Apparently I need help with AMD 1090t overclock

[Mothballs]

So I went digging, yet again.

"HyperX Fury memory is not compatible with the ASUS Crosshair IV Formula motherboard. All of our currently manufactured 4GB memory modules are Single Rank and that motherboard requires Dual Rank 4GB memory modules. We no longer support the ASUS Crosshair IV Formula motherboard as we no longer manufacture memory that is compatible with it."

Well F#@%.

"DRAM used to be made with low density memory chips, most all DDR3 these days uses higher density memory chips for higher performance and the older mobos can't run with the higher density chips."

That might explain a thing or two...

I think I have a set of older 2x2 OCZ 1600 and a set of 2x2 G.Skill, let me check if they're on QVL... Shouldn't hurt to mix and match them, they're both good @ 1.65v 1600 9-9-9-24, right?

[Mothballs]

Alright, all the ram is on the QVL, but the G.Skill is a kit I got used: It's supposed to be a 3x 2gb Triple channel kit, will this hurt anything?

OCZ is supposed to be 1.65v, G.skill is supposed to be 1.5-1.6v.

Should I hold back on this and try and find a fully matching set?

[Mothballs]

Fresh windows install. Graphics card is alive again (It literally said It wasn't found, and reported back a ground breaking 1 fps on userbenchmark's test)

Clearly I managed to corrupt the ever loving god out of the previous install. Go me. I:

Any advice to be had on the ram, or should I just charge ahead full with the Furyx and this fresh install?

Also switched the Bios, i was apparently running a Beta. I: I have proper Voltage control now.

Edited March 29, 2017 by Mothballs

[Mothballs]

Against better judgement I went ahead with the Kingston again.

 

Running Prime on anything above stock leads to a bluescreen. Uncorrectable error.I probably should have been mentioning this, but running Prime and failing results in a dramatic crash, and it wasn't until I began digging more that I realized that It's NOT supposed to do that. This definitely seems like a memory issue to me at this point, not a CPU or MOBO issue.

This brings me back to the other question: Can I run the mix of OCZ and GSkills RAM?

Part of the reason I'm going so in depth on this is 1; I need to tide myself over until the next system is in my budget and 2; When I get the new system I want to know exactly how to do this sort of thing so it doesn't go wrong when I do it then. This is a learning experience.

[Mothballs]

OCZ 1600 @ 1.65v  DDR3 running Small FFT. Crashes @ 3.6GHZ, Bluescreen.

I'm going to have to start writing down these error codes.

Trying 1333 seems to keep it stable for ten minutes running small FTTs, but the fact that the system keeps crashing instead of Prime throwing an error is really distressing and discouraging. This is looking more and more like a total disaster and a waste of five days total...

[Mothballs]

1.35 volts CPU ten minutes stable @ 3.8ghz with the OCZ ram @ 1333

Pushing my luck @ 4.0 leads to windows displaying the circle of infinite loading.

 

Trying straight 20x multi with stock bus speed @ 1.4v. LLC 50% gives me a .04 Volt Vdroop. Seeing how stable it stays... Expecting the parade of Bluescreens to resume...

 

 

[Mothballs]

Happily surprised. Ten minutes with no failures on small FTT, Vcore reading @ 1.368v 46C
Switching to Blend.... BSOD: clock watchdog timeout, so cpu error.

Retrying with LLC set to auto results in a .04 OVERVOLT. I: still within tolerances, but jfc, why. trying blend with that setting...

[Mothballs]

Would it have killed them for a 75% option on the LLC? 50% is under by .04v and 100% is over by .04v
Right in the middle would have been really nice...

Temps go from 46-50 and cycle.
Ten minutes stable on blend. Trying IBT for ten passes on high... Fails. I: No BSoD however.

[Mothballs]

Doing the crazy bit where I listen to @Phate.exe and running a 225 bus speed with an 18x multi.

