Overclocking IBT vS. Prime 95 My thoughts/Your Thoughts?

Hey guys,

So after my dabble with both of the above stability testing pieces of software I thought I would give you my opinions on both of the bits of software, my experiences as both a noob to overclocking and to writing reviews/blogs.

So let me first introduce myself for those who do not know me, this is my first serious dabble into overclocking, with the rig in my sig, enormous amounts of help from the wonderful people at bit-tech and serious hours of reading/tweaking and fiddling.

I am going to review/talk about two pieces of software, Intel Burn Test version 2.53 and Prime 95 64-Bit, both of which are designed to test the stability of your CPU under load to determine any flaws in your hardware or your overclock.

Lets dive head first into Prime 95

So when I was first reading up on overclocking as my parts were rolling in for my first build I read that Prime 95 was the be all and end all of stability testing, so instantly assumed this all I needed (aside temp monitoring software) to assist in my overclocking.

I stuck my rig together, followed this guide from the awesome bit-tech team to overclock my CPU to 4.5 GHZ. Boot back up and run Prime 95.

You are met with this screen:


Fairly simple this gives us the option of 4 different test types:

1. Small FFT
2. Large FFT
3. Blend
4. Custom

The way P95 works is by finding extremely large prime numbers, even the slightest instability will cause the test to fail.

I consider Blend the best function this program offers for OC stability as it tests both the CPU, RAM and the link up between the two. I hit custom and away we went, my system BSOD almost instantly, this indicated not enough VCORE, after some tweaking we had a stable system.

Now stable is subjective, I took a clean run of P95 for 24 hours to be stable, others may take longer, others shorter. Over this period here are my pros and cons for Prime 95:

Pros

1. Extremely thorough.
2. Customisable/Varied Tests.
3. Relatively easy to run.
4. Free.
5. Widely supported.
6. Well scaled across multi threads

Cons

1. Sometimes confusing to review/read.
2. Does not always test the system to its fullest.
3. Extremely long run time.
4. Boring to look at. (I know its not a real con)

Now onto Intel Burn Test

Now after a bios update to my board today I had to OC my system once more, I figured this was a good thing, allowed for some more tweaking, the possibility of less VCORE and a chance to try new things.

After finding out about IBT I thought it is only fair to give it a try so after a quick 1.7mb download we were away. Stuck in the settings from the above guide booted up and ran IBT. This is what we are met with:



Quite a friendly GUI, very very easy to read and quite a simple looking set up. Now although at first glance it doesnt look like we have the same options as above but we do. The stress level I read to be best was High, this gives a great impression of OC stability when couples with 20 iterations of the test. So I wacked it to 20, set it to high and away we went.

Instantly we can tell the program is away by the quite simple yet useful graphic of the flames, if it stops moving during the test it is a clear sign of an oncoming BSOD or crash.

Now with the settings from the guide I got an always instant lock up. Quick reboot and some more Vcore and away we went again, after 20 solid iterations on high and 10 on very high I took my system to be stable. This took roughly 15 minutes. So here are my pros and cons of IBT:

Pros

1. Very quick to run.
2. Creates more stress/heat than Prime 95.
3. Customisable tests.
4. Good GUI.
5. Useful Flame Graphic.
6. Lightweight.

Cons

1. Could be more customisable.
2. No ways of targetting different areas of system such as RAM.
3. No way of telling which core is unstable (Useful in AMD situations).

Overall my preference lies with Intel Burn Test mainly for the speed at which it detects instabilities and the excellent GUI.

Now for your thoughts? On both my review/blog and on your preferences and if there is any software I have missed?

I have added a poll about what your preferences are and I look forward to hearing your comments.

Thank you very much for reading, and for the comments I hope to receive.

Matt

 

I voted P95, as I've never used IBT.

I don't get the bit about IBT being faster. Faster to do what, exactly?

Does it find/highlight instabilities quicker than P95? If so, then that could be usefull. I know all too well that a 15-30 minute run on P95 does not prove a system is stable. Although, I've never known a PC to fail an overnight test, having successfully run for two to three hours.

 

Yes, it took my P95 almost 1 hours to find an instability that took IBT almost 8 seconds.

 

I like OCCT / Linpack. Will use IBT if OCCT is inconclusive.

 

I use IBT. Main reason is that I had some of my earlier overclocking tests pass 2 hours with P95 blend, and then fail 2 runs of IBT at maximum stress. IBT finds instabilities faster than P95.

