I bent over... and got an I7 920.

However, even at 1.25v and 3.9ghz, I'm running into ridiculous heat. I mean, on idle it's like ~50C or something
Click to expand...

I've said it before, I'll say it again. Prolimatech Megahalems one 120cm fan air cooled-I've got mine at 3.8ghz 1.36vCO cpu and it runs steady 27C on average idle and 41C under load in the summer and 20C in the winter in Atlanta. Chuck was concerned with high temps using the Zalman too, but he now has a Megahalems and runs 28C in the Summer in Atlanta also. It's the best air cooler you can get IMO especially when you OC.
 
41C under load??! Is that prime 8 threads small FFTs kind of load? Also, how fast is the fan spinning that's on the cooler... I don't want to put up with noise...
 
It's a scythe something or other that spins at 1600rpm and I never hear it-ever. 41C is with prime all 8 threads and 43C with intel burn test and that's the highest I've ever seen it get. I've double checked that with several different applications and even a physical temp probe and they all say the same exact thing. That cooler is a monster for heavy loads and overclocking.
 
I'm using the OCZ freeze, it seems to work better for me than anything I've tried even though I've read reports that say others do slightly better.
 
Dude I DO NOT have my shit up and running.

It's kinda ridiculous, but I even took a day off work to OC this shit (admittedly I stayed up really late the previous night) and it's STILL unstable.

However, I think today I made a big breakthrough as a result of a forum post by a person who's getting similar problems as mine, which seems to (so far) cure a specific problem I was having.

Hopefully it will work out - I'll make a detailed post about it if it's resolved.

I'm running Prime in the background right now hoping it won't crash :-D . Seems fine so far (15 mins+) so I have my fingers crossed... well not really (hard to type that way) but you know what I mean ;-) .

The "solution" is really, really, really, REALLY quirky and stupid and related to my specific mobo it seems like.

Kind of un-funny... I spent SO MUCH FUCKING TIME tweaking voltages and frequencies that I'm basically a fucking expert on the damn x58 platform by now... and now this little stupid thing might be the cure-all.

I'll let you know if it's stable. It's probably going to be by Sat. evening that I'm 100% convinced (gotta run a 12hr prime on in-place large FFTs and a 12hr memtest+ to be sure).
 
I think I finally got it stable. Will be good enough to play some 2142 tonight before I go to bed :-D .

I'll finish up some work around the case tomorrow.

fuuuuuuuu.png


I've tinkered with the memory a LOT and my conclusion is that it's just a piece of shit. It's supposed to work at 1600 8-8-8-24 with 1.65v. The weird thing is, I actually got it to pass memtest with that, but not prime. That just boggles my mind, and the only thing I could come up with is that maybe as a result of the CPU stress there is some voltage droop around the mobo or the QPI link or whatever. Increasing voltages didn't help though.

I think I was deceiving myself when I thought I was buying cheap performance memory. The kits in that range are all 9-9-9-24. I also wanted to run 1T on this memory, and people are recommending 2T, especially if unstable.

Fuck that. It's the last time that I'm buying OCZ Gold personally. I've had it with their shit. RAM prices are uncomfortably high right now, so I think I'll wait until they fall or I can find a good, proper kit (like Mushkin Redline) to be on sale/rebate or something.

I've got it stable at 1266 (so far), but it's shit and way under my expectations for the kit. At least it shouldn't affect gaming too much.

The mobo is not too impressive - the OC options are sparse (less memory timing tweaks, less voltage tweaks, less steps in voltage adjustment especially for VTT which is annoying), but for $110 or whatever I got it for, I'm not complaining. I haven't found anything wrong with it yet (knock on wood) so I'm keeping it. With a better memory kit, I think it will run far better (one that actually does what it's advertised to do).

The CPU seems to run very fine. At 1.216v for 3.8ghz, it's sitting pretty right now. Except those temps.

I thought my old Zalman cooler would cut it. I was wrong. It's little better over Intel's own shitty stock solution. I looked at the heatpipes and it only has 6 (whereas most "designed for i5/i7 cooling" heatsinks have 12...).

So, yeah - I'm heavily looking at the Corsair H50 actually. The idea of a pre-built water cooler somewhat appeals to me - quite a bit actually. The only concern I have is for loudness, and fan direction. Contrary to the instructions, when I buy one, I'll reverse the fan to pull air out and see how that works. I think it might be close if not just as good, hopefully. I'm searching for one for ~$60. Zipzoomfly sells the cheapest one I could find for $72, which is not too bad and if nothing better pops out in the next couple of weeks, I'll bend over.

I'm going to be looking at thermal compounds as well. AS5 might not be the best anymore from what I hear.
 
FFFFFFFFFFFUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUU.

There's something wrong, somewhere...

Maybe I'll try... again... tomorrow.
 
Soldier4Real said:
I'm running the OCZ Gold at 1600 MHz and have never had an issue. I'm at 8-8-8-24.
Click to expand...

Hey Soldier... what voltages is the OCZ Gold running at? Mine seems to be stable (for ~15 minutes at least so far) at fucking ~1.75 vDimm and ~1.41 VTT in the bios... which is a lot, is annoying, and is increasing my temps already before I even start overclocking the CPU (it's at stock right now - I'm just testing mem).

What's the nominal (in the bios) voltage for your OCZ Gold, and what's the real one (if you know it - from a readout or something) - and also, what's your VTT to run it at 1600mhz?

I think there's also an issue with my mobo drooping voltages so it might not actually be running at 1.75v/1.4VTT ... but I wouldn't know unless I bought tools to measure or something - it doesn't have sensors for those voltages... also, what's your VTT at (voltage)?
 
I'm not sure, I'll check it when I get home later. I'm pretty sure it's at the stock setting (1.65V). If you check my CPU-Z in my signature I was originally running 9-9-9-24 but have since tightened the timings up a little.
 
These modules are also ridiculously hot to the touch when prime is running. They feel like 80C+ (I need a thermometer)!!! That's kind of scary.

In related news, it looks like the best "traditional" thermal paste right now on the market is some Shin Etsu shit (edit: Apparently, it's from Phoenix, AZ lulz... just the name is Japanese?!). It's not even necessarily made for CPUs lulz.

Comparison tests:
http://vapor.skinneelabs.com/TIM/IndigoX/IndigoX.html

It's failry cheap so I think I'm getting some with the H50 when I get that...

I'll keep fiddling with this memory tomorrow :-/ . It's way late (again).
 
After reading some shit, I've decided to buy some Shin Etsu (turns out it's Japanese origin anyway, but "Microsi" is the US redistributor from what I understand) paste for reapplication to my current HSF and possibly the Corsair H50 if I don't like the temp drop (if any).

The cheapest place I found to buy the Shin-Etsu X23-7783D is here:

http://www.petrastechshop.com/shx2thco1gr.html

$10 shipped with First Class.

Note that there are different "shin etsu" pastes. Refer here: http://www.microsi.com/packaging/thermal_grease_x23-7783d.htm

The G765 is readily available everywhere (including on ebay for $6), but is arguably the "worst" of all of them - although it is still better than AS5 supposedly (on average).

Well, we'll see what happens when it comes. What I also like about it is that unlike the AS5, it doesn't have a "curing" period, so it should be good to go as soon as you apply it. It's also non-conductive and overall a better choice it seems.
 
I've decided to give up. I don't know what the fuck is going on, but my current config will have to do.

OK, so basically it's now been 5 fucking days where I've constantly tried and retried different settings to get prime stable on in-place large FFTs, and failed. This is not worth my time anymore, and since I believe it is because of the memory (or the Integrated Memory Controller on the i7 920 which would suck), I've bought a better 7-7-7-20 1600 Corsair XMS3 kit which I know is quality and will get me answers. I've had a shitty experience before with DDR OCZ Gold branded memory where again I had locking up / restarting / bsod problems while trying to run it at its rated speeds - and these problems went away by me buying Mushkin Redline (back in the day). I don't know if it's the combination of my mobo/mem, or if I just got a badly overclocking kit or what the fuck the problem is, but I'd like to try something else to be sure.

Basically, my cores are stable until like 3.8-3.9 ghz or so even at ~1.225 voltage. I've verified this by running prime small FFTs overnight already.

My uncore portion (the l3 cache/etc) of the CPU is also stable at 1.35 VTT - verified by dropping the memory multiplier and running prime in-place large FFTs overnight already at uncore 3200mhz.

Alright, now for the kicker. My memory checks out overnight on memtest86+ v4.0 (latest) at 1600mhz 8-8-8-24 1.65v (rated speed). However, with the VERY SAME settings, it makes the PC reset/freeze/bsod rather quickly at running prime95 in-place large FFTs. I've verified it is NOT the cpu vcore, because it's running at 2.6ghz.

The ONLY other thing which could be the culprit is the Uncore portion... but that is perfectly stable when the memory multiplier is dropped down to 2:6 (yielding 1200mhz DDR). The second I put the memory to 1600mhz or close to it, it fails Prime in-place large FFTs.

The second kicker: Prime blend runs fine overnight with the memory at 1600.

