You should be pushing me harder to get one so that i feel less guilty & i can say you lot made me do it :DQuote:
Originally Posted by Rammsteiner
You should be pushing me harder to get one so that i feel less guilty & i can say you lot made me do it :DQuote:
Originally Posted by Rammsteiner
I think it's the nb whom degenerated here during my 2.4GHz stability test. I ran prime95 for two days with a nb vid above 1.4V.
During first testings i found that 2.9GHZ required a 2.4GHz nb to be somewhat stable. Since i can not get 2.4GHz stable with a reasonable voltage even with 2.8GHz cpu frequency 2.9GHz is no go area here.
There is no need to use the same voltage for the nb as for the cpu. In CnQ the cpu vid drops to 1.05V while the NB voltage stays at 1.3V. Of course it is save to keep the nb vid at least at cpu vid level, but usually the nb requires less voltage than the cpu. I think DFI recommends that because a too low nb voltage is hard to detect. It can take more than eight hours to detect a too low voltage with prime95. I always tried to keep the nb voltage as low as possible because it has an impact on the cpu temperature.
I'm still hoping to find a good combination of multis whom will be stable >2.8GHz. The system is very snappy even at 2.7GHz but 200MHz is no good oc result in my eyes. The only thing required for that is change the cpu multi which is not a funny oc. I had better results with my 9600BE. He ran at least at 2.6GHz(+300MHz) stable (with 32bit osses). Also my 9500 was stable at ~2.55GHz (+350MHz).
I settled down with settings in sig.
If it's stable for a long while Ill push the HTT by 1Mhz (at 217 now).
Also Im wondering whether the poor memory timings in DDR1066 mode will be always like this or will we be able to select tighter timings with a better BIOS?
Especially the sub timings selection is crap.
Why would you feel guilty?:shrug:Quote:
Originally Posted by Final8ty
I almost went with an intel penryn but I didn't and I'm damn glad I bought a Phenom. Let me state on the record that I spent more to buy AMD.:cool: And yes I may consider Intel in the future. They aint' totally horrible, they just bore me, and I don't believe in benchmarks unless I can see the source code and see how it works and whether or not it is indicative of real-world performance.
With AMD you always gotta take the time to learn ithe chips secrets so to speak. All those review sites generally don't take more than a couple of days max to play with the chips. Then the ignorant sheep take the judgment passed down by said review sites (which may be bribed or biased, hey they're human) It was that way with the Athlons and the Phenoms are even more complex. We'll all be greatful once this chip matures and we're seeing wicked performance even at lower clocks. I wonder if its forgotten that AMD and Intel have different defintions of clocks etc. Intel has always been able to clock higher (if you could overclock it at all) but AMD performs better at less clock, IMHO. So my friends are we buying numbers or performance?:D
Rant, rave. Rant, rave.:shakes: :shocked:
Performance everytime, people forget who pushed the work per cycle importance over raw Ghz clock speed, who pushed for 64 bit CPU's and programs for the masses.Quote:
So my friends are we buying numbers or performance?
Now how long ago was it my end of line K6 550 had the beating of the Intel offerings, well since then until core 2 appeared, AMD dragged Intel kicking and screaming into the present, not long ago that was it?
Now when I see this;
http://img.villagephotos.com/p/2004-...Benchmark2.jpg
I wonder :rolleyes: I suppose I could write an article/review use the above and convince half the world I didn't use a selective benchmark but a real world one based on a real world application.
BTW it's been running at 2.9 since Tuesday of last week:up:
Something wicked about your 9850. I can't run stock at 1.249 or boot 2.8G at 1.3 yet you are at 2.9G. Someone got the cherry!
Maybe the tart and cream to go with it too.Quote:
Originally Posted by AlabamaCajun
I'm still running the settings with brother esau's 416 template, only change I've made is upping the multiplier to 14.5x, not pushed any voltages or anything yet.:up: .
They sent all the best CPU's to the UK ;), manchester to be specific LOL
3dmark Vantage time boys, will be installing and testing later with the Sapphire 3870x2's in Xfire X excited to see how they do ;)
Hmmm, lol, where my CPU doesnt want to OC very well, my GPU does! Gave it only a slight Vcore bump from 1.34 to 1.4 and my GPU runs 904Mhz/1251Mhz... Thus far, on stock cooling:D.
Now I got to fix my 2nd card, a resistor broke off while moving it:mad:. I think it's possible to solder it back on though, but Im not so good with soldering... actually not good with anything electric:ROTF:.
I even blew one GPU a long time ago while I tried to measure Vcore... I accidentally connected two resistors with each other, spark, smell... crap.
[EDIT]Sandra seems to be the only real program to pick up the RAM bandwith right? Which version/tests are DFI proof, aka not killing my board and/or CPU?
Hey guys will be getting some more material added to the Front page real soon my apologies for not doing so as of yet but been real pr-occupied with my cats health and nourishing him back to optimal state....its a process:)
http://www.abload.de/thumb/werwirdmillionr7y1.jpg;)
Gotta be kte's mouthpeace now, he answerd a few questions already. :)
Timings are determined by the imc on the cpu, the bios can not provide better ones. For 1066 mode it's (3,4,5)-5-5-15 the detailed timings can be found in the BKDG.
Sandra configuration dialog allows disabling of DMI stuff and alot of other readings.
@Tony: Have you posted the stepping for your phenom in your thread here? I might have missed that. Have you tried only one cpu?
