ASUS Maximus Formula SE

[cadaveca]

apparently you haven't tried using many settings... but anyways the problem is prime and not the motherboard. the code for doing SFFT's has a bug in it (blend is fine). I recreated the problem with <1.5 vcore.

Um, blend does not put the same load, nor gives the same temps as small FFT, and hence current draw. Try running 4 instances of TAT. Same issue. Are you measuring power consumption to confirm I'm wrong, and if so, how, or are you again just assuming things? Have you properly debugged Prime and found the bug, and notified the author of what it is, exactly?

[safan80]

Have you properly debugged Prime and found the bug, and notified the author of what it is, exactly?

In some OCs using prime SFFT won't load one of the cores correctly giving 75% only usage. it will load all the cores correctly if you kill prime and restart it. it will sometimes say 75% but it will be loading all the cores. I was only testing the 32bit version under xp. I was using 25.5a.

[cadaveca]

In some OCs using prime SFFT won't load one of the cores correctly giving 75&#37; only usage. it will load all the cores correctly if you kill prime and restart it. it will sometimes say 75% but it will be loading all the cores. I was only testing the 32bit version under xp. I was using 25.5a.

So, if it's prime as you say, why does it happen with my B3, and when I pop in my G0, without touching bios, it does not? If it was setting-related, or prime(software), then changing cpu should have no effect, no?

[safan80]

So, if it's prime as you say, why does it happen with my B3, and when I pop in my G0, without touching bios, it does not? If it was setting-related, or prime(software), then changing cpu should have no effect, no?

Why does it not happen with my B3 but it happens on my C0? I think the bigger question is why does it only happen on prime SFFTs and not blend, OCCT, cinebench, or video encoding? Why does the x64 version do SFFTs just fine? under xp the second 2 cores don't always max out. If I could create the problem in a different program then there would be some merit to blaming hardware but as it stands at the moment I can't.

edit: have you tried to do a custom run setting the Max FFT size to the max your cache size is? it runs at full load even under xp.

vista pic is just a pic of standard small FFTs.

[cadaveca]

edit: have you tried to do a custom run setting the Max FFT size to the max your cache size is? it runs at full load even under xp.

vista pic is just a pic of standard small FFTs.

Um, that's how I run prime. Check the stability testing thread from a few months back, I stated this there. I run prime 8k FFT to largest FFT that fits in cache, and I'm sure I'm one of the first to advertize running prime this way, because I'm more than aware that it's L2 cache speed capabilities that holds most core2 cpu's from getting high clocks.

I've also been playing with quads for about 8-9 months. Running prime is different on core2 than on AMD, becuase ofthe shared L2. With AMD, just running small FFT is generally enough for stability testing, especially since memorycontorller is contained within cpu.

I'm one ofthe few die-hards that insists on using prime for stability testing, and nothing else, becuase I'm more than aware of how to properly use it. I understand what causes what type of errors, and what needs adjusting in a system based on the errors it gives, and I do not call a system stable until it can runn a 2-3 day run of prime. 8-12 hours jsut isn't enough for me.

But like I said earlier, in respionse to your questions, the algorithm used in the small FFT test is the most cpu-intensive, while blend runs do not even genreate the same heat, and X64 allows the 32-bit program a different operating environment that does not put as much stress on the cpu(32-bit mode vs 64-bit), again, a function of the cpu.

It doesn't happen with your B3 because it's cache expires before current usage reaches the same limits as your C0, and OCCT, cinebench and videoencoding do not cause my system to draw even thsame power form the socket as prime does...however, when the issue is pushed to the extreme, I can get OCCT to do it too(my cpu shows issue @ 1.477, @ 1.625, OCCT has load issue as well)

Not alot of people understand how cache is limiting Core2. This is why Intel have gone to 6MB for new cpu's...the cache does not get thrashed as much, and can therefore supply the rest of the core @ higher frequencies before failing...the inverse is true for why AMD has issues getting higher clocks...

Fact of the matter is, I do undersand this issue, and it seems, at this point, even cpu's and how they run, far greater than you'd like to think.

Anyway, your pic is useless, sry. the issue is not shown except in processes tab. You show 4 cores loaded, but I can easily show the same...flop over to process tab(you are on performance, BTW, not like how my directions said...) and system process is using one of those cores. The issue is not getting load on all for cores, but getting PIRME loaded on all four cores. Also, none of your thoughts on the issue explain why lowering voltage will make the issue go away...I am stable @ 3.6ghz on G0 at a far lower votlage than what causes the issue, so it's currently not a "show-stopper" bug for me now...but only becuase I've understood the issue and took steps to correct it.

I found the real cause, and solution already thanks.

[safan80]

. Also, none of your thoughts on the issue explain why lowering voltage will make the issue go away...I am stable @ 3.6ghz on G0 at a far lower votlage than what causes the issue, so it's currently not a "show-stopper" bug for me now...but only becuase I've understood the issue and took steps to correct it.

I found the real cause, and solution already thanks.

My B3 didn't have the issue and I tersted all the way up to 1.65 volts.
none of your thoughts explain why lowering voltage the under 1.50 caused the issue for me on my QX9650. The thing is I don't believe you know what's going on. I can create the issue but it doesn't affect me on the oc that I use day-to-day. I'd rather not listen to you talk until you can provide proof that you know exactly what the issue is. Prime is the old benchmark for stability here. I can be 24/7 stable on prime and it wouldn't matter the system must be proven to be also stable in other programs like occt, & Linux kernel compilation.