Snograt's toasty PC take II

Remember that big ol' thread that pretty much introduced me to the delights of Tech corner?

It's still hot.

It's an eVGA 780i SLI (NOT FTW)

QX6700

4GB of Crucial Ballistix Tracer DDR2-800.

One GTX260-216

Two 500GB WDCs

One Asus DVD RW

Onboard Azalea.

Driving a Dell S2409W 24" monitor.

I've been severely told off by Rahja and others for allowing it to run that hot - but I'm out of options. It's in a Tt Armor+ with fans a-plenty and a monolithic Tuniq Tower 120 for the HSF.

Reckon I got a dud QX?

I think the memory is crap too - DDR2-800 at 4-4-4-12 only giving 5.3 on the (admittedly useless) Windows Experience Index.

Hmm, actually that doesn't look so bad...

Anyway, the reason for renewing this topic is that I've started to get BSODs regularly, in particular 0x1e and 0x109 - the latter of which sounds worrying - "Modification of system code or critical data structure"

Do you think that's an indication of the QX deteriorating due to overheating?

Oh, for the record - the temperatures at the top are IDLE - all four cores at 0-2C

I'm too poor to get a new proc

your vcore looks too high for me. it's stock vcore is 1.35V, making it go at 1.40V might be a little much.

btw, what speed are you running it at? the default is 2.66ghz (266.67mhz X 10).

Your computer is indeed way too hot.

I don't know that much, but isn't a Tt armor made for LCS systems? I have a Tt armor and I got an LCS inthere.

Also, you say you have a lot of fans. Check your cabling. Fans don't do a thing if cables are taking away your airflow.

If you can't find your answer here, I would suggest looking at overclocking forums. They have helped me out a lot!

I must be tired - I can't see the 1.40v vCore?
Core Voltage 1.350 V <- that one?

I'm running it at stock, moriz - 266.7x10 on the nose.

I'm running the memory slightly undervolted - its 2.2v default seems far too high to me.

My cabling is fine - only the one card and no PATAs makes for good airflow

Tt Armor+ are suitable for air or water cooled systems.

OK, let's eliminate the obvious. Did you put some sort of thermal compound between the cpu and the heatsink? Is the heatsink properly mounted on the cpu (not tilted and fully touching)?
It might be a good idea to remove the heatsink/fan (HSF) from the cpu and make sure all the surfaces are clean and you don't have anything (like dirt or grit) keeping the HSF from seating properly. Then, clean both surfaces, apply new thermal compound and reinstall the HSF.

That is fairly hot for a desktop, but those temperatures are no where near damaging (just look at how hot laptops get). I would advise pin pointing where the BSOD is coming from. It could be from your hardware or a buggy driver. Perhaps you can run different stress tests and see if they pull up any errors (memtest, prime, ect). If they don't then I'd update your drivers and see if they stop. The heat isn't a huge problem and I doubt it is the cause of the BSOD. I will concede though that those temperatures are hot for a desktop, but are within spec (Intel says 71 degree C).

What the exact name, model, etc of the processor, including speeds?

CPU Type QuadCore Intel Core 2 Extreme QX6700, 2666 MHz (10 x 267)

That's something I know I really should try next. Even in my huuuuuge Tt Armor+, getting a Tuniq Tower 120 in there was one heck of a squeeze - maybe it's not level after all. Oh, and it's got a nice layer of Céramique in there. Maybe I should go back to trusty ol' As5...

Even more detail, Tarun:
CPU Type QuadCore Intel Core 2 Extreme QX6700, 2666 MHz (10 x 267)
CPU Alias Kentsfield
CPU Stepping B3
Instruction Set x86, x86-64, MMX, SSE, SSE2, SSE3, SSSE3
Original Clock 2667 MHz
Min / Max CPU Multiplier 6x / 31x
Engineering Sample No
L1 Code Cache 32 KB per core
L1 Data Cache 32 KB per core
L2 Cache 2x 4 MB (On-Die, ASC, Full-Speed)

Oh, and the Tjmax on this proc is 100C but, on the whole, I think I'd rather not hit that

Oh, oh - as any eVGA 780i owner will tell you, their BIOSes are poo. Mine's the latest - P08 - and they seem to have given up tweaking it now. A board built with tri-SLI in mind and dual channel memory - yet they can't get it to run memory in EPP mode with all four slots populated, which is why I'm only running 2x2GB rather than the full 4x2GB that I own

I use the same board you do, just with a custom logic set to my specs, which would make my board run cooler, but otherwise the same. So... I can sympathize and directly assist with this I hope.

