IntelBurnTest - The new stress-testing program

[loonym]

Technically none of you guys are doing this right.

There are two ways to do this.

The first is to run with the smallest amount of ram possible. This places as much data into L2 as possible keeping the CPU working at as close to maximum as possible (small FFT's).

The second is to run with as much ram used as possible to create as much stress as possible on the memory controller, chipset, and memory.

Testing both ways is preferred.

This makes sense to me. Great explanation, thanks.

[WaterFlex]

Great test! Ty !

[JohnZS]

@JohnZS
As long as temps are below 70°C you really don't have to worry... Also note that actual games and programs never heat up the CPU as much so worry even less.
You can check that with some graph monitoring tool or by RealTemp's Max Temp values. Just reset them before playing some game or using intensive app.

Thanks RejZoR
Slightly off topic but do you think that IntelBurnTest will soon be the new standard in stress testing or would Prime95 8hr to 12hrs stability still remain the "Benchmark" for stability.
John

[Super Nade]

Wow! This runs about 10C hotter than anything else I have ever used! Kick-ass!

How do I interpret the results?

Q6700 @3.52 GHz
Windows XP Pro

[BeastNotro]

Ok this program gets Cores HOTTER then Orthos Prime.

I need to figure out why 2 of my cores always read 10C higher then the other 2.

Linpack 0-1 Cores 84C and 2-3 Cores 74C Max.

Orthos = 0-1 Cores 74C and 2-3 Cores 64C Max. (Were talking like 2-3 instances of 3 secs of a Max of 74C)

Here is my test results just done 6 min ago and all is OK Residual result numbers 5 passes all same.

Also ppl read the read me text ... only use what free memory is available to use.

[WaterFlex]

Super Nade, bad result. Your system is not stable.

[krille]

Essentially you just made a simple front-end for this [ Lets put your CPU into REAL test! Intel Linpack 64bit ] in order to fish for donations? Clever...

You should at least give Cronos proper credit in the first post. I doubt your efforts are worth a donation... I'm sorry, but to me it just comes across as astute and arrogant.

Either way, Linpack seems useful (as does your program), so thanks for that. It should at the very least automatically scan for errors (inconsistencies) though.

[Super Nade]

Super Nade, bad result. Your system is not stable.

What should I look at to determine stability? Which numbers are you looking at?

Now this may be an excellent way to stress out the PSU. Run this and FUR together...

[Kilyin]

What should I look at to determine stability? Which numbers are you looking at?

Now this may be an excellent way to stress out the PSU. Run this and FUR together...

Numbers in Residual (Norm) are supposed to match.

[WoZZeR999]

Program Creator, If you created a program that would automatically run 2 tests, 1 with low ram, and one with ~remainder free ram, and also check for errors, that would be donation worthy. As the program stands right now, I agree with krille.

[RejZoR]

Technically it could be done noob proof, autodetecting remaining memory, automaticaly doing both high and low memory utilization and automatically compare normals to see if they are the same and display it with nice "You're system is fully stable" text instead several lines of numbers and letters. Or simply a summary display of tested stuff and "Ok" next to it if passed and "Failed" if not...

[Bojamijams]

Okay I'm probably a noob and will probably be flamed but... explain to me the need for this program.

Because.. this thing will test your CPU higher then anything ever will and will heat it up higher then anything ever will (other then this program obviously).. so whatever you're going to be using, it will not heat up your CPU as much as this program will and your CPU will not be stressed as much as this program will make it. So whats the point?

If I'm stable for my games, my 3dmarks, my SuperPIs, and my folds, then.. in essence, am I not stable for my needs?

[WoZZeR999]

If your computer is stable enough, then you don't need to worry about it. This is more for servers where you need 100% data integrity, or you want a 100% stable system. Some people don't need 100% stability, some people want just another test. It's basically another tool similar to OCCT and prime, it just runs your system harder so you can find out if it's not stable faster.

[Bojamijams]

Well i guess my question is.. if you're stable under Prime but not stable under this... why heat up your CPU more by increasing voltage to get stable under this program when your system will NEVER be tested under this programs circumstances?

[WoZZeR999]

Small fft prime (what most people run) tests mainly the CPU and L2 cache. If you run prime in blend, that will test your memory controller as well. This will stress your memory controller and cpu, while also working it harder so you get a result faster.

[Bojamijams]

Well yeah you get the result faster, but like I said, under the kind of stress that it'll NEVER be put under.

[WoZZeR999]

If you have the test fail, then it could happen at any time. Again, this test is more designed for areas where absolute 100% stability and data integrity are necessary. For the rest of us, it's a new stability toy to play with. I don't care if my computer will never be pushed to it's limits for normal day use, but if I can prevent stability issues by quickly checking, I see no reason not to.

