3570K 4.5Ghz
Attachment 129697
Printable View
3570K 4.5Ghz
Attachment 129697
9216 was decided on because it is(was?) the maximum memory that LinX can address before suffering GFlop dropoff when HT is enabled.Quote:
Originally Posted by dr_dx
It is always best to use the most ram possible. If you do not, you are not testing stability. I have seen on a number of occasions myself where an OC I have will pass with 5120MB, but when set to 9216 will fail.
Above 9216MB stress drops drastically.
If LinX was the ultimate stability test proving an absolute stability in a system I would be more inclined ta agree with a "All memory" or as much as possible requirement. But, we all know that neither LinX nor Prime testing provides any such guaranty.
Therefor LinX testing with a time limit and reasonable amount of memory makes just as much sense to me :shrug:
Then, with such a debate now raging, how is one to know just what actually will constitute a valid qualifying run?
Also, without trying to fuel the conflagration further, is CPU-Z a requirement? It doesn't seem to get valid readings with this particular i7-3820/DX79SI combo. The Vcc, for example, seems permanently locked at 1.000 Volts and the RAM/SPD tabs are totally off the wall. I tend to stick with Intel's own Tuning Utility to get proper read-outs...
LinX is superior to any other stress program out there. That does make it 'the ultimate' stability test.
I have yet to find an overclock that will pass a 9216MB LinX that won't run 24/7/365, unless there's a stability issue in one of my GPU's. Which of course has nothing to do with my CPU overclock.
The majority of people who claim LinX isn't good are just flat out doing it wrong. They're either using outdated binaries, or not all memory/diagnostic mode.
CPUZ is a requirement.
How you're supposed to know what a valid result is? read the thread.
I have read the OP, which is where the "rules" have been initially stated. However, I have NOT read all 92 pages to see how many times things have changed since the author left.Quote:
Originally Posted by zoson
And I'll certainly be glad to use CPU-Z, since it will make my system look sooo much better and have everyone wonder just how I managed such a feat on battery power and the world's slowest RAM:
Attachment 129751
But that may be after I remove all of my Quad-channel RAM for Dual, since it's only 4 GB and I won't have to be hobbled by incredibly long qualifying times.
And thanks very much for crediting me with nothing but frivolous questions. I believe I'll pass on this offer...
Just to show that LinX 0.64 is not as stressful as prime95 27.7 in terms of vCore...
3770K @ 5.0GHz
LinX for 1 hour @ 1.416v... passed
http://i458.photobucket.com/albums/q...VY/501416v.jpg
LinX for 7 hours @ 1.448v... 1.416v-1.434v BSOD
http://i458.photobucket.com/albums/q...VY/linx7hr.jpg
straight away run Prime95 at 1.448v... failed in 1 hour +
http://i458.photobucket.com/albums/q...imefail1hr.jpg
bump up to 1.48v... passed 4 hours + prime95
http://i458.photobucket.com/albums/q...VY/prime50.jpg
I also kinda wonder where this magic figure of 9216 MB comes from, since there doesn't seem to be a way to set it in my copy of LinX. There is not a direct mem setting, nor does choosing any available problem size set it. I would second having to point that "requirement" out in the thread as well rather than just claim no one knows how to properly use LinX in the first place.
Again - since the author has left and there is no real control here, and no two people seem to quite agree on anything...
...what's the use?...
EDIT -
BTW, LinX's most useful feature is it's ability to run multiple instances (face it - it's just an interface to Intel's LinPack, a feature shared by myriad stress tests outside of Prime95) as well as setting the thread count and problem size. Here's an example of how I find it handy when testing a rig's stability, where I've tried to push not only the CPU freq but to push out the Turbo capabilities as well:
The rig's set for ops at 1.25x BCLK, and on the CPU I use 32x (4.000 GHz) for boot with the four available Turbo slots set for 37-40x (4.625-5.000 GHz)
Test 1 - Single-Core speed of 5.000 GHz using a small problem size to maintain speed on the CPU. (It's hard to "lock in" the true single-thread speed with the OS in the background, so I grabbed a shot with sensor lag spread between the two monitors showing both speeds as it shifted. The graph line shows it sticking to that range during the run.)
https://skydrive.live.com/redir?resi...021B73932!1091
Test 2 - Locking in the 2-Core stepping at 39x. A larger problem size for more use of RAM, but not so much that it kicks in hard enough to drive the system core count higher than needed)
https://skydrive.live.com/redir?resi...021B73932!1093
Test 3 - 3-Core speeds, and a little more headroom for RAM use without kicking in that fourth core.
https://skydrive.live.com/redir?resi...021B73932!1092
Test 4 - things get a bit more interesting here. A full eight threads, but with three instances of LinX doing the same while running simultaneously. Emphasises multi-tasking as well as multi-threading in one shot.
https://skydrive.live.com/redir?resi...021B73932!1094
But I could do all of the same things with Intel Burn Test. Or even OCCT - and get all of the features of LinPack as well as incredible logging features.
LinPack is LinPack, that's all...
You can set it in the Memory(MiB) textbox... Just unclick the 'all' button and type it in.
9216MB, as I explained before, is the maximum memory linpack can address while HT is enabled and maintain full performance... At least that's how it is on Gulftown and lower. I haven't had time to test my 3930k yet.
Also, owikh84... You're obviously not using updated linpack binaries, AND you're not using as much memory in the LinX test as your Prime95 test. So yeah, apples and oranges comparison from someone who doesn't know how to operate LinX doesn't prove anything other than your lack of skill.
