M3A overclocking

[justapost]

I finaly recieved my M3A mobo last week and as there is no oc thread for that board i'll start one.





The Asus FTP Site with Bios Updates is here or here.
Asus Forum(no Asus tech support in there).

Found another thread about the board at legit review here and another one at techPowerUp here.

If you have trouble with the onboard soundcard read this thread on majorgeeks.

There is only one Bios version (0301) available atm. Now this bios has very limited options to oc a new phenom cpu but it did not that bad with an AM2 cpu. The bios does not include the tlb fix.

The board uses an new clock chip (IT8716F-S 0737-DXS), for whom i could not find a tool to modify the ref HT under windows.
AOD allows only cpu core multi and ram timing modifications with this board.

With a phenom cpu it is not yet possible to modiy the cpu multi in the bios if cool'n'quiet is disabled.
Also i was unable to boot with prozessor voltage set to anything beside Auto.



In terms of ram timing modification this board has more to offer than it's ancestor M2A.

If an AM2 cpu is in use or an AM2+ cpu is in use it allows setting CAS/TRCD/TRP/TRAS.

If an AM2+ cpu is used it allows those settings


With M2A-VM it was possible to ramp up the cpu voltages up to 1,55V + 100mV. M3A has no +100mV option so the max cpu voltage is 1,55V.

Ram voltages can be bumped up to 2,25V in oposite to the 2,1V max the M2A-VM did.

So overclocking phenom's with this bios version is a no go atm. No tools under windows and very limited bios settings.

I tried to oc an BE-2400 (2,3MHz, 65nm, G0) via bios settings and found that the board can boot with an ref HT ~340MHz (M2A-VM ~300MHz).
With 1,55V cpu voltage i was able to reach around 3,35MHz (M2A-VM maxed out ~3,2MHz with 1,6V) with watercooling. Not really stable but i could run super pi.

11,5 x ref HT 292


10 x ref HT 320


Overall the cpu oc'ed with lower cpu voltages than on the M2A-VM.

At legit reviews i found those posts:

Well the 6400 has like no overhead for over clocking but I clocked it stable to 3.52 in like 5 minutes using AI overclock. Just set the AI overclock to manual, increase the FSB to the freq you want and the board adjusts appropriately. I can't say how it'll work for a larger over clock on that cpu. I did stick in a X2 4200 that runs at 2.2, did the same thing, bumped it up to FSB 240 Multiplier 11, gave me 2.64 ghz, had to bump the vcore one notch and drop hyper transport to 4 and it ran stable on orthos for an hour. I started with AI over clock, ran a little funky, so I manually bumped vcore and dropped hyper transport ran good after that.

Hey Ken, I dropped multiplier to 10x and pushed FSB to 360, got 3.6 stable 15 minutes then temps started to creep (55C), ordered a ThermalRight 120 Ultra and A Vantec Stealth 120MM today. Should solve Temp and noise problems. Had it up to 3.52 with 16x stable same thing temp creep. The IceTank cooler does fine until you put full load on. I'm waiting until the cooler comes in before I push it (for my sanity's sake).

One sidenote: The board works flawless with Linux, no noacpi noapic irqpoll options where required to boot. I like that.

Waiting for a new bios with better phenom oc capabilities now.

[justapost]

Tried to find the max ref HT for the phenom at stock.
Running prime95 now with 220MHz ref HT and it is stable since 20 minutes now with cpu ~56° (EDIT: failed after 25 minutes trying 215MHz now). For comparison on the M2A-VM it required 1,275V to run stable at stock and ~1,4V to run at 220MHz.

CPU-Z does not read the HT and NB speed correct and does not validate.



AOD does even worse



I did not yet tried to oc with a sata disc attached. M2A-VM just trashed my raptor again. I got it back from RMA and tried to boot linux one time with 220MHz ref HT. I stuck during sata controller inizialisation and did not want to boot. Found the disc had a few bad blocks so I'll send it back once again.

[twilyth]

subscribed - got my board a few days ago but haven't been feeling well enough to tear down the old machine and build the new one. M3A is not a top of the line board so it's not too much of a surprise that the options are limited. The board I really wanted would have been about $70 more and I just couldn't justify it.j

Good to know I won't have to disable the TLB fix - didn't want that anyway.

