IBT

paradisiac · 19 Jul 2011 at 20:08

what settings do you need for intel burn test?

currently i have it on the one below maximum stress and 50 runs. is this too much or not enough.
also, what do the actual numbers it does for each cycle mean? is higher better?

j.col · 19 Jul 2011 at 20:24

Wingzero30 will be along shortly, but here is some of his explanation.

to run ibt properly. close down ALL running programs inc. antivirus, internet, etc..

before running, open task manager and see the amount of 'free' ram under the performance tab.
for example i have 2550mb free ram so in IBT i enter a custom amount of 2500. (just below the free amount)
then i change the threads from auto to 4 (quad core), i have a quad core.
now every run i make i always get around 40-42 gflops. (consistent) (sorry glflops are a good indicator on how fast your cpu is running, i can get around 40gflops, but in a test if i was getting only 20 then the test is not running properly)

" I would choose 'free' memory as available memory will also make use of cache memory which must be storing data for other applications aswell. Windows may free up some physical memory from the cached memory during testing but not totally and i think the rest will be the virtual memory as a compensation.

Available = Free + (Cached - Modified)"

running just free memory will help you get higher gflops, as its just using free ram and not cached ram
this should also have 4 or 2 cores running at 100%"

this will give you more consistent gflops score, but will also make your temps higher, as more gflops, the more the cpu is stressed"

full details on running IBT properly http://www.overclock.net/intel-cpus/645392-how-run-linpack-stress-test-linx.html
__________________

MeddlE · 19 Jul 2011 at 20:24

20 IBT runs at maximum and 2 hours of OCCT using large data set are what I use.

The numbers in the first column are the time it takes to do 1 run, the second column is the GFLOPS, and the third column should all be the same. If the numbers in the third column differ then it has failed.

**Edited to add, what he said. ^^^^^^**

paradisiac · 19 Jul 2011 at 20:26

my Gflops are 119.** :confused:

is that good or bad?

j.col · 19 Jul 2011 at 20:28

paradisiac said:
my Gflops are 119.** is that good or bad?

LOL i wish i had 119gflops

i am not familiar with your cpu, but i would say yes its good

paradisiac · 19 Jul 2011 at 20:31

i havent set it up as you described, just changed the drop down box to the one below maximum and set it to 50 runs. there are all 119.** Gflops and 64.***secs. all 4 cores are 100% and temp maxed at 78c but averages 72c.

j.col · 19 Jul 2011 at 20:36

try about 5 runs the way i said, and see if you get higher gflops?
also done a quick google, and your score is good.
what is your cpu speed?

paradisiac · 19 Jul 2011 at 20:41

i forget to say its running @ 4.6GHz.

j.col · 19 Jul 2011 at 20:44

seems good, but as i said i am not familiar with SB cpu's.
also looking at your other thread, your temps are good as well.
are you noticing that you get higher temps with IBT?

paradisiac · 19 Jul 2011 at 20:51

i got slightly higher temps with IBT over prime, but the temps after about 30 run seem to of dropped down to the 60c's :confused:

not long now until its finished the 50 cycles, then i will try to do it the way you said, although im not sure 100% on everything (see parts in bold)

j.col said:
Wingzero30 will be along shortly, but here is some of his explanation.

to run ibt properly. close down ALL running programs inc. antivirus, internet, etc..

before running, open task manager and see the amount of 'free' ram under the performance tab.
for example i have 2550mb free ram so in IBT i enter a custom amount of 2500. (just below the free amount)
then i change the threads from auto to 4 (quad core), i have a quad core.
now every run i make i always get around 40-42 gflops. (consistent) (sorry glflops are a good indicator on how fast your cpu is running, i can get around 40gflops, but in a test if i was getting only 20 then the test is not running properly)

" I would choose 'free' memory as available memory will also make use of cache memory which must be storing data for other applications aswell. Windows may free up some physical memory from the cached memory during testing but not totally and i think the rest will be the virtual memory as a compensation.

Available = Free + (Cached - Modified)"

running just free memory will help you get higher gflops, as its just using free ram and not cached ram
this should also have 4 or 2 cores running at 100%"

this will give you more consistent gflops score, but will also make your temps higher, as more gflops, the more the cpu is stressed"

full details on running IBT properly http://www.overclock.net/intel-cpus/645392-how-run-linpack-stress-test-linx.html
__________________

* how do i disable MSE and the internet :confused:

* chose what :confused, i only have a drop down box for ram/stress

j.col · 19 Jul 2011 at 21:02

paradisiac said:
i got slightly higher temps with IBT over prime, but the temps after about 30 run seem to of dropped down to the 60c's
not long now until its finished the 50 cycles, then i will try to do it the way you said, although im not sure 100% on everything (see parts in bold)

