x264 stress test

Has anyone used x264 as a stress test?

This guy is really big on it: http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics

His claim is it's superior to Prime95 in proving stability while not driving up the temps so dangerously high. He calls it "the cool stress tester".

I downloaded and installed the version he linked with the built in loop feature but I can't figure out how to execute the stress test. All I can get it to do is play a movie clip which only uses about 2% of the CPU. What am I missing?

Looks like you just run the .bat file for your system.

edit:

Yep, run the necessary .bat file (x264 Stability Test (64bit + log).bat and fill out the questions for your system.

In the readme it recommends to run from a HDD and not SSD because of the writes.

We do, well, Dino does, for our motherboard reviews...

As Don said, its a simple run the .bat file... sort of like PiFast.

X264 definitely does not stress as much as prime. I can run x264 for 12+ hours on BLU-RAY encodes on a system that is not more then 1 hour prime stable. X264 as a stress test would not be a bad choice, depending on the software and how it utilizes the cores and loads them up will push a system harder then most people will ever push it.

Edit. Yeah, its not close to being as good as P95 or other stress testing applications.

If it's not fully stressing CPU then it's not testing it right.
During stability tests you have to check various scenarios like constant full load but also mixed load as sometimes when CPU load drops to ~0-10% then C stages or other settings are causing voltage etc drop and when it jumps quicky to full load then not everything will react on time ( that's why some boards have manual vrm frequency and other things ).
There is no perfect stability test. Every test is good for some purposes so for overall stability you have to use couple of programs.

I also still have no idea why suddenly all are afraid to use new Prime95 or other programs which are using new instructions. It's stressing CPU more for sure but if your daily applications are using these instructions then you won't turn it off. So you are testing stability but you won't check if it's stable when full CPU functionality is utilized ? Later you see guys who are complaining that after 5h Prime or anything else it's still crashing in some games.
Noone is saying to run 24h+ in new Prime95 but 30-40mins usually stress CPU more and finds any errors faster than 24h of lighter stability tests. So you spend 1h on CPU + 1h on memory/cache and it's done, instead of 2 days of tests which are not showing any special differences.

Other thing is that long stability testing. We are using desktop hardware and barely anyone is using it for real work but most users care about it like it was mission critical hardware performing 24h+ stability tests. Regardless of used CPU stress software it may damage something when it's running at 100% load for many hours. Current hardware is designed to work couple of hours per day under low or mixed load.

Man, I tried the new Prime95 this morning and temps shot up to high 90s inside of 5 minutes! I shut her down quickly and went back to version 27.9. This is with a Noctua D14, 4790k, 4.6 ghz, 1.26 vcore, cache ring 44x on a test bench. I'll wait till I put the parts back in my case and under water (Swifttech H320) before I try the new Prime95 again. The new Prime drives temps up about 8-10 C over the older version.

A while back I posted lowest vcore needed to pass 1 hour of several different stress tests on my 4790K , posted results with pics of each run here, post 12166:

But requires .07v more to run prime 28.5 than x264.
To run 4.8ghz for 1 hour:
prime 28.5 needs 1.35v
prime 27.9 needs/occt4.1 needs 1.33v
x264 needs 1.284v.

I can run x264 at 5ghz on mine, but no way is it close to prime stable at 5ghz. People like x264 because it is easy to pass and has low temps.

prime 27.9 has 10C lower temps, and makes vcore high enough to run 24/7 stable, but I use vcore for 28.5 just to be safe. At vcore that I can run x264 overnight, I will get random bsods surfing etc from too low vcore.

Most that use prime 28.5 for testing, delid first. If not delidded I would use 27.9.

Well ...... coming from AMD side of the forum I have always dealt with Prime and it has proven to be good and reliable for me with my encodes when I do have to do multiple movies as in my above statement where my CPU will be pushed for multiple hours at a time. That being said I would for me consider my computer pretty stable if I can run BD Rebuilder, my software of choice for my backup's for said 12 hours. I would consider X264 as a good choice compared to other choices out there with the one stipulation being that the X264 Benchmark can utilize the core similarly to what BD Rebuilder does.

