AMD Fx 4350 Weirdness with cooling and voltage.

[Xeratais]

Hello all. I am new here and I searched and have not found anything describing the issues I am experiencing 'or maybe my kid is just making me go insane.' But I am having a issue with my *cheap* Fx build.

So I currently have a ASUS M5A97 LE Rev._2.0 Motherboard and a AMD Fx 4350 that I just recently built and I can't seem to understand whats with the voltage scaling so high even when configuring the bios not to compensate. (I'd better describe that one but not at the computer atm)
Also when running above stock speeds the temperature spikes insanely high even with only a 500mv increase above stock voltage of 1.3200.
So with my voltage at 1.3700 with asus's load line junk (sorry) it usually sits in windows under load at about 1.4250 which still is not that high (ran my phenom 2 x4 965 BE at 1.5 for a very long time till my H50 busted.) So is there some way to stop this from happening. I had to down clock to stock speeds and undervolt to 1.2 volts to keep the temperature below 76c where as to before id easily hit 90+ and the system would shut down due to thermal tripping. I've even tried with a older amd stock heatsink from my 965BE. Also with load line calibration off system become unstable at 3.7 Ghz with the voltage set at 1.4500 volts and will immediately BSOD upon booting windows but will boot a Linux distro fine.

PS: forgive me if this has been posted by someone else. I couldn't exactly find the information I was looking for other than asus messing with stock CPU-NB voltages. I have that locked at 1.2500V

 

[ShrimpBrime]

You have a very obvious cooling issue.

Check that the heat sink is fastened tightly and the thermal paste is not compromised. It wouldn't hurt to remove the heat sink entirely and re-apply some TIM.

What cooler was it you're using?

 

[Xeratais]

Well both heatsinks were remounted both times. First one was the stock cooler that came with the cpu when I bought it. The second one was a older revision of the same cooler from AMD that came with my 965BE . The newer one that came with the FX cpu seems to be slightly larger but has a copper plate on the bottom that is relatively thinner but is slightly longer. The older one is slightly smaller but has more thickness. The older cooler is what I am using now as it seems to preform better. On both heatsinks I am not using the stock fans as they are noisy as **** and move little air. I have a 92 mm fan on it now that is thicker and runs at a lower rpm than the slim stock fans but also move more air. I have checked the mounting multiple times and have went through two tubes of TIM. first one was some pretty good stuff although generic *Micro Si* and now I am using cheap Fanner 470 or something like that in the meantime till I can afford a more beefy cooler setup. It is strange that it shows the same symptoms on both coolers although the older one seems to cool about 3-4 degrees better. still anything over 74 is a little too high IMHO thus why I am asking.

I can still check again however I doubt things will improve. <old amd heatsink with the AVC branding the Newer one don't have any markings and appears to be made by CoolerMaster

 

[trents]

What are you monitoring load temps with? Let me suggest AMD Overdrive. You should not be getting temps that high on less than stock voltages.

Does your motherboard have the latest bios?

What's your case ventilation like?

 

[Xeratais]

Well the plastic shrouding has been removed and the fan is actually attached with super glue. It preforms way better than the stock fan configuration. As for the case model Id haft to search that one up. It is made by Antec. It is a slightly outdated case/discontinued but has plenty of fan openings. one large fan on the top a rear exaust a side intake and two front intakes. Ram is samsung can't get specific details all i know is it is branded pacific sun and is running at stock speeds right now of 1333Mhz 8-7-7-28-31 and it is 2x2Gb (two dimms 2 gigs each) And my GPU is a EVGA Nvidia GTX 550 TI ~1000 Mhz Core 2000 Mhz Shader and the ram speeds on that is 2666Mhz (GDDR) not that the speed of the gpu is relevant to the specific issue at hand and a EVGA Nvidia 210 for my secondary display and as for the psu just a humble Corsair 750CX 750Watt since my Sparkle 650W blew up its main filter cap. I also have six hard drives. I know my western digital 1tb drive is a green drive but it runs relatively warm it's self with plently of cooling and the others are always around 28-32c depending on room temperature.

Ill try to get back to you on the case model

 

[Xeratais]

Its a Antec Two-Hundred V2 Mid tower

 

[trents]

I re-read your original posts and did some serious editing of my first post. Please check it again.

Did your Sparkle PSU blow while connected to the current motherboard? That can fry stuff, you know. I wonder if the motherboard was damaged.

 

[ShrimpBrime]

Low power P states
#1: 4100 MHz, 1.3V
#2: 3600 MHz, 1.2375V
#3: 2800 MHz, 1.0875V
#4: 2000 MHz, 0.95V
#5: 1400 MHz, 0.8375V

I had to down clock to stock speeds and undervolt to 1.2 volts to keep the temperature below 76c where as to before id easily hit 90+ and the system would shut down due to thermal tripping.

Right. I haven't read the processor's overclocked speed.

That heat Pipe cooler should easily cool the FX-4350 at stock. 1.320v is nothing.

Screen shots of Cpu-z and w/e monitoring software you're using would help tons. (wanna make sure this isn't VRM temps you're looking at)

Picture of the guts so we can look at your set up would be a 1000 words.

 

[caddi daddi]

I also think he's reading the cpu line and not the fx4350 line in hardware monitor.

