**Official DFI LanParty UT 790FX-M2R Review/Overclock/Guide Thread**

[KTE]

You're too rich for me

I got the GBT780G to boot 9850BE too but not the rest yet - but at that time MSI also booted it, few days back. I don't want to keep taking off the Xig, hard to mount and paste it all up properly with some boards, applying TIM good is tricky.

[Tony]

KTE

Made a mistake there you need Pin 65 VSS, I usually use 64x2 as this is 1/2 vdimm, sorry man

You can use the cap between the dimm slots that missing or the one next to the coil also, they are all so close its not worth worrying.

[justapost]

You're too rich for me

Most of the stuff i test here is for work or will end up there. But I can not reuse broken phenoms there, so every failed phenom hurts my pocket.

I got the GBT780G to boot 9850BE too but not the rest yet - but at that time MSI also booted it, few days back. I don't want to keep taking off the Xig, hard to mount and paste it all up properly with some boards, applying TIM good is tricky.

Good to heary you use TIM nowadays.

KTE

Made a mistake there you need Pin 65 VSS, I usually use 64x2 as this is 1/2 vdimm, sorry man

You can use the cap between the dimm slots that missing or the one next to the coil also, they are all so close its not worth worrying.

Tony, I managed to measure the vdimm at the missing cap point. I measure d ~1.87V with 1.8V set in the bios. You measured ~1.92V for 1.8V at the dimm.
Do you get a difference around 0.05V between direct and cap point measurement?
Also I measured the vcore with a DMM and i get a difference in the range 0.04-0.05V between DMM and CPU-Z values. Did you measure vcore on the DFI? If so what difference do you get? Mine seems pretty high.

[Brother Esau]

The 9850BE hasn't booted for 4 days on any board. Thought it might be board but new 740G and M2R gives same error. I'll just try debugging it after I've played with the 4450e/740G setup

On DFI, like with Abit, it maybe BIOS related.

Don't think so KTE sounds like the processor is a Bust. Either that or the Bios thats installed in it is not a recent one. Try installing one of your other Phenom and Boot and hopefully Flash Bios to 1/30/08

[KTE]

OK, thanks Tony and Achim

Hehe, ye, I actually still skimp on TIM many times.. bloody coolers are so efficient, its not my fault

Camp, CPUs are known to do this only and only when I clear CMOS, they do it brand new too. I have not flashed a new BIOS yet on the Abit, DFI or 780G, still stock - hence why I think its BIOS.

Little niece played a prank on me with plasticine clay, stuck massive pancake rolls of it to my seat and I've sat on it warm for an hour. Got up and its all runny, mouldy, and stuck all over me backside

Better go clean the damn thing, what a pain.

[aGeoM]

OK, thanks Tony and Achim

Hehe, ye, I actually still skimp on TIM many times.. bloody coolers are so efficient, its not my fault

Camp, CPUs are known to do this only and only when I clear CMOS, they do it brand new too. I have not flashed a new BIOS yet on the Abit, DFI or 780G, still stock - hence why I think its BIOS.

Little niece played a prank on me with plasticine clay, stuck massive pancake rolls of it to my seat and I've sat on it warm for an hour. Got up and its all runny, mouldy, and stuck all over me backside

Better go clean the damn thing, what a pain.

Hi m8



Running 9500 om 416 it's more stable. HTT=255, CPU=x11, HT=x9, NB=x9

[KTE]

OK. Check this out.. Andreas reports there are fake DFI boards being sold - watch out

[Brother Esau]

Saw that about a week ago but its good you posted it Tye for those looking to get a DFI should definitely be aware of this.

[Oese]

woohoo my ram is here and they run perfectly well together with my powerchip ic-based gskill

OC tests using new PROMos-based Corsair DHX will follow

[Rammsteiner]

Oh, nice going there.

Im having a little more succes clocking the Phenom now.

Last night I primed 13.5x200 at 1.35Vcore and everything else stock.

It BSOD'd after about 5 hours I think if I read the Prime95 log correctly.

Very pissed and disappointed I clocked the NB to 2.2Ghz, decreased Vcore to 1.3375V and threw Vnb upto 1.3375V too. It's still running after 8 hours.

