AMD FX8350+M5A99 Evo R2 -> Persistent System FREEZE

[celta]

I hoped to find help here at this community for a very distracting problem. I found a similar problem on this forum
http://www.overclockers.com/forums/...9FX-pro-r2-0-AMD-FX-8350-Vishera-(B-E)/page2?
which solution was a faulty CPU…I try to figure out if it is the same for me.

My system:
CPU: AMD FX8350
Cooler: CoolerMaster Seidon 120M liquid cooler
GPU: Sapphire R290X
PSU: BeQuiet! PurePower 730W
RAM: Corsair 1333 8GB DDr3
MBO: Asus M5A99 EVO R2

Situation:
I recently upgraded PC (Mobo+CPU). I wonderd that it was NOT stable at stock. I was able to get Windows running, but any kind of stability test -> FREEZE. BIOS update -> still same issue

By freeze I mean system is stuck as it is, any last sound gets repeated indefinitely and no entry whatsoever in windows error log. It mostly occurs when playing youtube or randomly during gaming.

Did some research and found out that some ppl had the same Problems. Deactivated crappy “features” like cool n quiet, turbo core, C1E, C6state…and fixed the CPU core voltage manually to 1.3 V (as suggested here in this forum) and set power phase control to Extrem in +DIGI settings. All power phases always on, though that the freezing might be connected to a lack of power at the CPU and all the weird swithing around of clocks and voltages...

Seemed fine so far. Took the multiplier up to x22 and still seemed stable. (back then, like 3 month ago, I did not do what you would consider proper stability testing, I ran AMD Overdrive stability test for 15min and seemed fine for me)

Up until now the problem seemed fixed.
I don’t know what caused the FREEZING to come back (maybe it never was gone, because I did not game very much)
I noticed in UEFI Bios that my RAM was running at 1066Hz..oO..changed it to 1333 what it should be -> 10sec blender -> calculation error FAULTY HARDWARE

So I assumed maybe there is smt wrong with my RAM

Downloaded memtest64+ as suggested in this forum and tested. After about 5h and 3 passed (RAM@1333 as it is supposed to be) no error. BUT zap -> whole monitor is a mosaic of about 1cm^2 large 16bit colored squares. Memtest crashed, but before that no error (it happened by coincidence that I looked at the screen right after the 5 hours at the moment it happened)

That led me to think maybe something is wrong with my graphics card. But after 30min of FurMark extreme burn with 4xAA and 5760x1080 resolution I think it is safe to assume that it has nothing to do with the freezing issue (the system did not freeze while furmark testing)

OK BIOS reset to stock -> as expected not stable as it was in the first place.
Turning off green stuff, setting power phases again…cpu voltage to auto and back to stock frequency (4Ghz) and RAM to 1333 -> blender FREEZE after 1 minute or something

I set the RAM to 1066 (so way below its rated value) and to my surprise: 60min blender stable. What even more interesting that the automatically set voltage was 1.22! Volt. In my opinion rly low.
Though well, problem fixed. So I went slowly up with voltage and multip (blender for 15min after each step.)
Settled at 4.3 GHz by multiplicator and voltage 1.28V.
Higher voltage or multiplicator gave me Freeze again.)
Checked 45min blender if it is stable -> no error (was eager to finally game DA:I again, so I kept the test short)
Finally back to gaming dragon age inquisition for hours -> no problem whatsoever

Though myself..well if its stable at the cost of an underclocked ram, ok…go with it until next upgrade…as long as no freezes and performance is sufficient…

Today in the morning, start pc, youtube video ….FREEEEEZE…F***G NOOO!!
Back to 1.22V and 4 Ghz -> boot again -> youtube -> freeze
Boot again -> FurMark extremeBurn in 15min -> no issue -> continue with blender -> 15 min -> no issue (did not have more time before work)…youtube -> no issue


Remarks:
There are several things that disturb me…

Why are out of the box settings with newest BIOS not stable?

