LinX error

setter · 25 Jan 2011 at 18:05

Been running LinX on my oc'd i7, with all ram selected, (4206), problem size 23428, first i was running the following.

21x193
1547 mhz ddr3
pcie freq, 101
cas 7-7-7-24
uclk, 3095 mhz
qpi link, 6965 mt/s
1.2750 vcore
1.312500 qpi
1.66 vdimm
1.84 pll
ioh, 1.2
ioh pcie, auto
ich, 1.2
ich pcie, auto

Tried today with 20x200, vcore raised to 1.28125, ram at 1600mhz, 7-7-7-24

uclk, 2208 mhz
qpi link, 7218 mt/s

All other voltages as above, ive ran LinX/IBT on the first settings, (21x193) in the past, however though i had a lesser amount of ram selected.

Previous passed ss.

Failed ss

Cpuz cpu + ram settings.

Thanks for any advice/input.

setter · 25 Jan 2011 at 19:59

Anyone?

setter · 25 Jan 2011 at 20:04

The first set (21 x 193) were tested for 5 hrs ono error free with blend. Just a few mins ago i tried LinX again with memory lowered to 4096, same error.

EDIT: Ran blend on the 20x200 settings, worker thread 5 error after 3 minutes, ram voltage is set at 1.66v, which should be ok, will try another bump on qpi voltage. Increased qpi voltage from 1.312500 to 1.31875, blend running error free so far. Scratch that, worker thread 8 failed after 9 mins.

Raised qpi to 1.32500, failed on thread 4 after 7 minutes.

setter · 25 Jan 2011 at 22:09

Could this be a case of more qpi? The 1.2750 vcore settings at 21x193 have been tested for 8 hours stable with p95 small fft and various passes of IBT/LinX a while back. Currently trying 20x200, small fft, 1.28125 vcore, running almost an hour error free.

setter · 25 Jan 2011 at 22:18

Thanks, currently running vcore at 1.28125, qpi at 1.32500, small fft is fine so far on theese settings, decided to start running at 20 x 200 so as to get a little more from my ram, hence the testing, blend fails on every attempt so far.

setter · 25 Jan 2011 at 23:48

Yep, ill have another mess about with it tomorrow. Tbh, it's stable enough in anything im using it for, mainly gaming and a bit of video encoding. The only crashes/freezes etc ive ever had are caused by my rubbish usb wireless adapter, what is the max reccomended on qpi for 24/7 use?

@game4it, im currently running the ram at 1.66v, 1T, 7-7-7-24, what should i slacken timings to? Thanks guys.

setter · 26 Jan 2011 at 00:10

Thanks, the 20x200 settings ran ok for an hour and a half with small fft before i stopped it, bit of a short run i know, will have to let it run a bit longer. Ill also have a play about with the ram timings etc.

setter · 7 Feb 2011 at 20:01

Bit of an update. Finally found time to have a tweak with settings. Upped vcore to 1.28750, slackened the ram to 9-9-9-27, (previously at 7-7-7-24), 1.66 vdimm, 1.325 qpi. Running prime blend for almost 2 hours error free so far.

Gonna let it run for another few hours, if all's well ill attempt to lower the ram timings again.

setter · 7 Feb 2011 at 20:42

Thats next on the agenda, providing of course that blend runs for a minimum of 8 hours, havent had any bsods when previously running p95 blend, just individual worker threads failing after a certain amount of time, anything from 2 - 10 minutes.

setter · 7 Feb 2011 at 20:50

Thankfully so far this evening it's running a lot longer than its ever done, (2hrs 30 mins and counting), used to run the cpu at 21x193, ram downclocked to 1547mhz, recently changed to 20x200, with ram at 1600 mhz, (ocz reaper c7).

setter · 7 Feb 2011 at 23:33

Scratch that one, failed on thread 7 after 5 hours.

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

setter · 8 Feb 2011 at 00:03

Thinking that it's possibly a case of more vcore, the increase earlier allowed me to run a lot longer than previously, seems like this chip also doesnt like a 20x multiplier. Will run LinX tomorrow and have another playabout.

setter · 8 Feb 2011 at 17:19

Cheers m8, will try another voltage increase, the ram is rated at 7-7-7-24, but ive been testing it with timings slackened to 9-9-9-27 1T, 1.66v, temps with blend maxed at 65c iirc. Expect about 5-6c hotter with LinX.

setter · 8 Feb 2011 at 18:07

Upped vcore to 1.29375 in bios, windows idle, 1.288, load 1.296, ram at 9-9-9-27, 2T cmd rate.

Ran a quick LinX test, (just 5 passes at the minute).

Completed without errors. Will try a 20 pass run after a bit.

setter · 8 Feb 2011 at 20:16

Thanks mate

, la femme is candice swanepoel.

Gave LinX a 20 pass run, again completed without errors.

Looks like im finally getting somewhere, ill let it run prime blend as well over the weekend for another extended run, just to be sure.

setter · 8 Feb 2011 at 20:45

Will try 100 loops on an overnight run, id also like to see about tightening up my memory timings at some point, hoping that the instability was a result of lack of vcore alone, as the ram is running at its rated speed/voltage (1600mhz) just with the timings slackened a bit. Read a few posts on other forums were people were saying that their cpu's required a bit of a jump in vcore when using the 20x multiplier as opposed to 21x.

setter · 8 Feb 2011 at 22:58

Thanks w3bbo, it's looking a lot better than previously, temps are still pretty low as well, on my older s775 quads, LinX/IBT tended to run 5-7c higher than prime.

setter · 8 Feb 2011 at 23:16

Ill certainly give it a try, thanks for your advice.