On the contrary, IBT with AVX or LinX with AVX (both linpack based) loads the 1155 chips more than anything else. Also, both allow using all available memory, which is good as a memory test.
IBT
- Thread starter paradisiac
- Start date
I can't answer that as I don't have 1155.
In my experience, Prime 95 was last of any use at all on my Opteron 165. Since then, redundant. As you've just seen, it sniffed out instability in a matter of minutes, which would have taken p95 probably 24 hours or more, or perhaps not even detected it at all.
- Joined
- 22 Mar 2009
- Posts
- 7,754
- Location
- Cornwall
so if it does 35 runs ok then i should be good?
I've not noticed this - in fact quite the opposite - what are your sources?
I run all my new builds with IBT (20 passes) and some have passed no problems but when i've subjected the same machine to a few hours of small ffts or blend tests they've failed.
I only use IBT for a quick guide as to whether a clock is stable as it wheedles out an obvious flaky clock very quickly - but i've always relied on long prime blend/ffts to assure me that it's stable (and even that may not prove to be the case.)
In my experience, anything above 20 runs on max memory with IBT or LinX with AVX is enough to prove stability. The temperatures should stablilize within the first 5 runs or so.
- Joined
- 22 Mar 2009
- Posts
- 7,754
- Location
- Cornwall
what is AVX?
Soldato
- Joined
- 1 Jun 2010
- Posts
- 7,058
- Location
- Bedford
j.col speaks the truth (and where the hell is wingzero??)
Firstly, yes, one hundred and nineteen billion floating point operations for second is pretty damn incredible. Just think about it for a moment
Read me: http://forums.overclockers.co.uk/showthread.php?t=18206940
However with SB I assume the calculation for GFLOPS is wrong. With C2D, you get 4 FLOPS per core, so 4*(number of cores)*(speed in GHz) = GFLOPS. If you're getting 119 GFLOPS that means the 2500K is doing 8 floating point ops per cycle (yowza!) giving you a peak output of 4*4.8*8=153.6 GFLOPS.
P.S. This is clearly why SB is so quick - it's doing twice as much work clock for clock.
I hear you guys
I don't need to add anything extra. You guys are amazingly knowledgeable in the use of IBT
As stated above SB uses AVX instructions which double the flops from 4 flops to 8flops per cycle hence you can attain GFlops values in excess of 100.
IBT makes use of multi-add operations.
- Joined
- 22 Mar 2009
- Posts
- 7,754
- Location
- Cornwall
I hear you guys
I don't need to add anything extra. You guys have become amazingly knowledgeable in the use of IBT
As stated above SB uses AVX instructions which double the flops from 4 flops to 8flops per cycle hence you can attain GFlops values in excess of 100.
IBT makes use of multi-add operations.
so it does AVX automatically?
its on 33 of 35 runs, so if the last 2 are ok then im good
its remained constant time and Gflops so i assume all is as it should be.
Just anecdotal I'm afraid, from my own tests. It stands to reason though, that the test which is producing sometimes 20 degrees more heat is going to be more intensive. Furthermore, linpack is inherently more sensitive to error as it is essentially running an algorithm to solve huge systems of simultaneous eqns (in the thousands of eqns, this is what the memory usage refers to). As such, a single miscalculation results invariable in a different residual.
I don't know how p95 works come to think of it. Though it is clearly a pretty insensitive test, as it takes many repeats of many thousands of iterations of the same prime number to spot an error.
so it does AVX automatically?
its on 33 of 35 runs, so if the last 2 are ok then im good
its remained constant time and Gflops so i assume all is as it should be.
It's certainly a good starting point - but the harsh reality is until the sytem proves itself stable in day to day usage over the coming days and weeks you wont know for sure.
Stress tests are great for early indicators of instabilty but some systems can pass every stress test thrown at them but then inextricably BSOD when browsing...
IBT is a front-end to the Intel Linpack benchmark http://software.intel.com/en-us/articles/intel-math-kernel-library-linpack-download/. The included matrix operations may very well include trigonometric functions in addition to multiplications and additions, I have not seen an exact specification of the Linpack subroutines included in the benchmark.
Soldato
- Joined
- 1 Jun 2010
- Posts
- 7,058
- Location
- Bedford
so it does AVX automatically?
its on 33 of 35 runs, so if the last 2 are ok then im good
its remained constant time and Gflops so i assume all is as it should be.
Do you have win7 sp1 installed?
I think SB only takes advantage of AVX when it is running software such as win7 sp1 that supports AVX.
Possible. Though I'm yet to hear of a single system crash in day to day use after passing a properly set up IBT stress. If you can find one I'd be very interested.
- Joined
- 22 Mar 2009
- Posts
- 7,754
- Location
- Cornwall
yeh windows 7 sp1 is all installed. fresh install this afternoon.
It's solving simultaneous eqns according to http://en.wikipedia.org/wiki/Linpack as I described above. If this isn't the case I'd like to know.
It seems like it is though, as e.g. in LinX you can choose the number of eqns explicitly.
Thanks, so it does not seem to include any trigonometric functions. This is a major difference from prime95, which needs trigonometric functions for the Fourier transformations. It would be good to have a prime95 version that used AVX---this could be the ultimate test.
- Joined
- 22 Mar 2009
- Posts
- 7,754
- Location
- Cornwall
35 runs without error
A refreshingly honest and informative answer.
I understand where you’re coming from - and due to the intensive (brutal) and invasive nature of IBT you would (from data) think that any chip that survived 20 passes should be stable.
But from my personal experience I’ve found that IBT is a brilliant tool for routing out unstable clocks quickly but then I move on to prime for ~12 hours to try and route out a more subtle problems.
But as mentioned – even then, after proving itself stable after 20 passes and hours of prime you (well, 'I' in this thread
) have the occasional machine that will decide to have a conniption fit because i've opened up a browser while listening to my favourite Radiohead track :/That is one i can't answer other than my own odd rogue system.
I can point to threads where they've passed IBT and then gone onto BSOD in every day usage - but i can't state whether IBT was set up properly.
I think they both have their place (i wouldn't dismiss either) - along with other stress tests - as CPUs are tempramental little sods, when clocked, and sometimes it takes something from 'left field' to make it blink. Whether that be 20 passes of IBT, 24 hrs of prime or just IE8 doing a search...
I'm quoting WingZero as he adds the substance/reasoning to my personal experiences...
Last edited: