IntelBurnTest - The new stress-testing program

[Singh400]

Managed to BSOD my PC with this. Left my voltage [a little] too high, and the temps went straight into the 70s. 10-20 seconds later, PC BSOD'ed and I had to reset everything.

This program ROCKS.

[vannguyen0]

so can you only claim stability by passing 100% of the passes using the most extreme test (using all memory)? If I can pass 20 passes using 1/2 the memory, but can only pass 50% of the passes using all of the memory, can I not claim my system as being 100% stable?

[Singh400]

so can you only claim stability by passing 100% of the passes using the most extreme test (using all memory)? If I can pass 20 passes using 1/2 the memory, but can only pass 50% of the passes using all of the memory, can I not claim my system as being 100% stable?

Yeah, that's my view aswell. I believe 10x with all memory is fine to test system stability. Some say you need 20x, some say 50x and some say 100x. I personally do 10x's.

[stealthbomber]

Read my above test, try >24 hours for good measure, my max so far without a single error is just over 15 hours before the first calculation error occurs!

Edit:

Now my system is capping out with a calculation error within 30 mins of running IBT, going to max out the NB voltage @1.7Xv and try again, think this POS nVidia board (my last ever nv board) is gonna fail anytime soon.

[mikeyakame]

Very interesting. Thanks for sharing this. So...changing the PCI-E frequency will change a strap? I'm not very knowlegable about this frequency. I will definataly give this a try though. I am still searching for the answer as to why I am BurnTest and Prime stable at 400 MHz but become Prime unstable with only a small increase in FSB to 425 MHz. (This is only 25 MHz above default on the FSB). No one else seems to have had any solid advice regarding the IBT/Prime discrepancy that several others have been experiencing as well.

It doesn't change a strap, what it may do though is create a situation unknowingly where the phase locked loop oscillator frequency adjusts enough to reduce jitter in certain scenarios or interference at certain frequencies. I've noticed this same phenomenon on my current Nvidia GTX280, certain frequencies alter stability. Assumed is that there is some kind of jitter or external interference that doesn't interfere or dirty the signal so much.

[Amurtigress]

Very interesting. Thanks for sharing this. So...changing the PCI-E frequency will change a strap? I'm not very knowlegable about this frequency. I will definataly give this a try though. I am still searching for the answer as to why I am BurnTest and Prime stable at 400 MHz but become Prime unstable with only a small increase in FSB to 425 MHz. (This is only 25 MHz above default on the FSB). No one else seems to have had any solid advice regarding the IBT/Prime discrepancy that several others have been experiencing as well.

It is just a guess that it could be an internal clock divider or something else the BIOS does when going away from the default clock. I wish I knew more about the internals, myself.

<edit> Erm, I just saw mikeyakame's post on this... </edit>

My first 7:20h run was done at FSB 433, last night I cranked my Q9300 up to FSB 460 with no voltage changes, and it ran stable for 7:30h again. At around 470 my board bails out with a FSB wall, but I knew about it before I bought it.

Currently running my Q9300 at 3455 MHz, and happily stable!

I'm looking forward to your results. Frankly, having such a CPU and hitting a wall at 425 is nasty. But the X38 chipset is known to not overclock well on 45 nm Quads. Hence my change to the P5Q3 board....I had the Maximus Formula myself. It crapped out at 436.

[Razmatazz]

Amurtigress, you had inquired about default memory settings. Here are mine.

CAS Latancy: 7
RAS to CAS delay: 7
RAS Pre Time: 7
RAS ACT Time: 20
RAS to RAS Delay: 4
Ref Cycle Time: 60
Write Recovery Time: 10
Read to Pre Time: 5

BTW, I tried increasing the PCI-E frequency by 1 as you had. My system became unstable and I couldn't get into vista at a FSB of 425 MHz. If I leave the PCI-E frequency at 100 MHz, I can do most everything at FSB 425 MHz . I am BurnTest stable but again run into Prime95 errors inside of a minute. I'll try playing again with some settings later this weekend.

[Bail_w]

Hmmm... the temps on my system does raise to about the 70s - but I do not experience any blue screens, random reboots, or system hangs. It finishes what it does and tells me that it failed 80% of the passes.

