I'm a LOSERclocker

[Cloot]

This is kind of a long story, but I'll make this short as I can. I 've probably built 10 or so systems in the last year all working to this day all AMD based (we all know why). I recently upgraded to an A7N8X Dlx couple o' 256 MB dual channel sticks of 3200 OCZ and a 2800+ (cost me about $800 CDN and some change from NCIX) few months ago. When I got the stuff I was currently in transition between apartments and I was stayin at a friends. So I spent the afternoon guttin my old mobo etc. from my case (A7A266 1800+ @ 140 x 12.5 (board wouldn't let me go past 12.5 and never posted past 140 fsb and any multiplier) ) usin a volcano 7+ with the sensor stuffed right close to the base of the heatsink... oh sorry I'm goin on... anyways so back to my guttin show. I proceeded to install my new stuff and took care in all the ways necessary to avoid some cataclysmic disaster like shortin her out somehow. Upon initial boot the power seemed to have to build up as the boards (A7N8X) LED would blink and the cpu fan and other fans fluxuated between the on and off states for about 10 to 15 seconds until the board finally beeped, however it would not run past the logo screen. I'm tryin to make this short... I had never seen this in 8 yrs, power goin on off on off on on off on on on off ON! you get the idea, I jumped on the power supply as my first perp but it's fan worked fine and the rest of the board just didn't get power... or something. For some reason right at that time my old (300W) PSU decided to retire I guess. Last thing I guy would go after in the process of elimintation when a brand new system is started for first time and it won't boot, even after it's tourettes like power up... a PSU with a functioning history be the last suspect no?. I'm doin all this on my friends kitchen table in his small apartment... anyways there is a point to all this. I made absolutely sure there was nothing contacting the PCB that wasn't supposed to whilst I reseated the ram, installed a new expensive PSU and reseated the CPU and HSF. Right there is where I make my first mistake (complacency) I wouldn't find out till later (little less than 2 months later.) Since then I settled into a place and began usin my computer as normal playin games mostly, readin this and that on the web. I'm the type of guy to have stuff like WCPUID and Regclean and just apps or utils that allow a guy more control over his computer than the average "computer store computer" buyin joe, I'm pretty aware of what I can do with my rig. Mistake number two (ignorance) was not knowing that the ASUS Probe utility does not actually give you the temp of your CPU but the ZIF socket it sits in. So I was getting a 21 deg C mobo and 43 deg C CPU reading frome ASUS Probe 2.19. Great everything is perfect I thought... just right. Months pass. Upon doin some readin I happened upon a post from a guy and I ended up at MBM 5 site. I though wtf I'll check her out, besides I just updated to latest bios. MBM 5 is pretty comprehensive and I was, and still am pretty impressed at it's overall functionality, everyone should know this stuff, you got big dials in your car for temps and whatnot, same thing pretty much no? What's a 70 yrs old lady gonna do when a dial on her dash is in the red? pop the hood? not my gramma. Point is temp dials are there for a reason, and it would serve us all to know if the rig is runnin hot for some reason or other... (*cloot <----- SLAPS ASUS*) Anyways so I configured MBM 5 so that the CPU temp guage was actually reading temps from the diode on chip. Sit down if your readin this while makin your health drink, mobo was 23 deg C, CPU was 76 deg C, 76 blistering degrees at idle... not f'n kidding at all. I was obviously shocked to say the least. When I build a new system I try to familiarize myself with the bios and in my own rig's case I was well aware of what I was doing in the bios. When I initially built my system I immediatley changed my stock barton 2800+ 333 FSB to run at 400 fsb as I know it's quite capable of this even with stock cooling... usually, (ram was at 100% or same spd as fsb). Now I was at 76 degrees celcius at idle with my head in my hends. I made triple sure that MBM 5 was reading correctly as I made my brother an identical system and MBM 5 showed a 50 deg C idle and 58 deg C load... warm but ok. So back to my machine, I said heck with it and ran prime 95 torture test with MBM 5 runnin and the temp soared to 84 deg C yes that's not a typo 84 degrees... hold... hahahahahahahahaha arrrrgggh. Keep in mind for nearly two months I had been playin games like Battlefield and Unreal 2 and MOH etc. games that can strap a DX8 system on an oc'ed board oc'ed vid card and oc'ed ram and 2800+ doin 2.2, CPU must a hovered aound that temp for hours and hours at a time. I let my rig cool down as much as possible before I shut it down and waited about 30 mins before crackin the case the sink was body temp by then. When I removed the Volcano 7(also oc'ed) I saw that the Arctic Silver I used seemed cooked kinda gooey and pasty. I looked on the bottom of the heatsink and I saw that the chip had actually been sitting way off to one side of the copper core on the sink. Yes, way back at my buddies place during all the confusion (my current excuse) I had put my heatsink on backwards so that the bottom of the Volcano 7 which has two elevations, the very bottom of course for the CPU and the slightly higher elevation made room for the housing for the ZIF socket retention mechanism. Yes my heatsink was probably barely touching the Arctic Silver and barely touching the die. I don't know what to say... I just don't know. Anyways this is really a testament to AMD, they make very efficient and very tough CPU's because I have since properly mounted the HSF and now my idle temp is 53 and load is 58 (room tmp is 23 deg C) it's been totally rock solid and I mean it, dead stable for the last couple months since my fiasco. I just can't belive I did that... and the friggin thing is workin great right now at 218 FSB x 10.5 runs any bench and wastes my brothers identical system. Can you belive that... I have to, I did it, f'n moron that I am or was. Anyways hats off to AMD. And if you read this whole thing you poor SOB sorry I had to taint your wonderful day with my sad a$$ antics. Learn from us dummies. That CPU will run hot... HOT

