[Dash8Q4]

Hi all,

In what order would you rate the available stress test programs out there? I just got a new 4790k but i dont want to stress test it with Prime or IBT. What else is there that will test my OC?
PC is 80% of the time just idling, the rest is slight gaming and video editing using Premiere Pro CC.

Thank you.

 

[Thready]

Prime95 is the best one. Why don't you like it?

 

[Lord Xeb]

Use Prime95... If you can pass 24 hours you can pass anything. What is wrong with Prime? Never had an issue and it has always netted best results. However, there is no "quick" stress testing program if that is what you are looking for. IBT is okay, however, even under the most stressful of conditions, you will never push your CPU that hard.

Personally, I run Small FFT for at least 12 hours, then run blend test with 90% of my ram used for 24 hours. If you pass both, your golden.

 

[Dash8Q4]

I dont like to stress the cpu to the extremes that prime or ibt do to satisfy myself of "stability". For me, "stable" is no blue screens during gaming or rendering videos. Hence, i am looking for a lighter option. I have read that synthetic processes like Prime and IBT are extremely hard on the CPU and way beyond what real world load would be in my case. I have used both Prime and IBT on my lapped 2500K but it died a month ago while i was revisiting OCing trying to find an "Offset mode" configuration for it at 4.7Ghz. I previously had it at 4.8 24/7 on manual mode and at times was able to boot into windows at 5.0Ghz. Somehow i mustve done something wrong trying to find an offset for 4.7(it worked at 4.6) so it died. Now i have the 4790k which i plan to keep for 3yrs+. I wont be overclocking it just yet but once i get my Cryorig R1 Ultimate and put in a couple of months use then i will see what i can find comforable.

Thank you for your replies.

 

[Lord Xeb]

I am sorry to burst your bubble, and I am going to be blunt here, but you are doing it wrong. The purpose of stability testing is to see how the CPU does under stress. If you are rendering something in Photoshop, you don't want your machine to crash do you? Hell, you could just be browsing the web, get a sudden CPU usage spike, crash. A lot of us here have a lot of experience in overclocking and stability testing (myself included in this). Entire communities swear by some of these programs, so there has got be something be done here right, riiiiiggght?????

So, the right way to do it is to stress test it using said programs. Rendering will push your chip about as hard as a blend test in Prime. So, run prime. Gaming is another thing all together, however, same thing, run prime. IF you pass 24 hours of that, you can safely say your chip will not crap out during a game or a long period of rendering.

Stability testing also goes into idle stability and even then, sometimes a CPU will pass with stress testing, but crashes on idle and you gotta do more tweaking.

Also, when stress testing, have Event Viewer open. Go to system logs and make suere you are not getting any WHEA errors. If you are, you are not stable.

EDIT: Read your edited post. That is the point of the stress testing programs. To push your chip beyond real world stress testing. If you build a wall that can take a strike from a semi, then you can safely assume you can crash a mini-van into at the same speed. If you want something stable, hit it with the hardest thing you got. If it passes, then you know that no matter what you throw at it, the CPU will be stable.

Furthermore, you are not going to be hurting the CPU in any way. I once left Prime running for a week once. Didn't harm anything.

Things just run a bit hot. Keep things under about 80C (I am a dare devil, so I push things to 85C) and your voltages under control and below safe limits and your fine.

However, if you want true stability, you can always not overclock as well. To be honest, if you are not going to do right, then you really should not do it at all.

EDIT2: FYI, overclocking will degrade your chip and shorten your life, so keep that in mind. Also, getting a Haswell chip that will do 4.7 let alone 4.5 at reasonable voltages is not very common and they run hot. I can tell you that the likelihood of you getting close to the clocks you want are going to be slim.

 

[Dash8Q4]

I'll have to find out then. Reason i will overclock it is for flight simulator X. So once i have that installed thats when ill get into tweaking again. Then i may consider running prime. I think you have contributed to the bug of OCing.

 

[error-id10t]

Or like 99% of Haswell owners, you visit the Haswell / DC threads and realise what people actually use for stressing.. just a hint, it's not Prime or IBT.

