Problem with IntelBurnTest

[courtney01]

I currently run windows xp pro sp1 32-bit version

I'm trying to run intelburntest, but it won't work for me. I chose no error detection, max stress (using max of 3014MB available memory), and 5 times to run the test.

It then says "You have tried to allocate more than 1.87GB of memory while in a 32-bit OS. Since it's not possible to do so, the allocation has been reduced to 1.87gb....[skipping some not useful data here]...no runs is allowed for lindata\x8\linpack32.exe: not enough memory"

 

[fubar569]

it should let you allocate up to 2000mb for burntest running a 32bit OS

instead of max stress, choose option 4 which is custom

it will allow you to input amount of memory...toss in 2000 and see what you get...

 

[courtney01]

it should let you allocate up to 2000mb for burntest running a 32bit OS

instead of max stress, choose option 4 which is custom

it will allow you to input amount of memory...toss in 2000 and see what you get...

Same problem

 

[fubar569]

icky...it said 1.87GB....try putting in 1700 or 1800 then...or try half stress (should allocate 1500mb or so based on your first post)...

i know mine allowed me to toss in 2000 when max stress gave me that error...and it works every time.

 

[courtney01]

icky...it said 1.87GB....try putting in 1700 or 1800 then...or try half stress (should allocate 1500mb or so based on your first post)...

i know mine allowed me to toss in 2000 when max stress gave me that error...and it works every time.

Thanks! It's working now. I can hear the fan speeding up. But realtemp isn't showing any values (when I put my mouse pointer over the realtemp window, there's an hourglass). Is that supposed to be the case?

 

[fubar569]

probably not...though i use coretemp and i make sure that TjMax is properly set for my chip...

so i really can't help ya with that one...lol

i also try to go for at least 10 passes with IBT...sometimes i'll do 500 or 1000 pass blocks to heat my room when it gets a bit chilly...don'tlaugh...it works...lol

 

[courtney01]

probably not...though i use coretemp and i make sure that TjMax is properly set for my chip...

so i really can't help ya with that one...lol

i also try to go for at least 10 passes with IBT...sometimes i'll do 500 or 1000 pass blocks to heat my room when it gets a bit chilly...don'tlaugh...it works...lol

Hmm, this thing ended fast. I did 6 runs and there weren't any problems. How many more do you recommend I do?

Is this the software you're referring to? http://www.alcpu.com/CoreTemp/

How do I setup TJ max properly? I don't know what TJ max is actually, apart from the clothing store

EDIT: I just noticed that the little icon for IBT says "intel core 2 duo". How do I test quad cores with IBT?

 

[fubar569]

it should load all 4 cores 100%...it's multithreaded...

yep....that's coretemp...

TjMax is part of how those software programs try to determine your actual temperature. To set it, you need to know the processor you have and it's stepping...CPUZ and other programs can tell you the stepping...

for instance...mine is a Q6600...G0 stepping...i found numerous references and was told that 90c TjMax is correct for this chip...you'll have to go into the options or settings for your temperature program and make sure TjMax is set to the correct value or your temp readings will be inaccurate...

 

[Casper312]

Hmm, this thing ended fast. I did 6 runs and there weren't any problems. How many more do you recommend I do?

Is this the software you're referring to? http://www.alcpu.com/CoreTemp/

How do I setup TJ max properly? I don't know what TJ max is actually, apart from the clothing store

EDIT: I just noticed that the little icon for IBT says "intel core 2 duo". How do I test quad cores with IBT?

It should load all 4 cores, open up task manager, if all 4 cores are up its fine.

For 24/7 usage, i do 20 runs. I also had the same problem with intelburntest when choosing max stress. Myself i do around 100 passes to ensure stability. Remember to also do some orthos/prime testing, to check if your ram/mobo is stable.

 

[unclewebb]

for instance...mine is a Q6600...G0 stepping...i found numerous references and was told that 90c TjMax is correct for this chip...

Don't be confused by Intel's press release last year. The numbers they released are called TJ Targets. For a Q6600 G0, their TJ Target is listed as 90C but the actual TJMax is usually higher. Most of the Q6600 CPUs I've seen have TJMax set to a minimum of 100C and some have TJMax set differently for each core. The actual TJMax for core 2 and core 3 might be as high as 105C for many Q6600 G0 processors and the TJMax for some cores on the 45nm Quads might be as high as 110C.

Intel has stated that TJMax is not a fixed number. Each processor varies during their manufacturing process. Intel has also stated that these sensors have slope error which means the sensor changes temperature at a different rate than the actual core temperature changes at. That error alone can amount to +/- 10% and then there's the problem that these sensors, especially the 45nm Core 2 sensors, can also get stuck at lower temperatures and not move at all.

