RAM testing software

[rootsrat]

I never read that anywhere that prime95 is not good for modern cpu's... The program is even regularly updated for new cpu instructions.

 

Btw rootsrat, IntelBurnTest is very good for finding out major stability issues in a very short time-span. Use max ram -/- 2gb and do about 10 runs. If it passes, you will in 9/10 cases also pass a 12hours+ prime95 test. IBT stresses both cpu and ram components more than any other program and is great for detecting full load issues. Just watch your comp during the 1st test to make sure your cpu doesnt overheat :)

 

Prime95 is better for rooting out general cpu/ram failures, but in order to detect them it'll need to run 12-24 hours.

 

Is IBT compatible with AMD CPUs?

[mothergoose729]

Prime95 is only dangerous because it is effective. It is good at utilizing the registers and CPU cache on your processor, which is why your process gets so warm after running it. On a server processor prime95 could run for weeks without error. On your PC, eight hours or so probably means you are ok, if you are overclocking you would want to test for longer.

 

Intel uses linpack to bin their processors on the assembly line. If you want your processor to really burn up, try intel burn test. In my own experience though, prime95 prolonged tests is more reliable. Turn up your AC though :p.

[Nearox]

I never read that anywhere that prime95 is not good for modern cpu's... The program is even regularly updated for new cpu instructions.

 

Btw rootsrat, IntelBurnTest is very good for finding out major stability issues in a very short time-span. Use max ram -/- 2gb and do about 10 runs. If it passes, you will in 9/10 cases also pass a 12hours+ prime95 test. IBT stresses both cpu and ram components more than any other program and is great for detecting full load issues. Just watch your comp during the 1st test to make sure your cpu doesnt overheat :)

 

Prime95 is better for rooting out general cpu/ram failures, but in order to detect them it'll need to run 12-24 hours.

Is IBT compatible with AMD CPUs?

https://www.xtremesystems.org/forums/showthread.php?197835-IntelBurnTest-The-new-stress-testing-program

 

This is the developers' thread. It is designed for any x86/x64 cpu. Officially it hasn't beent tested afaik for intel i7s etc but as they say in the changelog: Tested by many users on Core i7 9xx, 8xx, i5 7xx, i7 2xxx, AMD 10xxt.

 

I've used it tons of times on numerous CPUs. 

 

Edit: The reason it's called Intel Brun Test is because it is based on  Intel's Linpack, which is more difficult to use.

[MontyMM]

Like Aiyen says, the problem with Prime95 is self-evident really. You are deliberately pushing the CPU to artificially high limits (using techniques that will maximize heat and power draw), and also stressing the psu itself. If you are overclocking, you do so while exceeding the specified limits of the hardware, and likely cranking up voltage further.

 

"Torture testing" is just not an adequate name for it - "testing to destruction" might be better. If you are Intel's production line, you can afford and are happy to weed out and destroy the less than perfect specimens. NASA might take the same view of something they put into space. But for the average user, with one expensive CPU and a limited budget, this becomes less rational. Is it an effective mechanism to test for instability at most extreme conditions? You bet. Do you really want to do that? That's another question.

 

If you are an overclocker, you accept that you take risks with your hardware, but do you really need to maximise the risk in an artificial and extreme proving process? What is the potential cost/benefit? And, why would one imagine that there is necessarily a "good" way to do so?

[whocares7]

MontyMM,

 

This stress isn't harmful because it's short term. The modern CPUs and MBs have a different mechanisms for overtemperature protection like throttling when the CPU goes to the specified temperature which is certified in the official specifications of AMD/Intel. Same is for the CPU voltage which also specified by AMD/Intel as max safe for normal operation and if it's exceeded then could have potential risk for the CPU if there is not adequate cooling/not only CPU, but some vents in the case to create a good airflow to cool the other components and to lead the heat outside and the the normal process degradation under these condition but I don't think that the casual overclocker cranking up the voltage to these levels. The PSU are the most important thing, it must be produced by certified vendor, with high quality components like Japanese caps, not cheap voltage regulators with radiators on them,guaranteed power efficiency under load/20/50/100%/, not from some unnamed chinese garage company.

