Printable View
UPDATE: I reinstalled Windows x32 with a raid -0- no problems.. I guess I lost my raid -0- or it got hung up some how? But all is well now no need to RMA the board. PM sent you saaya with a quick question..Quote:
Originally Posted by Lu(ky
Thanks SAAYA ;)
your welcome :toast:
glad we could solve that problem quickly :)
i dont know what exactly happened, but this kind of coruption usually happens when:
memory is pushed way too high
chipset is pushed way too high
chipset gets too hot
a good friend (mikeguava) recommended me a tool called xcopy.
he uses that to restore borked and corupted windows installs, it restores and fixes the install within seconds or minutes.
go and give it a try :)
im sure you will corrupt a windows install or raid image again sooner or later :D
Yeah I may have to look into that. I think I have a idea what happen. After I tried for a OC not LARGE small vs the one I was at with my new memory. The system rebooted and the alarm was going off and usually it resets, but for some reason this one did not. So I reach across to reset the board. (Using a TECH STATION) And maybe I hit while the computer was running the yellow RESET CMOS button by accident. And then I hit the reset to restart and then it was hung up? Anyways very happy I did not have to send to RMA. Now more time having fun playing the my OC stuff..Quote:
Originally Posted by saaya
\Quote:
Originally Posted by Telo
thanks
Question on the Blackops board itself do you guys set the FSB jumper PIN to 333/400/450 etc? Or do you leave it on AUTO when you OC the board? For some reason I am having problems with HIGH FSB settings. I am now on a FRESH install with new G28 bios installed.
careful when using the FSB jumper to set the FSB strap! :D
it locks all other OC options in BIOS... you can oc using setfsb tho
G28 is a great improvement. I can use my OCZ 1800MHZ at it's rated settings now.
Whoever is responsible for this BIOS deserves a medal as it has saved me from replacing my RAM.
Ironically it is now my CPU that is holding me back. It's an early Wolfdale so it requires a bit more voltage than some of the newer ones. Well, I say a bit, but it needs 1.41 Volts just to run stable at 3.9GHZ :-(
hehe, yes, G27 was already good but G28 is really nice :)
1.4v for sub 4ghz, yeah that sounds like an early 45nm chip... but its still a good oc :)
i wouldnt really say you are cpu limited, i dont think theres a game or app where you need more than 4ghz c2d power :D
It is true, sayaa - may be tests only, for example Crysis gives me no difference in FPS either with 4.3 GHz and 4.0 GHz settings, it is dependable on GPU overclocking.Quote:
Originally Posted by saaya
Especially not as all I have been playing for the past week solid was Starcraft :DQuote:
Originally Posted by saaya
It's just the boasting factor I'm missing out on :rolleyes:
hehehe, i wish there was a hack to make starcraft playable at higher res and on widescreen... :D
Hi guys. Please forgive the length.
I was posting in the shared bios thread but thought i'd try here too, i'll copy paste a bit of what i already said there.
I'm watercooling a QX9650 and NB, using a 2x2gb Corsair XMS3 DHX 1600 kit (TW3X4G1600C9DHXNV) rated at 9-9-9-24 2T 1.8v stock. My goal is to get as high an overclock as possible while remaining 20hr+ prime stable. I'll try to put in bold any changes i recorded. I am on G28 bios.
Here are my settings for 3.8ghz, which is 20hr+ prime stable.
_________________________________________
CPU multi 9.5x
FSB 400
Ratio 1:4 cpu333
Memory clock 1600mhz
CPU Voltage multi 1x
CPU target voltage 1.41254v
CPU actual voltage 1.3994v
Vdroop compensation on
CPU PLL 1.508861v
VTT 1.35v
GTLref all auto
NB voltage 1.403v
NB GTL +2 step
DRAM voltage 2.15v
Dram pull up auto
DRAM pull down auto
DRAM Ref voltage +2
Timings 7-7-7-18 2T, refresh period ratio x1, all others auto.
_________________________________________
Now i'm trying to go up to 10x multi for that magic 4ghz. I can boot into windows etc no problem, but prime stability eludes me and cores started dropping after a minute or two. After increasing the cpu voltage to 1.46 (1.44 actual), cores are still dropping out on prime. I've had the NB voltage up to 1.7v, and the temp peaked at 47c, however it made no difference in prime so i returned to 1.5v. I could probably be lower than this but i wanted to remove any doubt - once i get a stable 4ghz i plan on going through each voltage and seeing which i can lower while still maintaining stability. I also returned the memory to its stock settings of 9-9-9-24 here.
