Overclock.net banner
1 - 19 of 19 Posts

hammadj

· Registered
Joined
·
621 Posts
Discussion starter · #1 ·
Okay so while attempting to OC my two HIS IceQ 7950's, I ran across something weird. Core clock is at 1100, memory at 1350 and voltage at 1175. I ran heaven benchmark and was getting 120-150fps. about mid way through the test, my temps jumped form 60's to 511 degrees in afterburner and my computer froze and I had to do a hard reset. Another weird part is that the lower GPU had consistently higher temps than the upper one. After I did a hard reset, my GPU 1 usage is fluctuating constantly. I havent experienced anything like this before so Im confused. Oh, and furmark doesnt run at all even on stock clocks. Program crashes before running. Pics of all the anomalies below:

EDIT: heres my 3dmark score. :http://www.3dmark.com/3dm11/6222764
Doesnt seem too impressive...



 
Discussion starter · #3 ·
Quote:
Originally Posted by cam51037 View Post

Well I don't think that's an accurate reading, because your GPU *should* be fried if it hits that temperature.

Do the cards run well on stock settings?
Haha, i would probably be fried with it if it were actually that temp. On stock they are 100% fine. Even with a clock of 975 with 1100mv voltage it runs pretty fine. Im not quite happy with 975 though.
 
Perhaps your voltage is too high on that card? I would test one at a time and see if you can replicate the problem.

thumb.gif
***1337 POST***
thumb.gif
 
Quote:
Originally Posted by hammadj View Post

Is it a coincidence that it went up that high just as comp froze? At idle the temps seem accurate (top one is 10deg higher)
Yes most likely, but with any Over-volted GPU, theirs always that chance when the PC froze it lost the Vcore limiters and shot up to 3.3V and was really 511C for about 0.0001 seconds.
wink.gif
 
Quote:
Originally Posted by hammadj View Post

So what should I do about it? Just give up on OC my cards
frown.gif
This is Overclock.net...so NO! lol Just take it easy on the Volts and clocks and go from there
thumb.gif
 
Quote:
Originally Posted by rabidz7 View Post

My macbook pro and powermac report 250 C. It is just a bad temp sensor.
It's a MAC and a Laptop, both those are going against it. So it might be 250...F
smile.gif
 
I think you can 100% attribute those readings as false... the solder that holds the chip onto the PCB melts at about 250 degrees so the chip would have fallen off if the temps were actually that high!

Coincidently, the maximum number of a signed 10-bit number is 511 so most likely (in my opinion) the entire memory space got set to 11111111 etc just before it crashed and that's what AB interpreted the data as
 
Quote:
Originally Posted by pharcycle View Post

I think you can 100% attribute those readings as false... the solder that holds the chip onto the PCB melts at about 250 degrees so the chip would have fallen off if the temps were actually that high!

Coincidently, the maximum number of a signed 10-bit number is 511 so most likely (in my opinion) the entire memory space got set to 11111111 etc just before it crashed and that's what AB interpreted the data as
Although i had to check my book for signed 10-bit number but it is indeed correct, just a glitch in one of the downstream devices or a glitch in the sensor itself
 
Quote:
Originally Posted by DaveLT View Post

Although i had to check my book for signed 10-bit number but it is indeed correct, just a glitch in one of the downstream devices or a glitch in the sensor itself
lol how dare you doubt my maths!

Code:

Code:
max +ve signed n-bit number = 2^n -1
...also just realised this thread was started over a week ago... probably everyone already forgot and moved on... nevermind!
 
Quote:
Originally Posted by pharcycle View Post

lol how dare you doubt my maths!

Code:

Code:
max +ve signed n-bit number = 2^n -1
...also just realised this thread was started over a week ago... probably everyone already forgot and moved on... nevermind!
tongue.gif
Although I never really understood much about digital so i did rather badly for digital electronics. That's why i had to check my book
tongue.gif
 
Quote:
Originally Posted by HPE1000 View Post

Over 950f, impressive
thumb.gif


What will you be cooking tonight?
tongue.gif
From the looks of it, Tahiti XT
 
... *cough*... *cough*.... made a bit of an error in my formula above! the maximum positive signed 10-bit number is still 511 but the general form for an n-bit number is actually

Code:

Code:
2^(n-1) -1
I'll just be whistling nonchalantly in the corner now!

edit... and now I think about it, +511 decimal is '0111111111' in signed binary so I guess my suggestion the entire memory got written as 1's is somewhat suspect too... either way it's still a convenient binary pattern
 
Quote:
Originally Posted by pharcycle View Post

... *cough*... *cough*.... made a bit of an error in my formula above! the maximum positive signed 10-bit number is still 511 but the general form for an n-bit number is actually

Code:

Code:
2^(n-1) -1
I'll just be whistling nonchalantly in the corner now!

edit... and now I think about it, +511 decimal is '0111111111' in signed binary so I guess my suggestion the entire memory got written as 1's is somewhat suspect too... either way it's still a convenient binary pattern
I didn't reveal your mistake just now
biggrin.gif
 
1 - 19 of 19 Posts