 

[Train Wreck]

Quote:

Originally Posted by error-id10t

Or like 99% of Haswell owners, you visit the Haswell / DC threads and realise what people actually use for stressing.. just a hint, it's not Prime or IBT.

Isn't it AIDA 64?

I read somewhere that the current version of Prime should NOT be used for Haswell

 

[Marc79]

Aida64, ROG Real Bench 2.4

I will run 3DMark Vantage cpu test, play cpu intensive games, etc.

 

[Lord Xeb]

Quote:

Originally Posted by error-id10t

Or like 99% of Haswell owners, you visit the Haswell / DC threads and realise what people actually use for stressing.. just a hint, it's not Prime or IBT.

Why is that?

 

[Dash8Q4]

Yeah now i recall reading somewhere not to use the latest version of prime. And most 4790k can do 4.7ghz. Thanks for that. I was planning on thoroughly reading haswell dc guides and that sort

 

[daman246]

personally the best stress tester is Running a few hours of x264 if it can pass 12hours of video rendering then your set in my book.

 

[Train Wreck]

Quote:

Originally Posted by lokito50

Yeah now i recall reading somewhere not to use the latest version of prime. And most 4790k can do 4.7ghz. Thanks for that. I was planning on thoroughly reading haswell dc guides and that sort

You're welcome

 

[TheReciever]

I'm pretty sure [email protected] is still amongst the top contenders for stability testing

 

[StrongForce]

Quote:

Originally Posted by error-id10t

Or like 99% of Haswell owners, you visit the Haswell / DC threads and realise what people actually use for stressing.. just a hint, it's not Prime or IBT.

yea smartass, couldn't you just tell him what to use ?

 

[JVene]

The reason there are threads warning against Prime95 is the AVX option.

AVX is a powerful instruction, and in real world applications where it's used it is mixed with a wide range of other instructions, but the use of AVX in Prime95 is unrealistic, repeating the instruction more often per unit of time than makes any reasonable sense. It also generates the most heat of any instruction set in the chip.

Personally I prefer Aida64. I've used OCCT, but there appears to be a huge spike in temperature at initialization, then it smooths out. Aida64 has also exposed instability when OCCT passed.

I hammer this point all the time, and you've touched on it with your point about "stable enough". Long term stability doesn't prove stability. It demonstrates stability, but doesn't prove it. Failure is certain proof, but passing isn't because it's related to proving a negative on a large population. To prove there are no green grains of sand you'd have to search all grains of sand, exhaustively. If you find one, you've proven there is at least one, but there isn't time to exhaustively test every grain to prove there are none. Similarly, out of the 1.6 billion transistors reported in the 4790K (I know, I've read 1.4 billion, but that's for an older chip) - there is a variation in transistors, and those few which require the most voltage are tough to identify (like grains of sand). Even if you run the exact same test, you're not actually using the same transistors in the same combinations each time.

I got into this field from electronics, back in the late 70's / early 80's. I can tell you that knowing you can only prove failure can be used to help find stability. Trying to ride to the highest possible overclock is actually contrary to all engineering principals. The first Cray-1 ran at 120Mhz in the lab, but it couldn't operate at that speed in the field. They had to run them at 80Mhz, and you can bet Cray was seriously disappointed, but he knew the engineering well. You have to have margins, and you have to base them on failure, not passing.

I can also attest that it's incorrect to say most 4790K's can reach 4.7 Ghz. Most can reach 4.5 Ghz, but by 4.6 Ghz a large percentage soak up to much power for reasonable cooling solutions to manage 4.7 or higher. Too many have cockeyed IHS mounts and have to be delidded just to sustain 4.6 Ghz at reasonable temperatures.

 

[idkface]

Prime95 26.6 seems to be the version people use on Haswell. I trust AIDA64/Realbench/x264/XTU. I always start with XTU and Realbench benchmarks. If there's a stability issue those two benchmarks seem to find it fast. After I'm stable on 10 or so runs of benchmarking I'll set AIDA64's stability test to run overnight and hope for the best.