So i turned my attention to the GTLref values, and decided to record each change i made and what difference it made in prime. I've been at this since yesterday with very little sleep :(
I started off with the same settings as Jonbass who has been helping out in the other thread.
+10, +5, +10, +10 - core 4 failed in 19 mins
+10, +5, +10, +5 - core 4 failed in 32 mins
+10, +5, +10, +4 - core 4 failed in 5 mins
+10, +5, +10, +6 - core 4 failed in 18 mins, core 1 failed in 35 mins
+9, +5, +10, +7 - core 4 failed in 30 mins, core 1 45 min+ stable
+9, +5, +10, +3 - core 4 failed in 17 mins
Up til this point i had been just randomly tweaking up or down the GTLref on the core that had been failing, mostly core 4, but eventually i found the post from LardArse about what the steps actually do, being each step adds or subtracts 0.0005x from the default value of 0.66x. So for example at VTT 1.35v and default GTLref my gtl value is [1.35 x 0.66] 0.891v, and +1 step would be [1.35 x 0.6605] = 0.891675v, +2 step would be [1.35 x 0.661] = 0.89235v etc. Please someone correct me if this is wrong.
With this knowledge i remembered alot of GTL discussion in the old 790i data corruption thread, so i dug through that and found this post from 5010:
".67x is 00mv(standard) in the 790iU Bios
65nm Quads like it abit higher; .68x - .70x
Duals + Yorkies like it lower; .66x - .63x or even less (as low as .58x according to LardArse)"
So if 0.66x is standard in the blackops bios, i thought i could start using my calculator to work out where i should be. Assuming the above is correct, the GTL settings for 45nm cpu's on the blackops should actually all be negative (if not default) - so far i had been using positive values so i planned on switching to negative, though i think all of the other people in this thread are using positive GTL values - maybe someone can clarify?
So i commenced testing, and is my log with GTL changes and prime result. I'll put the GTL values i changed each time in bold.
-20 -20 -20 -20 - C1 error, had to clear cmos.
-10 -10 -10 -10 - core 4 failed in 5 mins
-10 -15 -10 -15 - core 1 failed in 2 mins
-11 -15 -10 -15 - core 4 failed in 33 mins
-11 -15 -10 -20 - core 4 failed 17 min
-11 -15 -10 -25 - Windows crashes on load
-11 -15 -10 -24 - whole system freezes in under 1 min of prime
-11 -15 -10 -23 - Windows takes a long time to load, whole system freezes within 5 min of prime
-11 -15 -10 -22 - Boot ok, Prime runs, core 4 fails in 23 min
-11 -15 -10 -21 - core 3 fails in 20 min
At this point i put the NB from 1.5 back up to 1.67v, and further relaxed memory timings to 10-10-10-25, just to remove all doubt that either of these could be holding me back.
-11 -15 -11 -21 - core 4 fails in 32 min
-11 -15 -11 -19 - core 4 fails in 30 min
-11 -15 -11 -18 - core 4 fails in 22 min
-11 -15 -11 -17 - core 4 fails in 20 min
At this point i found some settings of other people who were using a VTT of 1.5v, so i increased VTT to 1.5v to see if it made a difference.
-11 -16 -11 -16 - core 4 fails 19 min
It didn't, so i lowered VTT to 1.45v
-11 -16 -11 -14 - core 4 fails 30 min
-11 -16 -11 -13 - core 4 fails 32 min
-11 -16 -11 -12 - core 4 fails 30 min
At this point i came across GMCG's settings, he also has a QX9650, but with a different set of corsair ram, though this isn't a ram issue so that shouldn't matter i thought.