So I can only imagine a few things here, if this proves to be stable. Heat does indeed have an effect on stability, Vnb might be better 1:1 with Vcore and NB should be clocked higher too.

No clue which of ther above arguments are true but thus far it works.

Could it maybe be, since at xHTT and ymultiplier, the total Mhz switched a few decimals (according to CPU-Z at least), the NB speeds goes sometimes maybe by a very small margin below HT speed? I dont know if it's possible, but if that's the case it might explain a few crashes/freezes. I dont feel like testing if it's true, maybe later on, but atm Im trying to get the best out of it.

Also, did anyone get some good HTT speeds yet on M2R and Phenom B3?

[Oese]

interesting approach simply lower vcore and up nb/vnb if settings dont work well i can see phenom clocking is very different from x2 clocking..

but regarding nb/htt speeds interestingly its the same more or less with x2: cpu clocks lower then htt clock do not work, i experienced this behaviour with crystal cpuid...

there, htt variability is strongly correlated with cpu variability in clockspeed.. if htt clock jumps up a few 1/10mhz, cpu clock will also..

[Rammsteiner]

interesting approach simply lower vcore and up nb/vnb if settings dont work well i can see phenom clocking is very different from x2 clocking..

but regarding nb/htt speeds interestingly its the same more or less with x2: cpu clocks lower then htt clock do not work, i experienced this behaviour with crystal cpuid...

there, htt variability is strongly correlated with cpu variability in clockspeed.. if htt clock jumps up a few 1/10mhz, cpu clock will also..

Yes, Phenom is a lot different lol. ALthough, despite by the poor results this far it's a lot more fun though! If only HTT could be clocked higher so we could get more varaibles for RAM clocking it would be awesome. Right now Im pretty much bottlenecked by DDR800 or DDR1066. DDR1066 isn't working very well and DDR800 is quite slow due to poor HTT OC-ability. I could run the RAM DDR800 250Mhz HTT 4-4-4, but there's noway to achieve that now.

It's indeed true that if HTT speed varies a little by a few decimals most affected speeds just change along with it. I dont know if it's even possible the HT might run, even it was only for a millisecond, by 0.01Mhz faster and thus crash the system. If it's possible, this might explain things very well though.

[Oese]

mmh i'd say that sounds possible indeed...

whats wrong with the ram though? i'd think it should easily work with 1066 divider setting and all timings entered manually??


otherwise, the htt barrier is a thing i don't understand really.. which part of the cpu might cease working if htt is high but everything else is in order by changing nb/cpu/ht multipliers?

[Rammsteiner]

It's working more or less yes, although not 'nice'. Timings etc are quite limited by changing to DDR1066 IMO which isn't worth it for me. Besides that, soo much hassling for me.

I think it's the board its self which is crapping out by high HTT with a Phenom due to bad BIOS or so? When everyone tried B2 rev they didnt get good HTT speeds and everytime a new BIOS came they could get it a little higher all the time. I know the board can do 292Mhz which i used for a while, but Im sure it can go even higher (for example, Brother had it 350Mhz). I tried 250Mhz (to get same RAM OC as with X2) and changed CPU multi to 10, NB multi to 8 and HT to 8. But it wouldnt boot while all things were basicly at stock speed.

Although I read HT multiplier settings are borked, so maybe the 8x didnt get applied, on the otherhand I booted up a few times with lower HT and saw speed was indeed lower in CPU-Z so I doubt that.

[Rammsteiner]

Ok, I stopped Prime after 9 hours. That's Large FFT's of course, smalls dont stress the whole CPU. Dont know if Large does, but at least it does test better than Small.

Anyway, rebooted and set Vcore and Vnb at 1.3125V and priming again. Ill work my way down like this. After that Ill try and get higher clocks again, but first I want to make a BIOS template myself for a XP64 installation and hope to see some 'links' among the variables (if there are any, but the fact that at x settings 2.7Ghz isnt stable, but after setting NB multi a little higher and set Vnb 1:1 Vcore and having as result it runs stable make me wonder).

