ABit AX78 Overclocking/Stability

[KTE]

Sorry for the delay LIKMARK, I've had many many issues and only made posts I can risk losing - lost my post to you about 8 times before I quit [unstable Phenom]

Thats a problem known with the Brisbanes rather than the board I think.

Yep.

I can't change nb multiplier with Phenom on this board (with BAR EDIT). Anyone got any ideas?
Example: changing NB multi with msr register on bus 24/3 from ***101 (1800MHz) to ***100 (1600MHz, seeing it change "on the fly" in cpu-z)
Reboots, and "bang" back to stock speed.

You can only change NB multi using the NB registers through tools like CCPUID/Bar_Edit, but they don't set it until after a restart [CPU limitation].

It looks like you have a BIOS limitation or a CPU limitation, as 9500/9600/9700 should be capable of dropping NB multi. I had one 9600BE which did not drop NB multi in the same BIOS which dropped NB multi for the second 9600BE, so its very possible

Has anyone else tested this with a Phenom on this MB?

NB multi is applicable with either BE Phenom and/or earlier/later bioses than bios #12.

I'm not sure if it is my Phenoms memcontroller or the board not liking ballistix, but they can't oc as well as with my Athlon X2 on Crosshair. Awaiting new ram to test with. Today I'm running 4GB Ballistix PC6400 @940MHz 5,0 4 4 12 2T, with relativly high latencies on the subtimings @2,0 vDIMM.

Cas 4 is almost a no go (at reasonable vDIMM), this was also the story when running Phenom on Crosshair (AM2), so I'm thinking its the Phenom memcontroller who is the sinner.

My memory ran 1173MHz 5,0 5 5 5 15 2t@ 2,20 vDIMM and 1100Mhz 4,0 4 4 7 2t tight sumbtimings @2,20 vDIMM with Athlon X2 6000+ and Crosshair.

It's a Phenom limitation, yup. They don't oc RAM well although 3-3-3-3-11 1T 450, 4-4-4-4-11 1T 500 and 5-5-5-15 600 is easy on them at 2.2v if MEMs can do it.

Edit: It's certainly not on par with a 790FX DFI board, but what could you expect from a board in this price range?

Your board is better for RAM/Monitoring than the DFI 70FX, try one of those and try getting some good stability and accuracy, or RAM compatibility and RAM oc, and you'll see why
Their BIOS has tons of good options but also tons of problems with basics settings and options. They are improving but since Nov'07, not enough.

Anyone know where the ganged/unganged memory mode values is on this board? See bios pics earlier in the tread.

No idea yet.

That is what I thought too, apparantly it is trfc set in ns on this board according to users at Abit forums. It waas set to 105 for dimm1+2 and 75 for dimm3+4, wich would have made sense if it were a delay due to dimm distance to imc.

Many 700 series boards set 105ns for the first DIMM and 75ns for the others at boot, default. More of a safety mechanism.

Think that is 95W ACP, about 120W TDP @stock. My 9500 is overclocked to 2,5GHz and runs at 1,35v. My whole system draws about 275W when playing, and 240W when priming. Well above 125W TDP on the cpu i think.

I do not think they would sell a mobo with full Phenom support unless it can handle it.

They wouldn't have added an 8-pin 12V connector if it couldn't support above 16A CPU draw

Your oc'd 9500 runs at minimum 126W TDP, if you take 95W TDP being stock. You're using the same hardware voltages and CPU/NB speeds as 9850BE does, and the current draw is the same, so is the end power.

Anyway, I'll talk with an Abit rep today and find out if 9850BE is officially supported and with which BIOS.

Which is the most stable BIOS ATM BTW?

Has anyone ran a 9850BE and if so, which BIOS allowed it boot?

I have tried 9850BE about 6 days ago and it gave me C1 errors at POST, not tried again. Don't have an X2 spare yet to flash to better BIOS and check.

[davidletterboyz]

KTE, I just found one person who has tested ax78 with 9850BE (been searching for days lol). --> http://forum.uabit.com/showthread.php?t=137048&page=2
He mentioned that there is no sign of PWM problem....so I guess it's your BIOS? BTW, he haven't stress that CPU enough to be 100% sure the PWM is a non-issue. He has problems with detecting the HDDs, so i guess it's still remain to be seen if the mobo can handle it under full load.

At the same thread, someone pointed out that he blew a MOSFET when ocing an x2 6000+ (125W TDP) within 4 days. But it has been running without any problem for a week after replacing the mobo (RMA).

[KTE]

KTE, I just found one person who has tested ax78 with 9850BE (been searching for days lol). --> http://forum.uabit.com/showthread.php?t=137048&page=2
He mentioned that there is no sign of PWM problem....so I guess it's your BIOS? BTW, he haven't stress that CPU enough to be 100% sure the PWM is a non-issue. He has problems with detecting the HDDs, so i guess it's still remain to be seen if the mobo can handle it under full load.

