Is Prime95 just pulling the wool over my eyes?

So, odd issue. Prime 95 runs in place ffts (mostly processor) perfectly for hours and hours and hours, no problems in sight. However, the blend test returns the "Hardware failure detected, consult stress.txt file.
FATAL ERROR: Rounding was 0.5, expected less than 0.4" error.

So I narrow down the possibilities. (the only reason i ran prime95 was because my system randomly restarted when browsing the web - after running hours of crysis without fail!)

One thing I did was bump my memory voltage from 1.64 to 1.65 to get it up to spec.

I ran memtest86 and it reported no errors at all. I ran the windows memory diagnostic tool and it also report no errors. My powersupply is MORE than adequate at 750w(antec) for a single GPU config. The 12V connection shows no less than 12.04v

So this testing tells me my cpu + ram are fine. My PSU seems fine too. There are no overheating problems and that wouldnt make sense because the restart happened doing idle web browsing. Do you think the memory voltage bump could have fixed it or could there be another underlying issue?

 

The memory bump could have done it, but what is this doing in the notebook forums (and in the sager and clevo section)?

 

lol i trust the guys opinions in this section more than anywhere else

 

The same thing just happened to me!

I ran RMclock for undervolting for a few months now and I had everything set up with a maximum voltage of 1.1250 on the T9900.

However last night while doing a random check-up I got the FATAL ERROR: Rounding was 0.5, expected less than 0.4.

So I disabled RM and started running orthos for the CPU and it was ok; Then on the Kingston hyper-x 5-5-5-15 it gave the same error @ 2-3 minutes...

Things I've done since the sistem worked right: updated all drivers, even card reader ones, installed motorola modem (junk I know), tricked XP into AHCI mode.

All my temps are normal: ACPI never above 75, CPU below 60, CPU+VGA ~70, HDD 50-60, all this in Celsius and under stress (orthos/3dmark2006)

So ? is the ram failing?? it's no more than 2 months old... all other components are new and I hever had an error in a game EXCEPT GTA4 wich gaver random blue screens (yes even without undervolting).

Any solutions greatly appreciated. Going for a memtest, brb.

 

That's definately a red light for bad memory, vadimpelau. Call up Kingston and get yourself an RMA.

 

ok... on the first mem test it just rebooted, it almost didn't get to start.

The second one I ran for about 30 minutes and seemed ok... but anyway I'm gonna contact the shop I bought them from.



I'm now testing each module on it's own, so here's a last picture before I send them in; the first one failed @ 8 minutes and the other one at 19 minures of orthos ram testing.
Also the ram temp is about 75C isn't that a little hot?

Just a noobish question, is that fsb:ddr ratio right?
How about the frequency? I sure as hell don't see 1066 anywhere

 

The ram is DDR (Double Data Rate) so 533MHz = 1066MHz effective.

Test each stick separately with memtest for a full pass (will take a long time, but it's pretty much a certain test)

 

KHX8500S3ULK2/4G this was the model number but I can't test them anymore because I already sent them in.

So what memory do you guys recomend?
Is another 5-5-5-15 try worthy or should I get a regulat 7-7-7-20 ?

 

If you sent it in already, you're going to get a good kit back. Kingston doesn't screw up with RMA's. 5-5-5-15 is better so just wait for it to come back.

 

Well... I actually sent it to the shop where I bought it from.
Kingston policy seems a bit different for europe:
https://shop.kingston.com/Web_RMA/rma_europe.asp?no_of_kinds=1

I just hope everything turns out fine.

 

What shop did you buy it from?

 

Some local shop, no big chain or well-known...

But this is kingston's rma policy for europe it seems.

 

New modules here...

I still get the errors, memtest doesn't even start!

Any other ideas??

 

5-5-5-15 timings are pretty tight for this chassis. Have you got any other RAM that runs at CAS 6 or 7 that you could try?

 

Nope :| but I was thinking to check them at the nearest shop and also check higher latency modules.

memtest (ran it 3 times) failed @ 35,% 40%, 85% and I checked the ram... they were as hot as the ACPI heatsink and no, I could not hold my finger on a module.

Should I try them again with proper ventilation?
*They pass in orthos with an open case but thy still fail in memtest 2.11*

Is there a way I could increase the latencies?

Can high temp damage them permanentley?

Also when I got this machine they worked perfectly... (tested with orthos NOT memtest)

THX

 

Well, i'll tell my bad experience too:

i tried these Hyperx module in a Clevo M860TU / Quadro FX2700M (i'm not a gamer)

1st kit : failed in memtest86+ v2.11 from the [1st pass - 5th & 6th test]. But when tested module by module, one get no errors and the other get errors from [2nd pass - 5th test].

i asked Kingston for RMA, they sent me another kit because i can't return only one module.

2nd kit : memtest86+ failed from [2nd pass - 5th test] with no difference if i install the kit or module by module.

Kingston sent me this reply :

WELL CLEVO LAPTOPS NORMALLY IS NOT CONSIDERED AS BRANDED, ITS RATHER A NON SPECIFIC SYSTEM

 

Hmm... so now a major ODM brand is concidered "off-brand"? I'll think twice about buying Kingston now.

 

Well I managed to bypass (in orthos) with memset (5-6-6-18) can't change CL but it seems it works well.
Memset with start on boot is a decent compromise if you don't want to risk your warranty by playing in the EEPROM.

I also tried spdTool but some values and voltages are obviously being read incorectly (latencies unknown, voltages 1.8 to 3.3) I changed CL related settings but with no effect.
I restored my original settings, hopefully nothing got broken due the bad read.

Thaiphoon burner seems the ultimate solution and I've now permanently written the EEPROM with JEDEC E standard (on the left, original timings on the right).
I have made two passes of memtest successfully mybe i'll do more when I have the time. I'm sure that the latencies can be lowered but that again would take a lot of time to tweak & benchmark.



Don't forget to backup you readings first! Best of luck.