Cant read any sensor data from mainboard with CPUID or HWMonitor…only Asus tools work …but at lame 5sec intervals (GPU sensors on the Rx290x are rly good, 0.1sec intervals with GPUZ)

The 5V and 12V rails at my PSU seems quite stable (not less than 11.75V). But the 3.3 Volt rail drops to 3.195 (but as far as I read is this a general issue with the bequiet pure power series that the 3.3 voltage regulation is a bit loose)

Why did the Mobo change 1333 to 1066 without notifying or anything? And 1066 seems semistable and the rated 1333 not at all?

I am using a custom built casing, VRM and Mobo are well cooled directly by a 20cm FAN and if I turn cooling into performance mode MB tempr does not rise above 40° and CPU not above 60° during blender. VRM heat spreader isn’t even getting really warm.
Mobo can rise to 48° when FurMark (the r290x is a thermal beast).
So I don’t think I have a thermal issue. But regarding the last results maybe warmer is better suggested by the stability after the FurMark burning?!?




Mobo and CPU is fairly new…RMA would be possible…but pointless if it’s the RAM or PSU after all…3 weeks without gaming pc in winter would also suck

I thought about rly full loading the PSU to rule it out as faulty: if it could handle blender on 7 cores (one core to feed the Furmark) + FurMark simultaneously it can handle anything, right?

Whats your opinion on that weird behaviour? I know here are a lot of people who have seen some strange hardware issues. Any Ideas on how I can narrow the fault or even fix it?

 

[rescuetoaster]

My opinion is that you probably need to manually set your ram's voltage to 1.5, and that it is undervolted which is probably the biggest reason for your instability.
Find your ram's specs, and manually set your voltage, timing, and speed. We'll go from there.

 

[WhitehawkEQ]

Check the ram voltage, I find what is set in the BIOS and what the VRM's put out are off by .01 to .05 volts, like if BIOS is set to 1.5V, the VRM's put out 1.46V. Problem is with tolerance with the parts used in the VRM's. I ran into this on 2 of my systems.

 

[Johan45]

I agree , it's likelt the ram that's the culprit. Even if it's not faulty it may just not be compatible with the board. Do you have 2x4GB or 4x2GB and is it all matched dual channel ram? It might be wise to upgrade the ram in any event since the FX will handle and like ram more in the 1866-200 range.
You could try setting the ram voltage at 1.55v if it's rated for 1.5v and set the CPU-NB voltage to 1.25v manually.

 

[Mandrake4565]

Give us a peek under the hood. Download Hwmonitor "non professional", Prime 95 and Cpu-Z. Open Hwmonitor and run Prime Blend 20 minutes. If the temps on the Cpu or the Package go above 70c and 62c respectively stop the test. Then post back here with screen shots of Hwmonitor and the following tabs in Cpu-Z, Cpu Spd and Memory.

 

[celta]

thank you for your fast reply!!

the RAM are a 4x2GB Corsair XMS 1333 Kit

I did rise the vcore to 1.3 V and the CPU/NB to 1.20 (i stayed below the suggested 1.25, because value was already yellow).
see picture below

I tested Dram Voltage with 1.5 and 1.55 V -> both blender freeze after 5min

I set all the RAM values to those I know and those I could find (dont know waht all the other values are or where I could find a reference specifically for my RAMs for them)



I did as suggested. Blender -> freeze after 5min : here the requested ss frozen to the screen

CPU_Z SPD


CPU_Z CPU Tab and HW Monitor


and the CPU_Z memory tab


I really appreciate your help with this tiresome issue

 

[Janus67]

Can you run memtest86+ with your memory and see if it freezes?


Also, someone more knowledgeable about AMD can probably answer this -- Does the OP's motherboard have the proper VRM section needed for an 8350?

 

[Mr.Scott]

Too hot.
Idling at 44c and load near 70c in 5 minutes.
A single 120 rad AIO ain't gonna cut it. Especially at that low a vcore.

