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

[Tony]

Guys its friday night...lots of numbers on a laptop keyboard LOL

Don't trust bios et al, I measured this at the dimm slot

[KTE]

He's talking about measured vDIMM, i.e. real vDIMM which is different to software and the only actual correct one. Thanks for the details Tony

Both MSI 790FX I tried were undervolting vDIMM but I don't remember how DFI board handled it.

Edit: just seen your reply now, nm.

[justapost]

Ups, thx for clarifing @ Tony&KTE

Tony I have a DMM here, is it possible you describe where i have to place the probes for a comparison? And of course thanks for pointing on this.

[KTE]

Nm, wait for his reply - but he is most likely measuring it from the MSOFET legs feeding the DIMM slots Achim.

[justapost]

Back on the Sapphire again. Flashed 416 bios and did a temp comparison.

307/416

Idle


Load/Prime95


Load/CoreDamage

No difference between 307 and 416 here. A note I manuals set CPU TEMP Adjust to +2.
Now close to 40°C seemed abit high. Well I had the USB-Stick plugged in.

USB/NOUSB

Temps dropped 6°C.

Better now. I'l give the lates sapphire bios a try now.

[Tony]

Nm, wait for his reply - but he is most likely measuring it from the MSOFET legs feeding the DIMM slots Achim.

nope at an actual dimm slot...bit dangerous for end users though me thinks

There is a missing cap you could use next to the coil on the edge of the board infront of the dimm slots, this is close to the dimm slot measurement also

[KTE]

Ah ok, thanks for the info

Yeah DIMM slot measurement is known more accurate than MOSFET.

[justapost]

Thank you Tony,
Localiced measure points

1.8V - 1.875V

EDIT: I'm on sapphire 415 bios now same temps same usb power consumption issue.
A note about Everest with a sapphire bios.
Everset displays PWM temp as CPU and CPU temp as Aux.

[justapost]

Attention:....."New" D.F.I LanParty UT 790FX-M2R / Phenom 4/15/2008 "Bios Template" Click Attachment Link Below!
[/code]

Bro, i tried your settings here. I found that my third core is the weakest few days ago on the M3A. Prime95 failed after a few seconds on that core.
So I upped CPU vcore to 1.27V, tried prime95 for a few minutes, closed it and started Sandra. The system froze during startup.
Upped vcore to 1.3125V and now Sandra works, including the Memory Latency Test.
But nevermind, thanks for sharing.

EDIT: Ran into prime problems at 1.3125 so i upped ram to 2.0V now sandra does not freeze my machine at 1.275V vcore.
I normaly used 5-5-5-16-18 75ns. So the cpu definately benefits from 5-5-5-18-23 195ns timings vcore wise.

I upped the NB VID a notch (2.2GHz) and set NB Voltage to 1.275V. Passed the sandra latency test here but failed prime95 on the third core after a few minutes. Now I'm puzzled.

Previously I ran 2.8GHz(cpu)/2.2GHz(nb)/533MHz(mem) at 1.3125(vcore)/1.275V(vcorenb)/2V(mem) stable for days. There must be an interference between the NB speed and the cpu or mem voltage.

Went back to previous settings and they also failed prime95 after few minutes. Upped NB Voltage to 1.3V and prime95 did not fail (20min atm).

First time I run a phenom with a higher nb voltage tha cpu voltage.

UPDATE:

Few hours later it turned out that the memory tweaking has no impact on stability.

1.27V failed after 1:30hrs
1.2875V failed after 2:00 hrs
1.3V is what i'm testing (it had failed last week after a few hours).

I had a similar situation with the 9500 while i was looking for the lowest NB voltage.

It was stable for a short amount of time with settings like this:

CPU 1.25V NB 1.25V
CPU 1.2625V NB 1.1V

But finaly fully stable at CPU 1.2626V NB 1.2V.

UPDATE
1.3V failed after 1hr
1.3125V stable for 9hrs

I keep vcore fixed now and will go lower with NB voltage (was 1.3V before). I ran CoreDamage for a while and i'm at ~62°C cpu temp now, 1-2°C higher that my results during bios comparison at stock.

1.2875V prime95 stable for 1hr
1.25V passed sandra and 1min() prime95
1.2V passed sandra and 10min prime95+3dmark06
1.1V passed sandra but failed prime95 immediate
1.15V passed 10min prime95+3dmark06, passed 10min PassmarkBurnIn, passed prime95 for two hours with an additional 3dmark06 run also passed sandra latency
1.125 failed prime95 after 2mins
1.1375V failed prime95 after 8mins during parallel 3dmark06 run
1.14375V prime95 stable for 3hrs

One lower setting to test
1.30625V(CPU)/1.14375V(NB) passed sandra and 10min prime95+3dmark06 ... failed after 6hrs

1.3125V/1.14375V failed prim95 after 90mins
1.3125V/1.15V prime95 stable for 4:15hrs at ~60°C cpu temp, then i lowered the fan speeds till the cpu temp was around 70°, prime95 failed after a few minutes.