I have only 1 9850, stepping will be posted when I next change the board out ;)
Thank you. :)Quote:
Originally Posted by Tony
Interesting Limits beside HT COF<=NB COF, BKDG p. 35
http://www.abload.de/img/pstate-limitsjc3.jpg
The funny thing is we ran the 9600BE at 100MHz which should normaly not be possible. It was slow as hell at 100MHz, 400MHz are good to go for surfing and such things. I postet how to tweak the p-states in the DFI Official thread around page #24.
I have Step B3 but date-coded 0810APAW. Looks like us Uncle Sam fanboys are getting the older stock :cry:
I'm going to do the 416 template tonight if I get the time to diddle with it.
@AlabamaCajun.....4/16/2008 is the same as 4/15/2008 except for a message in CPU Voltage string you can find the Tamplate you seek already done in its entirety on the front page where my 2800mhz O.C is with full Bios settings its in Attacment form at the Bottom of that post just edit the Bios Code Date;) .....Post #4 ...http://www.xtremesystems.org/forums/...61&postcount=4Quote:
Originally Posted by AlabamaCajun
Thank Bro,Quote:
Originally Posted by Brother Esau
About the RAM timings. The 2x2G GSkills I'm running are TRFC 127, I notice you have them laid back at 192 is that for 1066 or are you running 2x4s ? I'm only running in the 800 range until I know these 1000s will do 1066 before I push them.
I have the sticks in my sig:) What you do is load optimized defaults reboot let it build the DMI pool re-enter Bios set Memory on manual leave the rest at auto and just change the divider to 1066 if it boots it will run the sticks at 1066 divider but you have to do that step first get to desktop and see how it is re-boot in few minutes and change the rest of the settings primary timings trc and trfc and cpu Pwm and all that good stuff. Just start with 195ns set on all 4 trfc to begin with and then see what you can lower from there;)
why does it seem that barely anyone can get 3.0ghz+ stable while tony is rockin 3.5ghz? =(... the highest i have been able to get 24/7 stable is 2.956ghz, my cpu just HATES 3.0ghz... lol, im using 11.5multi and 257 HTT. i have been able to run 3dmark06 at 3.0 and 3.60 (with x12 multi) but i wasnt 24/7 stable with occt/orthos >.<... oh well, ill just have to wait until the 9950 or the deneb's until i can get my 3.0 :(
Quote:
Originally Posted by GunterFalstaff
Because you are using High HT:yepp:
no... thats not why... because using the x15 multi wont work either... ive tried 12/12.5/13/13.5/14... etc... none are stable at 3.0Quote:
Originally Posted by Brother Esau
I still think there's some fishy OC'ing behaviour in vista x64...might just be me thoQuote:
Originally Posted by GunterFalstaff
It might just be the trfc, I thought with most 2G per-stick you would use 127, 192 was for 4G sticks and 364 for 8G.
I'm at 2.6G tonight just piddling around but will test later. I did have to USE AOD and the BIOS setting was ignored (may have left the X12.5 setting :shrug: )
Update; I'm being owned by this beast. Reboot after a few minutes. Back on the Dec 2007 that came loaded. Moved the RAM to the green slots and holding steady. If this is stable then I may just have to stay where I am. As for the OS boot crashes, they seem to have come from a bad registry entry that I corrected with a reconfiguration of services. As for the reboots, that sucks and I can't have it. I'll keep up on what is going on until something sticks out as a problem.
Yes, that is you yeah.Quote:
Originally Posted by stocius
Cuase uder XP64 it's still strange:ROTF:.
At least your Phenom OC's that far man. Ive this stupid crap it wants to be arouns 2.7Ghz. Even if I would trhow 100MV on it it wouldnt move one 1Mhz:(
Just blew my theory, I thought the APBW was a better clocker :bsod:Quote:
Originally Posted by Rammsteiner
Update: Ran through the night smoking at 56-58C in a 75F room at 1.31V (Auto). Apparently lowering the volts results in unstable results. I thought these were supposed to run at 1.25V with headroom to go to 1.3. So far the only issue I saw was Firefox failed to start one time but ran my CAD roller coaster designer with build and simulator running never dropped out. Build is an intense program that creates huge matrices to build the structure that supports the tracks. You can edit this in 3D and it really sunk this machine to it's knees. The apps are all single threaded from what I've seen but so is Firefox. The coaster designer is just for edutainment and not critical but I develop software on this rig making it the box I need to be stable. The alternative is to move development stuff to the (errg sintel box) but I may have no choice if I can't get it solid. Next year I may just go with a workstation rig which means ECC ram but that's the loss you endure for stability. I still like to have fun with the OC and want to see this thing do near 2.8, putting it above the 5600 it replaced. Going back to the 5600 on the DFI board is still a tempting option and wait for the Dennys.
I just noticed what a crap post I made... lol. Full of typo's.
Anyway, this stepping doesnt seem to be a very good clocker at all.
Quote:
Originally Posted by Rammsteiner
I would tend to agree with KTE and say that right now due to growing pains with this New Phenom Technology there for the most part will be no consistency with Stepping / Overclocking Headroom like we saw on the end of Socket 939 Platform with X2 & Opteron Chips.
As always with AMD, newer revisions give better results. I know we're only on the 2nd rev which was mainly for TLB.
But it's just that the yields are so bad on this one. Voltage hardly matters.
:shocked:Quote:
Originally Posted by Brother Esau (Old Thread)
This might explain why my OS goes T**s up in minutes but I can run the olde proggies all night. :censored:
I'll have to do some checking but can a non phenom OC well with a 64bit os?