Even when an overclock passes 5 runs of IBT maxium stress, extreme mode, it is still not a guarantee of total stability. Folding will still find any small level of instability. I consider my systems stable only after passing IBT then also folding (Advanced wu) for 24 hours without a crash.

 

@lysol, IBT is linpack, same as OCCT.

IMO the best stability test is LinX, another linpack interface. Run it with Windows 7 SP1 and it uses the AVX instcution set and stresses your CPU like a mofo; nothing else will give you the same level of stressing or the same load temps. It takes a bit of time to use it properly (eg leaving system to sit for 10 mins before running, disabling unnecessary services, using all available memory etc) but the results are worth the extra effort. A good indicator of stability with LinX is GFlops performance - better stability equals higher GFlops.

I've had Prime fails even after six hours... it's all down to instability, but if it takes Prime that long to find it, it should be reasonably easy to correct. A 100% stable rig won't fail a stability test... ever. Period.

 

I have had Prime 95 run for 24 hours and not find a problem, yet IBT bailed on me within 10 minutes.

Another reason I use IBT is that PRime 95 only ever uses around 5GB of my RAM max. With IBT set at Maximum, it uses ALL my RAM.

IBT is IMO the only test of stability.

I generally set it to run 20 tests with stress level at maximum.

 

According to the OCCT site, the OCCT one is different, made by them. Then it has the linpack/ibt as a second option.
Even still, sometimes the OCCT end unexpectedly, or has an unknown error, so then I use the standalone ibt to recheck.
Though I haven't seen odd behavior since the v4 release.

 

I've been having issues with Prime95 just screwing with me, while IBT is fine, so I'm going to stick to IBT now I think...

 

Prime 95 doesn't stress your system using all your RAM, therefore it's not a reliable method of stress testing IMO.

Set IBT to maximum and it will use all available RAM no matter how much you have.

 

Yes sorry let me clarify, I ran both prime 95 on settings I knew to be slightly unstable, Prime 95 took close to 4.5 hours to fail, IBT took close to 10 seconds. It always has for me (and others) found the issues a lot quicker.

Any feedback on my write up? Just so i can learn really

 

Prime 95 and IBT does not stress all the full system bus. You may well be Desktop stable, but what about 3D stable, especially when running multiple cards.

This Asus specialist explains in more detail, and gives very good adivce. 9:30mins explains in detail what tests he uses.

Enjoy!

 

interesting! I'll give that a go. I notice though, that while stressing the whole bus, there's still very little RAM being tested there.

 

I use prime95 because i heard of it first. I figured it makes my CPU's go to 100% load and makes them hot. If the computer doesn't melt, I don't really care what program does it as long as it lets me know my cooling is adequate. That's all I use it for so I don't need IBT's fancy GUI.

 

First of all i want to make it clear I am not inciting arguments here

I wasnt praising IBT on its GUI solely but saying that I felt it looked nicer and also worked better which was a benefit for ME. Also 100% load and hot in prime95 is not the same 100% load and hot in IBT.

Both programs hit 100% load on my CPU (as it should) but IBT resulted in about 4 degree higher temps than P95, also IBT uses all of the RAM available which tests both the CPU, RAM and QPI Link (I think but might be wrong here)

 

You are correct.. QPI link will be stressed more using IBT.

Maybe IBT set to Maximum stress AND Unigene Heaven simultaneously would be the ultimate test

 

I think...my rig would wave a small white flag and just drop to its knees haha!

 

I'm going to try it out for a bit, the next time I push the CPU hard.

I have given it a quick run through and compared it with P95.

I didn't see a difference in max temp, but I did notice that the temps drop off every now and again in IBT; presumably as it ends one cycle and begins another. I've never seen that happen on P95.

Why not? We like a good argument here.

Actually, I'm no more attached to P95 than I am to a specific brand of CPU or GPU, I'll simply buy which I believe is better value at the that time. The only difference is they tend to cost a bit, and these apps are free to use.

 

Yes, as IBT ends one iteration, there is a drop in temps and load, im not entirely sure if this makes the test anymore taxxing on the CPU as it is asking more than just staying stable at 100%. It is now asking it to remain stable while going from 10% to 100% and also with the accompanying voltage drops.

Haha I only wrote this up as I was bored, had tried out some new things and wanted to get peoples opinions on both my review and my ideas!