His settings were:
Voltages: Vcore (BIOS) 1.4625V (shows 1.4464 current Vcore in BIOS), Vdroop=enable
V PLL 1.546582V
VTT 1.35V
GTL0 +2
GTL1 +2
GTL2 +2
GTL3 +3 (that is my CPU I guess, I added +1 to compensate core 4 failure in Prime95 and it worked)
V NB 1.438V
GTL NB +3
V SB 1.5V
V DRAM 2.099V, DRAM compensations:
+3 step
+3 step
+3 step
FSB 430 MHz
0 ps
800mV
100 MHz PCi
Spread = disable
So i copied his voltage settings which meant making these changes - i already had the same vcore, i raised the PLL 1.508861v to 1.546582V, lowered VTT from 1.45 to 1.35, i put my NB to 1.5v as i figured it's ok (temps are fine, not high enough to cause instability), changed NB GTL from +2 to +3. I kept my multi at 10x and fsb at 400 as i will be quite happy with 4ghz right now. His core 4 was apparently also dropping and was fixed with those GTL settings, so i copied them and continued testing:
+2 +2 +2 +3 - core 4 fails in 23 min
+2 +2 +2 +4 - core 4 fails in 20 min
+2 +2 +2 +1 - core 4 fails in 4 min
+2 +2 +2 default - core 4 fails in 2 min
+2 +2 +2 +5 - core 4 fails in 20 min
+2 +2 +2 +6 - core 4 fails in 33 min
+2 +2 +2 +7 - core 4 fails in 6 min
+2 +2 +2 +8 - core 4 fails 9 mins
+2 +2 +2 +9 - cores 3 and 4 both fail in 34 mins
So as you can see i'm not having much luck with these GTL's, though certainly changing them does seem to have an effect.
Is the fact that he's on bios G26 important at all? Would it be worth my while flashing that from current G28?
Otherwise my next course of action would be raising the core voltage up 1 or 2 steps, however i've been told that going over 1.45v with these cpu's isn't a good idea. My temps however would actually be ok there, and I see that GFTout and Sash are both using more than this, i think GFT is on 1.5v. If temperatures are fine with those voltages, is it safe to be there? Or is it possible for your temperatures to be great but for that voltage to kill your cpu alot quicker anyway.
If there's a way to keep my current 1.46v (1.44v actual) and get stability from these GTL's, i'm willing to spend weeks if not months to get it. However if someone with a little more knowledge on them can tell me if there's no chance then i guess i'll have to give up.
The last setting i just tried was one that LardArse suggested to Sash, which was GTLs of +12, +7, +12, +12, and core 4 (you guessed it) failed in 32 mins.
If you got this far then thankyou for reading, any help is appreciated.
Hi Xello. It looks for me that you have a temp problem - the limit of 32-35 min in all "successful" tests is pretty stable, also your 3.8 MHz stable result confirms it. How do you control your temps during a test? Please describe your WC system as well.Quote:
Originally Posted by Xello
Hi gmcg.
I actually got this system from a gaming pc company here in the uk, i don't have the WC know how or time right now to do an elaborate build myself, but this WC system is i guess you could say 'professionally' done.
I can tell you the following specs: D-Tek Fuzion CPU Block, Thermochill Triple Radiator (i thought it was a 320mm rad but i take it that is the same as a 120mm tripple rad), Laing DDC 12v with Modded Plexitop, 6 x 120mm Xilence Fans, 2.5 Metres of Tygon 1/2" Tubing. They also machined a new top for the NB block, it's delrin. Sorry if this isn't enough info.
What would make you think i have a temp problem though? My temps don't actually change (noticably) between 1.39v 3.8ghz and when i'm testing 1.44v 4ghz - i idle at 32c and load generally at 55-58c, it peaked at 61c during prime yesterday and it was a very hot day! I've read of people running this chip safely in the 60s and even 70s so i was thinking these temps were not bad.
