1090T overclock problems

Hi there I need some help to reach 4.0 GHz (that was my initial goal).

I have this system:
mb aSrock 890fx deluxe 3 (beta bios)
cpu amd 1090t
gpu his HD5830 1GB
psu corsair tx650
ram corsair dominator 4GB 1600MHz cl 9 9 9 24 2t
ssd samsung evo 840 120GB
hdd wd caviar green 1TB
cooler arctic xtreme freezer, fan full-on with one more fan attached on the back

The thing is that I cannot move from 3.8 GHz while still pass 10 ibt runs on 1024:

I've tried only rising the multiplier as high fsb seem to be instable on +3.6GHz..

3.2-all stock (1.325 vcore)-passed
3.3-all stock-passed
3.4-all stock-passed
3.5-all stock-failed

I had to increase vcore by one notch:

3.5-1.3375-1.1500(stock)-passed

from now on I've raised both vcore and cpu/nb voltage getting to these settings:

(clock-vcore-cpu/nb)
3.6-1.3875-1.1875-passed
3.7-1.4250-1.2250-passed
3.8-1.4500-1.3250-passed (+0.1 here!)
3.9-1.4500-1.3250-failed

also
3.9-1.4625-1.3500-failed
and here I stopped returning to the stable 3.8, so I stopped and starting some work on ram and cpu/nb freq.
I got it stable to 2800 cpu/nb and 7 8 7 7 20 1t @1600 on ram..

is it safe to raise over these voltages? I've read dolk guide and a couple others but these seems the max for 24/7..

Why have to overvolt it so high, others did much more with less voltage, do I have such a bad chip?

Is the cooling inadeguate? I have 21°C in idle (room 24°C) and max temp 47°C under ibt..

I noticed, since i moved it on the desktop, that the psu make some high pitch sound when under load (4years old), can this be the sign of degradation but also of OC instability?

The motherboard should allow some more overclocking room imho! But maybe this board isn't the best (they made 3 revision of it)..

Help me please, I'm pretty desperate

A couple things I see here , you're getting your CPU_NB voltage way too high. This will cause extra heat. Also these CPUs will usually do better with an increase in NB speed. So I would suggest dropping the NB to about 2600 and drop the CPU_NB voltage to at least 1.25. Leave the memory at stock timings till you get the cpu sorted out.Then give your 3.9 another go.

Johan45 said:

A couple things I see here , you're getting your CPU_NB voltage way too high. This will cause extra heat. Also these CPUs will usually do better with an increase in NB speed. So I would suggest dropping the NB to about 2600 and drop the CPU_NB voltage to at least 1.25. Leave the memory at stock timings till you get the cpu sorted out.Then give your 3.9 another go.

Click to expand...

Thank you, that was right! I made a test with 3.8-1.450-1.250-2600(cpu-nb) and was ok with 10 ibt runs with 1024MB and got 3°C less on the coremax!

Will report back for 3.9 or 4.0 tests!

3.9 with the 3.8 settings is not working in ibt, what next?

More vcore?

Did you drop the ram timings back to stock??

Cool , just take 1 thing at a time. You can tweak the others later but core speed is always king.

Johan45 said:

Did you drop the ram timings back to stock??

Cool , just take 1 thing at a time. You can tweak the others later but core speed is always king.

Click to expand...

Back to stock before testing, made another ibt run with no joy, I've got an error with the first iterarion...

Not every 1090T will run 4.0Ghz stable. There were a large number that stopped at 3.8/9Ghz-ish. OR they did go ahead and run 4.0Ghz with a BIG increase in voltage to the cpu and then they ran hot. Hotter than was usable day in and day out.
RGone...

Basically you're going to have to add some V_Coere and see where that gets you. Watch your temps 55-60c max for the core temp.

Can you post with pics attached of CPU-z tabs: "CPU", "Memory" and "SPD"? I think we need more info about your RAM frequency and timings and those three tabs will give us a load of info about your overclock settings. Do you know how to attach pics with posts?

Click on the Go Advanced button at the bottom of any new post window and then click on the little paperclip tool at the top of the advanced post window when it appears.