 

[WhitehawkEQ]

Does the OP's motherboard have the proper VRM section needed for an 8350?

Yes, Asus M5A99 EVO R2 is a 990 chipset motherboard.

 

[Mr.Scott]

Yes, Asus M5A99 EVO R2 is a 990 chipset motherboard.

That doesn't mean anything.
6+2 digi VRM. IMO, on the edge of acceptable. I wouldn't beat on it.

 

[celta]

i can explain the high idle temps...its the way i set it and has nothing to do with the cooling capabilities

the fan iddles low, to keep the noice down when I am not gaming and working. But the cooling ramps up fast and even faster beyond 60°. esp when I go from cold start to blender it overshoots but mostly keeps then at 68° at the point where I set 100% cooling. The Mobos default settings were 100% RPM @ 75° CPU

The average gaming temps hang around 60 deg. You think the threshhold is too high?

The Mobo offically supports the whole fx series....its not like the fx8350 is over high end CPU....I would say the board was even made for this segment of cpus

I will do another memtest tmr with the curent settings. at least they seem windows stable and had so far no freeze this evening (except the blender tests)

 

[Mr.Scott]

68c is still too hot. Most FX processors lose stability between 50c and 60c.
Your single 120 AIO isn't adequate to cool an FX processor.

 

[chrisjames61]

68c is still too hot. Most FX processors lose stability between 50c and 60c.
Your single 120 AIO isn't adequate to cool an FX processor.

Those are high temps. Especially for that low vcore. It would be interesting to see what his thermal margin is. You really need at least a 240 mm radiator to cool an 8350.

 

[Mandrake4565]

I have a M5A99X Evo which is similar to the board the Op has. It should be fine to about 4.5 with the 8350 on it. I do agree with the guys above about your temps. If the Socket temps is that high, I would touch the VRM/NB heatsink, while under load. I would bet it's really really hot, which may be the cause of your issues.

 

[Mr.Scott]

The Mobo offically supports the whole fx series....its not like the fx8350 is over high end CPU....I would say the board was even made for this segment of cpus

Be that as it may, real world tests/users say different.
Even if the FX8350 weren't a high end CPU, it doesn't change the fact that it's power hungry and runs hotter than any processor ever made by AMD......with a few exceptions.
The motherboard may indeed support the chip, but that doesn't mean it's a capable board. It sports a 6+2 power phase system. Most boards that can handle the chip adequately sport at least an 8+2 phase system.
So you have 2 strikes against your setup already.
While your board might handle the CPU at stock or a small OC, it doesn't change the fact that your cooling solution will not.

EDIT- @ Manny- I would thank you but.....no button. lol

 

[celta]

i was not aware of the power phase issue. it really concerns me that I did not stumble across that sooner. I just saw that the board supports all fx, by the fact that I choose a middle class one, it did not came to my mind that it could be a serious limitation. the blame is on me then for not researching that closer

I wasnt also aware that those CPU are THAT temperature sensitive. My old 965Be was hardly under 60° and stable OC for more than 5 years...
It is also hard for me to imagine that the enclosed stock fan of the fx8350 could keep the CPU even close in that range you proposed without sitting in a freezer.

i based my decision about the cooler on that seemingly well documented combination of seidon and fx8350 with some OC involved:
http://www.dvtests.com/cooler-master-seidon-120v-test-and-review/ and some other posts that they work fine together.

regardless I tried to challenge the temperature hypothesis. set cooling to maximum and did some testing again. same settings as before, 1.3core,increased CPU-NM, stock clock and RAM 1333 with adapted timings and voltage

Freeze again after 30sec of blender @42°C CPU and 40°C Package...that is way below the suggested critical temperature levels and strongly suggest something else is wrong.
remarkably I set RAM @ 1066 , everything unchanged
blender stable for 15min ...CPU temp up to about 60°C.

any thoughts on that? RAM probably just incompatible? can it be fixed? is it possible that memtest gives no errors even something with CPU RAm communication is wrong?