 

[Xeratais]

Soo much info lol. I actually don't have internet. I am at my mother-inlaws house atm but I will get back with pictures asap. And no I almost positively sure it is not vrm temps.. I know this board only has a 4+1 power phase setup but it should easily slap the crap out of my cpu. I have acutally fingerd the vrms while running ADIA 64 stress test (CPU FPU Cache) and they are barely warm as in I can hold my finger on any of the Fets longer than 7 second and still not sting. And I have used overdrive to check temps but as fx processors don't report temps in the usual manner (more accurate around 40c) overdrive measures with a thermal margin. and it frequently turns red meaning overheat. I have read the technical documents directly from amd on that one because when I first built the system I was confused about reading 16 degrees while idle with hardware monitor. Amd stated max temp to be 90c so I ran stress test and waited till it said I had 10c degrees before turning red and used that as a base to set the temperature offset in speedfan so when overdrive says it is at 10 degree thermal margin it reads 80 and at a 57 degree thermal margin it reads around 32-36 degrees.
Ive noticed while sitting in the UEFI Bios it reads the temp at around 46 degrees with a little fluctuation here and there. and Linux pretty much shows the same as windows.

As for the power supply being hooked up when that happened, Yes it scared the **** out of me. Computer shut off and my lamp on my desk flickerd then pop fizzzz sizzle and carcinogenic gasses spew out. I think it was just a bad cap all in all as it was a decent branded filter cap from Matsu****a (Panasonic) and I fixed that psu and it still preforms admirably. Ps that cap was on the mains side right after the bridge rectifier and after the main chopper fet. (I do know a little about smps) and that is no worse then having a over voltage on the mains.

As to the P-States thanks I did not know the actual voltage setting.... Wish they had something like Phenom MSR tweaker ot K10Stat for this architecture.

Funny now that you mentioned it. I have all the open fan spots in use.. Two front intake and the side intake reversed to remove heat directly from the gpu as the PSU is only rated for 30c at max wattage (i seriously doubt im able to pull 750 watts at full load) and then I have a rear intake right behind the cpu to help with temps as setting it as a exhaust seems to make this thing run hotter. Don't know why as this seems counter intuitive, and a top exhaust as just that. this case dose not have a opening below the psu for it to pull air from outside the case so I could count that as a exhaust how ever ineffective that really is considering the extremely low speed that its fan (yate loon) runs at.

And I apologize for how much I type. I just want to go into detail if that helps any.

As for the comment on hardware monitor. I aint using it anymore and I don't see a socket temp if that is what your referring to. I see a Aux temp that never really gets above 55. Also the cpu heatsink barely gets warm unless I stop the fan then it will burn my finger. So that tells me the heatsink is doing it's job.
I am beginning to think its a issue like it was with intel. I am just scared to try and delid if that's the issue because I broke two cpu's doing that lol thank god they were old and slowww, but I have never herd of that issue with AMD.
and yes latest bios and not a beta.

 

[Xeratais]

inside the case Idle (cold boot and apartment was 56 inside aint been home all day. and the voltage settings. Sorry for all the pictures.
Also failed to mention I was having this issue before the psu blew its guts.

 

[trents]

If when you stop the fan the CPU heat sink gets hot to the touch then you know that the TIM between the die and the lid is working well to transfer the heat to the lid. I don't think delidding is the answer. And if the heatsink feels cool to the touch when the fan is running then that subsystem seems to be working as it should. It's beginning to look like there is something wrong with the CPU itself or the CPU die temp sensor.

Do you have another CPU you can stick in the socket?

 

[Xeratais]

No I really don't. My 965 BE took a crap when my H50 AIO watercooler decided it was going to bust it's seal. Lost a gpu motherboard and ram with the cpu on that one. Hence building the fx system.
It also ran the same outside of the case.... I will be replacing the heatsink with something more better like a H60 or H100 from corsair... But judging from my past experience that will only net me a 10c difference if at all.

Luckly it don't thermal limit anymore and I have never had throttling. The only throttling Ive ever experienced was my GPU going into safemode by limiting the clock to performance 2D state of 405 Mhz but that was when I was overclocking to find it's max and also when I had a cheaper power supply. (not the sparkle or the corsair.. had the sparkle putting out exactly 12.2 volts under full load the corsair hits 11.88 so I had to downclock my gpu a bit and that was all tested with a multi meter... I don't trust software readings....)

Maybe I could borrow my friends thermal gun and probe around lol. I know for a fact the NB (not the CPU IMC) gets fairly hot after about 10 minutes of just having the system on hence why I have a small fan on it but that don't report a temp.

 

[trents]

So if you have not had any thermal throttling in all this that suggest to me the temp readouts are not accurate because we know for sure you have significantly exceed TJmax which for those CPUs is around 70c if I'm not mistaken. According to the pics of AMDOD you were exceeding Tjmax by 6c.

 

[Xeratais]

Ok So Maybe the bios is not reading it correctly or something else like a bad thermal diode in the cpu... Ok would make sense considering like you said exceeding by 6 degrees.... I know for a fact after going over 2 degrees past tjmax that it would throttle but as I said it don't. So makes sense to me. Mybe I should borrow a thermal gun and scheck thatway to be sure.... I can easily get one for like 60 bucks So again thanks for all the help.

 

[trents]

On ebay you can get a thermal gun for a lot less than $60. Less than $20 probably.