Strategic Double Clicking

[Mark Rose]

Originally Posted by storm5510

How are you using memory sticks in context with this project?

I have six of them. It's needless to say that I have never had one go bad.

I've installed ClearLinux onto them, for running mprime (Clear better than Ubuntu, and I get better performance). There isn't a heavy write load, but after many months I'm getting read errors, which leads to things crashing.

I don't think they were intended for heavy usage.

[storm5510]

Originally Posted by Mark Rose

I've installed ClearLinux onto them, for running mprime (Clear better than Ubuntu, and I get better performance). There isn't a heavy write load, but after many months I'm getting read errors, which leads to things crashing.

I don't think they were intended for heavy usage.

I see, and no, I doubt they were ever intended for heavy use. Is there a possibility of ClearLinux creating a ram-drive, and using that for read-write processes? This would really cut down the stress on the stick.

[Mark Rose]

Originally Posted by storm5510

I see, and no, I doubt they were ever intended for heavy use. Is there a possibility of ClearLinux creating a ram-drive, and using that for read-write processes? This would really cut down the stress on the stick.

The only thing doing any real writes on there is mprime saving every half hour. There is a minor bit of logging (couple MB per day).

Now these weren't high-end branded sticks. Still, ~5MB * 48 = 240MB per day, and after a year, is 87.6 GB. On a 4 GB stick, that's writing the entire thing 22 times, but I don't think these drives would actively be wear-levelling, and I wouldn't be surprised if some areas were hit harder.

[Madpoo]

Originally Posted by storm5510

@Madpoo: I was looking at your column headers in post 1589. What do Sus, Solo, and Mis mean?

Sus = suspect: count of unverified results from that CPU that had known errors during the run
Solo = count of tests by that CPU that haven't had a double-check at all yet
Mis = mismatch: tests from that computer that have been double-checked but mismatched

Those things can sometimes help determine the overall healthiness of a machine. Well, more the "suspect" results at any rate. High # of mismatches could just be from the people in this thread testing results we think were bad.

If a machine has more bad than good, sometimes it's helpful to see how many unverified results it has out there... if it has 1 bad, 1 good, and 1 solo result, yeah, sure, it's probably worth doing a double-check on that solo result. If it has 40 unverified results, I'd probably do a few cherry picked tests of it's smallest exponents, or break it down more by year/month and see if any trends develop. It could have just had a bad day and everything else from it is good.

Ultimately it's up to whoever might be interested in picking up any of these for an assignment... I just include it to help make that decision.

[storm5510]

Originally Posted by Mark Rose

...Now these weren't high-end branded sticks. Still, ~5MB * 48 = 240MB per day, and after a year, is 87.6 GB. On a 4 GB stick, that's writing the entire thing 22 times, but I don't think these drives would actively be wear-levelling, and I wouldn't be surprised if some areas were hit harder.

Actually, that's not a lot. I have a 64 GB Sandisk plugged into the front of a Blu-Ray player. It has documentaries on it. It generally sees three to five hours of playback per day. I've had it about six months. Strange as this may sound, I cannot concentrate in a quiet room.

I've always wondered if simply being energized a lot shortens their life, like a light bulb.

Originally Posted by Madpoo

Sus = suspect: count of unverified results from that CPU that had known errors during the run
Solo = count of tests by that CPU that haven't had a double-check at all yet
Mis = mismatch: tests from that computer that have been double-checked but mismatched

Well, I got one right: Suspect.

Solo is an interesting category. No verified tests. A person might think that there would not be many of these. Your latest list proves the opposite. There needs to be a lot more DC work performed. There are some which do not care for the amount of time it takes, and would not, or could not, shell-out for a high-end system, or GPU,

Off-topic: There is one little detail which might clear-the-waters a bit for those, like myself, who peruse the recent data. The "Type" column on both pages uses "C" to indicate a composite value. If it could use "CU" to indicate unverified, and "CV" to indicate verified, this would cut back on page reloads needed to return to the main listing on each page. Just a thought...