 

IBT failed on the second pass last time. Third this time...

CPUNB and HT both going up to 2700, voltage set @ 1.2V, LLC brings it to 1.25v. I:

Temps hit around 52C and it fails on second pass... Other peoples documentation tells me that bringing up NB should increase stability. Not seeing it.

[Mothballs]

On 3/28/2017 at 4:51 PM, 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).

There is definitely a memory issue involved here, and I'm trying different memory, different clocks, etc. I'm not sure what the heck I'm doing wrong that it spits out errors and crashes constantly on me. I'm trying different LLC, I'm trying Combinations of base clock and Multi, I'm trying Lower Ram timings still, I've tried bringing up the HT and CPUNB, and leaving them alone.

What's pissing me off the most is other people basically just have Prime spit out an error at you and stop a worker. I get full system crashes. There's practically a cult to this Mobo and CPU combo on other sites, this is considered a good, solid overclocker, this is supposed to be stable and relatively simple, but it friggen just hasn't been.

[Mothballs]

Stability in ibt seems to happen around 52c

 

Thermal wall? I know it's considered bad to go past 55 for stability. 

[Mothballs]

Holy shoot, I think I have the most of it figured.

IBT throws errors if the CPU gets above 53C, Because the CPU is getting too dang hot.

Prime finishes 10 minutes of blend and ten minutes of small FTT with the voltage dropped in the bios. Idle is 1.392, under load it's over-volting to between 1.404-1.428. Finished 10 passes of IBT on High. CPU/NB Voltage is sitting at 1.210 Idle to 1.244 Load.

All of this while windows tries to run updater processes in the background.

Alright, now, here's where I need to ask even more advice. 

I have it set, right now, at 18x ratio, 225 bus Freq, DRAM @1200, with the CPU N/B @2700 and the HT-L @2700

LLC is at full on the CPU NB and the CPU.

Voltage is on manual, set to 1.38125 on the CPU and 1.2 on the CPU/NB

DRAM is set to 1.65v

CPU spread disabled.
Virtualization Disabled.
C'n'C Disabled.

I've been running Prime95 for 10 minutes each on blend and small FTT, and Intel Burn Test on high. IBT throws an error if the core reaches 53C, no matter what. Thermal wall I'm guessing.

This iteration managed to pass both prime tests for ten minutes, and IBT on ten iterations, but The end goal is to get the DDR3 speed up, maybe try to force the Kingston set, and to get C'n'C Enabled and using offset voltages while staying stable through whatever you suggest for stability testing.

Lay it on me, what are my next steps in your oppinions?

[Mothballs]

Testing with Prime's Blend, Finally had a worker stop as opposed to a BSoD. Only failed on core 5

[Mothballs]

If nothing else this thread will be something that the next fool who decides to follow in my footsteps can look at.

Look forwards, weary traveler, and know there is light.

Alright, dumb aside, I'm sitting at a 1.44v under load from Prime on the CPU. CPU-Z's Validation puts it at 1.412  (http://valid.x86.fr/k24nch ) Temps are around 50c, but literally nothing I do real world should put this much strain on the CPU.

Going to test the other DDR3 later.

If I plan on turning back on Cool 'n' Quiet and using offset voltages, I am guessing I should run my final stability runs with those all on?

[SpaceGhostC2C]

5 hours ago, Mothballs said:

If I plan on turning back on Cool 'n' Quiet and using offset voltages, I am guessing I should run my final stability runs with those all on?

Good to see you are making progress!

I would keep the CPU vcore at manual mode, it seems to give you a more direct control on it (or at least it's easier for me to understand what I'm doing  ). If you pass stability tests as it is, I would try re-enabling C'n'C. I haven't had problems with it, and it's good to let your CPU "rest" when the PC is just idling. If that's how you want to use it, it makes sense to run the final tests with it enabled.