1.3125V/1.15V prime95 stable for 22:30hrs with cpu-temp ~60°C max and 23°-24°C ambient

[Oese]

guys i'd like to measure vcore with a dmm, i'm sure its described somewhere here but perhaps someone could point it out to me in short once more?

i found higher vcore with 4/16 as well.... cpuz showed 1.456v instead of 1.44v...

[KTE]

Good testing Achim.

Tony, vDIMM at pin 7?

[Brother Esau]

Back on the Sapphire again. Flashed 416 bios and did a temp comparison.

307/416

Idle


Load/Prime95


Load/CoreDamage

No difference between 307 and 416 here. A note I manuals set CPU TEMP Adjust to +2.
Now close to 40°C seemed abit high. Well I had the USB-Stick plugged in.

USB/NOUSB

Temps dropped 6°C.

Better now. I'l give the lates sapphire bios a try now.

Achim......I noticed the 2 times I tried 3/07/2008 that temps were much higher then other releases such for example 1/30/2008 so I shy'ed away from it for that reason. Where on the other hand the 3/12/2008 "Beta" Regulated temp readings marginally lower not drastically but still .....This is getting quite confusing and I have no Idea what is right and what is wrong pertaining to temp read outs I really wish that D.F.I would quit this Tom Foolery with this fluctuating temperature read out with Bios Release

[Rammsteiner]

Justapost, I printed Brother's BIOS template as well but for me that's a no go really.

Im priming now at 1.35Vcore, it just freezes or errors when running lower. No clue really how results could differ so much. 1.25V vs 1.35V is quite a significant change.

Ive Vnb at auto though, Vdimm at 2.12, all auto settings apart for TRFC, set those to 195. Auto settings results in 5-5-5-18, unganged DDR800. HT and Vht are stock as well.

I might have just bad luck with the stepping or I messed something up resulting in OC and Volt scaling.

Things I might check in future:
1) 2x DIMMs instead of 4x, maybe it affects stability
2) Testing how temperature and clocks affects each other
3) Try to get rid of freezing and find the weaker core, but the freezing seems to happen almost always.
4) Play with NB vs CPU ratio's both Volt and Mhz wise

[Brother Esau]

A good reason that you are requiring more Vcore is because when you first got it you leaned on the voltage too much....Possibly! I find since back with 939 Opterons and x2 overclocking is that with AMD Processors its Taboo to intruce High Voltage to the Processor off the front line I also find that if you start out with lower voltage and find the lowest it can run on at stock (Boot) you can actually train AMD CPU's to respond to and utilize lower voltage when Overclocking Most say that is a Myth but I have successfully done this with every AMD CPU I have since I read this 2 years ago Maybe its and I am just lucky but 6 AMD CPU"s later I really believe that its achievable from my results and what I have experienced at least.

[Rammsteiner]

Good to know, if it's true or not, Ill try it out with my next CPU. Can't hurt to try it out anyway.

[stocius]

A good reason that you are requiring more Vcore is because when you first got it you leaned on the voltage too much....Possibly! I find since back with 939 Opterons and x2 overclocking is that with AMD Processors its Taboo to intruce High Voltage to the Processor off the front line I also find that if you start out with lower voltage and find the lowest it can run on at stock (Boot) you can actually train AMD CPU's to respond to and utilize lower voltage when Overclocking Most say that is a Myth but I have successfully done this with every AMD CPU I have since I read this 2 years ago Maybe its and I am just lucky but 6 AMD CPU"s later I really believe that its achievable from my results and what I have experienced at least.

I know that worked on the old athlons, for some reason it allowed a stretch, I asked AMD about it and unfortunately I forget what they said. GRRR.

[Brother Esau]

Well think of it this way>>>>. You should really apply that logic regardless because of the Memory Controller residing in the CPU its self and memory controllers are much more sensitive then the CPU its self any day of the week.

[Oese]

OK found that vcore meas. point on some of those little ceramic caps next to the cpu socket.

with cpuz showing 1,456v its actually reading 1,475v at 1,4v * 101,23%

will lower that down and do some testing...

ok at stock with cpuz reading 1,440v its 1,458v on the dmm..

will flash 04/15 and measure again..

[Tony]

Good testing Achim.

Tony, vDIMM at pin 7?

http://pinouts.ru/Memory/dimm_ddr2_pinout.shtml

Pin64 VDD

[justapost]

Achim......I noticed the 2 times I tried 3/07/2008 that temps were much higher then other releases such for example 1/30/2008 so I shy'ed away from it for that reason. Where on the other hand the 3/12/2008 "Beta" Regulated temp readings marginally lower not drastically but still .....This is getting quite confusing and I have no Idea what is right and what is wrong pertaining to temp read outs I really wish that D.F.I would quit this Tom Foolery with this fluctuating temperature read out with Bios Release

What realy astonished me is that having an usb-memory stick plugged in results in ~6°C higher core and cpu temps in idle (not under load).
I don't remeber if i tried 130 bios in the past. However I tried the 9850BE in the M3A with the latest bios.
At 2.8GHz(cpu)/2.2GHz(nb) 1.325V/1.3V cpu temp was around 55°C after a few minutes of prime95. I had the window opened at those testings.
If I open the windows today i get a modes cold airflow over the board (not in a case atm) and the cpu temps drop to ~55°C also.
On the M3A i used the silicon thermal component coming with the xigmatec cooler on the DFI used AS5.
So based on these results the high cpu temps i get depend more on the cpu than on the board here, I'd need an other 9850e to double check this.
While I tested the M3A I had removed the heatpipes from the booard and applied AS5. It did not help with chipset and pwm temps.