Also not sure i completely understand 'controlling' my temps during the test, if you just mean what do i use to monitor, i use Realtemp for the cores and Everest for the NB, which peaks at 44/45, 46/47 on a hot day (on 1.5-1.6v)
I've been continuing GTL tests, here are my results from yesterday, continuing from the last one i done which was +2 +2 +2 +9 - core 3 and 4 fail both in 34 mins:
+2 +2 +3 +10 - core 3 fails in 10 mins
+2 +2 +1 +10 - core 4 fails in 31 mins
At this point i wondered if tweaking core 2 might affect core 4 because of the way they are on the die, so i tried changing that
+2 -3 +1 +10 - core 4 fails in 18 mins
+2 +1 +1 +10 - core 4 fails in 31 mins
It did change the time to fail, didn't fix the problem though :down:
+2 +1 +1 +11 - core 4 fails in 20 mins
+2 +1 +1 +13 - core 4 fails in 32 mins, core 1 fails in 34 mins
+2 +2 +1 +14 - core 3 fails in 32 mins, core 4 fail 34 mins
+2 +2 default +15 - core 4 fails in 17 mins, others stable for 1 hour+
+2 +2 default +16 - core 4 fails in 20 mins, core 1 fails in 34 mins
default +2 default +17 - core 4 fails in 32 mins
default +2 default +18 - core 4 fails in 9 minutes
default +2 default default - core 4 fails in 32 mins, core 1 fails in 2 hours
+3 +2 default -1 - core 4 fails in 17 mins, core 1 fails in 34 mins
+1 +2 default -2 - core 4 fails in 18 mins
And the run i started before i went to work today:
+1 +2 default -3 - core 4 fail 22 mins, other cores stable 8 hours+
So as you have noticed the 'successful' runs have all had steady fail rates at 32-35mins, i don't know what to make of that though, i'm out of my depth here i just hope the recording can help someone come up with a theory!
Thanks again.
:update:
My runs this evening:
+1 +2 default -4 - core 4 fails in 1 min
+1 +2 default -5 - core 4 fails in 5 mins
+1 +2 default -6 - core 4 fails in 33 mins
+1 +2 default -7 - BSOD after 20-25 mins, INTERUPT_EXCEPTION_NOT_HANDLED
+1 +2 default -8 - core 4 fails in 18 mins, core 1 fails in 37 mins
As you can see, core 1 was stable for over 8 hours on +1 +2 default -3, yet it failed in 37 mins on +1 +2 default -8, and it had the same GTL value both times. I guess this means that adjusting the GTL value for one core can affect other cores, and if that's true this could take years :mad:
:further update:
Well i caved in and started upping my core voltage.
Changes:
vcore from 1.456(1.44 actual) to 1.467(1.458 actual)
default default default default - core 4 fails after 32 mins
+10 +5 +10 +10 - core 4 fails after 33 mins
vcore from 1.467(1.458 actual) to 1.487(1.47 actual)
default default default default - core 4 fails after 2hr 42 min, others stable 8hr+
Finally some progress i thought, but isn't it rather insane that i can get a completely stable 3.8ghz at a mere 1.39v, and for simply 200mhz more i'm having to go above 1.47v?
I upped it once again from 1.487(1.47 actual) to 1.5 (1.48 actual). I hope someone will answer my question from before:
At 1.48v My idle temperatures are still 32-35c, and load is still between 50 and 55, peaking at 60 today.Quote:
Originally Posted by Xello
Basically, what do you go by for judging the safety - the temps or purely the voltage?
Take at look at this too:
http://img.photobucket.com/albums/v4...estar/2334.jpg
I noticed that cpu-z is reporting my core voltage approx nearly 0.4v lower than it actually is. It's version 1.46, is it just an error in the program or do i have a more serious problem? HWmonitor also reports 1v~. Also, might be worth mentioning that core 4 is always 5-10 degrees lower than the other cores. I'm sure it has always been like that, so i didn't see it as a problem - seen as core 4 is the one dropping out all the time though is this maybe something to do with that?
Thanks.
@ Saaya
Hey buddy, were you able to test the 2 sets of CellShock blue kits on the BlackOps yet? Don't mean to bother you, just want to know if it works before I buy the set.
Failing at 10x multiplier when 9.5x works fine seems to be CPU temp/voltage, and especially when failing occurs over a period of stress, it may hint at temperature.Quote:
Originally Posted by Xello
It looks fine, should be good enough for 4.2 GHz, not only for 4.0 GHz.Quote:
Originally Posted by Xello
Your temps are fine as well.Quote:
What would make you think i have a temp problem though? My temps don't actually change (noticably) between 1.39v 3.8ghz and when i'm testing 1.44v 4ghz - i idle at 32c and load generally at 55-58c, it peaked at 61c during prime yesterday and it was a very hot day! I've read of people running this chip safely in the 60s and even 70s so i was thinking these temps were not bad.
