LinX - A simple Linpack interface

[Johnny87au]

ok thanks dude, Ill stick to clicking ALL

[Chrysalis]

mine passed all 20 so I hope my stability problems on this board are behind me now, I had 40 gflops rating.

[TwwIX]

Sorry for being ignorant but could somebody explain "GFlops" to me.

Do i want it to be higher or lower?

Thanks in advance.

[Leeghoofd]

higher is better, though difference betweeneg 45 and 53 is not reflected in eg benchmarks like 3D06, Pi32, PCmark05 etc... on my I7 setup the Gflop rating is very dependant on the CPU features setting

[colore]

can this burn a laptop???

or in any way actually damage it?

[Zeus...]

Any stress test can cause damage 'if' the cooling isn't good enough and the components overheat. Doesn't matter if its a lappy, desktop or net-book.
But if it cant handle it is it any good? Plus some do have a throttle that decreases the voltage multipliers when things start getting too hot.

[sniper_sung]

Could you explain what changes has been made to linpack_xeon64.exe and linpack_xeon32.exe? The digital signatures from Intel has become obsolete. I don't suppose it's some clean actions?

[Mitchell]

There is a newer version of the Linpack Library on Intel's website, that's what I've been using for a while now rather than the one included in LinX 0.6.4.

Here are the Digital Signatures on both xeon64.exe and xeon32.exe with the newest Linpack.





For those running the i7 980X with HT on, using the latest Linpack should result in higher and more sustained GFlops.

You can download the newest Linpack Library from here:

http://software.intel.com/en-us/arti...pack-download/

Simply just copy xeon64.exe and xeon32.exe and replace the originals in your LinX/IntelBurn directories.

[zalbard]

Great tip, thanks.

[shadewither]

Great tip, thanks.

Maybe the 4.5G club needs a clean-up.
There are some bogus results in it, esp. the 6C12T ones.

For intel linpack, low Flops=less Stress, and the difference could be very big.
It is possible to pass linpack by preventing it from running efficient.

[safan80]

Please update Linx with the newer linpack files. If one uses LinX 6.4 without changing the Linpack files to the one listed above you get insane Gflops with Gulftown without having to fully load your cpu.

Old VS new Linpack files

[Big Myke]

When I run LinX after about 5 mins my videocard drivers stop responding and the screen flashes black continuously. What kind of sign would this be?

Win 7 64bit
759 classified
i7 920

[donmarkoni]

When I run LinX after about 5 mins my videocard drivers stop responding and the screen flashes black continuously. What kind of sign would this be?

Win 7 64bit
759 classified
i7 920

Try bumping VTT one notch higher.

[Big Myke]

Correct me if I'm wrong but isn't it ideal to run linx .64 in safemode to get the MOST memory stressed during the test? The only problem I expierence while testing in safemode is that realtemp or other temperature apps do not work in safemode.

[donmarkoni]

Correct me if I'm wrong but isn't it ideal to run linx .64 in safemode to get the MOST memory stressed during the test? The only problem I expierence while testing in safemode is that realtemp or other temperature apps do not work in safemode.

After booting into W7x64 I have 5GB available to stress test. IMO, it's enough.
If I start LinX and then stop it after 20 seconds or so, I have 5.5GB available, as Win releases some memory.
And that downside is a big one.

[safan80]

Correct me if I'm wrong but isn't it ideal to run linx .64 in safemode to get the MOST memory stressed during the test? The only problem I expierence while testing in safemode is that realtemp or other temperature apps do not work in safemode.

If you want to be able to stress more memory disable all the aero related stuff and extra services you don't need. If you don't want to do that install windows xp x64 it uses very little memory.

[Nullack]

I notice with using the new Intel versions of Linpack my dual 5630 Xeons are stressed more. Highly recommend a new version of this tool with those new compiles from Intel in it.

[PatRaceTin]

harder to pass new library

[jcool]

I found that running only the # of threads matching your physical cores actually puts more strain on Gulftown/Westmere-EP CPUs. For example, running 12 Threads instead of 24 on my Dual X5680 gives higher Gflop rating and produces higher temperatures than running with 24 Threads.
Apparently WIndows 7 is intelligent enough to put the load only on the physical cores if the # of threads don't exceed them.

Next I will follow Mitchell's great tip and update the Linpack librariers. Curious to see the difference

Edit: Nope, updating Linpack libraries is useless. I get next to no CPU load and 50 (!) Gflops rating
Running 12 threads is still the best option for dual socket 1366 and linpack.

[Johnny87au]

@jcool thanks for the heads up m8!

[jcool]

No problem mate, but see edit.. updating libraries does not solve the HT problem. Just run 12 threads... or prime x64 with all ram, which is VERY sensitive as well.

Edit: New libraries are more sensitive though. I just got an error with the same settings that passed on the old libraries

[sniper_sung]

About gflops plz see http://www.xtremesystems.org/forums/...d.php?t=256633

[jcool]

About gflops plz see http://www.xtremesystems.org/forums/...d.php?t=256633

Yes, we've known about that for a long time. But seeing as I normally run with HT, since WCG gets a 40% boost from it, I don't want to disable it just for stress-testing. Hence my tip that you can simply run LinX with the # of threads corresponding to the amount of physical cores in your system, and you get the same effect as with turning HT off.

Here's a shot with updated libraries, but otherwise identical settings to my earlier screenshot - Gflops are lower, but the program is even more sensitive now. Looks like I have some fine-tuning left to do

[Kondik]

About gflops plz see http://www.xtremesystems.org/forums/...d.php?t=256633

Afaik HT lowers per core performace but 'raises' the core count. Meaning a i7 with HT off will have performace rating of 100 on 1 core meaning 400 on four and with HT on Per Core PFRating drops to 75 but core count jups to 8. Meaning Anything that uses up to 4 Cores will be faster on a non HT's i7 but anything that uses more than 4 cores will benefit from it.
Bold explanation.

An 4 Core Usage aplication =

4 Non HT'd cores 400
4 HT'd Cores 300

An 6 Core Usage aplication =

4 Non HT'd Cores 400
6 HT'd cores 450

An 8 Core Usage Aplication =

4 Non HT'd Cores 400
8 HT'd cores 600

This is of course only in case the Aplication Code is superb and it really uses the cores.

Simply taken. Video Compression, WCG , Rendering ... HT will profit
Anything Else, HT will drop performace.

That's why we see a Gflops Drop in Linx with HT on.

I observed 4 GHz i7 930 with HT on gives me about 51 ~ 52 GFLOPS , HT off gives me around 58 ~ 60

I talk too much

[orlywutlol]

Hey guys I have a bit of a dilemma here. Currently I have my Q9550 running at a stable 3.825ghz (450x8.5) I can get into windows at 4ghz (471x8.5) but when I run LinX after the first pass.. cores 0 and 1 would drop 8-10c even though they are still under 100% load. Cores 2 and 3 would soon follow after a minute or so. Gflops would drop and times would sky rocket. Temps are around 77 77 72 72. Some say I am throttling but I have seen people run these temps under LinX without issues. However I did pass OCCT 1 hour without errors at 4ghz and temps never touched 70c. Am I at the limit of my board? What else can I do? Any ideas?



heres a screenshot

Real Temp shows only 78% load but the task manager and core temp show 100% load on all cores.