Apparently I need help with AMD 1090t overclock

[Mothballs]

I scoured countless guides. Searched countless forums. Double checked all my figures. Had six hours of stability running small FFTs in Prime95.

And yet, stability is still an issue. Crashed after about 40 minutes of Blend testing in Prime. Occasional BSOD at random. Numerous codes.

AMD 1090t CPU
Hyper 212 cooler
Asus Crosshair IV Formula Mobo
Gigabyte R9 270 Graphics card
Samsung 850 EVO SSD (2x 120gb)
WD Black 1tb HDD
Corsair HX620 PSU
Antec 900 case with Noctua fans

Bios settings are Auto excepting:

Turbo Core Disabled
CPU ratio 20
CPUNB 2600
HT link 2600
PWM full phase
CPU Voltage OCP Enabled
CPUNB Voltage OCP Enabled
CPU Offest + 0.090625
CPUNB offset +0.128125
CPU spread Disabled
Cool 'n' Quiet : ON

I'll give you a HW monitor Readout after posting this and running prime so if the system crashes again I don't lose this post. Will also grab pictures of the BIOS screens if you need them.


EDIT: HW monitor data attached

Two things I have noticed: In the Bios, if RAM is set to auto, it runs 1.7 volts into Kingston HyperX 1600 DDR3, which is supposed to be a PnP memory kit that runs @1.5 volts. I put it down to 1.5 Volts, and all the timings seem to be where they are supposed to be. Is this a possible issue here?

Secondly, now that I have an overclock on the GPU, I'm getting massive artifacting and other issues with the GPU's overclock, which was stable before now. I do wonder if I'm having issues with power delivery...


Anything that could help would be great right now.

HWMonitor.txt

[Mothballs]

I am thinking Voltage 5  0.12 Volts [0x8] (+3.3V) Might be the issue?

[Terryv]

What does HW monitor say about the power supply voltages?

[Mothballs]

Just now, Terryv said:

What does HW monitor say about the power supply voltages?

Voltage 0        12.00 Volts [0xE9] (+12V)
Voltage 1        4.97 Volts [0xE8] (+5V)
Voltage 2        1.27 Volts [0x6A] (CPU VCORE)
Voltage 3        1.68 Volts [0x8C] (VIN3)
Voltage 4        1.63 Volts [0x88] (VIN4)
Voltage 5        0.12 Volts [0x8] (+3.3V)

Voltage 6        1.68 Volts [0x8C] (VIN6)

Voltage 7        2.28 Volts [0xBE] (VIN7)
Voltage 8        1.64 Volts [0x89] (VIN8)

BIOS says the 3.3V is running at 3.4, confirms the 12v and 5v readings in HW monitor.

[Terryv]

18 minutes ago, Mothballs said:

I am thinking Voltage 5  0.12 Volts [0x8] (+3.3V) Might be the issue?

Based on what you posted, it could be the problem or HW monitor is misreading.

 

Does the +5V dip lower than 4.97 during stress testing?

[Mothballs]

4 minutes ago, Terryv said:

Based on what you posted, it could be the problem or HW monitor is misreading.

 

Does the +5V dip lower than 4.97 during stress testing?

Not that I have seen, no.

HWMonitor2.txt

Edited March 28, 2017 by Mothballs

[Terryv]

Do you have another known good PSU to compare?

[Mothballs]

Just now, Terryv said:

Do you have another known good PSU to compare?

I will have to dig the other one out of the other computer. It was slated to go to a friend, and the non modular cables didn't fit in my case well.

Stress Voltages:
Voltage 0        11.95 Volts [0xE8] (+12V)
    Voltage 1        4.97 Volts [0xE8] (+5V)
    Voltage 2        1.40 Volts [0x75] (CPU VCORE)
    Voltage 3        1.68 Volts [0x8C] (VIN3)
    Voltage 4        1.54 Volts [0x80] (VIN4)
    Voltage 5        0.12 Volts [0x8] (+3.3V)
    Voltage 6        1.87 Volts [0x9C] (VIN6)
    Voltage 7        2.28 Volts [0xBE] (VIN7)
    Voltage 8        1.64 Volts [0x89] (VIN8)

[Terryv]

Nothing jumps out besides +3.3. and it could be misreporting.