Why do you think there are black edition/K/ processors both from Intel/AMD with unlocked multiplier, some fancy enthusiastic MBs from ASUS, Gigabyte, Asrock, MSI or high frequency memory with advanced cooling, water cooling blocks, huge tower type air coolings ?

 

rootsrat, If you doubt in the effect of the synthetic applications and benches, you'd use some not poorly threaded games which stress heavily the hardware like Crysis or Metro 2033 or Last Light/with their bench tools which allow you to run up to 99 times/, stores and unloads huge amount of data in the RAM and VRAM and if there are some system instability, it'll show.

[rootsrat]

Good conversation here :)

 

Well, I will be testing my RAM first, then move on.

 

Re OC and Prime95: My previous CPU was AMD Phenom II Black Edition, with 1 or 2 cores locked - can't remember now, it's been a while. Unfortunately both were malfunctioning and I wasn't able to use them when unlocked, so I OC'd my CPU a little, and that's when I tested it with Prime95. I didn't to extensive 24-h test, but I ran it for about 8-10 hrs or so. I have a decent ThermalTake PSU, I'm sure it can take on a lot, or so my testing proved.

 

But anyway, thanks for all advice, it's much appreciated - and I learn a lot as well, as I'm not really an expert in these matters :)

[Nearox]

By the way what type of RAM do you have?

 

When I first got my DDR3 1600 ram and installed it, it ran at 1333 mhz. Most of the 1600ghz ram is basically a factory overclock of the default 1333ghz. However the BIOS settings may also still be at default. So when I cranked up my ram to 1600mhz as it was supposed to, I started to run into all kinds of small issues during the first few weeks.

 

Later I discovered that the RAM Voltage was still at default and not at the higher voltage required for 1600mhz. 

 

This is something to check, if you haven't already.

[rootsrat]

I have 8GB Corsair Vengeance sticks, but... I can't remember at what speed they run... I think they're 1600's. I think I did set them up as per spec, but I'll check again when I'm back home. I do remember having some issues when I first installed them, but I looked into them and reconfigured them in BIOS. Saying that, I updated my BIOS recently, so there is a possibility of the setting going back to default maybe.

 

Good point, thanks Nearox! Never thought of that!

[Nearox]

I also have corsair vengeance sticks. 

 

Selecting the XMP profile in BIOS only increased clock speeds for me but it didnt change the voltage accordingly. Might be my motherboard though, I don't know, just had to do it manually :)

[rootsrat]

I also have corsair vengeance sticks. 

 

Selecting the XMP profile in BIOS only increased clock speeds for me but it didnt change the voltage accordingly. Might be my motherboard though, I don't know, just had to do it manually :)

 

I haven't got XMP, as I have AMD CPU, and it's my understanding that you it's not for AMD.

 

::EDIT::

 

I do all the config manually in BIOS.

[Nearox]

Hmm I believe AMD has had XMP support for some time on their newer motherboards. I know some FM2+ boards got it. 

 

But yeah you don't stricly need it of course, just a convenience (if it works properly, which it didnt in my case!). 

 

The setting might be hidden by default:

 

https://smallbusiness.chron.com/enable-xmp-amd-board-37024.html

[mothergoose729]

Whocares is right. Cou are designed to be able to operate long term at high temperatures. A torture test lasting 8 to 24 hors is not going to significantly impact the lifetime of your processors. Personal philosophy; if I can make it break its not good enough, however I place a premium on stability.

[MontyMM]

Why do you think there are black edition/K/ processors both from Intel/AMD with unlocked multiplier, some fancy enthusiastic MBs from ASUS, Gigabyte, Asrock, MSI or high frequency memory with advanced cooling, water cooling blocks, huge tower type air coolings ?

 

Because there is a market for them. This doesn't tell us much about the risks of an individual overclock, and nothing at all about whether prolonged torture testing at maximum load is a good idea. The purpose of any stress test is to determine the point of failure, and the nature of that failure is unpredictable.

 

Cpus have had temperature cutoffs for many years, and these are really no guarantee of safety when overclocking. We can argue about how well cpus are capable of withstanding extreme conditions, but to suggest that applying an overclock and then artificially stressing the CPU for 24 hours does not imply increased wear and risk to the hardware is, I think, obviously untrue. In addition to the CPU and the PSU, the components of the motherboard are also vulnerable to this treatment.