Thanks

Yeah it seems they work fine.

On another note, the mod there showed how to access CPU/NB FIDs in the BIOS: http://forum.uabit.com/showpost.php?...9&postcount=15

[KTE]

BTW, I have clarified 4 months ago with Abit and they did not get any problems with running RAM at 1.8-2.3v with this board and Phenoms, even after I gave them instructions to test what I thought was highly likely a Phenom killer setting

[LIKMARK]

Sorry for the delay LIKMARK, I've had many many issues and only made posts I can risk losing - lost my post to you about 8 times before I quit [unstable Phenom]

No problem

It looks like you have a BIOS limitation or a CPU limitation, as 9500/9600/9700 should be capable of dropping NB multi. I had one 9600BE which did not drop NB multi in the same BIOS which dropped NB multi for the second 9600BE, so its very possible

I love inconsistency..... Makes things so much easier pinpointing...

It's a Phenom limitation, yup. They don't oc RAM well although 3-3-3-3-11 1T 450, 4-4-4-4-11 1T 500 and 5-5-5-15 600 is easy on them at 2.2v if MEMs can do it.

That's with two sticks right?

Your board is better for RAM/Monitoring than the DFI 70FX, try one of those and try getting some good stability and accuracy, or RAM compatibility and RAM oc, and you'll see why
Their BIOS has tons of good options but also tons of problems with basics settings and options. They are improving but since Nov'07, not enough.

Yea I know, really bad example there I saw that as soon as i wrote it. Its beginning to look like the last AMD chipset board DFI made.. crfx3200 or what it's called. Brilliant when you made it work, but pure hassle for most of the time.

They wouldn't have added an 8-pin 12V connector if it couldn't support above 16A CPU draw

That's what I'm thinking too, one of the reasons why I bought this board.

Your oc'd 9500 runs at minimum 126W TDP, if you take 95W TDP being stock. You're using the same hardware voltages and CPU/NB speeds as 9850BE does, and the current draw is the same, so is the end power.

I thought 95W was ACP, not TDP. TDP being about 120W for the 9500/9600

Anyway, I'll talk with an Abit rep today and find out if 9850BE is officially supported and with which BIOS.

Case closed

Which is the most stable BIOS ATM BTW?

I've only tried #12 so far.

KTE, I just found one person who has tested ax78 with 9850BE (been searching for days lol). --> http://forum.uabit.com/showthread.php?t=137048&page=2
He mentioned that there is no sign of PWM problem....so I guess it's your BIOS? BTW, he haven't stress that CPU enough to be 100% sure the PWM is a non-issue. He has problems with detecting the HDDs, so i guess it's still remain to be seen if the mobo can handle it under full load.

At the same thread, someone pointed out that he blew a MOSFET when ocing an x2 6000+ (125W TDP) within 4 days. But it has been running without any problem for a week after replacing the mobo (RMA).

Basically down to getting a bad board I think.

Thanks

On another note, the mod there showed how to access CPU/NB FIDs in the BIOS: http://forum.uabit.com/showpost.php?...9&postcount=15

Actually he shows a noob how to unlock cpu fid by disabling C'n'Q....

BTW, I have clarified 4 months ago with Abit and they did not get any problems with running RAM at 1.8-2.3v with this board and Phenoms, even after I gave them instructions to test what I thought was highly likely a Phenom killer setting

I'm running my ram at 2.2v ATM. Just to test this. (low vcore/imc - high vdimm)

Wouldn't mind if my Phenom died of this as it oc's horribly. Early version....

[KTE]

That's with two sticks right?

Yea DC Unganged or Ganged.

I thought 95W was ACP, not TDP. TDP being about 120W for the 9500/9600

That applied to their server lineup mainly, TDP for Phenom lineup was said to be what they stated. During user testings the 95W rated pulled less than 90W, so it kinda proves their ratings. I'm not exact about them all though, only have very limited data. I have not seen a Phenom 9500/9600 draw more than 89W at the worst though, so who knows. Usually they pull 68-75W DC at stock. The 1.25v rated 9500/9600 at 76A would still equal 95W maximum theoretical, which is their lowest bin IIRC.

Actually he shows a noob how to unlock cpu fid by disabling C'n'Q....

It shows NB FID option there, that's what I mention and the mod had a 9500. Do you see that option with your 9500?

[LIKMARK]

BTW confirmed from users in Abit forum that cpu-nb fid only is accessable in beta bios 13a. Then I need to flash it

[KTE]

Any link to it?
Or was it the one I posted?

[LIKMARK]

http://forum.abit-usa.com/showpost.p...5&postcount=18

Edit: Misunderstood there.. Yeah it is the same bios you posted.