Then consider this post a placeholder until I can generate a run in the morning. I'll edit it in so as not to add another to this already lengthy conversation...
We can see your memory usage in your prime95 screenshots and in both it's greater than your memory assignment in linx... So yeah, nice try. Also, all of the older members of this forum will laugh you off the forums for thinking you've done anything 'more' or 'better' than I have.Quote:
Originally Posted by owikh84
And if you are using the latest binaries, your overclock obviously has something wrong with it seeing as in the last two pages everyone at 5GHz is 2-4GFlops higher than your performance.
Lie more noob.
Lastly, if there were ANY truth to what you were saying, Intel would stop making LINPACK and just test with Prime95 instead. But oh yeah... They do make LINPACK, and do their own stress testing with it.
Just asking.......Quote:
Originally Posted by owikh84
Is the LinX memory over 7Gb and the Prime memory 6Gb?
Thanx
LinX >7GB, Prime95 7GBQuote:
Originally Posted by newhit
Dr_Dx | Core i7 3770K [4C/8T] @ 5300.3MHz | 1.488v | ASUS Maximus V Formula| Water | L221B009 | LinX 0.6.4 AVX w/ Linpack 10.3.11.019 | Liquid Pro 7168 RAM
Attachment 129812
Dr_Dx | Core i7 3770K [4C/8T] @ 5300.3MHz | 1.488v | ASUS Maximus V Formula| Water | L221B009 | LinX 0.6.4 AVX w/ Linpack 10.3.11.019 | Liquid Pro 9216 RAM
Attachment 129813
All else identical except tested RAM amount - GFlops are lower with 9216 RAM...Looks like 9216 is not the "best" amount for an IB...
That's within margin of error from boot to boot. 125.79 vs 126.46... Less than 1 GFlop.
Dr_Dx | Core i7 3770K [4C/8T] @ 5000.1MHz | 1.288v | ASUS Maximus V Formula| Water | L221B009 | LinX 0.6.4 AVX w/ Linpack 10.3.11.019 | Liquid ProQuote:
Originally Posted by zoson
Same Boot @ 5Ghz in order of GFlop performance
Attachment 129814
9216 is way down the list (there were actually 2 other mem settings that were faster - didn't have the room to show), but it's clear 9216 is not the setting to give max GFlops.
Zoson - You have yet to show us the post # of where Zalbard or the "community" decided on 9216 for the RAM setting. Several people have provided accurate measurements; all you have provided is hearsay.
It's time to put up or shut up. Otherwise folks will think you are just a TROLL.
While I'm still tryin' to get more than 20 minutes' stable here (a bit trickier than I imagined based on how well the rig runs with a 5 GHz top Turbo), I have to admit to being stumped on everyone's incredible through-put rates. I seem to enjoy but an embarassing 50% of that, and I had thought this system to be designed for such. Apparently not so successful at it, either.
I would be most appreciative if anyone would be willing to point out some specific posts or threads on this. You need not walk me through a hand-holding session - just point out where I can do some relevant research, if you would...
Do a thread search for 9216 you lazy bastard.
You're the one who's late to the party, not me. You have to do your research and homework.
It also looks like the reason why we chose 9216 may have been fixed. There are TONS of results showing the issue in this thread.
Lazy Bastard? Lying Noob? That's the pot calling the kettle black. Your search button broken? There are 15 posts in this thread (most of them recent) that have "9216" in them and none of them show where Zalbard or the "community" decided on 9216. You started a discussion in 2011 about it but no one agreed to it. Zalbard updated the OP in 4/2012; If 9216 had been decided upon, he would have put it in. And now you are giving people a hassle over something that isn't even an issue anymore but you were too lazy to verify was still valid.Quote:
Originally Posted by zoson
I am sorry that you can't get your 3930K onto the list but that is no reason to take out your frustration on everyone else that is trying to comply with the requirements set out in the OP. Try being helpful, instead of critical and calling people names. It's because of TROLLS like you that this thread is dying.
I agree on this, after I got my 3960X I saw it much harder to stabilize P95 than LinX. Well of course LinX GFlops may tell if something is really wrong, but usually I got a BSOD in P95 right after one hour or so even if LinX passed with flying colors. My 990X on the other hand was a different story, LinX was more stressfull.Quote:
Originally Posted by owikh84
On my 990X, after I upgraded from 6 to 12 GB, the max mem I could use with LinX was that 9216MB, otherwise I would get a radical drop in GFlops.
At no time did I say zalbard decided. just the community. And it's in there if you read the discussion. Cryptik, donmarkoni, newhit, xarot, and several of the big players in this club all agreed based on the results from our testing. This was all started by people noticing that there was a sharp GFlop dropoff between the 8192MB and 10240MB settings.Quote:
Originally Posted by dr_dx
lol, my 3930k isn't in the club because I didn't bother trying. It's an HTPC and I don't need it running that fast. It's got an air cooler on it... Way to attempt an insult and fail miserably.
Also, way to not address the fact that the whole reason why we decided on 9216 may be fixed now.
LOL@U. Keep trying, but you're just a failure.
Bumping the threat with Limpak updated libraries ver.11.0.1.005
You need to copy also the libiomp5md.dll file from redist directory for 32 or 64 Win u use in Linx folder.
ASUS P6X58-E WS, Water Cooled, i7-980X, 3045A974, 4.81GHz, 1.536V, 101.8569 GFlops
Very nice values. Just water?Quote:
Originally Posted by big.al.v
ha ha ha ha
I7-8086K