LinX - a nicer friendier Linpack

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DavidK21770

Member
Jan 1, 2005
40
0
0
Originally posted by: supastar1568
My recommendation to all.

Do NOT turn on error checking. When it askes for yes or no, say no, which is "n".

This is because, if it is on, you do not get your results until the test is passed. Which is fine if you are going to bed and won't look at your computer till the morning.
This applies to Intel Speed Test, not the program that is the topic of this thread -- LinX! LinX tells you what iteration you are on, gives you a progress bar, and stops on the 1st error.

So far, I like it much better than Intel Speed Test. My only complaint is that you can't have it remember to use the max memory. If you had more memory on one iteration and then reload it, it'll complain. And vice-versa. If you have less memory, it'll use what you had last unless you reselect. (And, I'm not sure that it's using all possible memory. Only that it's using what's reported as free. I'd like to be able to override like I can with Intel Speed Test.)

 

supastar1568

Senior member
Apr 6, 2005
910
0
76
Originally posted by: DavidK21770
Originally posted by: supastar1568
My recommendation to all.

Do NOT turn on error checking. When it askes for yes or no, say no, which is "n".

This is because, if it is on, you do not get your results until the test is passed. Which is fine if you are going to bed and won't look at your computer till the morning.
This applies to Intel Speed Test, not the program that is the topic of this thread -- LinX! LinX tells you what iteration you are on, gives you a progress bar, and stops on the 1st error.

So far, I like it much better than Intel Speed Test. My only complaint is that you can't have it remember to use the max memory. If you had more memory on one iteration and then reload it, it'll complain. And vice-versa. If you have less memory, it'll use what you had last unless you reselect. (And, I'm not sure that it's using all possible memory. Only that it's using what's reported as free. I'd like to be able to override like I can with Intel Speed Test.)

Yup, you are correct. Havent tried the thread topic LinX GUI
 

DasFox

Diamond Member
Sep 4, 2003
4,666
46
91
Originally posted by: aigomorla
Originally posted by: VirtualLarry
Does it fix the problem with IntelBurnTest's "maximum stress" setting giving errors due to out of memory?

why would you want to test max memory?

the whole point in linpack was to load up your cpu and have it do cpu intensive tasks, not both.

The good thing about this program is its fast, i shifted all my stress program to this.

I set it on custom settings, use 512megs of ram, and let it run 10 passes.

If it passes that your very good to go, i have yet for my system to fail crunching on WCG for 2 weeks now without a reboot.

Oh just a note, this program WILL PISS YOU OFF. It WONT pass where OCCT and Prime would pass 24/7. And YOU will need to compromise.

:X

Gillbot is scared silly to use it, and it PISSES me off from time to time because i'll get 1/10 thats not correct and need to retune.

Lastly, expect a 10-15C higher temp mark then OCCT and Prime using this program. So make sure you got your CPU temps in check or you'll throttle VERY fast.

I guess if you set the memory size then it bases it on that, not the Problem Size.

But what if you have 4GB of RAM, what amount should you test?

And you're saying on 'Times to run' you are only putting it on 10?

THANKS


Originally posted by: supastar1568
I will now use LinPack to test my future systems stability.

From what I had passed in Prime, I failed in Linpack. I was finally able to figure out that I needed +.05V on the FSB, and +.025V on the northbridge, the cpu voltage remained the same and I got 100 passes completed on Linpack. Obviously, I will never stress my system equal to 100 passes of Linpack.

Also, I have learned to turn error checking OFF. If it is on, the program will run the full amount of tests and give you a "stable" or "not stable". Turning it off, shows the results dynamically, and you are able to quit out of the program when you see an error, no point in having to wait till the end if you have even one error.

What do you mean, you don't leave, Stop on error checked? I thought it will just stop if it finds and error if you leave this checked?

Why do I get the feeling that LinX just pushes the system to much and it's just not a realistic stress test? Maybe I'm just thinking like this is all, but after reading what people are saying here, it makes me wonder, does it go way beyond the norm, and it's not really needed.

