Fickle setup helprequired.

[jb3nny]

Hi guys i have checked a lot of posts with FX6300 and its certainly reassuring. However my Asrock Extreme3 970 seems to be handcuffing me or my 400 Antec High current PSU (although it seems to be a champ).

So basicly the back story is i need to get FPS in Guild Wars 2 and i needed to do it on a budget. I believe my details are in my sig. I built this myself for under $600 and i did shop around. (Only things i def wish i had was a better mobo now.) But none of that is going to happen until i pay off my credit card... so which leads me too... help me get a decent OC (like 4GHZ?) that is rock solid.

I have tried all sorts of things i have read here and my main problem seems to be voltage drop and CPU Freq. drop in the midst of Prime which of course leads to workers stopping BSOD etc. At first i did just multiplier and Volt increases and i could game at 4.3Ghz but prime would fail. So i backed that off and tried adjusting NB and HTT and Ram etc etc... so now im confused and looking for step by step methodical pointers on what to focus on first, 2nd, 3rd, etc. I dont need super indepth explanations as i can do it all.. provide screens etc. but i need a safe game plan with realistic goals for a fickle MOBO. RIght now im tring a different approach it would seem and focusing on:
cpu 18.5 x Bus @ 215 (started at 230) i get errors all the way back to 215 where it is now on prime with the 5/6 worker.
Do i need more voltage on NB bus? Ram volt increase? Is doing the bus first a stupid idea? All green options are off Amd TCT, APM, cool n quiet, C1E CPU fan control is manual and pretty high.

I have tried CPU NB and Memory/ HTTLink values on at default and also recommended values from what i have read on this board.

Also temps never get above 62 ( even at 1.45v+ )
Should i start from scratch with def speeds and let you guys walk me through it?

Bring it on im game.
and thanks!

 

[caddi daddi]

start back at stock, list details about ram. the guys will be in shortly to get you going.
do a 30 min p95 run and post the 3 cpu-z tabs and hardware monitor.

 

[jb3nny]

rgr in progress will update.
Okay stupid question#1 turn off all the green options? These images are complete stock Bios setup.

Core speed is dropping as the multiplier is dropping to 15 every couple seconds. This must be the auto settings for overdrive and the like?

Prime resulted in a solid round with no errors throughout. Also CPUTIN has maxed at 54Deg Cel.
Let me know my first step/s and i will try tomorrow. Sick tonight so off to bed for me 8(

 

[trents]

"AMD TCT": Is that the turbo core option? Turbo core should be disabled.

Do you have an option called APM (Advanced Power Management)? The terminology might be different in your bios. It's a key player in the green stuff - the throttling of voltage and CPU frequency under load when power draw is exceeding the TDP envelope.

Why are you running your ram at 1333 when the CPU-z SPD tab "XMP-1600" shows that the ram is capable of 1600 mhz? You would likely need to manually set the timings.

You say you are hitting 63c under load. Is that package temp (what we also call "core temp") or is that CPUTIN (= CPU socket) temp? I don't see temps that high in your attached pics from HWMonitor. If that's package temp you have answered your own question in that 63c is probably past the threshold where the processor starts to cut things back to prevent damage. We tells folks running the FX CPUs that as a rule of thumb you want to limit package temps to 60c and socket temps to 65-70c.

Your Corsair H40 is not helping you here. The H80 or H100 would keep the CPU much cooler. Even a good air cooler would probably do better than the H40.

 

[jb3nny]

HI Trent thanks for the feedback, i did before i posted, have the bios settings tricked out but i reset everything to stock as per caddi daddi request. TCT is the turbo core technology i think it throttles the cores up and down. I will turn off the green stuff right now, load my ram's built in profile timings. The 63c was socket temp - package temp is usually so low that i think its wrong?

Okay so going to tweak the green stuff off... and set ram timings. You didn't mention to do anything with the bus or CPU or voltages so i'll leave that all alone post some more screens and we can go from there.

Prime happily is working well with cpuz and HWmon reporting rock solid numbers with nearly zero fluctuation. Package temp shows a max of 38c and cputin shows 57c
AMD TCT = disabled
APM = disabled
Ram @ DDR3-1600
Cool n Quiet = Disabled
C1E= Disabled
CPU Thermal throttle Disabled (is that good to do?)

Going to bump multiplier up 1 and run prime.

 

[Johan45]

Hi jb3nny, and
Not that I want to interfere with your OC because this info will be relevant to my suggestion.

So basicly the back story is i need to get FPS in Guild Wars 2 and i needed to do it on a budget.

At this point trents hit the nail on the head about the H40 and you'll probably get a slight OC with it. BUT Guild Wars will only utilize one core of that CPU.
I assume your BIOS will allow you to save multiple OC profiles.
So my suggestion would be to disable 1 or 2 modules and OC it as high as possible. This will give you some relief from the heat and probably give better results for your game. You'll have to experiment with the number of modules to disable but anything I have read about that game is it'll only use 1.
I have one game that gets a little squirrely with the eight cores running so I disable half and crank it up to 5GHZ and it just screams
Johan

 

[jb3nny]

Hi there Johan, valid points and i have heard of people disabling cores to increase overclock as the games don't utilize all the cores. I may have to do that. But for now I was rock solid at 18X and i think i'll try to up it to 18.5 unless someone has any other ideas.
Thanks!

 

[jb3nny]

Mult @ 18.5 seems good temps are up a bit but voltages are still stock. Temps hit CPUTIN 60c and package temp 40c
That passed a round of Prime w no issues. Going to up the mult. to 19.