 

I've got nothing besides trying another PSU

 

Hope it goes well when you try it

[Mothballs]

9 minutes ago, Terryv said:

Nothing jumps out besides +3.3. and it could be misreporting.

 

I've got nothing besides trying another PSU

 

Hope it goes well when you try it

Yeah, new PSU still reads .12v on the 3.3 rail.

Now reads 12.158V on the 12v
5.038V on 5 volt
CPU Vcore 1.404-1.416 Under stress

[Terryv]

Just now, Mothballs said:

Yeah, new PSU still reads .12v on the 3.3 rail.

Now reads 12.158V on the 12v
5.038V on 5 volt
CPU Vcore 1.404-1.416 Under stress

Has the system stabilized?

[Mothballs]

Just now, Terryv said:

Has the system stabilized?

I'll let you know in a few hours after running stress. MSI Afterburner still seems to run awful, so not looking great. /:

[Terryv]

Make sure to stress with the GPU at stock. That way we eliminate the variable of an unstable GPU overclock.

[Mothballs]

1 hour ago, Terryv said:

Make sure to stress with the GPU at stock. That way we eliminate the variable of an unstable GPU overclock.

Intel burntest spits out an error on second iteration on a High test. Completes 25 iterations Normal with no problems.

The error is not small. First two iterations are Result 3.816942e-002 failure was Result 1.#QNAN0E+000.

[Terryv]

at this point you'll have to swap parts one by one to find the problem

 

- memory (try using 1 stick first, then use different memory altogether if possible)

- CPU

- Motherboard

[Mothballs]

17 minutes ago, Terryv said:

at this point you'll have to swap parts one by one to find the problem

 

- memory (try using 1 stick first, then use different memory altogether if possible)

- CPU

- Motherboard

Another set of Kingston Fury X DDR3 Spat the same error with the same settings.

Letting the system auto-detect the RAM Voltage shoots it up to 1.812V under load but It passes the Burn test?

[SpaceGhostC2C]

3 hours ago, Mothballs said:

Two things I have noticed: In the Bios, if RAM is set to auto, it runs 1.7 volts into Kingston HyperX 1600 DDR3, which is supposed to be a PnP memory kit that runs @1.5 volts. I put it down to 1.5 Volts, and all the timings seem to be where they are supposed to be. Is this a possible issue here?

It is possible for both RAM voltage and CPUNB voltage to cause instability (if not high enough) when using OC profiles for RAM. Some motherboards will increase them on their own when they fail to boot. Now, 1600 doesn't sound too high, but the Phenom is an older DDR3 chip, so maybe it was, and fast RAM become more of an issue as you OC the chip itself.

So, what can you try: you'll probably have a 1333 or even 1066 profile for your RAM as well. Try use it that profile with 1.5v and see if this time is stable. If it is, you can then go back to 1600 and play a little with voltages to get it stable.

[Mothballs]

Dropping to stock shows that the GPU is still glitching. I think she's dead.

 

Passes ibt high 20 passes with 20x multi and nb @ 2600, HT stock. crashes in five minutes in primes blend.

1 minute ago, SpaceGhostC2C said:

It is possible for both RAM voltage and CPUNB voltage to cause instability (if not high enough) when using OC profiles for RAM. Some motherboards will increase them on their own when they fail to boot. Now, 1600 doesn't sound too high, but the Phenom is an older DDR3 chip, so maybe it was, and fast RAM become more of an issue as you OC the chip itself.

So, what can you try: you'll probably have a 1333 or even 1066 profile for your RAM as well. Try use it that profile with 1.5v and see if this time is stable. If it is, you can then go back to 1600 and play a little with voltages to get it stable.

i'll try downclocking the ddr3, but it's rated and known to be okay with ddr3 @ 2000. The NB is running at its rated max more or less.

 

I'm hesitant to up the cpu voltage more at this point, though there is more room...

 

 

[SpaceGhostC2C]

1 minute ago, Mothballs said:

 

Passes ibt high 20 passes with 20x multi and nb @ 2600, HT stock. crashes in five minutes in primes blend.

i'll try downclocking the ddr3, but it's rated and known to be okay with ddr3 @ 2000. The NB is running at its rated max more or less.