Well, i am suspecting my motherboard Mosfet or Northbridge is overheating. So i might upgrade to P45 to see if i can bring this sucker to 4.0ghz stable.

[stealthbomber]

I can't seem to get my rig stable past 16 hours even with such a low overclock, think the POS 650i board with it's 3 phase power supply just doesn't cut it, time for a new P45 board.

[Amurtigress]

Amurtigress, you had inquired about default memory settings. Here are mine.

CAS Latancy: 7
RAS to CAS delay: 7
RAS Pre Time: 7
RAS ACT Time: 20
RAS to RAS Delay: 4
Ref Cycle Time: 60
Write Recovery Time: 10
Read to Pre Time: 5

BTW, I tried increasing the PCI-E frequency by 1 as you had. My system became unstable and I couldn't get into vista at a FSB of 425 MHz. If I leave the PCI-E frequency at 100 MHz, I can do most everything at FSB 425 MHz . I am BurnTest stable but again run into Prime95 errors inside of a minute. I'll try playing again with some settings later this weekend.

Hi,

have you tried setting RAS to RAS delay at 8-10 yet? The 4 and 5 as defaults for this value seem to be ancient and wrong setttings that apply more for DDR2 RAM than DDR3. OCZ for example recommends 10. DDR3 RAMs cannot be that different from each other to seriously require such drastically different tmings....

As for PCI-E at 101, that's a pity. But it appears that the PCI-E Freq. is worth tampering with...I wonder if other values than 101 and 100 can bring stability?

[Singh400]

How vital is the PCI-E freq to a stable system? How does it relate to other settings?

[Amurtigress]

How vital is the PCI-E freq to a stable system? How does it relate to other settings?

Well, hard to tell. For some it doesn't change a thing, for me it meant utter stability. I believe it depends a lot on how much you're stressing your northbridge. Main factors are:

1) FSB clock
2) Number of memory modules, and CLvalue. CL9 modules stress more than CL7.
3) Type of CPU. Dual core vs. quadcore.

The stress factor can be seen indirectly by the required voltage on some chipsets. My X38 always needed 0.04V more when using four modules.

Also it can depend on wheter you're using a older Core 2 CPU (65 nm) or a newer 45 nm core. They just behave differently on the frontside bus.

[drminer]

The 32-bit version of Intelburn cannot be relied on to prove stability over the short number of passes many people are currently using. For example, I ran 100 passes of Intelburn (2 hours), using maximum memory, and it displayed an error only on pass 99! It doesn't seem like Intelburn is much of a time-saver for 32-bit systems, at least. Hours of stress testing is still required, it seems. Oh, well!

[Singh400]

Quick question, can it be added so you are able to run IBT for a length time. So aswell as passes, say run for 60 mins or 120 mins?

Well, hard to tell. For some it doesn't change a thing, for me it meant utter stability. I believe it depends a lot on how much you're stressing your northbridge. Main factors are:

1) FSB clock
2) Number of memory modules, and CLvalue. CL9 modules stress more than CL7.
3) Type of CPU. Dual core vs. quadcore.

The stress factor can be seen indirectly by the required voltage on some chipsets. My X38 always needed 0.04V more when using four modules.

Also it can depend on wheter you're using a older Core 2 CPU (65 nm) or a newer 45 nm core. They just behave differently on the frontside bus.

Ahhh I see, so you can leave this setting at default - and only try different values if no other values make a difference. Sort of last resort?

[Bandit_44]

Hey Guys,I had a Question asked today I really couldn't answer! A customer asked me ,Just how much he was taking out of the Life of his CPU (E8400) by having it 20% Overclocked! His temps max out aT 52 Gaming and is IBT and P95 Stable! How would you answer that? His IBT ran 120 times and was Stable all the way.Thanks!

[stealthbomber]

His system will be absolete and replaced well before the CPU fails.

[Razmatazz]

Hi,

have you tried setting RAS to RAS delay at 8-10 yet? The 4 and 5 as defaults for this value seem to be ancient and wrong setttings that apply more for DDR2 RAM than DDR3. OCZ for example recommends 10. DDR3 RAMs cannot be that different from each other to seriously require such drastically different tmings....

As for PCI-E at 101, that's a pity. But it appears that the PCI-E Freq. is worth tampering with...I wonder if other values than 101 and 100 can bring stability?