Look what I'm saying here with LinX is, hell you could place your box in a 100 degree oven if you wanted to do some stress testing, LOL, see the point? So is LinX going to a unsually high level of testing that is just not realistic benchmarking, way beyond what a normal PC will ever see in being used?
 

n7

Elite Member
Jan 4, 2004
21,303
4
81
DasFox, i get what you are saying, but you're looking at this the wrong way.

Fact: a properly set up stock system will ALWAYS pass Linpack (whether IBT or LinX) testing.

It's that simple.

So if your OCed system doesn't...well, sorry to say, but it's not really stable, at least not as Intel designed it to be.

You can argue levels of stability all you guys want, but it's really not that hard to understand.

Stock systems always pass, so if you cannot pass OCed...well, you can derive your own conclusion from that.

 

DasFox

Diamond Member
Sep 4, 2003
4,666
46
91
Originally posted by: n7
DasFox, i get what you are saying, but you're looking at this the wrong way.

Fact: a properly set up stock system will ALWAYS pass Linpack (whether IBT or LinX) testing.

It's that simple.

So if your OCed system doesn't...well, sorry to say, but it's not really stable, at least not as Intel designed it to be.

You can argue levels of stability all you guys want, but it's really not that hard to understand.

Stock systems always pass, so if you cannot pass OCed...well, you can derive your own conclusion from that.

Ok, well what I wanted to know was if LinX was realisitc benchmarking, and it's not going to some extreme measure that wasn't real world realistic was all.

The system it's going to be used on is stock...

If you have a 1GB, 2GB & 4GB system for each system what should you set 'Memory to use' at? Also what's the best amount to set the Times to run on for optimal stability, wanting to make 100% sure?

Also Stop on error checked is good?

THANKS
 

n7

Elite Member
Jan 4, 2004
21,303
4
81
Eh, i personally use IBT not LinX.

For RAM to use, you want to set it so it doesn't use more than the OS usage + Linpack usage combined obviously, just below would be good.

Originally posted by: aigomorla

why would you want to test max memory?

the whole point in linpack was to load up your cpu and have it do cpu intensive tasks, not both.

Yes & no.

I'm guessing you've never noticed this, but setting it to use more memory puts more stress on the CPU test actually.

I've tried with half memory used vs. all with my 8 GB, & temps gets way hotter with all memory tested than with less than total.

Something most people don't catch, but test it for yourself; more memory tested = higher CPU temps.

edit: i'm guessing it has something to do with problem size.

IBT automatically selects problem size i'd imagine based on how much RAM you are using.

I just finished comparing with LinX, but i select more memory to test, it increases problem size.

I cannot get it to set the same problem size for differing amounts of memory to test, so i imagine that affects things, as when i was testing max vs. 2 GB, i was getting 7-8C in load temps.

That still agree with my findings of more RAM tested = harder on CPU though, even if it's based on problem size, unless there's some way to force the same problem size for all RAM amounts.
 

DasFox

Diamond Member
Sep 4, 2003
4,666
46
91
Ok so what's the best amount of times to run it? 10, 20, 50, 100?

Really for 4GB of ram you could put it on like 3500?

THANKS
 

n7

Elite Member
Jan 4, 2004
21,303
4
81
Well, max is based off of what your OS is using, so you may be able to do 3500...LinX sets the max automatically...just use that IMO.

On my system right now, for example, it's settings max to 7031 MB, as the rest is being used by programs & Windows.

How many times is up to you. I run 25 personally, but that's like asking how long do you run P95?

Some people are happy with an hour, others want 8, others want 24+ hrs... etc.
 

DasFox

Diamond Member
Sep 4, 2003
4,666
46
91
Originally posted by: n7
Well, max is based off of what your OS is using, so you may be able to do 3500...LinX sets the max automatically...just use that IMO.

On my system right now, for example, it's settings max to 7031 MB, as the rest is being used by programs & Windows.

How many times is up to you. I run 25 personally, but that's like asking how long do you run P95?

Some people are happy with an hour, others want 8, others want 24+ hrs... etc.

Ok so LinX sets defaults, ok

Is LinX just CPU testing or ram too?

