. All empty.

It's an SMBus problem on the board. I never tracked it down, just got a new one.

Well, the heatsink definately fits better, but it had no effect on the diode readings. I did change the fan from the tornado to an 80mm panaflo. Temps increased about 4C in all conditions. Everything is just as stable, but now cpu temps are up to 51C after 9 hours of prime. I can live with that. I was doing some reading about older athlon XP cpu's, and evidently they had problems with the diode when they first came out. I have a week 24/03 barton, so it isn't the newest chip in the lot. Playing with the fan speeds more, I can tell that the diode IS functioning as it's fluctuation range definately increases with cpu temp. One of these days I'm going to pick up a duron or something to stick in the other ulltra infinity I have with the bad SMbus for the wife. When I do, I'll stick it in this board first to see if the diode works.

I doubt that EMI would be a contributing factor, mostly because I've seen other people's mbm displays with "normal" diode readings (1-2C over the core temp) on this same board. I'm more inclined to beleive that effects from my heatsink mounting are the culprit. I'm going to take a dremel to the heatsink where the caps interfere and see where that gets me. I think this is a problem beyond a sensor or A/D accuracy issue since the readings fluctuate so rapidly. I have mbm set to capture every 2 seconds and every reading is way different from the last in the ranges I noted in my last post. Unless the heatsink weren't mounted properly (possibility) heat dissipation in different areas of the core wouldn't be possible due to the large mass of the heatsink that is (supposed to be) in even contact with the entire surface of the core. This would pretty much average out any differences in heat dissipation between one part of the core and another. If it isn't mounted properly though, this may be a valid theory. More on this after the dremel....

I hear where you're coming from, but I have read in more than one place that the ultra infinity / LPB have a Vcore "droop" problem at higher voltages. I found a mod to correct it that involves replacing some SMD resistors, which I can do, but now that I have the board at a reasonable speed, I'm just going to leave well enough alone for a little while. I would like to know what's wrong with my diode sensor though. Right now with the fan on low, the winbond core temp is 41, the diode is bouncing around between 24 and 30. If I crank my fan up to high, winbond drops to 38, diode bounces between 18 and 30. Maybe the heatsink is pushing on the core in a bad way. I have it bolted down, but it hits three caps on the one side. Thermalright's website says it's compatible though using the neo backplate. Go figure... I'm saving my pennies for a swiftech water cooling kit anyhow.

Ok, so I got a new board. Right away I put Shinobi's 619XT bios in from the tmod CD and proceeded to OC. With my patriot "value" ram, I got it up to [email protected], 3-4-4-11 or [email protected] 2.5-3-3-11. I couldn't get x9.5 to work at 230, so right now I'm running 225x9.5, [email protected], 1.825V core, 1.7V chipset prime stable. Here are my new problems.... The CPU diode reading is all over the place. If the winbond sensor reads 43C, the diode is anywhere from 23-44C. I used the MBM5 settings from soundx98. http://www.dfi-street.com/forum/showthread.php?t=6798 Looking at my vcore in cpuz, it's all over the place. Set at 1.825 in the bios, cpuz reads it jumping around from 1.76-1.82. For something that you can adjust in .025V increments, a variation of .060V is kinda high, isn't it? I have read that these boards have a voltage stablility problem somewhere. Is there anything I can do about it?

Wow, thanks! if you get the rest of them in there and get it all cleaned up, this would make a really good sticky, especially if all the links to the files are good.

So I downloaded and burned Tmod's CD (very well done BTW) and there are all these different modded bioses to choose from for an Ultra Infinity. Is there a compiled list somewhere of all the NF2 bioses and what their features/changes/pros-cons/etc are? Which one would be best for my setup and why? I did some searching of these forums, but I didn't turn up anything.

Ok, so I just bought another Ultra Infinity. This one is new... well, refurbished, but IMO, that's better than new since you know that someone somewhere actually took this board and tested it. It should be here on Monday, so when I get it, I'll do a realyl good visual comaprison with my board to see if I can see something different.

Sending you a PM on the Ultra B

I might. I do want it gone. Ram swapping didn't do anything. I feel bad selling it to someone on here though. I mean, I'll definately disclose everything that's wrong with it, but still, I feel bad selling something that doesn't work right.