It looks fine either.Quote:
Also not sure i completely understand 'controlling' my temps during the test, if you just mean what do i use to monitor, i use Realtemp for the cores and Everest for the NB, which peaks at 44/45, 46/47 on a hot day (on 1.5-1.6v)
This is pretty interesting, I can assume that you have some kind of algae/dirt on your D-Tek fuzion in the region of 4th core? The cooling is good enough for 3.8 GHz but fails for 4 GHz locally?Quote:
+1 +2 default -3 - core 4 fail 22 mins, other cores stable 8 hours+
That is my only guess for now (or it can be a TIM application problem in that region).
Last time when I got quite similar problems I had to clean my D-Tek Fuzion and only by installing of quad nozzle kit I could get to 4.3 Ghz, without it I was able to get to 4.2 GHz max.
BTW don't worry about Vcore on CPU-Z screen, it is a known BlackOps bug/ incompatibility with CPU-Z method of measurement.
Thanks for the replies.
I finally got stable at 1.48v, all GTLs at default, and temps were actually not bad - again mostly between 50 and 60c, peaking at 63/64, however i decided that it was unacceptable to pump such a high voltage just for an extra 200mhz, even if it does feel good to be at 4ghz :]
I've returned to 1.38v @ 3.8ghz for now, thanks for all the help.
May be it is your particular QX9650 ... it looks really strange to trade only +200 MHz for +0.1V. Mine is L739A555Quote:
Originally Posted by Xello
Is there a way to find out the batch number with a program or something?Quote:
Originally Posted by gmcg
All i can see in cpu z is stepping 6, revision C0.
I returned to 9.5 multi, set the vcore at 1.44v and am going to up the fsb and see how high i can get from that, which is what i eventually intended to do anyway to pump my ram up a bit. Today's run 8hr+ stable at 8-8-8-20 with 410fsb (3.9ghz) so i should end up with a decent setup. What i was really concerned with was removing as much bottleneck as possible from a pair of 4870X2's when they come out in a few weeks, as this is primarily a gaming rig. I think this should do a good job of it :D
I took pictures of CPU before mounting it, I have no idea if any program can read the batch number anyhow.Quote:
Originally Posted by Xello
Yeah, it looks like a good idea.Quote:
What i was really concerned with was removing as much bottleneck as possible from a pair of 4870X2's when they come out in a few weeks, as this is primarily a gaming rig. I think this should do a good job of it :D
I would like to mention that BlackOps is a great board IMO for any using (I am quite satisfied with my water cooled 9800GX2, for example), and especially for 4870x2 or 2x4870x2 due to its high transfer rates. Look on those results for wPrime 32M:
http://forums.vr-zone.com/showpost.p...3&postcount=19
eva2000 has his QX9650 (also L739XXXX) running at 4.4 GHz using Asus Blitz Extreme with DDR3 memory @880 MHz CL7-6-5-18, his FSB is 1760 MHz, his wPrime 32M result is 9.735 s. My result is 9.015s with QX9650 at 4.3 MHz with DDR3 memory @860 MHz CL7-6-5-18 ... doesn't BlackOps completely rule here?
http://www.hwbot.org/result.do?resultId=745773
Agreed, i put my limitation down to the cpu and not the board. You win some you lose some with batches. For example my memory has suddenly decided it can be (410fsb) 1640mhz 7-7-7-16 1T tRD6 10hr+ prime stable :DQuote:
Originally Posted by gmcg
*keeps pushing*
Good job! BlackOps requires patience, but then you'll fall in love with it, as I did ;)Quote:
Originally Posted by Xello
Another small quirk i forgot to mention, i noticed it was happening to someone else early on in this thread too, returning to activity from Vista's Sleep Mode will 100% of the time cause the comp to for some reason reboot.
Blackops issue or something else maybe?
Cheers.
:Edit:
Well well, something wonderful has happened!
After all that time trying to get 4ghz stable from a 400fsb with 10x multi, and eventually needing 1.48v to do it, i'm currently several hours stable into a prime test on 420fsb with the 9.5x multi for what is reading in cpu-z as 3995mhz, and only at 1.44v. Not only that, but whereas before i had my memory at 1600mhz and timings loosened to 10-10-10-25 2T, right now i have the memory at 1680mhz with 7-7-7-16 1T tRD 6!
I can't think of any explanations, but i'm happy and will next up the fsb to 425 just to see if it will play ball :D
Hi! All
For Anyone Thats Interested
http://www.anandtech.com/mb/showdoc.aspx?i=3366
Best Regards
flanr