Yes the 6400BE flies on this board.Quote:
Originally Posted by AlabamaCajun
See the early posts in this thread;
http://www.xtremesystems.org/forums/...d.php?t=169812
I think, that a Phenom actually might be unstable on a 32bit OS where it is at a 64bit OS. Maybe it just takes longer for it to achieve a crash. But it doesnt make sense that a 32bit OS would give higher OC's because it uses less registers... In the end it would use some registers for something and crash anyway?
Ive no clue about the interface between hardware and software, so apologies for that. But I dont think it would (and should) matter.
I can vouch for the lower OC on a 64 bit OS.
Running 32 bit Vista 2.9 was prime stable, could run Seti all day without issue.
Running 64 bit Vista 2.7 max anything over and BSOD, and lock ups.....
Running 64 bit means 64 hammers nailing at once that might equate to more draw from the on die power rails.
i dont have any 32bit OS's to try it out on (all i have is 64bit Debian sid, xp64, vista64) but across all of these operating systems i am extremely stable up until 2956mhz... any higher and i am unstable no matter what voltage i try and no matter what htt im trying, but according to how some others are doing these seem to be good results, and i got a lucky chip =/. sucks tho, because i was really hoping for a 3.0GHZ chip and i got one JUST short of it... :( . when is the 9950 due to release? lol..
Quote:
Originally Posted by GunterFalstaff
Yeah you any everyone one else here bro:p:
@ALABAMACAJUN......You get those sticks to boot @1066 y doing what I said?
Windows Server 2008 Enterprise 60 days trial :up:Quote:
Originally Posted by GunterFalstaff
I haven't tried yet, I've got a couple of things going while it's stable. I may try again in the next few days. I was concerned these GSkills were not compatible but I've seen no real indication they would be. I've got 2 sets of these and one set of 6400s.Quote:
Originally Posted by Brother Esau
But the point is, Prime stable doesnt mean stable. I couldnt believe it my self, but it's true. The fact I ran 14.5 hours Prime stable was WORTHLESS, it would freeze anyway.Quote:
Originally Posted by Slagathor
If anything over 2.7Ghz BSOD's on 64bit, but runs Prime stable and such sorts at 2.9 on a 32bit OS... Really man, it still doesnt say anything. Best way to test stability is by using it 24/7, game a little, play music etc etc. And dont shut down the PC, leave it on for days while you do your 24/7 stuff. If it's still running fine after, then you can consider it to be stable pretty much yes:p:. If it crashes after all, then it's most likely you'll reach the same OC as on a 64bit before it stops crashing.
I mean, Tony's OC is done on 64bit after all;).
MRL does with Phenom too, one of the first and most important calculations performed at boot involves it :yepp:Quote:
Originally Posted by Rammsteiner
Attachment 77609
Ty for that KTE...
Although it's quite confusing lol.
tCL in MEMCLK's is just the tCL we read in say CPU-Z and set in BIOS right? Or do they mean the 1000/effective speed*tCL?
The normal BIOS set Cl ;)
That made my brain itch!:shock2: Great numbers though.
:coffee: :shock: :sofa:Quote:
Originally Posted by KTE
http://www.abload.de/img/mrl1y8f.jpg
http://www.abload.de/img/mrl2wsu.jpg
I think this training results are the cause for the different MRL values I got with for example different mem voltages.
^ Yep and they can differ each time - many times you get no boot with a too low MRL or too high NB/too low MRL combo; I know I have :yepp:
RAM and NB clocking looks very dependent on it here though.
I've been thinking before... could the C1 errors be a failure in the MRL setting computation at boot? :confused:
I dont know if that would be possible. Maybe by bad BIOS programming, but not by manual settings I think. At least I cant set a MRL value in BIOS. But I do know by saving in Memset at any default start up your system goes mad and you'll need to delete the save file in safe mode.Quote:
Originally Posted by KTE
Also, HT multi on DFI higher than default doesnt seem to work. Not that I care atm because my system made a big crash with using 240HTT and both NB and HT at 10x multi... 00ps:shocked:
I did I did, I ran ot for a while at 2900mhz...
Played a lot of Oblivion, 3DMark's, PCMark's, BF series, F.E.A.R, I still love F.E.A.R., it;s got to be the best game of all time.....
And I crunch Seti 24/7, olny BSOD's I got were cause by silly things I did, well, tried to do..... :p:
Maybe the CPU need'd more juice???
I even did some 3.0 Super Pi runs, and nary a glitch, 64 bit, 2.8 Super Pi dies....
Quote:
Originally Posted by Rammsteiner
Im happy for you mate. Although I still dont think it's really stable, sorry to say. I'd believe it if it was on for serious a week long 24/7 without any reboot and everyday used for normal stuff. And idle at night, or playing music (speakers muted) etc.Quote:
Originally Posted by Slagathor
What about crunching at night, cause I'll do it......
I have a 40GB HDD I can install on to do it.....
ok, so i tried out running my cores at different speeds for the first time... it was interesting >.<... i HATE AOD, random lockups and freezes and hangs all the time... anyways, it turns out i have a bad core #1... all of my other cores were able to successfuly run at 3.1 stable but the moment my core #1 was set at anything above 2.956... i lock up... =/. just my luck... im being held back but just one lousy core, lol
Seeing all this I did a little sploring. I tried 14X in the BIOS at 1.32V, C1 came on for 2 secs and it booted at 2.5 :p
Fired up AOD, slammed it to 14 on all four cores hit 1.32V and hit GO then Closed AOD.
Good! Still running, I usually don't make it this far.