Here's the latest version of RealTemp that lets you not only adjust for all of the issues these sensors have but it also lets you thoroughly test these sensors so you can see exactly what their temperature response curves look like from idle to full load.

It's also the only program currently available that reports core temperatures of Core i7 and Core 2 Quads in their correct physical order.

http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

For Nvidia users, it also includes GPU temperatures as well.

 

[Zero82z]

For Nvidia users, it also includes GPU temperatures as well.

Are you planning on implementing that for ATI cards as well?

 

[fubar569]

Don't be confused by Intel's press release last year. The numbers they released are called TJ Targets. For a Q6600 G0, their TJ Target is listed as 90C but the actual TJMax is usually higher. Most of the Q6600 CPUs I've seen have TJMax set to a minimum of 100C and some have TJMax set differently for each core. The actual TJMax for core 2 and core 3 might be as high as 105C for many Q6600 G0 processors and the TJMax for some cores on the 45nm Quads might be as high as 110C.

Intel has stated that TJMax is not a fixed number. Each processor varies during their manufacturing process. Intel has also stated that these sensors have slope error which means the sensor changes temperature at a different rate than the actual core temperature changes at. That error alone can amount to +/- 10% and then there's the problem that these sensors, especially the 45nm Core 2 sensors, can also get stuck at lower temperatures and not move at all.

Here's the latest version of RealTemp that lets you not only adjust for all of the issues these sensors have but it also lets you thoroughly test these sensors so you can see exactly what their temperature response curves look like from idle to full load.

It's also the only program currently available that reports core temperatures of Core i7 and Core 2 Quads in their correct physical order.

http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

For Nvidia users, it also includes GPU temperatures as well.

well then...

what is the correct TjMax setting? i take it there is no true way to find out? if so...how?

 

[unclewebb]

TJMax will forever be a guessing game. One user might swear that 95C is right for his processor while another user with the exact same model might swear that 100C is right. You know what? They might both be right.

rge and I over on XS came up with a testing procedure to try to see if a TJMax number is believable. He drilled a hole through the IHS of a few different processors he had on hand and mounted a Fluke thermocouple to get some numbers similar to the method that Intel recommends. He was then able to re-mount a few different heatsinks to gather some info and came to this conclusion:

To calibrate idle temps for core 2 duo: First set vcore 1.1V bios +/-~.03v and 6x333 mhz manually or leave speedstep enabled. Then set idle so it reads listed degrees above ambient for given cooling:

COOLING...............................IDLE DEGREES ABOVE AMBIENT
High end water...............................6C above ambient
High end air (true push/pull) ........6-7C
High end air (1fan).........................7C
Mid air (zalman 9500)....................8-9C depending on fan rpm
Intel stock cooler...........................10-11C

This was for an E8400. I did some further testing and found that at 6x266 for a Q6600, the difference in power consumption is only 4 watts greater at the wall so the difference in idle temperature can be at most 1C higher. With a Core 2 Dual Core chip with only 1MB of cache, the idle power consumption was 4 watts less than this baseline so the idle temperatures would be 1C less than this baseline.

For most any Core 2 based processor, Dual or Quad, 45nm or 65nm, you should see similar idle temperatures to the numbers in his chart give or take 1C. If your reported temperatures are way out of that range then you have to start doing some testing and try to determine how much of your problem is slope error related and how much is related to errors in TJMax.

I'll admit that this is not an exact science. The good news is that you don't have to be 100% accurate in TJMax to get very accurate core temperatures in the temperature range that most people operate their CPU at. If you check and calibrate your CPU at idle, your reported temperatures can be very accurate even with the far less than perfect sensors that Intel is using on their CPUs. Post a screen shot of a RealTemp CPU Cool Down Test and I'll try to show you what to look for when trying to understand what your sensors are telling you. Also give the low MHz / low core voltage test a try.

 

[Zero82z]

rge and I over on XS came up with a testing procedure to try to see if a TJMax number is believable. He drilled a hole through the IHS of a few different processors he had on hand and mounted a Fluke thermocouple to get some numbers similar to the method that Intel recommends. He was then able to re-mount a few different heatsinks to gather some info.

That would just give you the Tcase temperature, wouldn't it? It isn't exactly comparable to the temperatures within the cores themselves.

 

[unclewebb]

Yes, measuring at the geometric center gives you TCase. But measuring there and comparing to an IR thermometer and comparing to what the core sensors were reading gave him a lot of data to help him come up with his recommendations.

Here's a YouTube video which shows an example of some of his testing and shows the large gradient that exists between core temperature and TCase when running LinPack.

http://www.youtube.com/watch?v=q7ua2FFByfI

rge is the kind of guy that likes to let his CPUs breathe

 

[Zero82z]

Man, I would just about shit myself if I ever tried to drill a hole into the middle of my heatspreader .

I will watch the video shortly.