IntelBurnTest - The new stress-testing program (Now works on AMD CPUs!)

Printable View

Show 100 post(s) from this thread on one page

09-30-2008, 02:25 PM
Final8ty

Running Quad prime95 full blend & Burntest level 1 together now..just to be sure :D

Edit: substantial V-drop now. its really feeling it.
300MB left out of 8GB.
09-30-2008, 02:37 PM
AgentGOD

Here's a run from my AMD machine:
http://i37.tinypic.com/5vcgh2.jpg
09-30-2008, 02:43 PM
Final8ty

Quote:

Originally Posted by AgentGOD

I've never tested Phenom, sorry; so I can't give you official word on Phenom support.

I've only tested on an AM2 K8 (X2 4200+ OC 2.8 GHz), and S939 K8 (Opteron 165, OC 2.91 GHz). Both passed 100% with this version.

Keep in mind, the Linpack binaries were written by Intel(R) themselves. I removed that nasty limit that disallowed it to run on AMD machines.

For now, I suggest throwing out the extraneous value (first value) if you believe that it's a Linpack bug.

#EDIT:
Hmm, I think I've noticed a trend here. The bug seems to occur with Linpack 64-bit version (according to your screenshots, you were running 64-bit Windows). I've never actually put the 64-bit version to test on any AMD CPU (since I do not have XP X64 installed on any of my PCs at the moment except this Intel Q6600 one).

So, the tests I've conducted on AMD CPUs were only with 32-bit XP. There's still too many variables to conclude whether it's a Phenom-Linpack bug, or it's just a 64-bit Linpack and AMD bug. I know the binaries aren't corrupt since I can run the same patched 64-bit binary on my Q6600 perfectly fine.

Please do help me test guys!

I'm running the 64bit version on vista 64 & all my results align including the first one.

I wonder if it could be mobo related.
09-30-2008, 02:50 PM
AgentGOD

Quote:

Originally Posted by Final8ty

I'm running the 64bit version on vista 64 & all my results align including the first one.

I wonder if it could be mobo related.

Well thanks for reporting a success!

I know my new theories have no direct basis on evidence, but here goes:
1. It could be caused by a weak/unstable on-die memory controller! Remember, AMD processors are extremely sensitive to any memory controller/memory problems!
2. It could be caused by defect RAM! Remember, AMD processors are extremely sensitive to any memory controller/memory problems!
3. As you said, it could be motherboard related.
4. The binary itself is defect (I doubt it, since you ran it perfectly fine on your Phenom, and for me too on my Q6600).
09-30-2008, 02:51 PM
Particle

I don't know. I run the K9A2 Plat as well, but I don't know if we're using the same BIOS revision.

--

It seems like hardware is unlikely to be a problem here since both of my 9850 machines at stock exhibit the first-off problem. I wasn't the only one, either. What are the chances that we all have the same defective hardware causing an identical issue?
09-30-2008, 02:53 PM
AgentGOD

Quote:

Originally Posted by Particle

I don't know. I run the K9A2 Plat as well, but I don't know if we're using the same BIOS revision.

--

It seems like hardware is unlikely to be a problem here since both of my 9850 machines at stock exhibit the first-off problem. I wasn't the only one, either. What are the chances that we all have the same defective hardware causing an identical issue?

Remember B2 stepping and TLB-bug, anyone?
09-30-2008, 02:55 PM
Particle

What, you'd more easily accept massive hardware glitching than a software issue?

Besides, if we had bad hardware we wouldn't be getting a bad first run exclusively. It would be more random and unpredictable.
09-30-2008, 03:01 PM
Particle

I replaced linpack64.exe with linpack32.exe to test your 32/64 bit theory. The 32-bit one passes me. The 64-bit one always fails me. While that doesn't prove the point I've been making, it might be useful general debug data.
09-30-2008, 03:03 PM
AgentGOD

Quote:

Originally Posted by Particle

I replaced linpack64.exe with linpack32.exe to test your 32/64 bit theory. The 32-bit one passes me. The 64-bit one always fails me. While that doesn't prove the point I've been making, it might be useful general debug data.

Guess we'll have to see what Intel can do to fix the 64-bit binary.
09-30-2008, 03:04 PM
Slagathor

1 Attachment(s)

Heres a run I just did, I did the 5 pass's or what ever with half my ram.
If I should try a different settings let me know...... K...... :d
09-30-2008, 03:07 PM
Particle

Quote:

Originally Posted by AgentGOD

Guess we'll have to see what Intel can do to fix the 64-bit binary.