Time for a little CPUZ to see if it really went there BAM!, rebooted.
I flashed back to the Dec 2007 BIOS due to stock instability.
Should I go back to 416 with the Pheno in my sig or get a different one?
Conventional wisdom is to go with latest but in some cases a problem can arise. I may be able to get 2.8 with the optimal BIOS or just get a stable 2.7G. I would go the split different divider deal with each core but prefer to run them in tandem.
We have almost the same problem. My problem child core is Core 2.Quote:
Originally Posted by GunterFalstaff
My 9850BE just stopped working. This time lm-senors detect was the cause. I booted the board with an 5000BE flashed the Bios and cleared the DMI area. But no louck with the 9850BE, stuck at C1, even with DDR2-800 memory. The 9600BE works. :mad:
I had the same problem with my first 9850, i sent it back and it was verified dead. Now my luck seems to be getting worse cause my new 0811bpkw wont even hit 2.6ghz without bsod'ing. It may be this m2r thats almost got me ready to switch to intel but im not sure yet. Another weekend of cursing will be the tell.Quote:
Originally Posted by justapost
You also trashed your cpu with sensors-detect under linux? I saw one user reporting the same issue here. Thought it is a random occuring phenomenon, so i tried it. 2.6GHz does not sound promising. Intel slowly becomes an alternative here also. :shakes:Quote:
Originally Posted by The Weedman
I'm using lm-sensors, no problem for me on MSI K9A2 Platinum and Abit AX78. Looks like a board specific happening :(
I'm sorry for you, I know what it feels, mine was after benching in Sandra. :(Quote:
Originally Posted by justapost
+1 , my "new" 9850BE 0809EPMW only boots at 2.6GHZ and vCore 1.36v in Vista x64, more it's a no go, doesn't matter what I do. Vista x86 2.7GHz at same vCore.Quote:
Originally Posted by The Weedman
Yeah, I ran lm-senonsor succesfully on the M3A with the 9600BE. It's the detection routine for SMbus that killed the chip.Quote:
Originally Posted by KTE
Time for RMA, 2xPhenoms for AMD, 3xHD for Western Digital, that saves shipping costs. :)
@aGeom: Thank you, looking at your and weedman's steppings 810's seem to be the best clockers atm.
Tssk , Tssk , Tssk:shakes: Why is it that I do not have any of these issues with DFI 790FX-M2R or 2 Phenoms now? Lets look at the common denominator here.....Ummmmm....thats right High HT and too much voltage:yepp: When are you guys gonna listen? I keep saying High HT means moot and AMD CPU's have a memory when it comes to voltage applied and the Phenom needs to be approached with a light and artful touch:rolleyes:
Perhaps this is my fault as I have not been able to dedicate the time to posting the necessary precautions associated with overclocking Phenom with this board:( You CANNOT leave the default Vcore of 1.31 set in Bios and work from there by leaning on the Voltage in attempts of overclocking Phenom guys! First agenda should be to get it to run stock at the lowest possible voltage settings and train you're CPU to run at such. AMD CPU's have a memory and if you train them to run stock at @1.31v then after a certain period of time they will not run stock at lower voltage and even more voltage will be required to overclock the CPU in Return, which from the start what I have seen on this forum in regards to Phenom Overclockig leads to failure in a hurry! Take the advice and release these old mundain thoughts associated with K8 Overclocking (High HT) as it means ...MOOT...and should only be used to fine tune you established Base Overclock which is achievable via CPU , NB ,HT , Ram all being able to clock independently HT Overclockig is DEAD...Get Over It already please:yepp:
My cpu died on an nearly unoverclocked system.
CPU: 13,5x200 = 2700MHz at 1,3125V
NB: 10x200 = 2000MHz at 1.3V
MEM: 400 4-4-4-12-20 75ns at 2V
It froze as sensors-detect tried to find SMBus devices and gives C1 since then. I'm sure it would ahve done the same at 2.5GHz.
:rofl: man, we are not talking about 9500/9600 vanillas, HTT/HT are set @200/2000 default, mine died @:Quote:
Originally Posted by Brother Esau
CPU=15x200=3000 vCore=1.36v
HTT=200
HT=10x2000/1.27
NB=12x2400/1.325
This new 9850 one, after I pop it in, my first set was 2500@1.25v, didn't even boot, stoped after build DMI and backup CMOS.:yepp: :p:
Hey Achim:) Don't worry brother I am not honing in on you're current misfortune with my statement it applies to One and all:D But seriously guys Please take the advice and test my theory out I really don't need to be right for my own vanity sakes I mean that but what I propose holds true with the Past 3 Current AMD CPU's that I have owned and I think that if enough would be willing to get past the old way of thinking and possibly take this suggestion we may be onto finding out the way this works effectively and efficiently in a much shorter time frame seeing how Phenom User Base has exploded since B3 Release;)
Thanks for Listening..........Campbell:)
I was trying to test this one http://www.rebelshavenforum.com/sis-...;f=29;t=000042Quote:
Originally Posted by Brother Esau
;)
Quote:
Originally Posted by aGeoM
Well then perhaps you may be inclined to test my theory out next go around? Perhaps we need to do as this thread is intended to do Debug Phenom Problems and if that means lowering the Voltage threshold of what not to exceed then so be it and we will be doing our parts so others don't experience the same anguish:up:
Quote:
Originally Posted by aGeoM
Polygon is very well respected by me and I hold his knowledge in high regard but......My intent is to get facts on what is going on and my suggestion is a common sense one ...DO NOT...Run Unnecessary Voltage into a Electronic component when its not needed and you most certainly never hear me saying what I propose will yield Higher Clocks:clap: I Propose Logic through Simplicity Nothing More!