 

 

But it can happen. When OCing my FX cpus (not the same, but still AM3+), I couldn't initially run my memory at its rated speed (it did run fine at stock clocks). Using a lower JEDEC standard it was stable. In my case, a small bump to the CPUNB voltage allowed me to get it stable at the rated speed - even overclock it a bit past their rating (the CPU NB is essentially the integrated memory controller plus stuff).

[Phate.exe]

How high do you actually have it clocked?  I might have missed it but I didn't see your bus speed.

 

If it's at the stock 200, you're "only" at 4GHz, so I'd just drop the multiplier a bit and bring the bus speed up some.  If I'm wrong than just ignore this part of my post.  Mine really didn't like having the multiplier much over 19-19.5.

 

Mine's very happy at somewhere around 217 bus speed with an 18.5 multiplier.  Under load it only seemed to want 1.392-1.404 volts to do that, so I played with the LLC and got it stable at something like 1.425vcore.  LLC is goddamn amazing.  Haven't messed with the CPU-NB, so that's the next thing I'll screw with once I repaste my cooler.  Eventually I should probably put some faster ram in it, I'm only running 1333 (at higher speeds with the bus speed increase).

 

This is all with a Rosewill Quark 550W psu and a Crosshair V Formula-Z.  Previously, I had a very tough time getting it over 3.9GHz on my old M4A88T-m and Cougar 560W psu.

 

Mine is on water, but I did a garbage job of pasting the waterblock last time I had it apart, so an air cooler would probably do better.

[bcguru9384]

go into windows control panel

open power options

click advanced and change the pci link settings to no savings

see if that helps

[Mothballs]

38 minutes ago, Phate.exe said:

How high do you actually have it clocked?  I might have missed it but I didn't see your bus speed.

 

If it's at the stock 200, you're "only" at 4GHz, so I'd just drop the multiplier a bit and bring the bus speed up some.  If I'm wrong than just ignore this part of my post.  Mine really didn't like having the multiplier much over 19-19.5.

 

Mine's very happy at somewhere around 217 bus speed with an 18.5 multiplier.  Under load it only seemed to want 1.392-1.404 volts to do that, so I played with the LLC and got it stable at something like 1.425vcore.  LLC is goddamn amazing.  Haven't messed with the CPU-NB, so that's the next thing I'll screw with once I repaste my cooler.  Eventually I should probably put some faster ram in it, I'm only running 1333 (at higher speeds with the bus speed increase).

 

This is all with a Rosewill Quark 550W psu and a Crosshair V Formula-Z.  Previously, I had a very tough time getting it over 3.9GHz on my old M4A88T-m and Cougar 560W psu.

 

Mine is on water, but I did a garbage job of pasting the waterblock last time I had it apart, so an air cooler would probably do better.

Bus is still at 200, I was sorta hoping to get away with just Multi, but if Bus clocking will help, I'll sure as heck do it. Seems like I might need to squeeze more vcore, It's running at about 1.390~1.404s under load right now. LLC is on Auto, the IV doesn't have as good options as the V.

My ram options are both Furyx Ram, 2x 4 sticks @ 1600 and a pair of the same @ 1833. We'll see what happens.

I'll try this tomorrow. I'm busy for the rest of the night, but the help is greatly appreciated.
 

30 minutes ago, bcguru9384 said:

go into windows control panel

open power options

click advanced and change the pci link settings to no savings

see if that helps

Is this for the GPU? I'll try it after I get the memory and CPU stable. For now it can sit at stock.
 

51 minutes ago, SpaceGhostC2C said:

But it can happen. When OCing my FX cpus (not the same, but still AM3+), I couldn't initially run my memory at its rated speed (it did run fine at stock clocks). Using a lower JEDEC standard it was stable. In my case, a small bump to the CPUNB voltage allowed me to get it stable at the rated speed - even overclock it a bit past their rating (the CPU NB is essentially the integrated memory controller plus stuff).

Alright. I'll bump it back down and try to get stable on just CPU before trying anything with the Mem.

Tomorrow We get back to it. Thanks for the help so far.

[Phate.exe]

5 minutes ago, Mothballs said:

Bus is still at 200, I was sorta hoping to get away with just Multi, but if Bus clocking will help, I'll sure as heck do it. Seems like I might need to squeeze more vcore, It's running at about 1.390~1.404s under load right now. LLC is on Auto, the IV doesn't have as good options as the V.

