Core i7 & i5 (1156 & 1366): "Striving For Ultimate Stability" Club - Overclock.net

Clockadile Dundee · 10-09-09

Edit:

Minimum of 4 GHz i5 or i7, 1156 & 1366. Any base clock or multipliers you want to use to get there.

1. This is listed first, but you can open cpu-z last. You need to have main cpu and memory tabs (two incidences). Have an open notepad, and write the current date and OC.N name on there. You need these three things in each of up to 4 screen captures, with the test results. (LinX, 22h Blend, 4h large FFT, SuperPi 32M time). There is no reason you cant open these three after each test is done, right before the screen capture is taken.

2.LinX. You may leave ~1.2 GiB free for the OS. At 6 GiB, this means 25,000 + problem size. You may also use "all memory" as long as you have enough RAM free. The more GiB of RAM you are trying to stabilize, the more CPU VTT bump you will need(on Evga mobos)(A.K.A. QPI/DRAM VTT on Asus mobos).

Memory
2 GiB 300 runs
3 GiB 260 runs
4 GiB 210 runs
6 GiB 170 runs 25,000+
8 GiB 155 runs
12 GiB 140 runs
*OR Prime95 , see note A

3. Prime95 (Custom - Duration : 22 Hours) - Equivalent to Blend
-Min. FFT size (in K) : 8k
-Max FFT size : 4096k
-Memory to use (in MB) : Max available in OS -100 MB

4. Prime95 (Custom - Duration : 4 Hours) - Equivalent to Large FFTs
-Min. FFT size (in K) : 4096k
-Max FFT size : 4096k
-Memory to use (in MB) : Max available in OS -100 MB

5. SuperPi 32M time. Flexibility is allowed for achieving the optimal time, but please use the same CPU core clock or greater as used for the other tests.

*Notes
---------------
A : This may be run as an alternative to the LinX runs :
4. Prime95 (Custom - Duration : 22 Hours) - Equivalent to Small FFTs , optimized for i7
-Min. FFT size (in K) : 8k
-Max FFT size : 32k
-Memory to use : see below, check run in-place
-Must check "run FFTs in place"

That is all. Any questions or suggestions, feel free.

We all may still refine the requirements, but don't worry about that, test away, it will count.

Members:

DannyM
Core i7 920 @ 4200 MHz HyperThreading Enabled (8 Logical Cores)(21x*200 MHz)
>TjMax 75c-72c-72c-70c RAM = 800 MHz 8-8-8-24-59-2T

[LinX 300x]

[Prime95 4096k 4h] [Prime95 8k-4096k 22h][SuperPi 32M : 8m 55.845s]

GigaByte : Core i7 920 @ 4009 MHz HyperThreading Enabled (8 Logical Cores)(19x*211 MHz) RAM = 844 MHz 8-8-8-24-94-1T

[LinX 170x][Prime95 4096k 4h][Prime95 8k-4096k 22hr][SuperPi 32M : 9m16.348s][Linx 30x (150 ps cpu skew, 50ps IOH skew)]

Chicken Inferno
Core i7 920 3849B018 @ 4536 MHz HyperThreading Enabled (8 Logical Cores) (21x*216 MHz) >TjMax 82c-78c-78c-76c
RAM = 864 MHz 9-9-9-24-88-1T

(Awaiting final submissions)

[LinX 200x & SuperPi 1M : 9.038s]

------------------------------------------------------------------------------------------------

WARNING 1156 Users : Make sure you don't have the suspect 1156 socket before attempting these tests with 4.3 GHz + . Just because you don't have a Foxconn mobo, doesn't mean you don't have a Foxconn socket.
http://images.anandtech.com/reviews/...burnedpadS.jpg

AnandTech Article:
http://www.anandtech.com/mb/showdoc.aspx?i=3661

OC.N Discussion Thread
http://www.overclock.net/hardware-ne...heck-your.html

Here are instructions for freeing up memory before running the tests:

Quote:

Originally Posted by ChickenInferno

Running LinX through a single run of All Memory will clear up alot of RAM too. Let it finish 1 pass, stop it, Click All Memory off and then on to show how much more you have available, Subtract 100mb off of that number, input that into prime95.