[i43]

thanks for the program AgentGOD

well , can i read this stable system ? . still a lil confused about the result readouts .. maybe some smarter people help me to ''decode'' this TIA
( was using 1024Mb for Ram .. ? half of total coze have 2Gb )
sys spec..
CPU: E8400 (@4005Mhz) /445X9 / Vcore set= 1.2500 // Aircooled noctua NH-C12P
RAM: G.Skill F3-10600CL8D-2GBH (DDR3-1333 PC3-10600 1024Mb x2 CL-8-8-8-21 1.65v / voltage set 1.70
MBO: Asus P5K3-Deluxe wifi/ap bios v.1014
GFX: XFX 9600GT (512MB)
PSU: Chieftec 650W (CFT-650-14CS)
( naked /caseless ( table ) setup ALL AIR COOLED

bios setup: vCore=1.2500 ; CPU PLL-1.50 ; FSB Term v=1.40 ; DRAM v=1.70 ; NB v=1.40 ;SB v=1.20 ; LLC=Auto ; CPU GTL=0.63 ; NB GTL=0.61

btw .. on same settings,
OCCP 2.0.0a 1 hour Passed
Prime95 (25.6) Blend test fail ( second core dropped ) ~0.5 Hour run (no crash)
Prime95 (25.6) Small FFT test 1.5 Hours - fail ( second core dropped ) no crash

for common usage - seemsto be stable tough ..
other readouts visible on screenshot

[SoLoR]

I need to figure out why 2 of my cores always read 10C higher then the other 2.

i have same issue on my q9450, cores 0&1 are allways around 8-10 degrees higher then 2&3, remouted heatsink 2-3 times, even lapped it once, it doesnt makes a difference, then someday during prime95 test i noticed, that each time when core 0 or 1 will fail, failed core will starts to read temperature from other core, so i think its bugged/stucked sensors and i stoped caring about temperatures on core 0&1 and i just take 2&3 as only ones valid under stress... problem is linpac sometimes on hot day can get core 0&1 on 95 degrees and that turns on thermal throttling but like i said that exceptional on very hot day, when i get temperatures inside case over 40 degrees, so i dont mind much...

[WoZZeR999]

Yeah, you failed. Judging by the fact that you failed blend and small FTT, I would say you need a but more voltage to the cpu. Playing around with the NB GTL may also help a bit.
For Prime:
Small FTT = CPU test. The CPU holds all of the data it needs in L2 Cache
Blend Mode = CPU and RAM/NB. The CPU data cache cannot hold all of the data that it needs, so it makes frequent calls to the memory for new data.

If you can pass Small FTT for hours, but blend fails quickly, then you need to play around with your memory system. Adjust NB GTL's, up NB core or loosen RAM. If you fail Small FTT quickly, you need to play around with your CPU GTL's, CPU voltages or lower your FSB/Multi.

[MarlboroMan]

works awesome under linux using wine... my lappie stayed as hot as hell

[WoZZeR999]

http://www.netlib.org/benchmark/hpl/ You can run linpack in linux without the use of a "not emulator".

[JohnZS]

This could be my PC being unstable....but if I run 7 tests using 3300MB of RAM (as I am on Vista x64 SP1) all values are identical (which is good as that indicates stability), however if I run, then a few seconds later open CPU-Z and RealTemp while the test is in progress my 1st 2 results are completely different. Is this a sign of instability? Or is it because the applications I am launching whilst the test is in progress are causing some sort of conflict.
Please note that if I run Real Temp and CPU-Z and leave them open I can complete 7 tests with no errors, however launching them during the test causes 2 results to differ.
John

[WoZZeR999]

I'm not 100% sure, but if it has to move around memory to load cpu-z and RT, that may cause problems with the run.

[BeastNotro]

i have same issue on my q9450, cores 0&1 are allways around 8-10 degrees higher then 2&3, remouted heatsink 2-3 times, even lapped it once, it doesnt makes a difference, then someday during prime95 test i noticed, that each time when core 0 or 1 will fail, failed core will starts to read temperature from other core, so i think its bugged/stucked sensors and i stoped caring about temperatures on core 0&1 and i just take 2&3 as only ones valid under stress... problem is linpac sometimes on hot day can get core 0&1 on 95 degrees and that turns on thermal throttling but like i said that exceptional on very hot day, when i get temperatures inside case over 40 degrees, so i dont mind much...

No i think the problem is the IHS is concave in that area and it is causing a Air bubble or a GAP there in which the heatsink is not making 100% contact.

The sensors are not stuck as if you use RealTemp you can Test the Sensors for movement in which my move np's.

I lapped my Ultra 120 Extreme 2 days ago now and it dropped temps 5-6C from previously when not lapped at Idle.

But under full load it dropped maybe 2-3C, but it is still MID 60's-65C-^8C on average on those 2 cores where as the other two are 8-10C less.

I think i may have to lap the CPU, but i hate to do so cause of warranty, but i not care in one way as i can afford another one if need to be, but it would suck if it did die.

Now cause the Thermalright Ultra 120 Extreme is LOOSE when installed this can cause a 4C-5C increase in those cores expecially if the IHS is concave in that area .. now add heat to the issue and that is another 4C-5C increase making the total overall 8-10C increase on those 2 cores alone.

I may try to use a dime or penny or a 1/8 washer ynder the retention bracket to try and get it to be a tighter fit and see how that works out .. with new Paste of course.

I using AS5 right now, but i am going to try the thermalright Chill Factor stuff as in many reviews it is about 2-3C less then AS5 and NO BREAK IN period.