Have you tried the overdrive utility with it?

[justapost]

subscribed - got my board a few days ago but haven't been feeling well enough to tear down the old machine and build the new one. M3A is not a top of the line board so it's not too much of a surprise that the options are limited. The board I really wanted would have been about $70 more and I just couldn't justify it.j

I ordered it because i was curious about the new platforms power consumption. Plan to buy a more expensive board once the platform has matured.
On the M2A-VM the first bios supporting phenom (1501) had the same issue as the M3A, no cpu multi adjustment was possible. It worked with the nex version (1601).

Good to know I won't have to disable the TLB fix - didn't want that anyway.

I think it will be included in a future bios version, the latest M2A-VM (1603) bios had the fix applied and no bios option to disable it.

Have you tried the overdrive utility with it?

Yes and it is pretty useless at the moment. I added a few screenies in my second post.
It calculates a wrong ref HT (176MHz instead of 215MHz) out of a wrong read nb multiplier (11 instead of 9).

176 ~= 1800/11*215/200

Also it can only read CPU VID at the moment all other voltages are reported as zero.
It only allows cpu multi, CPU VID and Memory timing modifications.

[justapost]

Tried NbFid modification with WPRCEDIT.
It was already mentioned by thesavage in the amd forum, the board seems to run with an 11x multi. I don't believe it because high nb clocks are known to cause problems with phenom. I think the speed is missreported.



After setting it to 9x at least AMD power monitor and cpu-z showing correct nb speed values, the ref HT in AOD is unaffected.



If i modify and reboot it eighter resets the value to 00111h (7+4=11) or does not boot (<00100h).

LOL CPU-Z Channels # switched from Dual/Unganged to Single/Ganged for a short time.

[justapost]

There is a new Bios version(0406) available for
download.
alternative download

This bios includes the tlb fix. It can not be disabled via bios, but AOD and CrystalCPU works.

The cpu multi can now be modified (8-11 with 9500).



Also the bios allows modifying CPU and NB VID in the range 0.8-1.25V now but the CPU VDDC field has gone.




The memory timings submenu has changed, it does no longer show all these options


It does not allow to configure dct0 or dct1 but only both dct's.


The memory runs in ganged mode and there is no bios option to switch to unganged.

There is one nice new setting called process downcore. It allows disabling 1-3 cores.

Disabling 3 cores saves ~14W DC, two cores saves ~5W DC and one core saves ~3W DC in idle.

The NB multi is now at 9x and wprcedit shows the correct value 00101b. I tried 00100b with 1x HT but the system did not restart.

Prime95 failed after an hour yesterday with ref HT 215MHz and after half an hour with 220MHz (Ram-V 2,1V everything else auto).

Today i tried it with CPU VID and NB VID at 1,25V and it failed after a few seconds. With only one core enabled it's stable since half an hour with ref HT 230 MHz at ~41&#176;.

Guess I need more CPU VDDC to reach higher speeds so i have to wait for the next bios.

[noee]

Thanks for posting that new bios, seems to be a little friendlier with the G2 Brisbane (I've been trying a BE-2400 and the 5000+ BE).

Looks to me like this bios is set up to really work with AOD and do all the serious O/C from within Windows. I have had zero luck with AOD (beta version .14) on this board/cpu combo. AOD locks my machine no matter what every time I hit "Apply".

I was able to get 3150Mhz Orthos stable with stock volt (1.344) and DDR2-667, idle temps around 36C, load temps at 51C using a Zerotherm Butterfly, HT at x4. This was the 5000+ BE, 13x @ 240 fsb.

I know I can go higher but there just aren't many DRAM options (compared to even my Biostar TA690G). Do you guys know if I can set Max ASynch Latency in the Bios?

[Thesavage]

Thanks for the update justapost, thought i may pass on some more speculation it may also be that the pci bus is not locked i have had my drive fail whilst overclocking. Also the asus forum is reporting issues with the atheros lan.