Basically, the whole experience makes me scream WHAT THE FUCK inside. The only guess that I have is that because of some memory incompatibility or voltage combination, the uncore and IMC on the i7 920 don't like running with each other at 1600/3200 and produce a freeze after ~15 mins or whatnot. I hope this is because of the memory, and something that the IMC doesn't like about it, or perhaps the voltage - which will be resolved when I get the Corsair XMS3 kit.

There are other people who have had problems running the OCZ Gold memory at its rated timings on my mobo, which makes me hope that different/better memory will make everything go away. I'll post when I get the thermal paste and the new kit.

P.S. My stable config is 200bclk/3200uncore(just for kicks and L1/2/3 cache speed, only need 2400 in reality)/1200mhz mem 8-8-8-24-88-12-6-6-6-32-1t (fuck my asshole).
Voltages are 1.35v vtt, 1.64v mem, 1.225v CPU, 1.88v CPU PLL,everything else stock. I'm just testing memory right now, and will most likely up the multiplier to 19x soon (it's at 13x right now to make sure vcore is stable for prime and any problems are only mem). It should be just fine at 3.8.

Hopefully this will be improved to 4.2ghz with a 21x multi (with the new thermal paste and possibly the corsair H50) and 1600mhz mem with tighter timings (with a better kit) which is what I wanted in the first place and thought my current components would deliver.

P.P.S. To top the shit off, I've run into some ABSOLUTELY RETARDED problem where the installer for visual c++ redist doesn't finish since yesterday, outlined here: http://social.answers.microsoft.com...s/thread/f4dd5033-9bcb-4788-b100-5e98bd80f6bf
Currently, there is no solution really - and I'm glad I discovered this is happening early, so I can quickly reinstall win7. It would be SO FUCKING annoying to discover it later. BTW this means that every game which has an automated VC installer (such as the BC2 beta which I wanted to try on the new CPU yesterday) fails to install and cannot be played. Win7 reinstall shouldn't take too long, luckily, though.
 
I had the same problems with my MSI X48 motherboard. It turns out all the problems were due to the bios, flaky ones, all three they released in over the year I had the thing.

I don't think I'll ever buy another MSI motherboard again. The board was great, the bios was for shit. I've had Asus and Gigabyte ones before and never had problems, but the EVGA I have now is by far the the best MB I've ever owned and that is due in large part to their bios releases which seem to get shit right. There are no unexplained WTF moments at any time and they've released 15 versions with various updates and tweaks since I bought this a year ago.

Anyway, I don't think I'd worry about it if you found a setting that works stick with it and be glad it does what it can. The corsair memory I think you'll like. That's what I'm running now, and it will do 6-7-6-19 at 1600mhz comfortably. I never noticed a difference in benchmarks between that and the default 7-7-7-20 so I've just left it at the recommended timings. Still, it's very reliable memory.
 
Yeah, that was one of the suggested things in one of the threads I found about someone who was having a similar problem. The just of it was "downgrade to 1.20 from 1.30 and it will be fine." Problem is, now the bios is up to 1.90, so downgrading to 1.20 is not really an option. I'm not willing to fuck around with bioses at this point anyway.

It is true though, looking at my bios and the options which are available on other people's bioses with boards like DFI, high-end Asus boards, and high-end Gigabyte boards (to name a few) - there are just options which are entirely missing on my mobo. They're just not there. The basic options are, but things like clock skews, advanced memory timings, and even some advanced voltage adjustments are just not present here.

I was thinking about the possibility that there is some memory timing which is "hardcoded" in my bios, and not adjustable, and just touching that one little timing/adjustment would give me stability. However, again I hope that a higher-end memory kit will be able to "handle" the more aggressive timing/whatever is not adjustable in my bios and run fine.

I knew what I was getting into by buying this board though, and if it doesn't work, whatever. At the very least I'll have a more stable memory that I know won't error out on me while writing a file (the worst possibility because it corrupts the data on the hard drive).

And yeah - the whole issue is a bit ridiculous actually now that I think about it. Memory latency/frequency matters in very isolated cases, and in most games, it just doesn't matter at all - hardly. It's a bit ridiculous that I'm spending quite a bit of money to get another 400mhz processor and 400mhz DDR Memory speed with a possible 1 cas improvement and that I spent 5 days of my life trying to make this one work. If I went to a psychologist, they'd probably find the issue pathological, but hell... I guess I'm just a geek/perfectionist/whatever.

Even if it only gives me .5fps, just not knowing that everything underneath the hood isn't "maxed out" is enough to make me unhappy. And then people like roth are perfectly happy running everything 100% stock...
 
OK... so this is pretty insane.