 

7 hours ago, Mothballs said:

 

Voltage is on manual, set to 1.38125 on the CPU and 1.2 on the CPU/NB

If you can't get your RAM to work at the speed you want, you can try upping the CPU/NB voltage to something like 1.3v (I don't know the safe range for the 1090T, but I would assume 1.3 is still safe. I'm currently using 1.3v on the FX CPU, which defaults slightly below 1.2v. Whenever I mess up my RAM configuration and the PC fails to boot, my mobo bumps it to 1.4v on its own, which seems to work every time, but after undoing my crimes and dial it back).

[Mothballs]

4 hours ago, SpaceGhostC2C said:

Good to see you are making progress!

I would keep the CPU vcore at manual mode, it seems to give you a more direct control on it (or at least it's easier for me to understand what I'm doing  ). If you pass stability tests as it is, I would try re-enabling C'n'C. I haven't had problems with it, and it's good to let your CPU "rest" when the PC is just idling. If that's how you want to use it, it makes sense to run the final tests with it enabled.

 

That's what I figured, but at the moment C'n'C is greyed out. If I'm reading things correctly, I need to enable offset voltages on this mobo for it to work.

Something something, old tech.

 

Offset shouldn't be that bad to figure out, stock voltage is around 1.28~1.35 on core and 1.1x~ cpu/nb. Just add until you hit value X, where x is the known good value.

 

...After saving the good value in the oc profiles...

4 hours ago, SpaceGhostC2C said:

If you can't get your RAM to work at the speed you want, you can try upping the CPU/NB voltage to something like 1.3v (I don't know the safe range for the 1090T, but I would assume 1.3 is still safe. I'm currently using 1.3v on the FX CPU, which defaults slightly below 1.2v. Whenever I mess up my RAM configuration and the PC fails to boot, my mobo bumps it to 1.4v on its own, which seems to work every time, but after undoing my crimes and dial it back).

I think 1.3 is the limit on my cpu/nb if I am correct. Trying really hard to sit juuust at the the lowest voltage for stable overclocking.

 

Yeah, the 1090t is straight rated for only 1333 ram, because the cpu/nb is the integrated memory controller, and its nooooot great. Bumping NB should help with that in general. I'll try messing with memory now that I can get prime to last longer then ten minutes. Now that I'm not crashing, I feel a little more confident.

 

Still not sure how I feel about LLC shoving 0.05volts down the systems throat or going under by that. I sorta feel like the overshot could be worse if there's a brownout or something. Should probably get a UPS...

[-rascal-]

The 'Thuban' CPU's weren't know to overclock ridiculously far.

4.0 GHz ~ 4.2 GHz was usually the limit on air or AIO AND without the voltages going way too high.

 

I've used a ASUS Crosshair IV Formula, and Crosshair V Formula (non-Z) with my Phenom II X6 1090T (before finally going to a FX-8350). I was able to get 4.0 GHz out of my 1090T without too much issue, but 4.1 GHz require a significant jump in Core Voltage.

 

Are you able to post a few pictures of your BIOS settings?

I don't why you were running it in offset mode in for the first place.

 

If you push the CPU-NB and HT frequency too high, it WILL cause instability.

And I think the CPU-NB frequency for a stock 1090T is 2000 MHz.

 

You may need to tweak some of the other voltages as well to help with stability (e.g NB, SB, etc.)

I will try and dig up some of my old 1090T testing results/values and resources...

[Mothballs]

3 minutes ago, -rascal- said:

The 'Thuban' CPU's weren't know to overclock ridiculously far.

4.0 GHz ~ 4.2 GHz was usually the limit on air or AIO AND without the voltages going way too high.

 

I've used a ASUS Crosshair IV Formula, and Crosshair V Formula (non-Z) with my Phenom II X6 1090T (before finally going to a FX-8350). I was able to get 4.0 GHz out of my 1090T without too much issue, but 4.1 GHz require a significant jump in Core Voltage.

 

Are you able to post a few pictures of your BIOS settings?