[Brother Esau]

Thats the reason why I have not posted quite as much in regards to Overclocking Phenom because its very complicated in the sense that the Fast way (Instant Gratification) is the half assed way! My personal opinion although not written in stone and it certainly does not apply to everyones train of thought or approach is that Phenom requires a light Artful touch as too much Brute Force well you know (Bammo) So if you guys don't see me posting too much about this right now is because I have allot to Read up on and I want to take my Time and gather up as much conclusive & concise evidence as I can before I give recommendations or put examples forward that others will follow perhaps

I really think that More is Not Better in the case of Phenom and I also believe that a successful overclock can be achieved through low voltage on all components with Ballance and a focussed Artful light touch :

Also too....I refuse to Push on this Phenom until D.F.I gets the Volatge Table in Bios Sorted out with this Over Volting Situation going on!

D.F.I.............I know you are watching this thread now can we PLEASE finally get serious with these Bios Releases so you can stop wasting our time Unnecessarily Running US around in Circles !

[Rammsteiner]

Good post mate.

The more Im getting into testing the more I think my previous posts were a good solid example of posting... as always.

Anyway, can you comment a little on your HTT OC results if you tried it yet? Cause I really can't seem to get a lot out of it. Although now I think about it I used a very little Vcore for the test but still, 207Mhz is crap. Even MSI, which was crap in the very beginning of Phenom and 790FX chipsets, is doing a lot better now (at least, it looks like it).

[Brother Esau]

I have already stated my Opinion on HT Overclocking with Phenom like 3 times already But will do it again......HT Overclocking on Phenom Means (MOOT) and should be avoided unless you wish to dial in you're Overclock Incrementally and get some added performance and speed by Increasing the HT Bus after you have a Established Base Overclock working Its only my opinion but lets think about this for a second shall we?

HT = Can Be Independently Clocked
CPU = Can Be Independently Clocked
NB = Can Be Independently Clocked
Ram = Now has Dividers to be set from 400MHZ - 1066MHZ to Establish (Base Overclock)

HT Bus = Should be used to Tweak you're Base Overclock Established By Independently Overclockable Components .......NB - HT - CPU - RAM

[Oese]

Thats right but only applicable to black edition (and thus high-priced) cpu's. furthermore, when fx series hits the market again, we wont see unlocked cpu's any longer i suppose... then, we'll need htt clocking also with phenom to keep the spirit of oc alive which is clocking lower priced cpu's to or above their fx counterparts.. thats why we buy dfi boards to can do that and tweak it the way it works!

i would love to get say a 9550 phenom at a 145€ price tag and clock it to 3,2ghz and i don't understand why this isnt possible at the moment..

besides, with x2 i get my board stable at 366mhz ht no problem, so whats going on there??

but i truly agree in that "artful" approach, overclocking always had this component to me... especially if thinking about ram oc it was always finding the balance between timings and speed and ht and that special subtiming that kept the ram away from the last 10-20mhz.. i loved this and especially with dfi boards one had the opportunity to really get the most out of any ram and of course cpu.

but i will like phenom oc when my wallet and wife comes down to it i knew from the beginning the more art thats needed (or better the more art and skill does pay oc-wise) the better the experience will be...

btw, many former intel users now switch over to phenom because they like this oc-approach more then an easy "more volts and more fsb" thing.. and i fully understand.. only the performance was keeping them away from phenom the whole time. but now as phenom gets better gaming scores clock per clock they're all back again...

[justapost]

Oh, nice going there.

Im having a little more succes clocking the Phenom now.

Last night I primed 13.5x200 at 1.35Vcore and everything else stock.

It BSOD'd after about 5 hours I think if I read the Prime95 log correctly.

Very pissed and disappointed I clocked the NB to 2.2Ghz, decreased Vcore to 1.3375V and threw Vnb upto 1.3375V too. It's still running after 8 hours.

So I can only imagine a few things here, if this proves to be stable. Heat does indeed have an effect on stability, Vnb might be better 1:1 with Vcore and NB should be clocked higher too.

No clue which of ther above arguments are true but thus far it works.

Could it maybe be, since at xHTT and ymultiplier, the total Mhz switched a few decimals (according to CPU-Z at least), the NB speeds goes sometimes maybe by a very small margin below HT speed? I dont know if it's possible, but if that's the case it might explain a few crashes/freezes. I dont feel like testing if it's true, maybe later on, but atm Im trying to get the best out of it.