[Evax]

Im getting one of these soon for my phenom 9600, is it ok if u dont want to overclock ?

[madfaze]

Im getting one of these soon for my phenom 9600, is it ok if u dont want to overclock ?

yup

[noee]

I don't know about the K10 cores, but the G2 Brisbane loves this mobo. Just messing around, I can undervolt the CPU and RAM at stock speeds quite nicely for an extremely quiet, low-heat setup. (Antec Solo w/Seasonic 430).

[justapost]

Thanks for posting that new bios, seems to be a little friendlier with the G2 Brisbane (I've been trying a BE-2400 and the 5000+ BE).

Looks to me like this bios is set up to really work with AOD and do all the serious O/C from within Windows. I have had zero luck with AOD (beta version .14) on this board/cpu combo. AOD locks my machine no matter what every time I hit "Apply".

Nice, thought AOD works only with phenoms. What can be modified if you use an AM2 cpu?

I was able to get 3150Mhz Orthos stable with stock volt (1.344) and DDR2-667, idle temps around 36C, load temps at 51C using a Zerotherm Butterfly, HT at x4. This was the 5000+ BE, 13x @ 240 fsb.

Do you use DDR2-667 ram? Have you tried to find the max ref HT with the 5000+ BE?

I know I can go higher but there just aren't many DRAM options (compared to even my Biostar TA690G). Do you guys know if I can set Max ASynch Latency in the Bios?

U can't even the options 0301 had with a phenom did not allow to set the max async latency.

Thanks for the update justapost, thought i may pass on some more speculation it may also be that the pci bus is not locked i have had my drive fail whilst overclocking. Also the asus forum is reporting issues with the atheros lan.

Np, the ftp site allways has the bios files a few days before they show up on their website.
Is this hard drive problem a known issue with SB600? Damaged three sata disks on an M2A-VM. Had no issues with an old 20GB ide hd.
The atheros lan works without problems but i used an wxp install cd with most hotfixes and latest driver pack stuff integrated. The atheros driver coming with driver pack is the same version asus packed on their cd.

Im getting one of these soon for my phenom 9600, is it ok if u dont want to overclock ?

I don't know about the K10 cores, but the G2 Brisbane loves this mobo. Just messing around, I can undervolt the CPU and RAM at stock speeds quite nicely for an extremely quiet, low-heat setup. (Antec Solo w/Seasonic 430).

Tried to find the lowest possible CPu and NB VID with the phenom.
Ran stable with 1,1875V CPU VID and ~1,05V NB VID.
I tried the AI Gear tool from the driver cd but it froze my system.
I agree with you, it's a nice board in terms of energy efficiency. Also it has an higher max ref HT as it's predecessor m2a. I could reach ~100MHz more oc with the be-2400 that on the m2a-vm.
As for phenom it is too early to say anything final about overclocking, bios need to mature here.

[KTE]

I ordered it because i was curious about the new platforms power consumption. Plan to buy a more expensive board once the platform has matured.
On the M2A-VM the first bios supporting phenom (1501) had the same issue as the M3A, no cpu multi adjustment was possible. It worked with the nex version (1601).

I think it will be included in a future bios version, the latest M2A-VM (1603) bios had the fix applied and no bios option to disable it.

Yes and it is pretty useless at the moment. I added a few screenies in my second post.
It calculates a wrong ref HT (176MHz instead of 215MHz) out of a wrong read nb multiplier (11 instead of 9).

176 ~= 1800/11*215/200

Also it can only read CPU VID at the moment all other voltages are reported as zero.
It only allows cpu multi, CPU VID and Memory timing modifications.

Justapost, thanks for the feedback.

Those monitoring readings are incorrect because your BIOS>Phenom support is bad. Those AOD readings are only displaying what you're BIOS is telling it.
Your BIOS is not the first, every bad ASUS BIOS I've seen which lacks support for Phenom runs the NB multiplier async with the core multiplier, and so that messes Phenom monitor reading up in ALL software but CPUZ (however CPUZ won't validate in this state because it detects errors).

The real clock speeds in the above AOD screenshots (faulty ones) are NB MHz=HT MHz in CPUZ and CPUZ CPU/RAM is correct.