[KTE]

Np, thanks. Will try flashing and booting the 9850

[Jethro]

Just a heads up, when i first got the board I removed and inspected the heatsinks thoroughly and found the PWM sink to be arched and not flat. It had 4 mosfets in the middle of it barely making contact with the TIM material and there is no way it wouldve cooled properly. I took some time to flatten it but its a pain even vise mounted b/c of the shape of the ends.

[Jethro]

BTW are you guys just doing standard floppy bios flashing? No special flags etc? Its been along time since ive had a mishap even flashing in windows. Just wanna make sure

[LIKMARK]

You'll need /py /sn (just type awdflash /? an a list will show) maybe more as it is beta bios. You got something to add KTE?

[Jethro]

Thanks for the heads up on the Phenom compat! Look forward to more info's really leaning towards a B3 triple if they tend to hit at least 3gig.

[KTE]

I usually either use:

awdflash <name>.bin /py /sn /cd /cp /cc /cks /R

Or just type:

runme

[LIKMARK]

Lol

I'm gonna flash this thing tonight, hoping for better mem support too

[LIKMARK]

Anyone else having trouble flashing this one in DOS? I get "unknown flash type" message... using AWDflash 8.67

Edit: Being unable to flash my bios ATM, I tested som mem overclocking just to test with two sticks. As you know I can't even boot with Memclk 1066 with four sticks. With two sticks it is another story. http://valid.x86-secret.com/show_oc.php?id=344303

DDR2 1215MHz actually is new record for me . Adding two more sticks seriously impedes ram overclock with Phenom and a untweakable bios. (for ram at least) Can't even boot over 1000MHz

[Cheffy]

just an update on the power situation:

the pwm is fully capable of huge wattage, i was just loading it past 170w without issue (v-core drooped about 0.01v from 1.55 to 1.54) the high temps im seeing are not from the fets: its actually the coils that are getting very hot, but these are simple analogue devices so its no problem as far as i can tell - i just assume that the one closest the thermal diode is unusually well thermaly bonded to the motherboard

the random shutdowns - that was ocp on the +12v rail.

[Master of None]

BIOS13.B02 seems safe enough. It can be found here:

http://forum.abit-usa.com/showthread.php?t=122393

It will just be the BIN file. Download BIOS12 final

ftp://ftp.abit.com.tw/pub/download/bios/ax78/

and extract the files. Rename the Beta BIN to the same name as the official BIOS12 BIN. Replace the official 12 BIN with the now renamed Beta.

Use Dr.DOS to create a boot floppy

ftp://ftp.abit.com.tw/pub/download/utilities/drdos/

Copy the five files to the boot floppy, boot with it, and type runme.bat at the A prompt.

Clear CMOS after flashing.




tRFC

I have no idea why one dimm gets a slower timing than the other three. We've been setting them manually to 105ns or 127.5ns, depending on dimm size. I found some info on tRFC on a Samsung pdf. For some reason the tRFC you can set and run in BIOS is one step lower than the Samsung info suggests it should be. Maybe you need the slower tRFC if you run four dimms. I set 127.5 for my G.Skill 2 x 2GB kit of DDR2 1000 running at 480MHz.

[KTE]

The older BIOSes on the MSI RD790 and the ones on DFI/Sapphire AFAIR also default to 105ns and 75ns at boot regardless of what values you do set. But the new MSI BIOSes can set 75ns for both manually and it works so it looks like a BIOS bug.

Auto always defaults to DIMM1 105ns and DIMM 2 75ns on all boards that I've seen yet with 2x1GB.

Higher tRFC is needed for higher density RAM or when you have higher number of DIMM slots occupied or when you're oc'ing HT ref with X2 and Phenom.

[LIKMARK]

Thank you Master of None. Managed to flash it with the batch files you provided (by cdrom). Strange that it didn't work by regular method. Well well. Disappointingly there is still no cpu-nb fid option for me, so it must be cpu limited. That's too bad. Gonna test the bios a little tonight.

[LIKMARK]

Lol!!

That was one strange BIOS!!

First the positive things: (After some quick testing last night)

A few new settings, whatever SMART ram is. Probably some C'n'Q related application for ram.
Auto is default in most settings.
AOD support is greatly improved (2.0.17):
HT ref clock is included, making overclocking easier from AOD.
Voltage control for everything is applicable. With bios 12 there were only cpu VID and NB VID, now everything is included. Very good.

The not so good things:
No matter what I set cpu FID to, it chooses cpu FID 5,5. And yes, C'n'Q is disabled. No one else using this BIOS noticed this?
I know I'm not supposed to trust AOD voltage readings, but as I understood it VDDC is vcore, and set in bios to stock 1,25v, AOD read it as 1,55v. Set in bios to 1,35 AOD read it as 1,65v. That is quite much for Phenom IMO. Temps were low though, so I'm thinking it's AOD reading wrong values.