GigaByte · 10-09-09

Large FFTs tests the IMC more than blend as blend has alot of small FFT mixed in, though Memtest 86+ is still king for IMC and RAM testing. Do you want all of those done or..? I suggest the 200 runs with LinX be broken down like so:

2GB RAM: 260 runs all memory
3GB RAM: 205 runs all memory
4GB RAM: 165 runs all memory
6GB RAM: 120 runs all memory
8GB RAM: 90 runs all memory
12GB+ RAM: 50 runs all memory

Regarding Prime95 I suggest the following:

Large FFTs: 14 hours

OR

Blend: 12 hours

AND

Small FFTs: 15 hours

To make it clear a submission with 6GB RAM will include the following:

1. 120 runs or more of LinX all memory
2. Prime95 Large FFTs for 14 hours or more OR Prime95 blend for 12 hours or more
3. Prime95 Small FFTs for 15 hours

dham · 10-09-09

I believe a prime 95 blend for 12 hours is the most adequate solution. Throw in a LinX all memory and call it a day.

FtW 420 · 10-09-09

I'm all for stability, but days of 12 hour & 24 hour stability runs? I wanna torture & burn my chip benching, not just stability testing.
Does memtest do much testing on memory that the other apps don't cover? Also a big part of stability, prime says lots of ram tested in blend but do any of them completely test the memory?

FSF-Foxhound · 10-09-09

woot finally a club that is honest.

updating post in a bit, OP use a google spreadsheet

typoknig · 10-09-09

I would go 24 hours Prime95 on Blend settings for a few reasons:

1. Blend is the default setting for the Prime95 stress test.

2. 12 hours of Prime95 is not enough. I have had failures past 13 hours of testing before (I posted about it in the last i7 stability thread).

3. Linx is ok for a quick test, but it cannot hold a candle to Prime95. One reason is that Prime95 is crunching ONE monster number that can take months to calculate fully. Linx is crunching lots of small numbers (well smaller than those used by Prime95) so the test stop and start a lot, giving the processor a chance to breathe. You do not want to let the pressure off your processor if you are "stress" testing it, that would defy logic.

The only problem I can see with running Prime95 is the verification. I do not know any way to show that you actually ran Prime95 for a specific period of time, the results.txt file will not cut it, that can be easily duplicated. Does anyone else know a way?

GigaByte · 10-09-09

Quote:

Originally Posted by FSF-Foxhound

woot finally a club that is honest.

updating post in a bit, OP use a google spreadsheet

Well don't run your 2 day stress test yet

Its not finalized what the requirements are yet.

FSF-Foxhound · 10-09-09

Quote:

Originally Posted by typoknig

I would go 24 hours Prime95 on Blend settings for a few reasons:

1. Blend is the default setting for the Prime95 stress test.

2. 12 hours of Prime95 is not enough. I have had failures past 13 hours of testing before (I posted about it in the last i7 stability thread).

3. Linx is ok for a quick test, but it cannot hold a candle to Prime95. One reason is that Prime95 is crunching ONE monster number that can take months to calculate fully. Linx is crunching lots of small numbers (well smaller than those used by Prime95) so the test stop and start a lot, giving the processor a chance to breathe. You do not want to let the pressure off your processor if you are "stress" testing it, that would defy logic.

The only problem I can see with running Prime95 is the verification. I do not know any way to show that you actually ran Prime95 for a specific period of time, the results.txt file will not cut it, that can be easily duplicated. Does anyone else know a way?

you use both LinX and prime95, because they stress the cpu differently. It is needed.

edit:
oh I'm in canada right now, so I've got plenty o time to wait.

typoknig · 10-09-09

Quote:

Originally Posted by FSF-Foxhound

you use both LinX and prime95, because they stress the cpu differently. It is needed.

edit:
oh I'm in canada right now, so I've got plenty o time to wait.

Almost every time I pass a Linx test I fail in Prime95. I think you are right about them stressing the CPU differently, Prime95 stresses the CPU a lot and Linx just kinda stresses the CPU. When I can run Linx for 2 hours with problems bigger than 25k and pass, then run Prime95 for 10 seconds and have my machine crash, well, I think that shows which program is doing a better job of stress testing.

Plus if the is the *ULTIMATE i7 STABILITY CLUB* then the test to get in should be crazy hard to pass.

GigaByte · 10-09-09

Every time I got Prime95 going over 9 hours I never failed LinX, up to 50 runs. Same thing vice versa.

		Overclock.net - Overclocking.net > Intel > Intel CPUs
Core i7 & i5 (1156 & 1366): "Striving For Ultimate Stability" Club

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)