If you set 9x multi on Phenom, it'll boot up just fine and AOD will read the correct values. If voltage/memory control is missing in AOD, it's because of your board BIOS lacking support for it.

If you can change CPU/NB VIDs, then again this is a BIOS feature. However Sammi did say that the CPU VID is usually locked (1.20/1.25). Bad supported ASUS boards would show high MHz for Phenom and high volts and high and low VIDs but in actual fact were not changing anything at all (AOD will show this properly). So keep testing to see what the state is.

AOD works with X2 CPUs just fine.

[justapost]

Justapost, thanks for the feedback.

Bounced back.

Those monitoring readings are incorrect because your BIOS>Phenom support is bad. Those AOD readings are only displaying what you're BIOS is telling it.

You are right, my description was inaccurate. To be correct the bios writes the wrong value into the registers and AOD reads those wrong values correct.

Your BIOS is not the first, every bad ASUS BIOS I've seen which lacks support for Phenom runs the NB multiplier async with the core multiplier, and so that messes Phenom monitor reading up in ALL software but CPUZ (however CPUZ won't validate in this state because it detects errors).

Uhm must add a screenie for the latest bios. That one writes the correct valuest to the registers, so AOD and CPU-Z report the correct values now.

The real clock speeds in the above AOD screenshots (faulty ones) are NB MHz=HT MHz in CPUZ and CPUZ CPU/RAM is correct.

Well i could set the ht multi with 0301 bios and it was reported correct in cpu-z. I remember tictac mentioned the register storing the ht multi once in an barcelona thread but i did not check that value with wprcedit.

If you set 9x multi on Phenom, it'll boot up just fine and AOD will read the correct values.

nb multi u mean?

If voltage/memory control is missing in AOD, it's because of your board BIOS lacking support for it.

Memory settings work in AOD. The bios lacks a field for CPU VDDC.
But in terms of ref HT it is AOD/Clockgen/Systool lacking support for the clock chip.

If you can change CPU/NB VIDs, then again this is a BIOS feature. However Sammi did say that the CPU VID is usually locked (1.20/1.25).

Who is sammi could not find a user with that nick here.
Wasn't it macci saying CPU VID is locked in AOD? Anyway i can set values below 1,2V for CPU VID in bios.
I'm pretty sure that the new bios modifies CPU and NB VID and not CPU VDDC and NB voltage. So the labels (prozessor and northbridge voltage) are wrong, as the bios does not yet show up on the website i assume it's experimental wip atm.
Both fields are locked upward at 1,25V. The first bios had only a CPU VDDC field labeld as prozessor voltage going up to 1,55V (but nothing beside auto worked).

Bad supported ASUS boards would show high MHz for Phenom and high volts and high and low VIDs but in actual fact were not changing anything at all (AOD will show this properly). So keep testing to see what the state is.

AOD works with X2 CPUs just fine.

Bios settings prozessor voltage 1,2V prozessor-NB voltage 1,1875V


AOD shows 1,2V CPU VID but no CPU VDDC and allows CPU VID modification.
Also added wprcedit and cpu-z to show correct multi readings.


Bios settings prozessor voltage 1,1875V prozessor-NB voltage 1,175V


AOD shows 1,1875V CPU VID


Found that it is possible to set a lower ht multi with AOD in the range 6-9 with a 9x multi or in the range 1-5 with a 4x multi in bios.
Could not measure a difference with Cinebench10 OpenL but the systems power consumption dropped ~2W going from 9x to 6x.

[Thesavage]

justapost have you tried the crystal cpu tlb fix on this board with the latest bios? Also i found the performance plunge with this new bios is so bad that in the trackmania series the game has lost 50&#37;+ frames. There also seems to be no difference with aod overdrive and the tlb settings.

[noee]

Nice, thought AOD works only with phenoms. What can be modified if you use an AM2 cpu?

Just about everything for the memory. Not so much for the CPU. But, everytime I try to apply settings, it freezes the machine. So, while AOD may work with X2 cpus, it doesn't work with X2 cpus and the M3A.