Justapost, I printed Brother's BIOS template as well but for me that's a no go really.

Im priming now at 1.35Vcore, it just freezes or errors when running lower. No clue really how results could differ so much. 1.25V vs 1.35V is quite a significant change.

Ive Vnb at auto though, Vdimm at 2.12, all auto settings apart for TRFC, set those to 195. Auto settings results in 5-5-5-18, unganged DDR800. HT and Vht are stock as well.

I might have just bad luck with the stepping or I messed something up resulting in OC and Volt scaling.

Things I might check in future:
1) 2x DIMMs instead of 4x, maybe it affects stability
2) Testing how temperature and clocks affects each other
3) Try to get rid of freezing and find the weaker core, but the freezing seems to happen almost always.
4) Play with NB vs CPU ratio's both Volt and Mhz wise

I'm testing in ganged mode atm. With the 9600BE all my testings ran equal stability wise with ganged and unganged mode. I plan to try unganged mode once i found the lowest possible voltages (cpu,nb,mem) for ganged to see if it's the same. After that i'll try to find the lowest nb voltage required for 1.8/2.2 and 2.4GHz at 2.8GHz cpu frequency.
Next i'll lower the cpu voltage to 2.7GHz and test if i can get a stable system with lower nb ferquencies.
I also plan to check nb voltages at 2.8GHz with mem in 800 mode.
During my initial testings with the 9850 i found that the nb (at 2.2Ghz( required alot more volts to be stable at 2.9Ghz then on 2.8Ghz. That is an somewhat odd result. Getting 2.9GHz stable is a difficult thing with my cpu so i'll inspect these nb/mem/voltage issues in the save below 2.8GHz area.
Another note, in the M3A prime95 always failed on the thired core when i tried to get 3GHz stable. I disabled core three and four, the board has a bios option for that and voila 3Ghz cpu and 2.2Ghz nb ran stable at 1.35/1.3V. All four cores where not even stable at 1.525V. I guess it's mainly still a lottery thing how far the cpu can oc.

EDIT: I'll add my lowest voltage finds for nb and mem to my previous post to keep the results together.

[Brother Esau]

What realy astonished me is that having an usb-memory stick plugged in results in ~6°C higher core and cpu temps in idle (not under load).

Well its just a Real Fact as it Happens when my Epson Printer is Turnned on the CPU Temps climb just as you are experiencing and have done so since the beginning with very early Bios

[justapost]

OK found that vcore meas. point on some of those little ceramic caps next to the cpu socket.

with cpuz showing 1,456v its actually reading 1,475v at 1,4v * 101,23%

will lower that down and do some testing...

ok at stock with cpuz reading 1,440v its 1,458v on the dmm..

will flash 04/15 and measure again..

Can you take a picture and post it with marks please ? I don't have a camera here atm but i will be able to post pics tomorrow if you don't have one at hand at the moment.
You guys seem to use cap's as an shortage for capacitors, that baffled non native english speaking jap here.

[Oese]

with 04/15 its reading 1,440v in cpuz and 1,436v with dmm at 1,4v stock vcore btw.. so cpuz is indeed pretty accurate, but there are these certain discrete steps.. 1,44v, 1,456v and so on..

you can measure vcore at these points marked red...

[Rammsteiner]

Wicked, 2.8Ghz (didnt test HTT much yet) is the max for me too. I managed 2.9Ghz although it froze everytime after about 1h and 20m at the same test. Even at 1.45V it didnt help, either Phenom just doesnt go higher or we miss something/BIOS needs another update/whatever.

[justapost]

Wicked, 2.8Ghz (didnt test HTT much yet) is the max for me too. I managed 2.9Ghz although it froze everytime after about 1h and 20m at the same test. Even at 1.45V it didnt help, either Phenom just doesnt go higher or we miss something/BIOS needs another update/whatever.

2,9GHz work alot more stable here with nb at 2.2Ghz. Exspecialy if play with ref HT's. With 2.0GHz nb i run into stability trouble at ~205Mhz. I could sort out these harsh instabilities with 2.2GHz. For 2.9/2.2Ghz I need ~1.45/1.4V with temps in the 70°C and above area. I don't feel comfortable with those temps so I did not yet test with longer prime95 runs. That 2.9/2.2Ghz setup ran a three hour benchmark session stable however.
I can also run at 3.0/2.4Ghz at 1.525/1.5V with temps close to 80° after a few seconds of prime95. This is a non benchable setup.
That Xigmatec cooler i bought comes close to T120 Extreme and gives me ~5°C lower temps than the BigTyphoon 120 at ~60°C temps.
But for 3GHz I really need good watercooling for further testing.