Phenom 9950 Overclocking

[RASTAVIPER]

Hey guys, look at this!!!
21,56sec in Superpi with 1,3 VCORE!
http://www.hwbot.org/result.do?resultId=788776

Is this possible??

[RASTAVIPER]

I just cannot get below 22 on Pi..I will just wait for Deneb.. I am even folding on core 3 I ran Pi on Core 1

To get under 22sec with 9950BE isn't so difficult..
Trying to go under 21,6 is the really hard point!!This is the top of the notch for me!!

[ChrisB17]

Chicken patty. Did you see my pic on post #344? What do you think is wrong?

[Chicken Patty]

Chicken patty. Did you see my pic on post #344? What do you think is wrong?

do you have anything like cool and quiet enabled or anything? I would trust what CPU z says anyways, which in this case is correct.

[Rammsteiner]

Lol. You flipped the cable the wrong way. Floppy-drives start to blink green when powered on / cable wrongly attached

Flip cable and the problem is solved.

now that I remember, I went through that once, and it was the cable inserted wrong. Flipped it and it worked again.

Imagine I wasnt even drunk when I did all that It's actually quite an achievement to make so many ups. If you wanted to make such ups you wouldnt even be able to do that

Anyway, I decided to try more Drive Strength testing. I just started AOD to see what the auto settings are;

CPU odt 150Ohms, DQS 1.00, Data 1.00, memCLK 1.50, Addr/CMD 2.00, CS/ODT 1.50, CKE 1.50.

Now although Windsor is a complete different design, I got a very good result in the end, although not stable, with those settings; DQS 0.75, Data 1.00, memCLK 1.50, Addr/CMD 1.00, CS/ODT 1.25, CKE 1.25. Tweaking those settings gave me DDR1140 from no POST --> no boot --> succesfull SuperPI 16M run. Maybe Ill try it again now with Phenom cause something is simply wrong if cant get DDR1000 with 5-5-5 timings while those are DDR1066 sticks.

Also I noticed HT multiplier cant be higher than 10x, no matter what you set in BIOS.

[ChrisB17]

do you have anything like cool and quiet enabled or anything? I would trust what CPU z says anyways, which in this case is correct.

C&Q is disabled as far as I know. And I know this has been said before but wth is this lol. I want my PC to run for awhile and im kinda paranoid about my pc What is TMPIN3? Like what sensor? NB, SB, CPU, PWM ?

[mongoled]

My first ever quad core system, started playing this afternoon



CPU Speed: 3200mhz
CPU-Z: 1.36v

Will post further details once I get home

Would like some feedback on the temps

Cooling is using a IFX-14 with 2 low speed (1000rpm) 120mm fans

[Rammsteiner]

lol, I hope that Temp is wrong

Small update, I got my NB and HT at 2.4Ghz Priming for 2.5h stable now! I noticed that both at a multiplier of 10 has the biggest stability for both general stability but as well for RAM. My RAM is now at a 5:3, so DDR800 to rule out RAM.

Now running 12.5x240. Lets hope it's stable for another few but as well I can get my RAM to run at 240 2:1 4-4-4, that would be awesome. I do have to run my Vnb at like HEX 09, that's like 1.3625V I think. I also noticed that after pushing my 2nd core only to +4 ACC it became stable, so eventually Im able to drop the Vnb back to HEX 10 (1.35V) or even less.

[merkk1]

ChrisB17
I have the same thing . I think it a bad temp sensor . I can hold any of the 3 heat sink SB,NB and the power i can hold it for how every long i want without getting burn so if it was 128c you would not be able to do that and the temp never moves so i'am thinking bad senser.

[Jazzman]

My first ever quad core system, started playing this afternoon


CPU Speed: 3200mhz
CPU-Z: 1.36v

Will post further details once I get home

Would like some feedback on the temps

Cooling is using a IFX-14 with 2 low speed (1000rpm) 120mm fans

Under heavy load, the temps look right to me for the type of cooler you have. I am at 44-46C under water with OCCT stress. I did find a large reduction in temp by disabling the "CPU Tweak" option in the bios. It is denoted in AOD as a red circle around the green button in the upper right corner of the AOD display.

This is a fun quad to tweak. ACC is a great addition to tweakablity.

[Chicken Patty]

TMPIN 3 is supposed to be the V-regs I think. However most of the sensors are bad. Mine stays steady at 22ºc and doesnt move at all, meaning it is bad. So dont worry.

[Chicken Patty]

mongoled

nice job dude. Is that the 12w version 9950??

[Titan7171]

lol, I hope that Temp is wrong

Small update, I got my NB and HT at 2.4Ghz Priming for 2.5h stable now! I noticed that both at a multiplier of 10 has the biggest stability for both general stability but as well for RAM. My RAM is now at a 5:3, so DDR800 to rule out RAM.