Just did it. No change. I had it running on my desk with a different PS (brand new Antec SP 350) with just the ram, and a different video card. I should have pulled out the sticks of ram one at a time since the SMBus goes through the dimms. I keep forgetting to do that. Hrm.... I'll try it when I get home and put the board back in my rig. I had the ram in my A7N8X though, and I didn't have any problems with it in there. I verified that the voltages being read by MBM and the bios are correct. I'm just going to get a new board. I've spent too much time messing with this one. Anybody want an Ultra Infinity with a bad SMBus? Works perfectly at default speeds. $30.

Ok, so it does look like you have some devices on your SMBus. You have APIC off though, so it may not reflect what it would in my board if it were running correctly. Thanks alot though NEO. Could someone with APIC on please post an SMB dump from MBM?

Ok, so I've made some headway in troubleshooting my board... I've done some more research (actually I found this out when I first started looking things up, but I overlooked some important things that lead me on the wrong path). Here's what I know... [TRAIN_OF_THOUGHT]The Attansic ATXP1 is the overclocking controller. Datasheet (pdf) It recieves data from the SMBus at startup and modifies the VID (CPU Voltage ID, or the CMOS levels that tell the PWM circuit what Vcore to set) and the FID (CPU Frequency ID, or the CPU freq multiplier CMOS level that tells the CPU and chipset what FSB multiple the core should run at). Ok, so this component controls the Vcore override, and the multiplier override. Now you say, "but you can adjust your multiplier, can't you? You never said anything about not being able to adjust your multiplier..." And then I say, "I never tried, but now that I have, no I can't." So.... Now here's where I am now... The ATXP1 has 12 GPIO (General Purpose IO) pins. What I don't know is weather DFI used some of these pins to control Vdimm and Vchipset. We have what, four different choices for the chipset voltage, and eight choices for Vdimm? (I don't remember, but that sounds right) so that's two pins for the chipset, and three for the ram (00, 01, 10, 11 and 000, 001, 010, 011, 100, 101, 110, 111 respectively) with seven pins left over for whatever. So, theoretically, this component COULD control the voltage overrides for the core, chipset, and ram, as well as the multiplier that I can't change. BUT If this component were bad, would it necesesarily pass through the VID and FID that it gets from the CPU? (certain CPU pins are tied high or low internally to provide the CMOS output VID/FID to tell the mobo/chipset what it's default Vcore and multiplier are. Wire mods like the war mod tie some of these bits to Vdd or GND to make them high or low, thus modifying the Vcore or multiplier.) The answer is, I don't know, but probably not. I don't think it's that smart. So what would make the ATXP1 not adjust the Vcore and multiplier (and maybe the Vdimm and V chipset if it really controls those things) but still pass the VID to the PWM and the FID to the CPU and CS? Well.... what would happen if the ATXP1 weren't receiving the SMBus at all? Well, I'm not an electrical engineer, but if the SDA and/or SCL were open, then the configuration registers would remain in default positions. If this happened, then the VID and FID would be passed through. SO Other things happen on the SMBus, besides the control of the ATXP1, right? Isn't the SMBus where the data for fan speed, temperatures, and voltage feedback transmitted amongst other things? Well, I can read fan speeds, I can read temperatures (except for the diode... don't know why) and I can read voltages. So the SMBus is working in general, right? So it just isn't getting to the ATXP1. Even if the Vdimm and V chipset aren't controlled by the ATXP1, they may be set by something else on the SMBus that's further down from the ATXP1 and consequently also removed from the bus. [/TRAIN_OF_THOUGHT] I did a SMB dump in MBM5 just to see if I could see some other devices, but all it showed was the southbridge at register $5100. Could someone else open MBM5, open the settings window, and under General>System Info, click the debug dumps, and then SMB dump, and hit the copy button when it's done and paste it here please? I'm curious to see what a working SMBus looks like on an Ultra Infinity / Lanparty UB. I'm going to pull my mobo again now, and see if I can see anything that could be causing the SMBus to not be getting to the ATXP1. I have a feeling that I'm going to be taking my rig to work and checking it out with a scope. One of the engineers told me to take one home with me, but I felt funny about taking home a $3000 O-scope. Ideas?