Yes, it was the first thing I realize, on B3, he didn't like voltage at all, when testing 3.2GHz it booted to Windows vCore@default+107%~109% and didn't after it.Quote:
Originally Posted by Brother Esau
I know it was not high volts or HTT, was something else that overcomes my knowledge.
Hmm from what i have experienced with phenoms, they tend to have an upper voltage limit. Above you don't get it stable. So the top-down approach for overclocking will not work. Bottom-up is a long way if you really want to find all lowest voltages. Also knowing the lowest possible voltages does not help if you get in the idle freezing area, which seems to start at a individual cpu frequency.
I guess the fastet way is to test each core separate at say not more than 1.35V for B3's till the system starts to freeze idling and then use the multi of the weakest core.
During my first tests with the 9850BE it looked like the NB requires higher volts with higher cpu frequencies. This was a false impression it only took prime95 longer to fail with lower cpu frequency, so the nb voltage requirement is really independant from the cpu frequency. At 2.9GHz the system was alot more stable with a 2.4GHz NB frequency, but i could not get a stable 2.4GHz nb with a reasonable voltage (stopped at 1.4325V). So I think it's a good idea to test the capability of your nb first.
:2cents:
If someone finds a fix for the idle freezing I'd be glad to hear. :shrug:
I have Experienced no such Issues of Freezing or anything Esle for that Matter since O.C Phenom;) My system Runs 18hrs a day at a minimum:cool:.....Overclocking Minds Since...1986...:usa::surf::usa:
Just to prove my theory that Phenom can overclock with Low Voltage and requires a light and artful Approach, heres a 4hr Prime Run.
DFI LANPARTY UT 790FX-M2R (BIOS 4/15/2008 "Beta")
AMD Phenom 9850 Black Edition "B3"
Tuniq Tower 120
OCZ Reaper HPC (PC2-8500) 5-5-5-18 2X2GB
POWERCOLOR AX3870 X2 CROSSFIRE CATALYST 8.3
Auzentech XPlosion
Areca 1231 ML {Intel IOP 341} ~ 6 ~ 250gb Hitachi Sata II Raid 6 ~2~WD 74gb Raptors ~ Raid 0 ~
Plextor 750A
SilverStone OP1000
CM Stacker STC-T01
Code:>HT Link Control:
HT Link Width......................... - 16 bit
HT Frequency.......................... - 2GHZ MHz
IH Flow Control Mode.................. - Disabled
HT Link Tristate...................... - Disabled
2x LCLK Mode.......................... - Disabled
Unit Id Clumping...................... - Disabled
>CPU Feature:
CPU DID.............................. - 1
CPU FID.............................. - 0C
CPU-NB DID........................... - 1
CPU-NB FID........................... - 06
AMD CPU stepping..................... - P0
>DRAM Configuration:
MCT Memory Timing.................... - Manual
2T Mode.............................. - Auto
Cas Latency Time..................... - 5
Pre Charge Delay (tRAS).............. - 18
DRAM RAS# Precharge.................. - 5
DRAN RAS# to CAS# Delay.............. - 5
Bank to Bank Cmd (TRRD).............. - Auto
TRC.................................. - 23T
TRTP................................. - Auto
Write Recovery (TWR)................. - Auto
Write-Read Command (TWTR)............ - Auto
TRFC0................................ - 195ns
TRFC1................................ - 195ns
TRFC2................................ - 195ns
TRFC3................................ - 195ns
Trdrd................................ - Auto
TrwtT0............................... - Auto
Twrrd................................ - Auto
Twrwr................................ - Auto
CS/ODT Pin Fine Delay................ - Enabled
Bank Swizzle Mode.................... - Enabled
Auto Tweak Performance............... - Enabled
Optimal Performance Mode............. - Enabled
>ODC Control Mode:
DRAM Drivers Weak Mode.............. - Normal
Output Drivers Comp................. - Normal
>Channel A:
CPU On-die Termination.............. - 300 ohms +/- 20%
DQS Drive Strength.................. - 1.00x
Data Drive Strength................. - 1.00x
MEMCLK Drive Strength............... - 1.00x
Addr/Cmd Drive Strength............. - 1.50x
CS/ODT Drive Strength............... - 1.50x
CKE Drive Strenght.................. - 1.50x
>Channel B:
CPU On-die Termination.............. - 300 ohms +/- 20%
DQS Drive Strength.................. - 1.00x
Data Drive Strength................. - 1.00x
MEMCLK Drive Strength............... - 1.00x
Addr/Cmd Drive Strength............. - 1.50x
CS/ODT Drive Strength............... - 1.50x
CKE Drive Strenght.................. - 1.50x
>PWM Settings:
PWM Control.......................... - Enabled
CPU VID Control...................... - 1.23750v
CPU Vid Special Add.................. - Auto
CPU NB Vlotage Control............... - 1.23750v
CPU NB Vid Special Add............... - Auto
DRAM Voltage Control................. - 1.80v
SB PLL 1.2V Voltage.................. - 1.20v
NB Core Voltage...................... - 1.200v
NB PCI-E Voltage..................... - 1.120v
NB HT Voltage........................ - 1.270v
Genie BIOS Settings:
Timing Mode.......................... - Manual
Memory Clock value or Limit.......... - DDR 1066
PCIE 3,4 State....................... - Auto
PCIE 2 State......................... - x2
CPU TLB Cache Fix.................... - Disabled
AMD K8 Cool & Quiet control.......... - Disabled
CPU Clock Ratio...................... - Auto
Spread Spectrum...................... - Disabled
CPU Host/HTT Clock................... - 200mhz
CPU Host/HTT Boot Up Clock........... - 200mhz
CPU Host/HTT Adjustment Gap.......... - 5mhz
PCIE Clock........................... - 100mhz
Bro your chipset voltages are too high. My board ran stable with those at the lowest possible values. :) Four hours of prime95 do not mean that much, I ran prime96 for nine hours at 2.9GHZ/2.2GHz before it failed , 2.8GHZ/2.4GHz failed prime95 after ~13hrs. At 2.9GHz however i had the freezing issues, the setup i primed for hours froze after a few minutes just surfing the web, repeatable.