Now running 12.5x240. Lets hope it's stable for another few but as well I can get my RAM to run at 240 2:1 4-4-4, that would be awesome. I do have to run my Vnb at like HEX 09, that's like 1.3625V I think. I also noticed that after pushing my 2nd core only to +4 ACC it became stable, so eventually Im able to drop the Vnb back to HEX 10 (1.35V) or even less.

thats great how did you determine core 2 was weak??

[merkk1]

For me it core 4 by prime it always the frist core to drop when i run prime so i set my acc to +2%, +2%, +2% ,+4% on core 4. Core 4 in prime = core 3 in AOD

[Rammsteiner]

thats great how did you determine core 2 was weak??

Everytime time I try a new setting, hoping to improve overal stability, I noticed that fairly quick one worker thread in Prime 95 would error while the others kept on going. Although I let the other 3 threads do 3 more tests, if there was indeed a huge instability the others should have error'd quite fast.

So I loaded up Core Temps and noticed 'Freq #1', aka core 2, had the lowest load. Although it's hard to determine since when once core fails all cores seem to divide the load from the other 3 cores over all the 4 cores again lol But Core 2 showed the slightest load.

Defining which core failed by Prime 95 is not very accurate since it's done randomly, thread 4 could be done by core 4, but as well by core 1. It depends how many programs, tasks and services you fired up. Core Temp seems to be directly aimed at a core, so Freq 0 is Core 1, Freq 1 is Core 2 etc. You probably already know this, but just explaining to people who might actually not know this already.

I wasnt sure though whether Core Temp was aiming at specific cores but I thought it helped me in the past as well. So I set ACC one level higher for Core 2, resulting I just passed 6h Prime 95 Blend stability. That's enough, now back to the RAM

After Im done with the RAM Ill try and slowly decrease Vnb since I first thought the Core 2 error was because of a lack of Vnb. Probably it doesnt make sense, but I only changed NB speed this time and it error'd while 12x250 with both HT and NB multi's at 9 it ran stable for 6h as well, now it didnt.

This somehow leads me to the conclusion that a weaker core on the edge of stability might become instable when you OC the NB.

[ChrisB17]

I touched all the heatsinks and none of them are HOT. Just barely warm. My case has great airflow so all my hardware get cool air thrown at it. So that temp must be messed up.

[Chicken Patty]

I touched all the heatsinks and none of them are HOT. Just barely warm. My case has great airflow so all my hardware get cool air thrown at it. So that temp must be messed up.

trust me it is dude, don't worry about it.

[ChrisB17]

Ok thanks for all the help. I have got to say this is a very helpfull place.

[Chicken Patty]

Ok thanks for all the help. I have got to say this is a very helpfull place.

yeah man. Thats what we are here for. Its amazing how far this thread has gone since I started it. I havent had much time to post anything new so its been taken over by everybody else heheh.

[Sentential]

What kind of load temps are you guys getting because I just noticed something, all of you with ACC on are getting temps more akin to Core 2 Quad chips (ie 60-70c*)

I wonder if ACC is bumping any kind of internal VTT or other voltage up to get them to clock higher?

With as high as 1.6v I am getting under 60c* load in an 80F room which is unusual because my C2Q would normally load 70+

[Chicken Patty]

What kind of load temps are you guys getting because I just noticed something, all of you with ACC on are getting temps more akin to Core 2 Quad chips (ie 60-70c*)

I wonder if ACC is bumping any kind of internal VTT or other voltage up to get them to clock higher?

With as high as 1.6v I am getting under 60c* load in an 80F room which is unusual because my C2Q would normally load 70+

3.2 GHz at 1.4125v I get about low 30's idle, and no more than 45ºc under full load. Room temp is about 75ºf

[mongoled]

Under heavy load, the temps look right to me for the type of cooler you have. I am at 44-46C under water with OCCT stress. I did find a large reduction in temp by disabling the "CPU Tweak" option in the bios. It is denoted in AOD as a red circle around the green button in the upper right corner of the AOD display.

This is a fun quad to tweak. ACC is a great addition to tweakablity.

Thanks for the feedback and assistance Jazzman, I have to look at this CPU Tweak. There are so many options in the bios and so little info on the net regarding many of these!

I forgot to mention, the mobo is a DFI FX with SB750

nice job dude. Is that the 12w version 9950??

Oh how i wish this was 12w, LOL :P, yup its a 125W version, I bought two, both 0836, one was CPCW and one CPMW, i knew straight away that the CPMW would be better

Over 9hrs prime SmallFFT completed, now to blend, heeeheehehee

[Rammsteiner]

lol, last night I got like mad at my PC.

Some issues dont make any sense.