* how do i disable MSE and the internet
* chose what :confused, i only have a drop down box for ram/stress

sorry, just disable your antivirus for 1 hr or so,
just make sure you have no internet browsers open. ie. firefox, chrome, IE, etc...
the main thing is to have no programs running in background.
also just running free ram. ie ram that is not cached by the pc will improve your gflops ( or supposed to

)
in IBT enter a custom amount of ram, and dont use max, high etc...

paradisiac · 19 Jul 2011 at 21:07

ok, will disable as much as possible, think i will have to kill MSE in task manager as i dont think it has an option to disable it. its pretty much passed 50 runs on my settings though, so is that a good indication it should be stable?

joeyjojo · 19 Jul 2011 at 21:20

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.

119 as a percentage of 153.6 is 77%, which is pretty good. The larger the problem size (memory usage) the more the GFLOPS will approach the peak of 153.6, until you start using too much and caching kicks in, whereupon it will plummet. You should experiment and see how high you can get it. Aim for 80-90% i.e. 123-138 GFLOPS by giving it more mem.

30 to 50 mins is recommended in the guide, which would be about 30-50 passes if you're averaging a minute. The time per pass is strongly related to the problem size (memory usage).

P.S. This is clearly why SB is so quick - it's doing twice as much work clock for clock.

j.col · 19 Jul 2011 at 21:31

thanks joeyjojo, typical, get a IBT thread and no wingzero.
just had a quick look on ocn and he score seems kosher

here are some results
hoss331: Intel Core i7 2600K @5.2GHz, RAM 2134MHz 7-10-7-27GFLOPS:140
grishkathefool: Intel Core i5 2500 @3.8GHz, RAM 1600MHz 9-9-9-24 2T GFLOPS: 96.4883
jonesey17: Intel Core i7 2600K @5GHz, RAM 1600MHz 8-8-8-20GFLOPS:130.8
kyle2194: Intel Core i5 2500K @4.7GHz, RAM 1333MHz 9-9-9-24GFLOPS:115.7
lethalrise750: Intel Core i7 2600K (HT OFF) @4.8GHz, RAM 1600MHz 9-9-9-25GFLOPS:125.3
ballathefeared: Intel Core i5 2500K @5GHz, RAM 2154MHz 9-10-9-39GFLOPS:133.78
simpletech: Intel Core i7 2600K @5.2GHz, RAM 2132MHz 7-10-20-28GFLOPS:138.5
nagle3092: Intel Core i5 2500K @4.6GHz, RAM 1866MHz 8-8-8-24GFLOPS:120.9
koven: Intel Core i5 2500K @4.9GHz, RAM 2134MHz 9-11-9-28GFLOPS:70.9
tribeca: Intel Core i5 2500K @4.8GHz, RAM 1600MHz 9-9-9-24GFLOPS:123.3

and the fastest FASTEST CPU
hoss331: Intel Core i7 2600K @5.2GHz, RAM 2134MHz 7-10-7-27GFLOPS:140 :eek:

and these are with the ram set on high and not custom
http://www.overclock.net/intel-cpus/869018-intel-amd-gflops-data-thread-looking.html

paradisiac · 19 Jul 2011 at 21:57

ok, your way seems to take 143.***secs and is 121.*** Gflops.

i did however get a BSOD so have upped the ram voltage and vcore. giving it another go.

betaspin · 19 Jul 2011 at 22:26

The 115-120 GFlops on an overclocked i5 2500k are achievable only with IBT including AVX (advanced vector extensions), see http://software.intel.com/en-us/avx/. IBT (in fact, Linpack) with AVX produces way more heat (about 10 deg) than prime95 with small FFT.

joeyjojo · 19 Jul 2011 at 22:27

paradisiac said:
ok, your way seems to take 143.***secs and is 121.*** Gflops.

i did however get a BSOD so have upped the ram voltage and vcore. giving it another go.

Wicked

Did you test with Prime? I wonder if it would have appeared stable...

paradisiac · 19 Jul 2011 at 22:50

joeyjojo said:
Wicked

Did you test with Prime? I wonder if it would have appeared stable...

i did about 30 mins prime to check temps were ok.
IBT is def making the temps go higher than prime. on second run (with settings suggested and a bump in volts) temps are still good and its about 20 out of 35 cycles stable. hopefully will do all 35 without error.
if it does, will do a prime test tomorrow and then save settings and go for 4.8GHz+

joeyjojo · 19 Jul 2011 at 23:02

No point in the prime, IBT will find instability when p95 won't. Prime 95 is literally useless. Apart from wasting time I suppose.

paradisiac · 19 Jul 2011 at 23:04

i thought IBT wasnt the best to rely on for 1155 chips