NOTE: this is not the stable clock, the last stable clock is 3.8 with x19 multiplier..
Ram is on stock freq e clocks

Attaching not linking images in the forum...

...allows more people to view the image aNd keeps the image in place beyond the time outside links cancel storing the image.

In order to attach screenshots of INDIVIDUAL images as suggested, first crop and capture the images with Snipping Tool found in Windows Accessories or equivalent. Then click on Go Advanced, a button at the bottom of every new post window. Then click on the little paperclip tool at the top of the Advanced post window when it opens. Clicking on the paperclip tool brings up the file browser/upload tool and the rest is fairly obvious.

Uploaded on site and Edited the message to link the site directly..
I woke up this morning and it was freezing outside, around 10°C, so I opened the windows and tested higher voltage 1.750 vcore..
"I love the overclock smell in the morning", but with no joy ibt still have errors in computation with 3.9GHz
Mind that I had cores on 16°C on air and cputin at 30!
So heat seems not to be the problem for 3.9GHz

You have two pics of the SPD Memory Tab but no picture of the SPD tab as shown earlier. I doubt you have the ram too tight but without glancing at the SPD tab...no way to know for sure.

1. You may have shetty silicon that will not buzz up over 4.0Ghz. Like said there are those out there. For certain.

2. Were it my rig, I would drop the multiplier by 0.5 and raise the FSB/HT Ref Freq to 210 as a start and make sure you DROP that CPU_NB from right at 2600Mhz back to something on the order of 2200Mhz or so and don't let the ram speed or the HT Speed get too far above 2000Mhz. That means you are going to have to look and adjust accordingly and not just depend on the board to do as suggested. It won't.

There is "plenty" of time to adjust ram speed, timings and CPU_NB speed; IF you could touch 4.0Ghz. Plenty of time for dealing with the smaller things. Don't know if moving to lesser cpu multiplier and pushing on the FSB will touch 4.0Ghz but it is all that is left if you tried 1.75Vcore and the cpu would not budge.
RGone...ster.

Canister, I asked for the Memory and the SPD tab. You posted the Memory tab pic twice. The SPD tab will tell us what the manufacturer's recommendations are for voltages and timings at various frequencies. Need that info still.

We find that often we can get a little better overclock on lower voltages if we use a combination of the multiplier and the FSB in overclocking. Right now you are using only the multiplier.

I woke up this morning and it was freezing outside, around 10°C, so I opened the windows and tested higher voltage 1.750 vcore..

Click to expand...

10c may still not be cold enough to run that much voltage. You may need to get it really cold in order to do so like -10c. Additionally when trying to push higher hitting it with a voltage sledgehammer may not be the best method. Sometimes pushing the voltage slowly will help you get to the next X mhz.

Sorry for the wrong screenshot, I've uploaded the right one now.
The vcore in the previous post was a mistype, 1.4750 was the correct one!
Ram are all on stock! @ 1.635V that is the suggested one, In dual channel on the slots suggested on the MB manual..

EDIT:
I've read that maybe the max ram speed supported by my MB is 1333, I've rolled back to it, and lowered the cpu-nb to 2000, still no joy with 3.9GHz but also it seems that I've lost stability @ 3.8..

The memory at 1600 really shouldn't be an issue with the CPU or the board and the cpu will probably behave better at higher clocks with the higher NB speed. Like RGone said earlier not all X6 will run 4.0 with decent voltage and at 1.475v I'm sure your temps are starting to climb. Have you read Dolks guide here for the PhenomII overclocking? It's very informative and will show the relation of CPU and NB speeds. http://www.overclockers.com/forums/showthread.php?t=596023

I've read that maybe the max ram speed supported by my MB is 1333, I've rolled back to it, and lowered the cpu-nb to 2000, still no joy with 3.9GHz but also it seems that I've lost stability @ 3.8..

Click to expand...

A few things just so we can all be on the same page. In post #10 you stated that 3.8 was your last "stable" OC. What stress test are you running and for how long were you running it to consider it stable?