cloot

 

[shiyan]

Paragraphs, PLEASE!

we don't mind the long posts so much, but please use paragraphs!

 

[Steven4563]

thats alot off words lol

 

[stui_nz]

Nice story Cloot!

So what was wrong with ASUS probe - did you ever get it working properly?

Heatsinks sure do suck when you put them on backwards...

and Welcome to the Forums!!1

 

[Tyberius]

Ummm, I cant believe I read that whole thing, my brain hurts! Take your ritalin boy! I'll give an abridged version....
I built a system, something was wierd, kept turning on and off, my PS dies. Found out months later my temps were really high, after playing many games, opened case and found heatsink was on backwards. Wow, amd CPUS can operate at nearly the suns surface temperature and still function. Props to AMD...Yeah!

 

[Heatsinkguy]

What if you were doing word proccessing and it suddenly shut down?
Well with my intel i can just take of my heatsink and it will underclock to 200mhz and still giveme time to save my documents. props to Intel....Yeah!!!

 

[Steven4563]

thats a nice thing to be able to do take your heaysink off whos stupid enuff to do that apart that idiot on the reviews that time

 

[Heatsinkguy]

Tom is not a idiot...he is the greatest person ever..NOT
But he was showing what would happen if one were to take the heatsinks off.

 

[shiyan]

I just finished reading the entire story.

Quite entertaining actually. Thanks for that.

just remember to use paragraphs the next time.

nice overclock actually. nearly 2.3ghz is quite decent.

 

[onx]

yeah wow 84C but AMD rates its 2500+ to 85C so that musta been the difference

hehe jk

yeah welcome to the fourms Cloot!

 

[Foxie3a]

I used my curser to help me keep track of where I was in my reading..lol

WELCOME TO THE FORUMS

 

[Deliximus]

props to amd to put every chip saving feature knwon onto k8!! YEAH

84C, eh? hahahha haevn't beat my 103C yet on my old tbird!! alive!!

 

[Steven4563]

ive got a good idea

lets get some hardware and lets see how hot u can run it with it still being alive

i have run my t-bred B how it is in my sig with just the heatsink on no fan only into the bios didnt let it get to high 68C was high enuff for me

 

[shiyan]

my record was a 2.26b at 95~100C for 2 days...

it still worked.

 

[Cloot]

oh ya paragraphs and transistor counts

I'm curious about something... I was readin that there is some 58 million transistors (Barton), which are basically on and off switches that can switch between being on and off 2.2 billion times a sec... that's pretty unreal, so prove it, really... is there some way off actually being able to tell if two identical Bartons are actually using the the exact same amount of transistors.

Is it not concieveable that one Barton actually has a couple more useable tranny's (transistors) than the other. Besides benchmarking the two, is there and electron microscope capable of slowing down a recording to the billionths of a sec then watching the tranny's actually move? I've never heard of such a thing, so I guess my question is how the heck does intel know they have 107 million tranny's on a P4 and not 105. Or that all 107 million are working, not say only 105 million working tranny's.