I don't why you were running it in offset mode in for the first place.

 

If you push the CPU-NB and HT frequency too high, it WILL cause instability.

And I think the CPU-NB frequency for a stock 1090T is 2000 MHz.

 

You may need to tweak some of the other voltages as well to help with stability (e.g NB, SB, etc.)

I will try and dig up some of my old 1090T testing results/values and resources...

I've been going off this: http://www.overclockers.com/step-guide-overclock-amd-phenom/ And numerous other forum posts. 100% on the fact that getting too hot means instant fail on tests. My cpu will consistently fail @ around 53c I've found, just from watching temps while running IBT. 100% Failure rate at that temp, no getting around it.

 

I'll post Bios Pictures, yes.
It was stuck in offset because I apparently had a Beta BIOS loaded and I refused to not be on offset. It was annoying. Also, apparently on the most recent bios you need to have offset enabled to have C'n'C running, so that's a goal for after getting a stable base.

I'm going off the CPU/NB chart on that guide more or less, though even he states that sometimes you need to go higher then I have it.
 

 

Side note, In the long run, is a Liquid cooler better for CPU longevity when going for high overclocks? Like, I know you can do well with Noctua coolers, but the major difference between Liquid and Air coolers is that an Aircooler doesn't hold it's heat after the chip cools down, where water does. This means the shift from full load temps to idle temps takes longer with liquid as you have to dissipate the liquids thermal load as well. This means more violent temp shifts with Air compared to liquid, which should mean more stress.

I almost feel like that question deserves a thread... Or an LTT Video.

[-rascal-]

I found my old 1090T thread on another forum.

I went with the Base Clock overclocking at my first time around, rather than adjusting the multiplier

(eventually got my 4.0 and 4.1 GHz overclock with multiplier overclocking)

http://forums.legitreviews.com/viewtopic.php?f=24&t=31907

 

The 'skier' guy is a forum moderator (doesn't look like hes been on since 2014 though), but he made an overclocking log for his own Phenom II X6 1055T. It's a locked multiplier CPU, but some of the things can still be applied to the 1090T -- especially if you are adjusting the Base Clock.

http://forums.legitreviews.com/viewtopic.php?f=24&t=28671

 

For the sake of stability testing, turn off any power-saving features (e.g. Cool N' Quiet, C1E, etc).

That way, your CPU will run at constant frequency, and hopefully, constant voltage.

I don't know if the huge voltage swing (1.2# V to 1.41 V) is causing a problem...and that your overclock is not 100% stable yet.

 

You are correct, the recommended safe limit for the CPU-NB is about 1.30V, but keep is lower if possible.

Keep the Core Voltage below 1.5V -- ideally, 1.4V ~ 1.45 (1.47 or 1.48Vif cooling and temperature allows).

 

Use the LLC settings that will give you the most stable Voltages.

For an example, if you set your Core Voltage to be 1.42V in the BIOS, use the LLC setting that would hold it at 1.42V when the CPU is fully loaded (e.g stress testing).

 

HWMonitor should give you a good read-out of your minimum and maximum voltages.

In fact, I found the temperature readings to be quite accurate for AM3 CPUs.

Plus, it has labels such as "CPU", "Mainboard", "SB", and "NB" instead of..."TMPIN0", "TMPIN1", etc -- at least for the IV and V Formula motherboards.

[Mothballs]

3 minutes ago, -rascal- said:

I found my old 1090T thread on another forum.

I went with the Base Clock overclocking at my first time around, rather than adjusting the multiplier

(eventually got my 4.0 and 4.1 GHz overclock with multiplier overclocking)

http://forums.legitreviews.com/viewtopic.php?f=24&t=31907

 

The 'skier' guy is a forum moderator (doesn't look like hes been on since 2014 though), but he made an overclocking log for his own Phenom II X6 1055T. It's a locked multiplier CPU, but some of the things can still be applied to the 1090T -- especially if you are adjusting the Base Clock.

http://forums.legitreviews.com/viewtopic.php?f=24&t=28671

 

For the sake of stability testing, turn off any power-saving features (e.g. Cool N' Quiet, C1E, etc).