THANKS
 

DasFox

Diamond Member
Sep 4, 2003
4,666
46
91
Sorry double post, no one answered my last...

LinX is just for CPU stress testing correct?

THANKS
 

n7

Elite Member
Jan 4, 2004
21,303
4
81
No, not really.

If you are setting it to max RAM, it stresses that & NB heavily as well.

For me it seems P95 Blend/Large FFTs find FSB issues faster, but that's not going to be a common issue for most people.

HCI Memtest is still likely the best memory/NB test overall though.

LinX/IBT will run your CPU hotter than anything else though, assuming you are using the max RAM test.
 

AdamK47

Lifer
Oct 9, 1999
15,170
2,830
126
This is a VERY nice OC testing program. I do notice it's not using 100% of the CPU. It jumps around all 8 logical CPUs. Perhaps this and Prime95 together would be great combination.
 

dguy6789

Diamond Member
Dec 9, 2002
8,558
3
76
My machine passes this no problem. Temps never go above 60C! Q9550 at stock voltage ftw!
 

AdamK47

Lifer
Oct 9, 1999
15,170
2,830
126
The temps don't get hot at all for me. The highest I've seen with this is 66 degrees and that was for a split second. Most of the time it was between 60 and 63. With Prime95 using 8 threads the temps are stable between 70 and 73. These are temps recorded with SpeedFan. Add around +10 for RealTemp temps.
 

ch33zw1z

Lifer
Nov 4, 2004
37,734
18,004
146
Originally posted by: AdamK47
This is a VERY nice OC testing program. I do notice it's not using 100% of the CPU. It jumps around all 8 logical CPUs. Perhaps this and Prime95 together would be great combination.

Pretty much how I do it too. So far so good :thumbsup:
 

VirtualLarry

No Lifer
Aug 25, 2001
56,229
9,990
126
Originally posted by: AdamK47
The temps don't get hot at all for me. The highest I've seen with this is 66 degrees and that was for a split second. Most of the time it was between 60 and 63. With Prime95 using 8 threads the temps are stable between 70 and 73. These are temps recorded with SpeedFan. Add around +10 for RealTemp temps.

I've noticed that too. I find it strange that some people claim that linpack causes temps higher than prime95, because I just don't see it.
 

n7

Elite Member
Jan 4, 2004
21,303
4
81
I've found temps higher with LinPack than P95 on every CPU i've tested with...not sure why not everyone does, but i know it's true for most people.

I use that + P95 + HCI Memtest for stability checking though...i don't rely on just one program.
 

dguy6789

Diamond Member
Dec 9, 2002
8,558
3
76
Just going to confirm that temps are MUCH higher when using this program with all of the available memory versus the default.
 

n7

Elite Member
Jan 4, 2004
21,303
4
81
Originally posted by: dguy6789
Just going to confirm that temps are MUCH higher when using this program with all of the available memory versus the default.

Thanx for confirming what i'd noticed.

I suspect many people are not choosing max RAM when testing, which results in less stress, likely due to the problem size.

 

ch33zw1z

Lifer
Nov 4, 2004
37,734
18,004
146
Linkpack definitely gives me higher temps by at least 5C, just confirming others findings.
 

krose

Senior member
Aug 1, 2004
513
15
81
Linpack does not currently support hyperthreading so you will never achieve 100% load on core i7. That's why your temps are not as high as Prime. Text
 

nyker96

Diamond Member
Apr 19, 2005
5,630
2
81
Originally posted by: dguy6789
Passed 100 passes on maximum memory usage. Took just over 11 hours. :)

congrads, that is very stressful for any system. I think your rig has earned a bit of vacation after this run. :]
 

DasFox

Diamond Member
Sep 4, 2003
4,666
46
91
Originally posted by: n7
No, not really.

If you are setting it to max RAM, it stresses that & NB heavily as well.

For me it seems P95 Blend/Large FFTs find FSB issues faster, but that's not going to be a common issue for most people.

HCI Memtest is still likely the best memory/NB test overall though.

LinX/IBT will run your CPU hotter than anything else though, assuming you are using the max RAM test.


LinX/IBT?

Huh what's this?

Max ram test, where?