Basically, I was just restarting yesterday, and I heard this noise coming from the case (can't really explain what it sounded like, but sounded like something shorting out or a chip burning out). Then, I smelled this smell... can't really describe it, but I suppose it's a bit like burned plastic.

I shat my pants, and disassembled everything, but it was late in the night (or early in the morning if you will) and I just looked at the PSU, mem, and CPU before I went to bed. I didn't find anything visually wrong with them, and they didn't smell. So I turned to the mobo. It seems like the smell is coming from the CPU PWM area (to the left of the socket), so I'm guessing one of the voltage regulators burned out or something. I'll look at it more when I take out the mobo today.

Needless to say, the mobo is going back to newegg for a refund (- shipping and restocking fees though, but it's FAR better than nothing).

I've never in my life had a mobo burn out on me until now and I don't really know what to think about this... The PSU is a quality Corsair PSU, so I highly doubt that was the issue there. I was also not pushing any voltages which are extreme, or even outside of Intel's specs!!!

I can only hope that it was just a really bad, lemon mobo.

At least it burned out within newegg's refund period. It would've sucked some major balls if I just lost it.

Damn this shit. I had a lot of fun playing Shattered Horizon yesterday, as well as some L4D2 thrown in for good measure, and enjoying the performance.

I really, really, REALLY hope the burnout was the mobo and it didn't take anything else down with it. I'm gonna cry if I have to get a new CPU/mem/vid card/mobo... or have to RMA these ones.

So... I'm now looking at mobos again, and this time I think I'm going to look at the bios options for whatever I get in a lot of detail. And sending this piece of shit back to newegg.

This has been the most annoying, problematic upgrade that I've ever had.
 
Shit I hope not too. That's what happened to my MSI x-48. I was so fed up with it at that point I just went to the X-58 instead. I really liked the setup I had with the x48 too, but enough was enough.

If it is the MB, then it's definitely a good thing it's within the 30 day period. I know you probably don't want to consider it, but a better MB would be a good thing.

This one from EVGA is their lower end, but a damn fine board that runs flawlessly for $200.00 at Microcenter ant $239.00 at Newegg with a $20.00 mail in rebate.

http://www.microcenter.com/single_product_results.phtml?product_id=0312963

http://www.newegg.com/Product/Product.aspx?Item=N82E16813188049

I can tell you from personal experience that this MB will run 3.8 an all auto settings and 4.0ghz (before the turbo boost) with just a slight voltage increase and that is it! Anyway, good luck and I hope it's not a fried MB cause that just sucks.
 
I looked at that one, but instead opted for this one:
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128423

It looks like it was recently released (~1 month ago), I like the chipset cooling setup on it better, and it looks like it's already had a couple of bios updates, one with support for hexacore intels (already) and another with "improve memory capability" which is I'm guessing more options for memory adjustments. Also, I like the look of the power/phase setup on it a lot more than the EVGA.

I don't mind tinkering with it if it eventually runs right. The problem with this one that I had was that even after tinkering, it didn't. Hopefully like I say it was just bad luck.

I was on the edge of buying the DFI X58 UT one, but it seems like there's a problem with it and D0 processors, and that's why likely the board has kind of disappeared from retail lately. It's like a quiet discontinue. People say it runs very, very well with C0's though...

I also got it very cheaply at $10 off + $6.60 bing cashback from Superbiiz (former ewiz) = $194 total. Incidentally, that's exactly where I bought my last DFI Nforce 4 mobo 3 years ago, which has worked perfectly since then.

Well, I really really hope when it gets here, everything will finally start running flawlessly. I'm going to order a corsair H50 from zipzoomfly for $62 (they have a rebate going on right now)...

This upgrade has proven a lot more costly than I expected GRRRR.

Edit: Link to newegg edited, was wrong
 
I took out the old mobo - and indeed, the smell seems to be coming from the PWM area. It still smells, cold, and 24 hours later - it's like an acid-y, molten plastic-y smell. Visually, I can't see any burnouts though. It doesn't help that the mobo is black, but there was a small component (not sure what it's called, it's not a capacitor) which looked a bit black - but it might have been a soldering side-effect.

I'm now fairly sure that something in the PWM on the mobo simply decided to blow up.

I hope the new setup will start running flawlessly at first boot - something which I expected the current one to do, but instead I spent 5 days tweaking it for it to not work right in the end and then even blow up in my face.

With the corsair H50, I'm hoping for 4.2 with very tight memory timings at 1600mhz and why not even higher... as long as it's under 80C load :-D .

I can just imagine the person at newegg checking the RMA opening the package and being like "WTF" as the smell of burnt plastic/acid hits him/her in the face.
 
You must log in or register to reply here.
Back
Top