[moebius]

Originally Posted by moebius

Horrible result! I reacteded accordingly and have lowered the DRAM latencies and I started another Double-check.

Seems to be solved.. This Double-Check was successfully again.

44009333

[Madpoo]

Here's a new "top 100" bad/good and then re-ordered by exponent size:

exponent	Bad	Good	Unk	worktodo
41696293	1	1	2	DoubleCheck=41696293,72,1
41719147	2	2	2	DoubleCheck=41719147,72,1
42733279	1	1	2	DoubleCheck=42733279,72,1
43045133	1	1	2	DoubleCheck=43045133,72,1
43344919	1	1	2	DoubleCheck=43344919,72,1
43592701	1	1	2	DoubleCheck=43592701,72,1
43642481	1	1	2	DoubleCheck=43642481,72,1
43882621	1	1	2	DoubleCheck=43882621,72,1
44013859	1	1	2	DoubleCheck=44013859,72,1
44448641	1	1	2	DoubleCheck=44448641,72,1
44514427	1	1	2	DoubleCheck=44514427,74,1
44588273	1	0	2	DoubleCheck=44588273,72,1
44742949	2	2	2	DoubleCheck=44742949,72,1
44842979	1	1	2	DoubleCheck=44842979,72,1
45286627	1	1	2	DoubleCheck=45286627,72,1
45324871	1	0	1	DoubleCheck=45324871,72,1
45325499	1	0	1	DoubleCheck=45325499,72,1
45470323	1	0	1	DoubleCheck=45470323,72,1
46105831	1	1	2	DoubleCheck=46105831,72,1
46113547	1	1	2	DoubleCheck=46113547,72,1
46119443	1	1	2	DoubleCheck=46119443,72,1
46571467	1	0	1	DoubleCheck=46571467,72,1
46905743	1	1	2	DoubleCheck=46905743,72,1
46956491	1	1	3	DoubleCheck=46956491,72,1
47084671	1	1	2	DoubleCheck=47084671,72,1
47668079	1	1	2	DoubleCheck=47668079,72,1
47747389	1	1	1	DoubleCheck=47747389,72,1
47796143	1	0	1	DoubleCheck=47796143,72,1
47839741	1	0	2	DoubleCheck=47839741,72,1
48066199	9	8	2	DoubleCheck=48066199,72,1
48095123	1	1	2	DoubleCheck=48095123,72,1
48165989	1	1	1	DoubleCheck=48165989,72,1
48170179	1	1	2	DoubleCheck=48170179,72,1
48194101	1	1	3	DoubleCheck=48194101,72,1
48523327	1	0	2	DoubleCheck=48523327,72,1
48525839	1	1	2	DoubleCheck=48525839,72,1
48628457	1	1	2	DoubleCheck=48628457,72,1
50040247	1	1	2	DoubleCheck=50040247,73,1
50536769	1	1	2	DoubleCheck=50536769,73,1
50537183	1	1	2	DoubleCheck=50537183,73,1
50801873	1	0	2	DoubleCheck=50801873,73,1
50847029	1	1	1	DoubleCheck=50847029,73,1
51628531	1	1	2	DoubleCheck=51628531,73,1
51725251	1	1	1	DoubleCheck=51725251,73,1
51900467	1	1	1	DoubleCheck=51900467,73,1
51907969	1	1	2	DoubleCheck=51907969,73,1
51945079	1	1	3	DoubleCheck=51945079,73,1
52078027	1	1	2	DoubleCheck=52078027,73,1
52105457	2	2	2	DoubleCheck=52105457,73,1
52146673	1	1	2	DoubleCheck=52146673,73,1
52503727	1	1	2	DoubleCheck=52503727,73,1
52562707	1	1	2	DoubleCheck=52562707,73,1
52638863	1	1	2	DoubleCheck=52638863,73,1
52736503	2	2	1	DoubleCheck=52736503,73,1
52902649	2	2	2	DoubleCheck=52902649,73,1
53035573	1	1	2	DoubleCheck=53035573,73,1