You're a rather snarky one, aren't you? I can tell you're a programmer like myself. I get pissed off when people say my software has bugs before I even know if they're right or not. I also get ultra-defensive and fall into that all-too-easy "That's Impossible" zone.

Remember that Intel didn't implement AMD64 extensions identically to AMD. Their 64-bit version of linpack may just be incompatible. *shrug*
09-30-2008, 04:14 PM
Final8ty

Quote:

Originally Posted by Particle

I don't know. I run the K9A2 Plat as well, but I don't know if we're using the same BIOS revision.

Im using the performance bios that is in my Sig that has AutoXpress implemented.
http://www.xtremesystems.org/forums/...6&postcount=17
09-30-2008, 07:02 PM
Loser777

I'd like to see if this triggers any TLB problems with the B2s.
09-30-2008, 07:19 PM
AgentGOD

Quote:

Originally Posted by Loser777

I'd like to see if this triggers any TLB problems with the B2s.

Aww don't be mean.. I'm an AMD fanboy.
09-30-2008, 07:20 PM
AgentGOD

Quote:

Originally Posted by Slagathor

Heres a run I just did, I did the 5 pass's or what ever with half my ram.
If I should try a different settings let me know...... K...... :d

You actually entered higher memory than you have available at the time... so half the time it would have been paging to the disk. Look at the "free memory" box at the top... it tells you how much RAM you have free.

Also, try using "maximum stress" if you can. See if it's stable then. If so, you're another person who was successfully able to run 64-bit Linpack test on AMD Phenom (in this thread).
09-30-2008, 07:31 PM
Particle

Quote:

Originally Posted by Final8ty

Im using the performance bios that is in my Sig that has AutoXpress implemented.
http://www.xtremesystems.org/forums/...6&postcount=17

I gave the official 1.5 a shot. It seems to allow me to cold boot at stock settings finally! Woo! It used to reset itself 4-5 times and complain about bad OC settings. heh What a great mobo. Also interesting is that my performance is higher in in this app (verify off mode). ~30 GFLOPS instead of ~24 GFLOPS
09-30-2008, 07:51 PM
Particle

Hey, some of you guys who have ran 64-bit on a K10 without issue, can you post your CPUID and microcode codes?

Mine is an 0810 GPBW, 0100F23H/01000065H.
09-30-2008, 08:23 PM
Final8ty

~37 GFLOPS here :)
10-01-2008, 03:23 AM
Slagathor

1 Attachment(s)

Used Maximum stress and passed again...
10-01-2008, 04:05 AM
Final8ty

So NO to error detection will show you your progress as you go & give you GFLOP's. & check the align your self.
10-01-2008, 11:29 AM
AgentGOD

I wouldn't know what to tell you guys then... since it's working perfectly fine for some, but not for others.
10-03-2008, 09:06 PM
Daveburt714

AgentGOD: Thanks for your work on this software.. :up:

It seems to run fine on my machine, but I am running a 32bit OS (see sig).

CLICKABLE:
http://i30.photobucket.com/albums/c3..._IntelBurn.jpg

I haven't done enough testing yet to see if the settings are actually everyday stable, but it definetly heats things up more than 4x PrimeBlend did
and I don't have to run it for hours on end stressing every other component to check my OC...

Excellent work man! And Thanks Again.... :D
10-04-2008, 06:35 AM
Particle

Quote:

Originally Posted by AgentGOD

I wouldn't know what to tell you guys then... since it's working perfectly fine for some, but not for others.

Yeah, it's very weird. I mean, I know it's out of your control to fix--it's something odd between the 64-bit binary and (some?) K10s.

I just wish we had more to go on. I'm perfectly happy to discard the first result and just check them myself though. Thanks for the work you put into it.
10-04-2008, 06:47 AM
MomijiTMO

Ummm wow you don't seem to be hitting the INSANE temps we do. I guess if 149 are stable then the first is bogus. For me a fail gets a large number [factor of 10 bigger than the others]
10-04-2008, 03:46 PM
Final8ty

Quote:

Originally Posted by Daveburt714

AgentGOD: Thanks for your work on this software.. :up:

It seems to run fine on my machine, but I am running a 32bit OS (see sig).

CLICKABLE:
http://i30.photobucket.com/albums/c3..._IntelBurn.jpg

I haven't done enough testing yet to see if the settings are actually everyday stable, but it definetly heats things up more than 4x PrimeBlend did
and I don't have to run it for hours on end stressing every other component to check my OC...

Excellent work man! And Thanks Again.... :D

how come my G-FLOPS are higher than yours when your cpu is running 200MHZ faster & running your ram at 1066 while mine is on the 800 setting + BUS speed.

Show 100 post(s) from this thread on one page