I the above quote you state you seem to have lost stability at 3.8. When overclocking it is best to only make 1 change at a time. This way we know what may have caused the instability. Changing the memory and NB freq in one step may be the issue. As Johan pointed out, having the Nb Freq raised in relation to the Cpu OC may help stability. It's likely the drop in memory spd did not cause instability. Additionally if the OC was stable at 1600 Mhz on the Ram at 3.8 when trying for 3.9 unless you're using the FSB to OC, the Ram spd it is unlikely to be the cause of instability.

Canister said:

Sorry for the wrong screenshot, I've uploaded the right one now.
The vcore in the previous post was a mistype, 1.4750 was the correct one!
Ram are all on stock! @ 1.635V that is the suggested one, In dual channel on the slots suggested on the MB manual..

EDIT:
I've read that maybe the max ram speed supported by my MB is 1333, I've rolled back to it, and lowered the cpu-nb to 2000, still no joy with 3.9GHz but also it seems that I've lost stability @ 3.8..

Click to expand...

Did you roll back the RAM voltage as well, to 1.5 or so I mean?

Did you also lower the HT Link back to 2000? On the Thuban core CPUs you don't want the HT Link frequency to be higher than the CPU/NB frequency. Also, in my experience it's not always possible to have the HT Link frequency run neck and neck with the CPU/NB frequency when the latter gets up into the 2600 range. Sometimes the HT Link will just not go there but it will usually go to 2400 with no problems. Too much separation between CPU/NB and HT Link frequencies can also cause problems.

Canister, have you tried using the FSB to overclock or a combo of multiplier and FSB?

Johan45 said:

The memory at 1600 really shouldn't be an issue with the CPU or the board and the cpu will probably behave better at higher clocks with the higher NB speed. Like RGone said earlier not all X6 will run 4.0 with decent voltage and at 1.475v I'm sure your temps are starting to climb. Have you read Dolks guide here for the PhenomII overclocking? It's very informative and will show the relation of CPU and NB speeds. http://www.overclockers.com/forums/showthread.php?t=596023

Click to expand...

I've read that thing a lot! Temp are still under 50°C @ 1.4750V, but I rolled back to 1.4500 because I was not able to get to 3.9 stable..

Mandrake4565 said:

A few things just so we can all be on the same page. In post #10 you stated that 3.8 was your last "stable" OC. What stress test are you running and for how long were you running it to consider it stable?

I the above quote you state you seem to have lost stability at 3.8. When overclocking it is best to only make 1 change at a time. This way we know what may have caused the instability. Changing the memory and NB freq in one step may be the issue. As Johan pointed out, having the Nb Freq raised in relation to the Cpu OC may help stability. It's likely the drop in memory spd did not cause instability. Additionally if the OC was stable at 1600 Mhz on the Ram at 3.8 when trying for 3.9 unless you're using the FSB to OC, the Ram spd it is unlikely to be the cause of instability.

Click to expand...

I'm using ibt to test stability, 10 run wit 1024 memory, 20 for "rocksolid"..

Yes atm I'm using only the multiplier, but maybe ram speed was a problem as 1600 on the manual of the MB it's listed as "OC" setting and 1333 is the last normal one.

trents said:

Did you roll back the RAM voltage as well, to 1.5 or so I mean?

Did you also lower the HT Link back to 2000? On the Thuban core CPUs you don't want the HT Link frequency to be higher than the CPU/NB frequency. Also, in my experience it's not always possible to have the HT Link frequency run neck and neck with the CPU/NB frequency when the latter gets up into the 2600 range. Sometimes the HT Link will just not go there but it will usually go to 2400 with no problems. Too much separation between CPU/NB and HT Link frequencies can also cause problems.

Canister, have you tried using the FSB to overclock or a combo of multiplier and FSB?

Click to expand...

The voltage was 1.635 in bios that seems the default for them, never touched it, Am I wrong? Should I go for 1.50? The label on the stick says 1.63v!
HT was always on 2000! Never touched it, dolk said that maybe there should be some stability bump with it on thubans with higher settings..
I've tried to work with the fsb or with both multiplier and fsb but lost stability with ibt by reaching 3.6GHz