Overclocking CPU/GPU/Memory Stability Testing Guidelines

[Idontcare]

I tried doing that and yes I also see errors in Memtest86+ and even in Windows Memory Diagnostic tool. Weird thing is that I get it when using both sticks of RAM. When I try them one at a time, they are error-free. I already tried RMA'ing them and when I got the new kit, I still experience the same.

When you try them one at a time, did you try both dimm slots individually?

You might have a bad dimm slot.

 

[kevindd992002]

When you try them one at a time, did you try both dimm slots individually?

You might have a bad dimm slot.

Yes, I tried both dimm slots individually and individual sticks are error-free for all 4 dimm slots.

 

[BassBomb]

my linx doesnt seem to activate as i run it?

 

[Idontcare]

Yes, I tried both dimm slots individually and individual sticks are error-free for all 4 dimm slots.

Then it must be something related to the capacitive load presented by all four dimms being populated.

If true, then the way to address this is (1) more Dimm voltage, and/or (2) slow down the ram speed, and/or (3) loosen the ram timings, and/or (4) cool the dimms better.

From what you've said above in your posts, none of these are possible for your laptop. So I have no ideas on where you should go from here. 🙁 Sorry.

 

[busydude]

my linx doesnt seem to activate as i run it?

Wow! I have the same problem. Recently upgraded my processor, could not activate LinX. I kept it running for more than 10 hours, never stressed my CPU.

 

[kevindd992002]

Then it must be something related to the capacitive load presented by all four dimms being populated.

If true, then the way to address this is (1) more Dimm voltage, and/or (2) slow down the ram speed, and/or (3) loosen the ram timings, and/or (4) cool the dimms better.

From what you've said above in your posts, none of these are possible for your laptop. So I have no ideas on where you should go from here. 🙁 Sorry.

Wait, I don't have all four DIMMs populated. I only use 2. What I meant was that I tested all four DIMMs with one stick each time during stability testing and they all passed. They won't past when I use two sticks at dual channel though.

 

[pieguy]

I need a little help, I'm getting frustrated. I thought my 2500k OC was stable as I was passing everything I threw at it with Prime95 and no problems with Skyrim/BF3 gaming sessions. But I want to be certain it's stable so I stumbled upon this thread. When I use LinX per the instructions in the OP, CPU-Z is reporting a vcore that is lower than what I've set it at, and unsurprisingly it fails or I get a BSOD.

For example, for 4.6GHz I'm trying to test stability at 1.34v, I've tested both setting it manually and with a + offset. I get the 1.34v with basically everything I throw at it except when I fire up LinX it drops to 1.32v and no wonder it fails. Anybody know what's up?

 

[Idontcare]

The voltage will drop a little when the CPU is under load from a phenonemon called "voltage droop" or Vdroop as it is referred to.

The same thing is what happens during electrical brown-outs in hot summers when everyone turns on the Air conditioners.

Most mobo's these days have something called "load line calibration" or LLC which is designed to counter the Vdroop so it is minimal.

A Vdroop from 1.34 to 1.32 is not bad. But you can do two things to counter this. Increase the voltage with your + offset another 0.02V or so, or find out if your mobo's BIOS lets you adjust your LLC in any way.

 

[pieguy]

Ok Idontcare, thanks. I've seen vdroop mentioned a lot and I don't think some people quite understand what it means so I didn't know if that was my problem. I had dialed in my vcore and LLC setting (LLC medium on my ASUS board) so that I was getting enough idle voltage and then enough voltage when it ramped up under load of Prime95 and gaming.

So it threw me off when I tried LinX because I never saw any vdroop when running different tests with Prime95 and also my temps with LinX are around 8c higher and pushing me into uncomfortable territory.

 

[Idontcare]

Yep, that's why Linx is considered to be a good stress tester, it will certainly stress your system to its limits 😉

 

[pieguy]

Thanks for this thread and for your help. I think I've got my 2500k @ 4.6GHz stable to the "gold standard" 😀

 

[kevindd992002]

@Idontcare

I'm not sure how much RAM would be need to test with HCI Design Memtest. Is it the value under "free" or "available" of task manager? I wouldn't want the pagefile to be accessed frequently during the stability test since I am on an SSD and read/write operations would degrade the SSD if permitted.

 

[Idontcare]

@Idontcare

I'm not sure how much RAM would be need to test with HCI Design Memtest. Is it the value under "free" or "available" of task manager? I wouldn't want the pagefile to be accessed frequently during the stability test since I am on an SSD and read/write operations would degrade the SSD if permitted.

When you first open HCI the default input value is "All available memory" (or something to that effect), just click "Start" and the program will automatically handle this for you.

This method won't get you to the point of testing all the memory, but it gets you close enough.

You can get closer by repeatedly starting, waiting a few seconds, stopping, and restarting the test. This will convince Windows to put more of the ram contents out onto the pagefile so HCI can test more available ram.

 

[kevindd992002]

When you first open HCI the default input value is "All available memory" (or something to that effect), just click "Start" and the program will automatically handle this for you.

This method won't get you to the point of testing all the memory, but it gets you close enough.

You can get closer by repeatedly starting, waiting a few seconds, stopping, and restarting the test. This will convince Windows to put more of the ram contents out onto the pagefile so HCI can test more available ram.