Do you use DDR2-667 ram? Have you tried to find the max ref HT with the 5000+ BE?

I'm using some Patriot DDR2-800, cas 4, etc.... running 2T. No, I haven't tried to find max ref HT yet, I was going to wait it out for the next BIOS and maybe the next AOD.

The 5000+ BE starts to get nice and warm at 1.4V and full load. I've got a TI-128 I may try out, but the Zerotherm seems like a decent unit.

[justapost]

justapost have you tried the crystal cpu tlb fix on this board with the latest bios? Also i found the performance plunge with this new bios is so bad that in the trackmania series the game has lost 50%+ frames. There also seems to be no difference with aod overdrive and the tlb settings.

With TLB fix: Everest Memory Read 5622MB/s


Without TLB fix (disabled via Crystal CPU): Everest Memory Read 6943MB/s


With TLB fix (switched to ggreen/black mode via AOD): Everest Memory Read 5568MB/s


Wonder what that last mode does.

Just about everything for the memory. Not so much for the CPU. But, everytime I try to apply settings, it freezes the machine. So, while AOD may work with X2 cpus, it doesn't work with X2 cpus and the M3A.

Ah ok, so no ref HT modification with X2 also.

I'm using some Patriot DDR2-800, cas 4, etc.... running 2T. No, I haven't tried to find max ref HT yet, I was going to wait it out for the next BIOS and maybe the next AOD.

Strang, i have two different sets of G.Skill DDR2-800 one is better in terms of ref HT the other one in terms of latencies. Both have no problems with an ref HT ~320 without tweaking max async latency. Max async latency changes here if i use different dimm slots btw.

The 5000+ BE starts to get nice and warm at 1.4V and full load. I've got a TI-128 I may try out, but the Zerotherm seems like a decent unit.

I only have a zalman reseraptor 1 plus. Well the temps dropped from 84° to 70° under full load with 1,55V.
Aiming more for silence here. I overclock only for curiosity.

[KTE]

nb multi u mean?

Yep.

Memory settings work in AOD. The bios lacks a field for CPU VDDC.
But in terms of ref HT it is AOD/Clockgen/Systool lacking support for the clock chip.

Yeah thats good and bad. You have no Core Voltage control, as though the BIOS engineer didnt have a clue what the VID was and thought it was the VCore. VID alone wont get you anywhere in terms of oc Im afraid.

Who is sammi could not find a user with that nick here.

Macciś name.

Wasn't it macci saying CPU VID is locked in AOD? Anyway i can set values below 1,2V for CPU VID in bios.

Yep, upwards locked to ~1.20-1.25.

I'm pretty sure that the new bios modifies CPU and NB VID and not CPU VDDC and NB voltage. So the labels (prozessor and northbridge voltage) are wrong, as the bios does not yet show up on the website i assume it's experimental wip atm.

Looking at your ss, I agree.

Found that it is possible to set a lower ht multi with AOD in the range 6-9 with a 9x multi or in the range 1-5 with a 4x multi in bios.

Try AquaMark3 with HT changing. It may show you a difference (in the GFX score).

[justapost]

Yep.

Does not boot if i change it, already tried that with both bios versions.

Yeah thats good and bad. You have no Core Voltage control, as though the BIOS engineer didnt have a clue what the VID was and thought it was the VCore. VID alone wont get you anywhere in terms of oc Im afraid.

I guess the engineer wanted a bios to play with VID's and was too lazy to change the labels. The bios is only available via ftp and I'm happy that they share the beta's this way.
VID's aint that bad for dc.

Macciś name.

That simplifies things.

Yep, upwards locked to ~1.20-1.25.

As does reading.
Looking at your ss, I agree.

Try AquaMark3 with HT changing. It may show you a difference (in the GFX score).

Aquamark3 ends with a blackscreen here could not fix that issue. Tried Cinebench10's OpenGL benchmark, but found no difference.
I'll try specview.
I plan to measure my system components power consumption once again on the weekend with the M2A-VM and the M3A.
Made a few changes on my setup during my last measurements and the hd i used is on rma right now, that adds an inaccuracy i want to avoid.