You plan to go above 2.8GHz? :rolleyes:
During my first testings i first had freezing issues at 2.9GHz with NB speeds below 2.4GHz. I thought i could fix the freezings with an 2.4GHz nb but this turned out to be impossible due to the high voltage the nb would have required.
Looking at my testings with nb at 2.4GHz i think >1.4V for ~30hrs was too much even at a cpu temperature around 60°C. I used the bottom-up-lowest voltage method here and started at 1.3V.
After that test I started to realize I get those idle freezings even at 2.8GHz in winxp 32bit. So I think i'd better tested the nb speed via ref HT increasement with an 11x multi.
I use the bottoms up method also.Quote:
Originally Posted by justapost
I find that drinking Captain Morgan before trying to overclock makes those pesky reboots less painful.
:-P
(Sorry... had to be said.)
I'm at 2.55G using 204x12.5 and Auto.
VCore 1.3 OV'd to 1.31 with a 12.19V supply.
HT on 1.8M
All else on auto.
I'm going to attempt to lower the voltage over a few days to around 1.275. 1.25 has never worked but may be attainable over time is Bro Esau is right about the learning part. I just want to know where the magic is?
Lol, that might also help. ;)Quote:
Originally Posted by keithlm
So is AOD a must when overclocking phenoms? Can you reach the same limits just using bios adjustments? Sorry I didnt see the answer to my original question in the thread
What do you mean chipset voltage is to high? Which setting? If you mean CPU-NB Voltage I set it the same as CPU Vid as suggested by DFI Engineers and have yet to have a issue. Also I find since doig that I was able to kick it down 1 notch with CPU CoreQuote:
Originally Posted by justapost
I will OC more when my cat is better and I have time right now it has to wait!
If you can go down with the cpu voltage after you upped the nb voltage it will fail, but after a longer time. That's what i found with the 9500 on the M3A. NB and CPU voltages operate completely independant, assuming that the cpu temperature does not exceed 60°C.Quote:
Originally Posted by Brother Esau
I have no clue why DFI-Engineers recommend the same settings for CPU-VID and NB-VID. I mean that is the benefit from split power planes. In CnQ if the cpu enters p-state-1 the cpu voltage will drop to 1.05V and the frequency to the half of the stock frequency, while the NB stays at stock frequency and voltage.For 2GHz NB-frequency my 9850BE required 1.175V NB-VID.
But I talked about the four SB/NB ... Voltages at the bottom of the PWM page. My system ran perfectly stable with all those at the lowest possible values, but i have a not so power consuming hd2600xt here.
@Jimmy411: Hmm how (un)usable AOD is (with the DFI board) was posted here a few times. It does not read the correct actual values and applies stock voltages to the cpu and the memory, that causes system hangup's if your already bios tweaked system does not run at those voltages.
You can enable some AutoXPress features via red-button mode whom can not be tweaked via the bios. Also you can easily modify single cpu multis with aod. Otherwise you'd need to tweak the p-state registers of the cpu to archieve the same effect. AOD's stability test and auto overclock function both do not find the idle freezing issue many phenom users get here.
Overclocking via the bios is definately more reliable atm.
Hey Achim, Just wanted to say thanks for the Avatar change!
Lemmy Rocks... :D
I'm buzzed again, but I kinda have to agree with Brother E...
Ya have to be gentle with Phenoms.
Camp, lowering CPU voltages is exactly what looked like it killed my first 9600BE and since then I am no way going to try it. I started to overclock using lower than stock NB voltage and 1.28v CPU voltage. It went off pretty sudden with C1 errors and never restarted; the last changes I made before it happening for a week under load were lowering NB volts to 1.1v :(
I would be careful to void undervolting the NB, the common base in all my CPUs giving C1 errors after a while is undervolting the NB - overvolting CPU sub 1.5v is not the problem as long as you can keep the heat around 25°C below maximum core allowed temps.
There's nothing wrong with my current 9850, its the board that blew under load ;)
As for "burn-in": that's a myth, electrically its impossible but there's a big difference between a common engineer and an engineering professor, mountains to moles difference, the latter are the ones who mainly design new CPUs and even they will not be able to explain a lot of CPU happenings we observe. What they "expect" is many times not what we "observe".
Find me one microprocessor engineer who can explain idle instability we all observe.
Find me one who can explain x setting stable, reboot, x setting failed, reboot, higher y setting stable.
Find me one who can explain CPU death after 8 weeks of stock use.
Find me one who can explain the C1 errors on clear CMOS.
Find me one who can explain IMC sudden death when undervolting.
I'll let you know in advance: none. They are strange unexpected behaviors.