Haven't had a chance yet to play with the "other" memory timings. You mention that OCZ recommends a RAS to RAS delay of 10. I tried searching their site but couldn't find information regarding adjusting the additional, not usually listed, DDR3 settings. If you could point/link me to their (or others) info regarding these settings, that would be great.

[Deathspine]

Haven't had a chance yet to play with the "other" memory timings. You mention that OCZ recommends a RAS to RAS delay of 10. I tried searching their site but couldn't find information regarding adjusting the additional, not usually listed, DDR3 settings. If you could point/link me to their (or others) info regarding these settings, that would be great.

Anandtech had a review of the Asus Striker II Extreme motherboard with a tweaking guide. They recommended a tRRD (RAS to RAS delay) -[Auto, 1-15] "We recommend you set this valve to 1 and leave it there" This was not specific to OCZ but DDR3 in general. They did mention OCZ in the review and tested on this motherboard.

[WaterFlex]

Nice test

[vega22]

Hey, guys, i have a strange issue here...
With the rig in my signature (with a moderate OC), i can pass hundreds of IBT tests @ 10 pass. 100% Ok. But when i try to pass one only test @ 200 pass, just get errors. Sometimes I get only one hit and 199 errors!
It makes no sense. Help, please.

[Amurtigress]

Haven't had a chance yet to play with the "other" memory timings. You mention that OCZ recommends a RAS to RAS delay of 10. I tried searching their site but couldn't find information regarding adjusting the additional, not usually listed, DDR3 settings. If you could point/link me to their (or others) info regarding these settings, that would be great.

You have to look for tRRD on Google. I don't have it ready right away myself, but you can trust me on that.

*** Edit:

http://www.ocztechnologyforum.com/fo...d.php?p=277964

The cycles have to double for most DDR3 timings as compared to DDR2, since the clock seemingly doubles. Matter of factly, DDR1-400, DDR2-800 and DDR3-1600 have the same base clock of 400. Yet the number of clock cycles that we are messing with, the latencies, have to roughly double internally to achieve the doubled throughput as compared to DDR2.

It is in the nature of DDR3.

Most DDR2-800 RAM runs at 5-5-5-15-5, DDR3 RAMs of the same class at 1600 have 9-9-9-24 timings as JEDEC defaults programmed into their SPDs.

Trouble is, the FIFTH timing that ASUS gets wrong, is NOT programmed into it. It's not JEDEC standard. Thus ASUS' BIOS is stuck with the 5 for RAS to RAS delay unless you set it manually. Please see my post on the P5Q3 thread for it, including the screenshots.

http://www.xtremesystems.org/forums/...&postcount=661

[Shakti]

Thank you AgentGOD for this quick program, just one wish i have: Could you offer a choice of color for the text used? I get eye cancer from this red color, dark blue would be extremely welcomed.

[WaterFlex]

I`ve passed IntelBurnTest 300 times with no errors. But I have 2-3 errors in Win MemTest 3.7

[Dua|ist]

I`ve passed IntelBurnTest 300 times with no errors. But I have 2-3 errors in Win MemTest 3.7

Linpack is good at CPU stability checking because of it's highly optimized code using all of the CPU's available resources/modules, however, it doesn't put same heavy load at memory and northbridge, no matter how much memory it consumes. Because of that Linpack often passes OK where Prime/OCCT (and eventually MemTest) fail due to RAM/NB instability.

[platojones]

Linpack is good at CPU stability checking because of it's highly optimized code using all of the CPU's available resources/modules, however, it doesn't put same heavy load at memory and northbridge, no matter how much memory it consumes. Because of that Linpack often passes OK where Prime/OCCT (and eventually MemTest) fail due to RAM/NB instability.

Agreed...Linpack tests will help you get very close to a stable overclock quickly. It's most useful in quickly forcing issues with the CPU volts/frequencies that other tests OCCT or Prime95 would take hours to uncover. Once it's Linpack stable, you still need to OCCT/Ortho/Prime95 (pick your favorite) your rig to catch memory/chipset issues and tweak it from there. So for me, 20 cycles of Linpack + 24 hrs of Prime95 blend is probably as good as you can get in declaring the system stable.