As I said I got 3Ghz/2.4GhzNB stable with the ram at 5:3. With 2:1 it would error or even reboot. I tried lots of Vdimm untill I noticed my RAM was 70degrees Ive no clue how or why, my Vdimm wasnt that high. I figured the reboots came from too hot RAM, so I lowered Vdimm and instead of reboots I got errors. However, errors never went away. At some point I wanted to be sure my setup, excluding RAM, was still stable... It wasnt while I only tweaked RAM after the last stability check

Spend till 03:00 in the night figuring it all out and finally I just set all ACC to +4 and set Vdimm at 2.11V and the RAM still at 2:1. That was stable again. Now this morning I decided, with very loose timings, the RAM at 2:1 and it already runs for a very long time... Ive no idea what's wrong, but I think Vdimm might be incorrect with this motherboard? I cant give a conclusion why my RAM would run 70 degrees before with like 2.2Vdimm and it's only 30 degrees under full load with 2.1Vdimm.

Also I found some interesting about tRC, namely if it's loser you need a lot less Vdimm while it doesnt really kill the performance.

Anyway, I noted my all my settings this time so if I land in a labyrinth like last night again Ive a solid base to get back to. Now lets hope I wont have to mess a lot with timings anymore and get my 4-4-4 without too much Vdimm

[justapost]

lol, last night I got like mad at my PC.

Some issues dont make any sense.

As I said I got 3Ghz/2.4GhzNB stable with the ram at 5:3. With 2:1 it would error or even reboot. I tried lots of Vdimm untill I noticed my RAM was 70degrees Ive no clue how or why, my Vdimm wasnt that high. I figured the reboots came from too hot RAM, so I lowered Vdimm and instead of reboots I got errors. However, errors never went away. At some point I wanted to be sure my setup, excluding RAM, was still stable... It wasnt while I only tweaked RAM after the last stability check

Spend till 03:00 in the night figuring it all out and finally I just set all ACC to +4 and set Vdimm at 2.11V and the RAM still at 2:1. That was stable again. Now this morning I decided, with very loose timings, the RAM at 2:1 and it already runs for a very long time... Ive no idea what's wrong, but I think Vdimm might be incorrect with this motherboard? I cant give a conclusion why my RAM would run 70 degrees before with like 2.2Vdimm and it's only 30 degrees under full load with 2.1Vdimm.

Also I found some interesting about tRC, namely if it's loser you need a lot less Vdimm while it doesnt really kill the performance.

Anyway, I noted my all my settings this time so if I land in a labyrinth like last night again Ive a solid base to get back to. Now lets hope I wont have to mess a lot with timings anymore and get my 4-4-4 without too much Vdimm

Memory support is not as good as on 790FX on 790GX. The AsRock790GX overvolts the memory ~0.1V maybe asus does the same. Also had alot of trouble with cheap kit's whom required an too high tRC that K10 can offer.
KTE infestigated memory on 790GX abit more here.

[Rammsteiner]

Memory support is not as good as on 790FX on 790GX. The AsRock790GX overvolts the memory ~0.1V maybe asus does the same. Also had alot of trouble with cheap kit's whom required an too high tRC that K10 can offer.
KTE infestigated memory on 790GX abit more here.

But Ive 790FX.

Anyway, although idle freezing is solved, I think there's still something not working correctly. I dont know whether this is a stupid BIOS setting, BIOS issue or hardware issue.

Thx to other people on this forum, but also my own experiments, I noticed the following obvious things;

Trying to OC HTT using a 8x multiplier for at least NB made it impossible to POST, using 9x or 10x did work.
HT multiplier does not work above 10, if you set it higher it will stay at 10.
Vdimm acts weird, for me it effected not only (un)stability for RAM, but also for the CPU.
My RAM got very hot and caused reboots without a clear reason (Vdimm wasnt high). Lower Vdimm caused errors instead.

Now weird thing Im struggling with;

When running RAM 2:1 240Mhz I cant get it to run stable. Vdimm does not matter, Vnb does not matter and timings do not matter. Also often when using normal timings only one thread fails. If I adjust ACC for that specific core and start Prime Blend again, another thread fails for another core. At some point you even have to adjust ACC for cores you already adjusted only 15 minutes ago Either it's not the actual core causing issues or ACC isnt working as it should.

Thus far I sort of managed to get the whole setup stable as long as I run my RAM 5:3. I do have to tweak Vnb a bit more since I messed up HEX values for it. Default was 14, I tried 12 --> unstable, 10 --> unstable, 08 --> stable, but later on I noticed 10 --> 08 is actually a huge jump, since I forgot about 0F-0A, it's now at 1.45V. But well, I dont mind my RAM running at 5:3 240Mhz, but please with some normal timings and not 5-5-5 crap. My RAM might even be able to run that speed at 3-3-3 although I never really tested is for stability, only SuperPI 32M. But somewhere there's a timing and/or Voltage which needs to be adjusted to make it able to run at 4-4-4 even.

Although SB750 solved a lot regarding clocking, 9950BE 125W does not suffer from idle freezing, this platform is for me still not a very easy OC'er like skt 754, skt 939 and Windsor were. There it was quite fast clear what had to be adjusted to get a stable OC, if it didnt work out it was clear you simply hit a wall. But Agena... seriously lol