In studies, no one is a demi-god, they are humans and they learn things through common laws. Some things are just not in our knowledge unless we have experience of it, simple as that. Everything is proven or negated by practical empirical results, not theory or what someone believes.
Needless to say, I've had these relations within my family for years now, so it's not like I'm going to listen to someone who has less knolwedge, expertise and experience on figuring out microprocessors. Strange, unexplained CPU happenings are still occuring with Phenom. :shrug:
I've changed onto Abit board to checkout;
a) if the PWM blow with a 9850
b) if I still get idle instability..
Hmm but you have also tried high voltages and frequencies on the nb before you tried the lower settings if i remeber correct. Also you already had the freezing issues before. Can you remeber a phenom whom had the freezing issues right from the beginning without having applied +1.4V to the nb?Quote:
Originally Posted by KTE
I'm not talking about my dead chip, he definately died due to some lowe level io action caused by sensors-detect.
For me the freezing issues started after i tried to find the nb-vid for 2.4GHz stable and had the system run for ~30hrs with nb-vid's above 1.4V. My 9600BE has had no freezing issues but was simply unstable at reasonable voltages above 2.6GHz. I never tried high nb clocks there and used low nb-vid's.
Not on the 2nd 9600BE ;)Quote:
Originally Posted by justapost
Undervolted first to find max stable & max bench with stock voltages.
Last 9600BE and this 9850BE. I've not given this one more than stock NB volts yet, its always 1.3VID which gives 1.288-80v under load.Quote:
Also you already had the freezing issues before. Can you remeber a phenom whom had the freezing issues right from the beginning without having applied +1.4V to the nb?
That's why I never pushed the NB on this one, left NB at maximum 2400MHz and mostly 2000MHz because for me, both are easily load stable at stock volts.
However, I'm not sure what is stable with NB anymore.. XP, Vista and Linux are showing different levels of stability for each setting :confused:
Quote:
Originally Posted by KTE
In short Tye its new technology and we are all learning even the engineers so there is no stead fast rules applied as of yet but we will work on that I reckon now won't we?
Theory is subjective for debate as you don't have to be right or wrong thats the beauty of theory :rofl: I have had quite a time i the past overclocking Psychologists minds with that one when they try using them $10.00 Readers Digest Power Words when talking with me at parties and such:rofl: they think they are so superior with all this B.S Psycho Babble they Utter you know figuring everyone is as dumb as a bag of bricks compared to them and what not .....So I see it as my duty to shut them down with that Theory sheet myself:stick::hump::D
KTE, you are right :yepp: Engineers come in many flavors just like carpenters. Architect designs a new house, framing carpenter puts up the walls, finishing carpenter trims it out. Neither one can really tell you what that house is going to do until it's lived in. Come back a few years later and find the garage closed in, gutters added and a door replaced. Things that were experiences of the home owner are somewhat off what the designer envisioned and he never lived there. I don't think there is enough time for chip makers to fully test CPUs today. The need for more power grows daily and competition once bitten turns up the heat.
I don't believe in the magical powers of burn in either but I can't rule out the first few hours you run a new MOBO that it doesn't change. Heat does change things and semiconductors are no exception but this happens within a couple of hours not something you can make happen over a few days. Violins and Guitars age with time and climate changes getting mellower over the years. Chips just are what they are. When CPUs are mounted they go through several tests that stress then under voltages and heat/cool cycles. By the time we get them, they are already "broken in". What I am seeing here through experience and the posts here is a BIOS that grows into out settings and chip settings. I'm taking an approach I used to with my old P4, take it up slowly and in increments and leave AOD alone. AOD was good to let me know where the potential is, now BIOS tweaks can get me to 2.7G+, my goal.
Last week I tried just setting the clock up to 206 and met the BSOD.
Yesterday I started the slow climb with just 204 and Auto Volts.
Today I went in and dropped the VCore to 1.28 and dropped HT to 1800M, saved then reset right after the Save to DMI.
This time I went in to BIOS and raised the clock to 205 and touched nothing else.
Finally the VCore reading in CPUZ is actually 1.28 and speed 2.56G (2.558... as these CPUs seem to have a lot of errors in speed).
"I now see tunnel at the end of my light"
Crazy, good that i asked. I had the impression you are somewhat addicted to high NB clocks but have missed that you did not need that much voltage for em. My 9600BE however never ran above ~2.1GHz. Don't get me wrong, i', not a fan of an extreme undervolted nb. I try to find a limit where it starts to get unstable and normaly use 0.05V-0.1V above that for 24/7 use.Quote:
Originally Posted by KTE
The chip stoppen working with 2GHZ nb at 1.3V however.
I contacted the lm-sensors developers and they might blacklist the board for smbus scanning. Meanwhile i'll try to read the sensors-detect code.
Anyone have any idea what DFI is doing about these breakable-glass boards?
I guess having all these tweeks has it's drawbacks.
i'm sorry if this question is inappropriate in this thread. i got my dfi 790fx dk m2rs board a few days ago. i finally got a chance to fire it up. i'm in the process of testing stability. i can't help, but notice this high pitch sound coming from the pc. anyone have this problem with this board? is there a fix? the sound is deafening and i'm thinking of an rma to newegg. please help.
AFAIk there's no fix. I think you can RMA it though, but there's a chance your new board has the same issue though.
i see, :(. dare i say it, i'm going to rma for refund and get a different 790fx board. what's the alternative to the dfi? msi? gigabyte? please advise.
There is NO Alternative to DFI! jk,lol. I've seen some DFI users witch to Asus but haven't any results yet. Gigabyte or MSI should be ok.Quote:
Originally Posted by dng
does the M2R have the high pitch sound? i wonder if i can even find the M2R anymore.