That way, your CPU will run at constant frequency, and hopefully, constant voltage.

I don't know if the huge voltage swing (1.2# V to 1.41 V) is causing a problem...and that your overclock is not 100% stable yet.

 

You are correct, the recommended safe limit for the CPU-NB is about 1.30V, but keep is lower if possible.

Keep the Core Voltage below 1.5V -- ideally, 1.4V ~ 1.45 (1.47 or 1.48Vif cooling and temperature allows).

 

Use the LLC settings that will give you the most stable Voltages.

For an example, if you set your Core Voltage to be 1.42V in the BIOS, use the LLC setting that would hold it at 1.42V when the CPU is fully loaded (e.g stress testing).

 

HWMonitor should give you a good read-out of your minimum and maximum voltages.

In fact, I found the temperature readings to be quite accurate for AM3 CPUs.

Plus, it has labels such as "CPU", "Mainboard", "SB", and "NB" instead of..."TMPIN0", "TMPIN1", etc -- at least for the IV and V Formula motherboards.

LLC is being an issue this whole time. 0% is worthless. 50% is under by .04v, MAX is over by .04-.05v, Auto sorta just does whatever. Those are the only settings. Really wish there was a 75%. Friggen annoying. Just shoved it on max right now and let it run lower when not under load. I'll try it on something like Overwatch and see if it stays stable while being used at lower load. 

So Don't do stress with C'n'C on, even though I plan on having that on? That seems a little counter intuitive... I can see doing all the basic stability hunting with it off, but I'd think I'd try and stabilize it with that on for the final pass...

HWMonitor is a champ, and has given me so much guidance, it is a gem of a program.

[Mothballs]

Ten minutes with both Blend and SmallFTT. Dropping IBT because of unrealistic heat output. 

Still not super happy with the LLC shooting voltages up that high.. CPU is set @ 1.40v in Bios, but as we can seeeee...

Gonna try the Kingston ram Again, because I'm a sucker for punishment. (also because I paid for it and the computer normally boots with it, so..?)

[Mothballs]

Ha. Haha.

Alright, looks like the Kingston Set is good to go. Still performing realllly below average, but We'll see what we can do about that.

Big spike in FTT Test heat however, up to 52C. I wonder if I can bring down the CPU/NB Voltage?

[Mothballs]

Got offsets and C'n'C Enabled. Ten minutes good on both. I'll post the Bios settings Soon, mostly for archival and contrast to where I started. Too often have I gone looking for something only to have the thread where people started talking about it just end with "I fixed it, Nevermind."

Short and sweet of it, I just moved the offsets until It matched up with what was working on Fixed, opening then running the test to get a feel for it. The Heavy LLC seems to matter a whole lot less in offset mode, though i'm definitely going to keep an eye on it for the next while. The Kingston ram works fine, just waaaay under expected stats for that RAM kit. Not a huge deal to me, but yeah, it's a factual thing.

To everyone who was yelling at me that It was the voltage, yes, it was the voltage. I'm an Idiot and focused on too many things at once instead of the fact that I probably had too low of a Vcore.

It's twin, which isn't getting an overclock, will probably stand in for it while I Stress test for finalization.

[Mothballs]

Forgot to turn off sleep mode. System went into sleep mode mid test, scared the ffff out of me. Refused to resume from sleep, booted back up fine.

I love how the individual cores register as 2c lower then the room at Idle. I:

[Mothballs]

Here's the new kicker. Moved it across the room to a different outlet and Powerbar. Dropped voltages by .03 under load. I: The outlets are on different circuits, etc.

I think I'm gonna get a UPS as opposed to a power bar, my power is shit here anyways.