BTW, are you still waiting for the BE-2300 and the 9600BE?

[Thesavage]

There is a small patch for aquamark that i think fixes the black screen and also exiting without a score. But not for 64bit o's.

[KTE]

Waiting to get back home and test this CPU I have justapost.

[justapost]

justapost have you tried the crystal cpu tlb fix on this board with the latest bios? Also i found the performance plunge with this new bios is so bad that in the trackmania series the game has lost 50&#37;+ frames. There also seems to be no difference with aod overdrive and the tlb settings.

You are right the performance drops.

Crystal Mark MEM (2,2GHz 800 5-5-5-15)
0301: 24025
0406: 20538

FritzChess 4.2
0301: 10,73
0406: 9,86

I've already tried the patch (replaced the direcpll.dll) after that it did no longer fail during runtime, but it never shows results.

In SpecView Perf there where no differences between 9x and 5x multi.
Tried it with 3dmark. Very small differences in scores with 5x and 9x multi ~4870+-10. But between 1x and 5x the difference is valuable ~4400 vs ~4700.
So AOD can really change the HT mult. I thought those changes require a reboot, after reading that post from tictac.
With an muli in the range 1-5 in bios the multi can be changed between 1 and 5. With an multi above 5 in bios the multi can be changed in the range 6-9.

Saw someone found a way to fix the atheros driver over at the asus forum.

[justapost]

Ran a few benchmarks at stock ref HT (200MHz) with CPU multi 11x and Ram at 800 5-5-5-15-21.

1. Bios Version/TLB-Fix

Mem: Single (1 Module)

3DMark06

Version (TLB-Fix): Score/SM2/SM3/CPU
0301(off): 4806/1592/1996/3240
0406(off): 4802/1589/1995/3237
0406(on ): 4780/1586/1993/3116

Equal performance with fix off.

FritzChess

Version (TLB-Fix): Score
0301(off): 10,75
0406(off): 9,71
0406(on ): 9,63

Performance drop with new bios and with fix.

2. HT Multi

Mem: Ganged (4 Modules)
TLB-Fix: On

3DMark06

Multi: Score/SM2/SM3/CPU
1: 4449/1507/1799/3143
5: 4780/1582/1993/3170
6: 4783/1589/1990/3150
9: 4801/1593/2000/3157

Multi can be changed in the ranges 1-5 and 6-9 with AOD.

3. Number of cores

Mem: Ganged (4 Modules)
TLB-Fix: On

3DMark06

Cores: Score/SM2/SM3/CPU
4: 4801/1593/2000/3157
3: 4699/1594/1997/2558
2: 4488/1595/1999/1786
1: 3901/1586/1995/903

Speedup in CPU Score: 2 cores +97%, 3 cores +183%, 4 cores +249%

[Thesavage]

Thanks for the update justapost, yes its a strange bios after a fresh windows install (downgrade from vista 64 to xp 64) trackmania nations is now faster on the new bios than on the old but trackmania united has plunged nearly 30 fps. Also HL2 lost coast is nearly 30fps slower on the new bios. I also found 3d mark was unaffected by the change in bios. Is there no way to "port" the bios from the m3a32 deluxe series onto this board?
Did you dload this dll for the black screen at the end of aquamark http://aqm4.invisionzone.com/index.php?showtopic=445 (2nd post)

[justapost]

Thanks for the update justapost, yes its a strange bios after a fresh windows install (downgrade from vista 64 to xp 64) trackmania nations is now faster on the new bios than on the old but trackmania united has plunged nearly 30 fps. Also HL2 lost coast is nearly 30fps slower on the new bios. I also found 3d mark was unaffected by the change in bios. Is there no way to "port" the bios from the m3a32 deluxe series onto this board?

Hi Savage,
I saw your post on the asus forum. Also realize that bios 301 runs mem unganged and 406 runs ganged. So i did those single channel benchmarks. But even there fritzchess dropped.

Did you dload this dll for the black screen at the end of aquamark http://aqm4.invisionzone.com/index.php?showtopic=445 (2nd post)

I tried a few version, downloaded from various locations. None of em worked. Tried the location you posted, same situation.