53035649	1	1	2	DoubleCheck=53035649,73,1
53156867	1	1	2	DoubleCheck=53156867,73,1
53191093	1	1	2	DoubleCheck=53191093,73,1
53232449	1	1	2	DoubleCheck=53232449,73,1
53329681	1	1	2	DoubleCheck=53329681,73,1
53364041	1	1	1	DoubleCheck=53364041,73,1
53520547	1	1	2	DoubleCheck=53520547,73,1
53520613	1	1	2	DoubleCheck=53520613,73,1
53783837	1	1	2	DoubleCheck=53783837,73,1
54033407	1	1	3	DoubleCheck=54033407,73,1
54106807	1	1	2	DoubleCheck=54106807,73,1
54479993	1	1	3	DoubleCheck=54479993,73,1
54497749	3	3	2	DoubleCheck=54497749,73,1
54542801	1	1	1	DoubleCheck=54542801,73,1
54674621	3	3	2	DoubleCheck=54674621,73,1
54998431	1	1	2	DoubleCheck=54998431,73,1
55008181	2	2	2	DoubleCheck=55008181,73,1
56141483	1	1	2	DoubleCheck=56141483,73,1
56152133	1	1	2	DoubleCheck=56152133,73,1
56389639	2	2	3	DoubleCheck=56389639,73,1
56389973	2	2	3	DoubleCheck=56389973,73,1
56477437	1	1	2	DoubleCheck=56477437,73,1
56492083	1	1	1	DoubleCheck=56492083,73,1
56565331	1	1	1	DoubleCheck=56565331,73,1
56845619	1	1	3	DoubleCheck=56845619,73,1
56875199	1	1	1	DoubleCheck=56875199,73,1
56983049	2	2	3	DoubleCheck=56983049,73,1
57110591	1	1	1	DoubleCheck=57110591,74,1
57164689	1	1	3	DoubleCheck=57164689,73,1
57278267	1	1	2	DoubleCheck=57278267,73,1
57291181	1	1	2	DoubleCheck=57291181,73,1
57591487	1	1	2	DoubleCheck=57591487,73,1
57591773	1	1	2	DoubleCheck=57591773,73,1
57629809	1	1	3	DoubleCheck=57629809,73,1
57771949	1	1	1	DoubleCheck=57771949,73,1
58084207	1	1	1	DoubleCheck=58084207,73,1
58123129	1	1	1	DoubleCheck=58123129,73,1
58123693	2	2	2	DoubleCheck=58123693,73,1
58567319	1	1	2	DoubleCheck=58567319,74,1
58595063	1	1	2	DoubleCheck=58595063,73,1
58778611	1	1	2	DoubleCheck=58778611,73,1
58952743	1	1	2	DoubleCheck=58952743,73,1
59571931	1	1	2	DoubleCheck=59571931,73,1
59573699	1	1	2	DoubleCheck=59573699,75,1

[kladner]

Taking these:
41696293 1 1 2 DoubleCheck=41696293,72,1
41719147 2 2 2 DoubleCheck=41719147,72,1
Registered. ETA about 52 hours for each, concurrently on 2 worker windows, with 2 cores each. i7-6700k, 4.3 GHz, Kingston 2666 MHz RAM at 3200 MHz.

[storm5510]

DoubleCheck=xxxxx,42733279,72,1

Successfully registered with the server. 60 hours on my HP.

[storm5510]

DoubleCheck=xxxx,43344919,72,1

Also registered this one. The first is 23% complete.

[kladner]

Originally Posted by kladner

Taking these:
41696293 1 1 2 DoubleCheck=41696293,72,1

Matched.
22 hours to 41719147. I am back to running all 4 cores on one worker, after a week with two with 2 workers. The difference seems negligible. In fact, 2 workers with 2 cores each was hitting 4.28 ms/it, while 1 worker with 2 cores gives about 3.8 ms/it. All 4 on the same range exponent is doing 2.11.