I did not test any memory performance last night, this I'll test tonight and upload some pics if I find the time

[KTE]

I've been in touch with Abit.

The official BIOS Abit AX78 ships with is BIOS11 and they tested and were able to replicate my issues.

BIOS11 does not support 9850, neither me or Abit were able to boot it.
BIOS11 does support and boot Phenom X4 9750 and X3 8400 though.

The advice was to contact the retailer or the local distributor so they can help update to the latest BIOS as it supports 9850/9650.

I have a non-booting Phenom again anyway.

[Deliximus]

I got this mobo a few weeks back and it has been good except for a few quirky things that I can live with. Here are my specs first of all.

Specs:

Phenom 9500 w/ Thermalright Ultima-90 HSF
OCZ Reaper 2x1GB DDR2-800 on the furthest slots.
WD 2 x 320GB B3s RAID 0
External sata2 controller, WD 750gb
XFX Geforce 8800GT
Xfi-XtremeMusic
Intel Pro1000 GT Nic
Corsair HX 620W
Antec 900 Case

Everything installs fine (dualbooting XP and Vista Ultimate, both 32bit), everything runs well in windows, the weirdest things occur with the booting process. Currently, if I shutdown my computer and cold-start it, it will not boot and I can hear the 8800GT’s fan going at full. I eventually am forced to power off the PSU, let the red light go away, and the start it all back up, everything is fine then until next shutdown.
Then, when I get into windows, the USB Logitech mouse receiver for my VX Revolution doesn’t respond in the front USB ports. I have to take out the USB receiver and plug it back in and everything is fine. A soft reboot will not result in this problem. I am on BIOS 12 and not going to move to BIOS 13 beta considering all the weird issues. I’d thought I’ll just throw that out there. Anyone notice it takes a long time for the monitor to detect the signal during bootup. I usually see my screen when the RAID setup kicks in.

I had the Silverstone Zeus 650W and Crossfire 3870s before going to the Corsair and Geforce 8800GT. This boot issue happens to both combos. Weird….

[Master of None]

AX78, 9500, 2 x 2GB DDR2 1000

BIOS13 Beta2 -

Smart RAM has three options, Disabled, Table 1 and Table Best. It sets an ECC profiles more or less. It set CAS wrong but everything else correctly for the one set of G.Skill I tried. None of us will ever use it because other things must be set to Auto for it to work.

I can drop the multi without issue. I've run 239 x 10.5 with 32-bit XP and now 240 x 10 with 64-bit Vista.

The highest FSB, HTT, or External Clock that I and another Mod can boot with is 240MHz. 241 is a no go and we get the “CPU is over running” message and no matter how many times we enter BIOS and save we can’t get rid of it.


AMD OD -

I'm not sure if VDDC is really Vcore. My base Vcore is 1.25, I added +.15 and CPU-Z reports 1.392v under load. AMD OD reports 1.70V for VDDC.

Reported CPU temps - SpeedFan, AMD OD, and CoreTemp all report the same temp. SpeedFan only reports one temp, CoreTemp reports one and three 0's for the others in the system tray.

I was able to set 242MHz External Clock, 245MHz and the system froze. I wish I had tried this with 32-bit XP because I’m running 2 x 2GB G.Skill DDR2 1000 and 250 x 10 with the 1:2 memory ratio is where I wanted to be.


64-bit Vista -

I was running 239 x 10.5 with XP, running Folding@Home, and completed 28 work units without the first problem. I tried installing Vista with my proven overclock and I got a BSOD very early on. I set default speeds and the install completed without issue. After the OS was installed, updated, and patched I tried 2.5GHz again. No Go. I set 240 x 10 with the same voltages I used for 239 x 10.5 and it is working well. I've completed 3 work units without issue. So the reports that you can’t overclock as far with 64-bit Vista are true for me. Or at least you can’t overclock as far with the same voltages that work in 32-bit XP.


AbitEQ -

The only thing I was really using it for was monitoring the PWM temp. Another Mod was having trouble with it in 64-bit Vista so I installed the latest version, 2.0.0.6, from the AX78 Global page. It installed but I got the "Log in as administrator and run once so you don't see this message" with no reported values. I uninstalled it and it borked the system. I had to do a Repair by booting with the Vista DVD. It was quick and fixed the problem. I don't believe this version supports the AX78, I couldn't find the files for this MB anywhere but I could find them for other boards.

A couple of nights later I tried with the version that ships with the MB, 2.0.0.1. I installed it with UAC disabled and it installed and even reported temps. It also caused two Blue Screens in short order even though it wasn't running. I did a system restore to just before I installed it and the system is solid once again.