How much of the RAM is "All available memory" for the program?

But I want to use multiple threads (8 instances of Memtest), this way is more efficient and faster right? With this, I have to go manual inputting of RAM amount.

The technique you suggested is the one I am moving away from since I don't want the pagefile to be used (it is located on my SSD) as temporary RAM.

 

[Idontcare]

Disable your page file before running HCI. When done running HCI, reenable it.

I run 8 instances, the first 7 I set to use 1/8 of my installed ram, the final (8th) instance I leave to run "all available" which ends up being something less than 1/8.

 

[kevindd992002]

Ok. In that way, where will Windows get its memory if I disable pagefile at all and all RAM is tested by Memtest?

By 1/8 you mean exact powers of 2 like 1024MB for my 8GB kit?

 

[kevindd992002]

When I set 1024MB for 7 instances and All Available for the 8th instance, the 7th and 8th instances give me a message that my version of Windows cannot handle contiguous RAM blah blah. I need to manually put a lower amount of RAM for me to proceed.

 

[kevindd992002]

Anyone?

 

[Idontcare]

When I set 1024MB for 7 instances and All Available for the 8th instance, the 7th and 8th instances give me a message that my version of Windows cannot handle contiguous RAM blah blah. I need to manually put a lower amount of RAM for me to proceed.

I don't know what that means or why you'd be getting that error. Your best advice here is to see if the developer of HCI has a support team that you can email. If you do, and they help you, please post here to let us know the solution so others can profit from the knowledge.

 

[kevindd992002]

I don't know what that means or why you'd be getting that error. Your best advice here is to see if the developer of HCI has a support team that you can email. If you do, and they help you, please post here to let us know the solution so others can profit from the knowledge.

Here's the exact error I'm having: http://www.tomshardware.com/forum/22119-45-application-memory-allocation . I get it with "ALL" my computers and I use Windows 7 x64 Ultimate. I think many people experience it?

Here's a pic of the error:

I don't think HCI Desing Test still has a support team.

Do you personally run this test without any pagefile?

 

[Idontcare]

Do you personally run this test without any pagefile?

I do not. I launch all 8 instances of memtest, starting the first 7 as using 1/8 my physical memory.

Then the last one I start as "use all available" and stop it soon after, then restart, then stop, then restart, and so on a few more times.

What I noticed is that this convinces windows to page out much of its kernel and services such that the "free" memory for HCI memtest keeps expanding.

Then when I walk away and leave the 8 instances running, I am testing some 98-99% of the physical ram.

 

[kevindd992002]

I do not. I launch all 8 instances of memtest, starting the first 7 as using 1/8 my physical memory.

Then the last one I start as "use all available" and stop it soon after, then restart, then stop, then restart, and so on a few more times.

What I noticed is that this convinces windows to page out much of its kernel and services such that the "free" memory for HCI memtest keeps expanding.

Then when I walk away and leave the 8 instances running, I am testing some 98-99% of the physical ram.

When you say "convinces windows to page out much of its kernel and services such that the "free" memory for HCI memtest keeps expanding" you mean it will continually use the pagefile during the duration of the test?

 

[amdhunter]

So, umm. I've been testing my 4.5GHz 2500k for about 5 hours with every single one of these running at the same time. (100 for IBT, Prime running 4 instances, OCCT @ Infinite)

I am going for 24 hours. It's either going to crash and burn, or make it without a hitch.

Stupid? Temps seem to be peaking at 65C. Voltage is a modest 1.32v.

 

[truth_benchmark]

For OC stability testing, I run HandBrake (1GB, AVI to MP4) in the background while running Resident Evil 5 benchmark tool (variable benchmark mode)

If it doesn't crash and the video is encoded successfully, I consider it to be stable for my intended purpose which is gaming and video encoding.

Video encoding is the most demanding CPU-based application I use. I use Resident Evil 5 because it can utilize more than 2 CPU cores


By the way, I also use TessMark sometimes. I noticed that it loads the 3 CPU cores to ~90% while the remaining core is around ~30% CPU usage only. I'm using anA8-3870K without a discrete GPU

 

[Whitestar127]

Hi, I'm doing my first overclocking ever. It's an i7-2700K (special offer, actually cheaper than a 2600K).

Anyway, I have 8GB GSKILL memory, so using HCI Memtest I started 8 instances of it and set each to use 1024MB. Does that sound correct? Problem is I think I only have about 6400MB free RAM or so, so when I started the eighth instance it was really sluggish and didn't seem to have any progress at all percentage wise.

When the other instances was at 600% the eighth instance was at 0.something. Is this normal behaviour? Should I set up the test in a different manner?

On a side note, my 2700K seems to be rock solid at 4.4GHz (12 hours Prime and 25 cycles in Linx) but simply refuses to touch 4.5GHz. Even the auto-tuning stopped at 43 CPU ratio and 103 baseclock, getting it to 4429MHz. And when I tried to just set the ratio to 45 it crashes immediately when running Prime. Simply bad luck in the CPU lottery?

Not that it's a big deal really. 4.4GHz gives a nice boost in Skyrim.

Other hardware: Asus P8P67-Deluxe, Corsair H60, Radeon 5970, 8GB GSKILL Sniper RAM
OS: Windows 7, 64-bit