LinX failing at stock clocks on 2600K

Ice009

Weaksauce
Joined
Jun 22, 2008
Messages
93
Hey guys, turns out I might have a problem with either RAM, CPU or MB.

I just got my first SB CPU (2600K) and MB (Asus P8Z68-V Pro) a couple of weeks ago and was using offset to overclock. I was only going for 4.5-4.6Ghz overclocks. Volts under 1.38V most of the time. Went to about 1.42 a couple of times as I was learning to use off set. Didn't really do any stress testing at that voltage.

Once I worked out what I was doing after some reading on the Asus forums, I set it up to run 4.5Ghz and ran Prime95 blend and that passed so I left it for a week or so. I decided to try to go a bit higher to 4.6Ghz and this time I ran Prime95 blend, but used a custom RAM size of 6500MB (I have 8 GB of RAM) which was suggested to me by a member on another forum. Running it like that I get an error within 5 seconds. I then decided to run Intel Burn Test and it also failed. LinX same thing.

Usually I only run a few passes of LinX or Intel Burn Test then go to Prime95.

Anyway, I decided to load optimized defaults and put everything back to stock and I keep getting an error within 5 seconds of running Prime 95 Blend with custom RAM size. LinX and Intel Burn Test both fail too on stock settings.

I never actually tested at stock settings when I first got the CPU, MB and RAM so it was probably like that from the start.

So what do you guys think is the problem? From the quick Google search I did it seems to indicate that it's a memory problem. I'm running Memtest 4.20 right now, but so far there is no errors at all.

I'm stumped. Not sure if it's the RAM, CPU or MB.
 
I'm more inclined to think it's your RAM.

Run HCI memtest and see what happens.
 
I'm more inclined to think it's your RAM.

Run HCI memtest and see what happens.

What the difference between HCI Memtest and Memtest+ 4.20 that I am using?

OK, downloaded HCI and tried it run it. It recommended I run at least two instances of it with 2047MB of RAM for each one. I'll do that and get back to you.

Edit : I let it run 120% coverage with two copies running 2047MB each and there is no errors.
Edit 2 : Just started run a 3rd run with 2047MB and got an error within 5 seconds.
 
Last edited:
What the difference between HCI Memtest and Memtest+ 4.20 that I am using?


Almost all the other MemTest utilities force you to have a USB/CD and has to be done during boot/pre-OS loading, while HCI Memtest allows you to do it within a Windows desktop environment not making your computer a space waster while it's doing its thing. Probably less efficient, but it'll get the job done if done correctly like the others given the time and proper understanding of what it's doing (how many copies to run based off free system RAM).

OK, downloaded HCI and tried it run it. It recommended I run at least two instances of it with 2047MB of RAM for each one. I'll do that and get back to you.

Edit : I let it run 120% coverage with two copies running 2047MB each and there is no errors.
Edit 2 : Just started run a 3rd run with 2047MB and got an error within 5 seconds.


Yeah, most likely a RAM issue then. I was having no problems with the CPU overclocking myself (Q9550) and getting stable boots, Prime95, and IBT tests, but every 24-48 hours the system would randomly BSOD and restart.

Turns out because I was only changing the vcore and FSB for overclocking, and leaving the RAM to 1:1 with the processor clock, my memory couldn't take the higher frequency without manually adjusting the voltage. I raised it only from 1.5 to 1.52 and ran some 8 hour Prime95 tests and IBT on maximum and so far the past 2 weeks I haven't had a single random crash, so I'm guessing it solved the problem.

I'm betting the issue is you need to juice up the RAM's voltage a tad bit. It gets thirsty too :p.
 
You should be running enough instances to completely fill up your RAM, that way you get accurate results.

Since you had errors show up, it's almost definitely your RAM. There's a slight chance that it could be your motherboard or a bad memory controller on your CPU, but you can easily figure that out by using only one RAM stick at a time. It seems like one is good while the other isn't.
 
I just ran Memtest86+ overnight for 11h 25m and that still didn't get any errors.

Not sure if I am using HCI Memtest correctly. I've got 8GB of RAM so do I just look in task manager and see how much free RAM I have and run that amount? Do I input the free RAM I have or the amount I have available?

For some reason it won't let me run 6500MB in one shot. I got to run it over multiple windows. Why would that be?
 
What timings/speed are you running your ram at? What sticks are they?
 
Using this kit http://www.gskill.com/products.php?index=355

I manually set the timings to 9-9-9-24 1T 1.5V 1600Mhz.

I've tried changing it to 2T and upping the volts to 1.58V, then even lowering it to 1333Mhz, but none of that made a difference. I also upped the Vccio to 1.15V, again no difference.

Anything else I should try?
 
Last edited:
Pull up task manager, see how much free memory you have. In task manager, that should be Free + Cached. Divide that number by four, since you have four cores, and you want testing done at the max speed possible. Launch 4 instances of HCI memtest, and use the number you got to test your memory. That way, your CPU can test the memory at the fastest speed possible.

HCI memtest seems to be able to stress your memory more than Memtest86+, and detect errors faster.
 
I just ran Memtest86+ overnight for 11h 25m and that still didn't get any errors.

Not sure if I am using HCI Memtest correctly. I've got 8GB of RAM so do I just look in task manager and see how much free RAM I have and run that amount? Do I input the free RAM I have or the amount I have available?

For some reason it won't let me run 6500MB in one shot. I got to run it over multiple windows. Why would that be?

Memtest86+ does not stress the RAM enough to show up errors. It can take days for memtest86+ to show up erros that HCI memtest will show up in minutes.

The free version of HCI memtest is limited to testing 2GB per instance. That is why you have to run multiple instances at the same time to test all of your RAM.

You really should be able to run 3 instances with 2GB each, plus a 4th instance with at least another 1GB. Windows should swap out some of what is being used to the swap file.
 
Hey guys, looks like it's the RAM.

I tried 8GB of G.Skill Sniper 1600Mhz 1.5V Ram and Prime95 blend with 6.5GB and even 7GB passed, LinX passed, HCI Memtest also passed. So it's the RAM for sure then right?

Thanks for recommending HCI Memtest Tsumi.

It looks like Memtest86+ is a waste of time running it for hours and even days, and it still not picking up errors. Is it worth buying the Deluxe version of HCI?

One other question - Does G.Skill require a receipt for warranty? I just downloaded the G.Skill RMA form and it asks invoice number and other details that I don't have. I bought the RAM from a trading forum and didn't get a receipt. I've RMA'd Crucial and Corsair RAM before and don't recall needing a receipt for either of those brands as they went off of the serial number.
 
Is it worth buying it? Well, ask yourself is it worth the convenience of not having to manually set up the tests each time. For me, it's not. It doesn't take all that much time to set up each test, and the dialogue boxes aren't that annoying.

If they do require it, then you are out of luck. You can try emailing them to see if they will accept it without a receipt/invoice. I know Patriot doesn't require a receipt, same with Corsair.
 
Back
Top