 

[trents]

So far so good! Yes, CPU thermal throttle would need to be disabled in this case but keep a close eye on both package temp and CPUTIN, especially CPUTIN since it is quite a bit higher than package. I think disabling cpu thermal throttle would be safe as long as your CPUTIN doesn't exceed 70c. Be aware, however, that max "stable" temps are usually somewhat lower than max "safe" temps so you may encounter instability before you reach the unsafe threshold temp.

Most overclockers find that using a combination of FSB (aka, HT Reference) and multiplier allows for using less core voltage to get a stable overclock at a given CPU speed. I would suggest lowering your ram back down to 1333 level while putting your CPU multiplier to stock. Also, lower both the HT Link and the CPUNB frequencies (may be expressed as a multiplier) below stock to make some headroom since those two frequencies, like ram frequency, are "tuned" to the FSB. Then increase the FSB to a level that takes the ram back up to around 1600 mhz. I think that will be a FSB of about 242 mhz but you will have to tweak with that a little. Then go back and start increasing the CPU multiplier to get your overclock back to max, adding vcore as needed and as temps allow. After each .5x increase in the mulitplier, run a 20 minute Prime95 blend test with HWMonitor open to check for stability and temps. When you think you have about reached your max overclock, run a longer Prime95 blend test to confirm stability, longer meaning at least two hours.

 

[Johan45]

You do have more headroom but your socket temps are getting up there. Do you have any fans that you could mount to the side door and blow some air onto that area. When using the closed water cooling you're taking airflow away from the socket area(cpu,vrms and msofets) this could explain the high temps. Usually that reading is only~10deg over the package but you're approaching 20.

 

[jb3nny]

Okay went into the bios
set cpu to 17.5x 240 4.2Ghz
ddr-ram 1599
nb 1920
htt 1920 (rolled them both back below stock then the FSB brought them back to just under stock speeds)
all stock voltages.

Prime crashed 2 times then i promply BSOD.
I am guessing that FSB is just too agressive? or should i increase nb voltage or something like that?

 

[jb3nny]

Just increased the NB/CPU voltages from 1.175 to 1.1875
Prime BSOD right away

About to try 1.200v on the NB if i BSOD then i can only surmise that i cannot run this kind of FSB. *update BSOD

 

[Johan45]

If you're going to set the HT at 240 you'll need to back off your multi you can't expect to jump from 3.7 to 4.2 like that. Start at maybe 220 and 16.5 it'll bring you to just over 3.6 then bump up the HT a bit at a time.

 

[jb3nny]

Well i did think jumping right into 4.2Ghz was a little crazy but i wanted to follow trents instructions.
Okay so before i go ahead and perhaps try what Johan says i did go into the bios after that last BSOD
FSB=230 x17.5
NB freq x9.0 2070 mhz
1.175v (back to stock)
HT bus speed 2070 mhz
DDr3- 1532

*update got a worker stopped fatal error less than 4 rounding error so thats not stable. going to reboot and edit settings as per Johan suggestion.

 

[jb3nny]

Okay.. edit bios settings
16.5x 220 = 3630mhz
HT FSB 2200
NB Freq= 1980
ddr3 -1466 mhz

That seemed stable going to up the FSB to 225

 

[Johan45]

You know you were stable at 3.7 so you have a voltage setting. When you change the HT you will have to find a setting for the other frequencies. For now maybe you should re read trents last post. I wasn't being specific just telling you that you can't make a jump like that. Trents did say to back off the multi to stock settings.

 

[jb3nny]

Hi Johan at 16.5 x225=3712 mhz
HTHT Bus 2250 mhz
NB 2025mhz
ddr3 1500 mhz
i am running prime and it seems stable.

Just to clarify about trents post you mention he did say to run cpu at stock speeds which is what i did (17.5x <-stock 240FSB <-- not stock but what he recomended =4.2ghz)
So unless i am missing something. But its no big deal, i just rolled it back to 16.5 like you mentioned. I know you said you were not being specific but you said to try 16.5x 220 which seems stable and now 225 seems stable too so its all good.

 

[jb3nny]

Okay prime seemed good with the above config. I rebooted and increased the FSB to 227
16.5x 227=3745.55 mhz

Going to run prime. I sense i am nearing my FSB ceiling.

 

[trents]

Well i did think jumping right into 4.2Ghz was a little crazy but i wanted to follow trents instructions.
Okay so before i go ahead and perhaps try what Johan says i did go into the bios after that last BSOD
FSB=230 x17.5
NB freq x9.0 2070 mhz
1.175v (back to stock)
HT bus speed 2070 mhz
DDr3- 1532

*update got a worker stopped fatal error less than 4 rounding error so thats not stable. going to reboot and edit settings as per Johan suggestion.

I gave you a bumb steer. When I said to put the CPU multiplier back to stock I should have said lower it to bring the CPU frequency back to stock levels to begin with. Then start increasing it in .5x increments. My bad.

 

[Mandrake4565]

jb you need to have a little patience, you are going to have a difficult time getting stable if you just keep punching numbers in and hoping for the best. If you are going to use the multi to Oc then pust it the way you were .5 at a time a V Core when necessary and test. Then do the same with the FSB, set multi to stock bump fsb to 205, keep RAM, HT Link and NB Freq in line and test. Some boards just can't do big FSB numbers, you need to figure out what the limit is on yours. Using the process I said above I found that on my ASUS M5A99X EVO board I cannot set the FSB above 230, no matter what. On my Chrosshair V Formula I haven't found a limit yet.