Some of you will be laughing at this, but hopefully I've able to make you see what I'm getting at. How do we really know what we got here? I'm not sayin I'm not happy with my rig I'm ecstatic about it's performance but I'm not the type of person to just believe what people say even if they are waaaay smarter and work for a multi billion dollar semiconductor giant like intel. I want a bench that tells me how many tranny's I have and how many work flawlessly. Probably askin for a million dollar piece of software. I don't know... but I want to.

 

[Cloot]

by the way Asus probe works great... if you wanna know your ZIF socket temp/ temp on non die side of CPU... appearantly ASUS had some stability problems when trying to access the on die temp sensor so they left that out of all versions of the probe

 

[Zuzzz]

I think the moral of this story is to NEVER play with computer parts at you 'friends' house. Your friends might think its neat but you could have toasted a chip. GLAD it didnt go nuke

Z

 

[Cloot]

somebody stop me... please
also I was readin about chip architecture and whatnot and I was particularily interested in one part about voltages and "gate leakage" which seems to indicate that a faster clock speed doesn't a faster CPU make. Theoretically a 2.4 GHz chip is faster than 1.8, but because you may have to up the voltage in order to achieve a 600MHz overclock you are effectively pushing more voltage through the tranny and causing the dielectric to have to deal with trying to resist more voltage in an off state, sometimes causing a semi off or leaky state in that tranny and thus casuing errors.
Which means that the instructions have to be run again on that bunch of bits. Theoretically you may have cause your sucsessfully completed operations/sec to drop. Theoretically. The reason I say all this is because intel and AMD openly discuss this problem with voltage, so concievably they already have found the perfect coefficient between clockspeed and Sucsessfully Completed Operations/sec. Just not for every single individual proc only it's gen. as a whole. Or the avg. of a batch. Does SANDRA help us here? We may have high mem bandwidth high CPU spd but if the arithmetic and Multi media bench don't score high could this be an indication of what I'm referring to? AMD rep said once and I quote "we sell CPU's based on performance not clockspeed."

So it would seem to me we're already gettin a CPU that has the best ratio of clkspd/SCO per sec. except for a few anomalous CPU's of exceptional quality. Is more GHz faster? look at intels 3.2 GHz with 800 MHz fsb compared to AMD's 32 bit 3200+ a 400 MHz fsb and a 1 GHz CPU handicap and it's the 3.2's direct competitor... that's a huge theoretical difference there... ok I'll shut up for a while now

 

[bafbrian]

I good to see you don't it's a short story.

 

[Gnufsh]

Re: oh ya paragraphs and transistor counts

I'm curious about something... I was readin that there is some 58 million transistors (Barton), which are basically on and off switches that can switch between being on and off 2.2 billion times a sec... that's pretty unreal, so prove it, really... is there some way off actually being able to tell if two identical Bartons are actually using the the exact same amount of transistors.

Is it not concieveable that one Barton actually has a couple more useable tranny's (transistors) than the other. Besides benchmarking the two, is there and electron microscope capable of slowing down a recording to the billionths of a sec then watching the tranny's actually move? I've never heard of such a thing, so I guess my question is how the heck does intel know they have 107 million tranny's on a P4 and not 105. Or that all 107 million are working, not say only 105 million working tranny's.

Some of you will be laughing at this, but hopefully I've able to make you see what I'm getting at. How do we really know what we got here? I'm not sayin I'm not happy with my rig I'm ecstatic about it's performance but I'm not the type of person to just believe what people say even if they are waaaay smarter and work for a multi billion dollar semiconductor giant like intel. I want a bench that tells me how many tranny's I have and how many work flawlessly. Probably askin for a million dollar piece of software. I don't know... but I want to.

I think that if 2 million transisters stopped working, you'd have one dead CPU. Unless they all had to do with the L2 cache, you could disable that and have a castrated CPU. Transisters in a CPU aren't like brain cells, they can't make new connections between each other. Each has a specific function. They are arranged into logic gates, which are layed out to form the different functional units of the processor. If a transister stops working, it's function will no longer be perfored and you will start getting errors. Not that transister count means all that much. The P4 has more trransisters, but the Barton does more work per clock cycle on average, and it's smaller size makes it cheaper to produce.