Where I think the X264 Benchmark might differ is on the length of time where it is putting a load on your core with its short clips compared to the 30 to 50 GB size of a BLU-RAY disk. Also does it fully utilize all 8 cores, (FX 8350) fully as BD Rebuilder does.

Just my thoughts ...... for what its worth.

rge said:

A while back I posted lowest vcore needed to pass 1 hour of several different stress tests on my 4790K , posted results with pics of each run here, post 12166:

But requires .07v more to run prime 28.5 than x264.
To run 4.8ghz for 1 hour:
prime 28.5 needs 1.35v
prime 27.9 needs/occt4.1 needs 1.33v
x264 needs 1.284v.

I can run x264 at 5ghz on mine, but no way is it close to prime stable at 5ghz. People like x264 because it is easy to pass and has low temps.

prime 27.9 has 10C lower temps, and makes vcore high enough to run 24/7 stable, but I use vcore for 28.5 just to be safe. At vcore that I can run x264 overnight, I will get random bsods surfing etc from too low vcore.

Most that use prime 28.5 for testing, delid first. If not delidded I would use 27.9.

Click to expand...

I thought Devil's Canyon had a better TIM application than the original Haswell and somewhat eliminated the need to delid.

trents said:

I thought Devil's Canyon had a better TIM application than the original Haswell and somewhat eliminated the need to delid.

Click to expand...

roughly the same temps, and same 20C decrease with delidding.

No real refresh on the "Refresh" then, huh?

yeah, the main difference was just you were guaranteed 4.4ghz with 4790K (since the chip's stock turbo to 4.4 on all cores they way mobo's were working), and most would do 4.6ghz. Basically removed the dud 4.2 to 4.3 by default.

rge, have you delidded any Haswell's? I see where some people are popping the lids off by squeezing the chip in a vice.

Yep, I delidded my 4790k using vice only method back when I first got it. Just put masking tape on vice so vice teeth didnt mar cpu...then put IHS up against one side of vice and slightly tilt cpu so pcb catches other side of vice, and slowly turned til popped off, held cpu in my hand whole time to keep from slipping. And my 4790k temps decreased 22C at 185W load (prime 28.5 with 1.3v) after delid then using liquid metal for tim1, still just pk1 or nth1 for tim2. I posted before and after prime shots in another thread. And that was over 6 months ago, my temps still same.

Are temp and not volt limited trents?

EarthDog said:

Are temp and not volt limited trents?

Click to expand...

Hard to say. I'm stable at 4.6 ghz at a vcore of 1.55 but stress testing temps with top air are exceeding 90c. So to go any higher on the overclock I would have to add more vcore which I can't do because of temps. It's so cotton picking hard to find any solid info on what are safe volts on these Haswells. Seems to be no consensus. It's on a test bench right now being cooled by a Noctua D14 but when I put it in the computer it will be cooled by a Swiftech H320 which may be able to lower temps enough to get to 2.7 stable but temps would still be on the high side I got a feeling.

rge said:

Yep, I delidded my 4790k using vice only method back when I first got it. Just put masking tape on vice so vice teeth didnt mar cpu...then put IHS up against one side of vice and slightly tilt cpu so pcb catches other side of vice, and slowly turned til popped off, held cpu in my hand whole time to keep from slipping. And my 4790k temps decreased 22C at 185W load (prime 28.5 with 1.3v) after delid then using liquid metal for tim1, still just pk1 or nth1 for tim2. I posted before and after prime shots in another thread. And that was over 6 months ago, my temps still same.

Click to expand...

rge, can you give me a link to that post. I'd like to see the pics.

1.55v? Wow. Don't delid as you are already WAY over what most would consider a 24/7 voltage (1.35).

Maybe update your Sig with your new hardware?

EarthDog said:

1.55v? Wow. Don't delid as you are already WAY over what most would consider a 24/7 voltage (1.35).

Maybe update your Sig with your new hardware?

Click to expand...

Sorry, I meant 1.255 vcore. MSI Z97S SLI Plus motherboard. 16gb of 1600 mhz gskill.