Also, did anyone get some good HTT speeds yet on M2R and Phenom B3?

Hi Rammsteiner,
Thank you for the update.
I have this issue you have at 2.8GHz at 2.9GHz. Alot more stability with a higher NB speed.
I'm still in the process of verification. I have found the minimal vcore and vcorenb required for 2.8Ghz/2.0GHz.
At those settings I had one prime95 failure. I ran CoreDamage parallel to prime95 and upped the temps to ~70°C and it failed after a fe minutes.
I reran it at those higher temps and it was stable for about an hour again.
At 60°C this setup primed fine for 22,5hrs. I don't want to inspect this temperature issue right now. My aim is finding the lowest stable voltages for different NB speeds at 2.8GHz.

I use the following procedure for stability testing.

1. Sandra Memory Latency (that triggers a freeze here very quickly)
2. CoreDamage for a few minutes to see what temps i will get
3. Prime95 together with 3dmark06, to generate a very high power consumption
4. Prime95 longtest
5. So called WISE testing (Surfing, Openoffice, Antivir) for a while.


Lowest voltages for 2.8Ghz/2.0Ghz
CPU-VID: 1,3125V (Bios) 1,351V (DMM)
NB-VID: 1,15V (Bios) 1,144V (DMM)
CPU-TEMP: 63°C coredamage, 60°C max prime95
Pime Time: 22hrs

Lowest voltages for 2.8Ghz/2.2Ghz

Same setup as before, 2,2GHz NB speed starting with 1,275V NB-VID.

First results, with CPU temps at 63°-64°C (max, coredamage) 25°C ambient.
1,275V: passed 1-3, skipped 4-5
1,225V: passed 1-2, prime95 failed after 3min
1,25V: passed 1-3, primed for 1hr
1,2375V: passed 1-3, primed for 1:15hrs
1,23125: passed 1-3, prime95 failed after 12min
1,2375V: passed 1-3, prime95 failed after 1hr
1.24375V: passed 1-3, prime95 failed after 4:34hrs
1,25V: passed 1-3, prime95 stable for 12hrs

[Oese]

Update here as well!

Prime stable 4,5h with 6GB Ram @ 1,424v cpuz (stock vcore) 04/15 (which is 1,458v dmm under load^^)

[justapost]

Update here as well!

Prime stable 4,5h with 6GB Ram @ 1,424v cpuz (stock vcore) 04/15 (which is 1,458v dmm under load^^)

Thanks Oese,
You get a higher difference in vcore now it seems (0.034V).
How did you get those measurement points? Did you just try or have you found some sources how it's measured on other DFI boards with digital PWM?
Have you measured your vdimm? If so what results do you get?

[aGeoM]

HT = Can Be Independently Clocked
CPU = Can Be Independently Clocked
NB = Can Be Independently Clocked
Ram = Now has Dividers to be set from 400MHZ - 1066MHZ to Establish (Base Overclock)

HT Bus = Should be used to Tweak you're Base Overclock Established By Independently Overclockable Components .......NB - HT - CPU - RAM

How can you clock HT independently?
- Ooh, I know..., we have an option in BIOS to do that.
Does it work?
Yeah, underclocks nicely.

[Oese]

did not measure vdimm yet..

read something about where to correctly measure vcore in general ( at the positive leg BEHIND the smoothing capacitor) and because these i knew were the capacitors (read somewhere else) i thought about it and it seemed to be correct..

[justapost]

did not measure vdimm yet..

read something about where to correctly measure vcore in general ( at the positive leg BEHIND the smoothing capacitor) and because these i knew were the capacitors (read somewhere else) i thought about it and it seemed to be correct..

I'm not an expert in this area, but it can be that the voltage we measure here gets a drop from the voltage regulator chip.

Voltage Mod Guide
http://www.aoaforums.com/forum/hardw...mod-guide.html

Voltage mod on the M3A
Voltage Regulator
http://www.xtremesystems.org/forums/...&postcount=210
http://www.xtremesystems.org/forums/...&postcount=212
Pic from CRFX with the measurement points and shadable resistances.
http://www.xtremesystems.org/forums/...&postcount=400

May I nag you to measure your vdimm to have more data for comparison.