First off, ditch Ceramique. Ceramique is created for TEC and Phase Change (and LN2, not that it is common at all) cooling solutions. Silver can't hold off against those low temps, but silver is better for standard LCS and air based cooling solutions. AS5 = best with air cooling.

And yes, RAM won't run at EPP with all 4 slots populated. That is an unfortunate truth. 4GBs is plenty anyways.

As for the CPU Snoggy, I think you need to try to replace it. I think you just might have a QX with a lot of gate leaks, maybe a fault pMOS? Hard telling without doing lab testing on it, but my hunch is, the processor is faulty on some level of construction.

In addition Snoggy, your RAM is running at way too high a voltage. Bring it down to 2.0v, no higher. Only the Ballistix 1066 RAM runs at 2.1-2.2v (mine runs at 2.2v and that is with a subtle overclock even)

I also am apt to believe your GPUs are running hot because of the hot air the processor is putting off? Use Everest and grab temps from the cards. I would like to peek at your vBIOS code to peek at your delta clock values, voltages, and memory timing tables, as well as to check it for code integrity issues too. I can always grab a skeleton vBIOS from our network and fill it in with your manufacturers codes and settings, with my own custom uncore settings (I can't mask core changes). I will be able to mask the vBIOS and flash process using my dev tools, so it won't void the warranty either if we need to flash it over.

Let's fix this issue once and for all.

Thanks for the assist, Rah.

Tell me exactly what data I should pull from Everest and I'll bung it on here (using some arcane way around the unusual attachment size limitations)

I was pretty concerned with those RAM voltages too - the thing is, that's what the BIOS automatically sets them at - here's the SPD:

Snog, your temps are fine. The max for your CPU is 64.5° C.

Bear in mind Tarun, the temps he listed are idle temps. His load temps are what I am worried about... I honestly think the CPU needs to be RMAed.

RMAed - the company I bought it from went bust

Hmm, time for a load test, methinks...
__________________________________________________ ________
Interesting - if nothing else, my temperatures are very stable.

I just went and vanquished Ferndale (a hobby of mine >.<) and the results were as follows:

Idle: 49;48;42;47
Load: 59;58;52;57

That's GW with everything on full, 1920x1080, 4xAA

An exact, stable rise of 10C.

GW is hardly a "load test" for a multi core cpu. You need something that can devour all the cores if you really want to call it a "load test". But then again, at how high those temps already are, i don't think it could handle a real 100% load. As i expect you'd probly get way too hot.

I wouldn't dare run a proper load test, Blackhearted. SuperPi would eat my computer.

That shouldn't be happening Snog... SuperPi is demanding, but on a system without issues, it should be fine. You need a new processor, first and foremost (even if it isn't the real cause of this issue, it is still not in great shape)

In addition, you need to test your memory (using Prime95 is a great way)

What settings? Torture test/blend?



[Edit] No, not torture test. Almost immediate BSOD with 0x7e and it took me several panicked reboots before I realised that it had made itself run on startup. I was getting PFN_LIST_CORRUPT, several 0x7e and, on one memorable occasion, a blue screen with the error 0x0000000000000001.

A 16-bit error code?

PFN_LIST_CORRUPT refers to memory (RAM) if I recall correctly.

Damaged DIMMs or damaged northbridge (specifically memory controller)

Step 1: Try other RAM if you have some around

Step 2: RMA your current RAM with Crucial

Step 3: Pray

Step 4: Replace the motherboard via RMA... ewwww

Hmm, one good thing. It's Crucial RAM and an eVGA motherboard - both apparently good companies when it comes to returns.

Logic (and hope) tells me it's the RAM, though. I think I was having problems with the old board (a 680i SLI) but that's behaved itself perfectly since I dumped it on my son.

at least RAM is fairly cheap (compared to a new mobo) I like eVGA's customer support, never had to deal with Crucial.... best of luck Snogzy!

Gives me the option of getting some 1066 I suppose. Minimal gains (if any) but likely to be happier to clock. Probably look to G.Skill or others.

http://www.evga.com/forums/tm.asp?m=213483&mpage=1&key=

did this to my 780i some time ago, and it helped with overall high temps. however I do have the replacement fan set to blow, not suck.

as far as crucial mem goes, my faith is starting to wane. had 5 sets of ballistix tracer 1066 burn up..no matter the clock/volt(s).