Edit: VRMs are not hoter than the palm of my hand when fully loaded...

 

[chrisjames61]

GSkill is the go to ram of a lot of the veteran overclockers on this site. At least on the AMD side of things.

 

[RGone]

So okay, where are we...???

Asus M5A99 EVO R2 >> This would not be recommended motherboard for seriously overclocking an 8 core FX processor. With very good cooling and fans on the VRMs it might support 4.4GHz overclock. But the key is "might". Each system is different due to cooling in the case and overall temps in the computer room.

Cooler Master Seidon 120M Liquid CPU Cooler Review
Mon Feb 11, 2013
As far as entry level water cooling units go, this is as entry level as you can get. The Seidon 120M is batting in the same field as the Antec Kuhler 620 and Corsair H55. Not one of those AiO coolers mentioned in that sentence above would be considered adequete to overclock an 8 core FX processor. Not nearly heavy-duty enough cooling as experienced generally forum wide here.

This is probably the ram kit you have >> XMS3 — 8GB (4x2GB) DDR3 1333MHz C9 Memory Kit (CMX8GX3M4A1333C9)<< You did not give the part number which is very important. This is more middle of the road ram and not entry level but not top tier either. Plus it is a 4 stick kit and that is the hardest ram configuration to setup and have work the easiest. 2 stick kit far easier to use.

I cannot see if your bios for ram settings has READ to READ Delay setting. If it does what is the 'numbered' setting.

And by all means set the ram voltage to at least 1.55V since that has been KEY for many users in ensuring stability. I also would have no issues with setting my own CPU_NB voltage to at least 1.25V or higher. AMDs own 'papers' for overclocking their FX processors use 1.35 to 1.4V and show it in their overclocking guides.

So where the heck are we? IF you do not have READ to READ Delay setting in bios I most likely would suggest you get this ram if available where you live. It just works. It has no big-arse heat spreaders to get in the way of other cooling components and having used a couple of kits myself and suggested to others here in the forum and their memory issues disappear...well I would still suggest it to any "friend". To my enemies I might suggest otherwise. Hehehe.

G.SKILL Ares Series 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1866 (PC3 14900) Desktop Memory Model F3-1866C9D-8GAB

 

[Mandrake4565]

And by all means set the ram voltage to at least 1.55V since that has been KEY for many users in ensuring stability. I also would have no issues with setting my own CPU_NB voltage to at least 1.25V or higher. AMDs own 'papers' for overclocking their FX processors use 1.35 to 1.4V and show it in their overclocking guides.

Agreed,

 

[celta]

thank you for your detailed reply RGone. I would not buy a 4 kit again, just statistics speak strongly against that approach and your recommendation is a good start.

I alredy learned a lot from the preceeding posts, especially about stable temps of the fx...

I was aware that the seidon was entry level, but regardless i assumed it was sufficient. further i assumed that i would handle some OC but taking into account your information i can forget about that. I built the whole system in a custom laser cutted plexiclass case and space was a serious consideration (wanted gaming machine, as flat as possible).
the seidon fitted perfectly and the inlet temperature to the cooler is certainly lower than in a usual case (draws air directly from enviroment, not only case-inside) playing to my favor...probably the NB, RAM and VRM cooling is better than in a usual case, because of the 200mm FAN directly blowing from enviroment to component at the cost of a higher general enviroment temperature.
but the seidon now gives me serious troube...any compact powerful AiOs? or one with longer tubeing to move the hole thing outside of the case?

but thats more smt for a casemodding forum not OC first of all i have to get it stable as it is.

RGone, here ist the second page where READ to READ is listed. any idea about the values?
I will try incresing the CPU NB voltage and RAM further, maybe even to 1.6V for RAM?
Can I find somewhere all the other specification to set manually for my RAM? tried to find them fith the model Nr you posted, no luck