Mine doesn't but some M2Rs do and yeah you might have a hard time what with DFI discontinuing the board. I'll look around for you, if you want.Quote:
Originally Posted by dng
The squeal happens sometimes just bad luck of the draw is all and all of them do not do it thats utter rubbish;) You can get the D.F.I LanParty UT 790FX-M2R Here.......http://www.ncix.com/products/index.php?sku=27484&vpn=LanPart%20UT%20790FX-M2R&manufacture=DFIQuote:
Originally Posted by dng
thank you all for your help, but i might go with the msi. still have 14 days to decide if i want to do the rma. do you know why the high pitch sound? i really the dfi board. i had dfi lan party boards in the past withing the nf4. i switched to foxconn for the 590 and now i'm going back to dfi for the 790fx. if not for the annoying sound, this board is pretty sweet.
As said before I posted my dead cpu issue in the lm-sensors mailing list. Seems there is an similar issue already existing with the DFI Lanparty NF4 SLI Expert.
I mounted an X2 5000BE in the mobo and tried sensors-detect again. This time the system froze but booted without problems afterwards. :rolleyes:
With help of jean from the mailinglist we tracked the problem down to an unknown SMBus device at 0x2e. This device seems to exist at the same address in the DFI Lanparty NF4. Probing that device causes the system to freeze here and to reboot on the NF4.:shakes:
http://bugzilla.kernel.org/show_bug.cgi?id=5889#c18
ANy good bios updates laytely or is 415 still the best so far????
How bad is the sound anyway? Is it nticeable over loud fans? I'm starting to hear things I may not like, but then again it's most likely nothing.Quote:
Originally Posted by dng
it's loud even with loud fans and an open case. i wonder if a bios update will help. i'd much rather find a fix then do the rma. listening to this sound for a long period of time, can give me a headache or permanently damage my ears. i have another question, i migrated my 5000+BE and 4 sticks of crucial ballistix pc6400 from my foxconn 590 sli board to this dfi board. with the foxconn, i can oc to 236x14 with memory speed at 944 and 1:1. i can run the timings at 4-4-4-12 on the foxconn, but on this dfi 790fx board, i can't at all even with high vdimm. of course everything is left at default on both boards. the only thing i changed on the foxconn is the 4-4-4-12 timings. i have to run at default dram config on the dfi at 236x14. sorry, it's my first time using this chipset and the dfi bios is pretty intimidating.
Quote:
Originally Posted by justapost
Alright thats what I was looking for. Ive never been a fan of software based overclocking.
Anyone know if DFI has any plans to release a bios that enables the features enabled in red-button mode?
I was going to pick up a M2RS or non S but then SB750 was announced so Im holding off longer.
Quick question.:confused: Has anyone tried lowering 3 cores to say 2.0 and then Seeing how far they can push the remaining core? or is that not a posablity. Seems to me as far as gaming goes that most games only really push one core hard and the rest are just used to 25 ~ 50% so you would gain alot more with one Real fast core and 3 smaller backup cores. My 9850 wont be here till tuesday so i cant test anything. Leme know please =)
Game
Ive those noises too and other BIOS's dont fix it. Even worse, when your PC crashes there's a chance the sound is even worse untill you have it shutdown properly a few times.Quote:
Originally Posted by dng
The sounds are different each time, when Prime95 just completed a test the sound changes etc. If I open Memset the board makes a very scary sound IMO.
But thus far it simply works, and with some nice speakers almost plaing music 24/7, a HD3870 fan fixed at 70% it's barely noticeable for me. Or maybe I got used to it, in the end I got used to a Vantec 92mm Tornado fan too:ROTF:
Thanks for the details Achim :toast:
I have not had those weird noises on any board yet with Phenoms BTW.
I am running 2700/2800/2700/2800 stock voltage stable nearly 20 hours now ... :D
Benched 2800/2900/2800/2800 stock volts too.
Looks like I'm making headway, one core is most likely weak but the rest strong -- that's the one that causes a freeze at the 10-18 hour mark. Trying to figure out which one.
Next setting will be 2800/2800/2700/2800 and you know why... ;)
KTE: i've been following your thread on the msi plat. looks like i might have to get this board instead. the high pitch sound is unbearable even if the pc is loud and i have background noises. it's the kind of sound that just pierce thru everything. any advice on the msi board? what's the latest bios? please let me know. thanks.
Yeah I've heard the sound before, I know what it sounds like. :(Quote:
Originally Posted by dng
Latest BIOSes: http://www.msi-forum.de/thread.php?threadid=35629
Best is 1.13B though from November 16th, has all oc options you need with Phenom but for downcore/single core control.
Most stable and with most compatability is BIOS 1.40.
Board will be fine below 1.525v loadline real on a 125W CPU -- after that you are better off with other better, more costly PWM boards.
For Phenom, either get fully compatible 1066 RAM or have some 1066 RAM with some 800 RAM nearby, as when you clear CMOS//pull power plug with something not officially supported, Phenom might give C1 error at startup until you put some 800 rated RAM in there.
Advice? If you have Phenom, keep voltage as low as you can whilst overclocking - by that I mean clock one core at a time before moving up whilst leaving NB stock or very close to stock. Or you might be stuck oc'ing and figuring it out for 6-7 months like us. :p:
All boards are suffering from similar Phenom related problems, mainly BIOS, apart from the PWM screeches.
I'm on Abit AX78 for now, not the MSI.