My ram options are both Furyx Ram, 2x 4 sticks @ 1600 and a pair of the same @ 1833. We'll see what happens.

Your ram is definitely better than mine.  I have 4x 4gb sticks of XMS 1333 (although it runs just fine somewhere around 1600).

 

I'd turn the multiplier back down to 17.5-18 or so and set the bus to like 210 to start.  That puts you in the 3675-3780 range.  Bump the bus up bit by bit (Asus AI suite is pretty handy for this) until something goes wrong.  When I'm just trying to find settings that work, I'll usually just use AI suite, Cinebench, and Firestrike.  Cinebench loads down the CPU enough to let me know if voltage is going to be a problem at all, and I can see an immediate result to compare if I'm gaining more by just chasing high clockspeeds or if I should chase more bus speed.  From what I've read, I'm actually leaving a shitload on the table by not messing with CPU-NB speeds.

 

If you REALLY want to chase bus speeds, I'd drop the multiplier down to like 14 and see how high you can increase the bus speed before something freaks out, then dial bus back by a few notches and add some more multiplier and voltage back in.  On my old motherboard I started having problems with the bus speed around 230.

 

For LLC, the V only has auto/low/medium/high/extreme (actually I think there's probably a custom option, I'm just running medium or high).  LLC is pretty easy to figure out, set a safe voltage like 1.45, load the cpu down and note the vcore under load.  Bump the LLC setting by one and repeat the load test until you have acceptable/no/very little Vdroop, and then turn the vcore back down.

 

I'm at work currently, but either when I get home tonight or tomorrow when I wake up I can mess with my machine again and report back with more specifics.

 

1090T's are monsters when you get them over 4GHz.  After I remedy my cooling issue, I think I have another couple hundred mhz in it.

[bcguru9384]

actually it covers any pc component that runs on a pci lane.  gpu hard drives memory etc

link state setting under power options (where you choose powersave mode perfomance mode)

click advanced button under performance tell link state to not worry about power savings

may bring the 3volt back to 3volts instead of .12volts

[Mothballs]

16 hours ago, Phate.exe said:

Your ram is definitely better than mine.  I have 4x 4gb sticks of XMS 1333 (although it runs just fine somewhere around 1600).

 

I'd turn the multiplier back down to 17.5-18 or so and set the bus to like 210 to start.  That puts you in the 3675-3780 range.  Bump the bus up bit by bit (Asus AI suite is pretty handy for this) until something goes wrong.  When I'm just trying to find settings that work, I'll usually just use AI suite, Cinebench, and Firestrike.  Cinebench loads down the CPU enough to let me know if voltage is going to be a problem at all, and I can see an immediate result to compare if I'm gaining more by just chasing high clockspeeds or if I should chase more bus speed.  From what I've read, I'm actually leaving a shitload on the table by not messing with CPU-NB speeds.

 

If you REALLY want to chase bus speeds, I'd drop the multiplier down to like 14 and see how high you can increase the bus speed before something freaks out, then dial bus back by a few notches and add some more multiplier and voltage back in.  On my old motherboard I started having problems with the bus speed around 230.

 

For LLC, the V only has auto/low/medium/high/extreme (actually I think there's probably a custom option, I'm just running medium or high).  LLC is pretty easy to figure out, set a safe voltage like 1.45, load the cpu down and note the vcore under load.  Bump the LLC setting by one and repeat the load test until you have acceptable/no/very little Vdroop, and then turn the vcore back down.

 

I'm at work currently, but either when I get home tonight or tomorrow when I wake up I can mess with my machine again and report back with more specifics.

 

1090T's are monsters when you get them over 4GHz.  After I remedy my cooling issue, I think I have another couple hundred mhz in it.

My LLC gives me Auto, 0%, 50% and full. And it's apparently being very inconsistent. My CPU/CPUNB voltage adjustment is only giving me offset control in Bios, and when I bump up the offset in bios at full calibration it actually went down from where it was at auto. Annoying.

I dropped back down to a 18x multi, and it seemed stable, dropped the ram down to 1333 @ 1.55v, seemed okay. Bumped up offset voltage with LLC on 100%, immediately went from 1.332v under load to 1.308v under load. I:

System threw an error I thought I was free of, Jraid IRQL error. I don't have any raid setup. /: It used to do this at random before I did a clean install, thought I'd ditched it.