 

Given that some degree of risk is implicit in doing so, the question is what value you derive from taking that risk. If all goes well, you learn that your CPU is currently stable at performing in that particular torture test condition for prolonged periods. But how valuable is that information, weighed against the wear and risk?

 

The fact that an OC might be stable in those particular torture test conditions does not guarantee that it will be stable in all real world conditions; it is entirely possible for an OC to pass a torture test, yet still cause instability in a game or other situations.

 

Also, the CPU might fail the torture test, yet still be stable enough in the real world use cases you actually need.

 

I would recommend using a few run-throughs with synthetic benchmarks and something like 3d mark to determine if the CPU is stable enough for your requirements.

 

The idea that overclocks should be tested with prolonged torture conditions is a piece of received wisdom that is worth questioning, IMO, if you do not regard your hardware as disposable.

[Spock]

I'm not that much into the specifics of hardware, but a friend of mine got angry when I ran Prime for a night on the hardware I bought from him ("Man don't **** up your new hardware, I mean it!"). Although it's only premium components he said an extended stress test is not needed because your CPU will never be stressed like that and it can decrease lifetime expectancy of hardware because everything: voltage regulators, L1 L2 L3 caches, memory controller, every single ALU, chipset, RAM is used like normally only in months.

 

The advice I'm handling my hardware after: run Prime for 1 hour max, 3 for your final setup and then reduce the FSB clock by 1 or 2.

Memtest seems to be less aggressive, but 3 or 4 passes should suffice.

 

If you are overclocking on the new intel chipsets watch out for your north bridge (qpi)! Mine got to 107 degrees until I installed a Xigmatek Porter (best NB cooler on air available) with a 80x80 Noctua fan. Use Speedfan to find the right temp, the measurement can be off though.

[mothergoose729]

Why do you think there are black edition/K/ processors both from Intel/AMD with unlocked multiplier' date=' some fancy enthusiastic MBs from ASUS' date=' Gigabyte, Asrock, MSI or high frequency memory with advanced cooling, water cooling blocks, huge tower type air coolings ?[/quote'']

 

Because there is a market for them. This doesn't tell us much about the risks of an individual overclock, and nothing at all about whether prolonged torture testing at maximum load is a good idea. The purpose of any stress test is to determine the point of failure, and the nature of that failure is unpredictable.

 

Cpus have had temperature cutoffs for many years, and these are really no guarantee of safety when overclocking. We can argue about how well cpus are capable of withstanding extreme conditions, but to suggest that applying an overclock and then artificially stressing the CPU for 24 hours does not imply increased wear and risk to the hardware is, I think, obviously untrue. In addition to the CPU and the PSU, the components of the motherboard are also vulnerable to this treatment.

 

Given that some degree of risk is implicit in doing so, the question is what value you derive from taking that risk. If all goes well, you learn that your CPU is currently stable at performing in that particular torture test condition for prolonged periods. But how valuable is that information, weighed against the wear and risk?

 

The fact that an OC might be stable in those particular torture test conditions does not guarantee that it will be stable in all real world conditions; it is entirely possible for an OC to pass a torture test, yet still cause instability in a game or other situations.

 

Also, the CPU might fail the torture test, yet still be stable enough in the real world use cases you actually need.

 

I would recommend using a few run-throughs with synthetic benchmarks and something like 3d mark to determine if the CPU is stable enough for your requirements.

 

The idea that overclocks should be tested with prolonged torture conditions is a piece of received wisdom that is worth questioning, IMO, if you do not regard your hardware as disposable.

That is definitely true, prolonged torture testing is just one of many tests you aught to run on your hardware before you overclock. I used to overclock and benchmark a lot more. Had something like 5k posts on the OCN (and still do I guess, but I never go there anymore). If you want to be sure things are ok you do need to download a couple bench marking suites and run those tests a few times, and also torture test. Stacking torture tests is a good idea, because just because prime95 seems ok doesn't mean that linpack won't run into problems, ect. Its exhausting to do, and once you go through all the proper paces to be 95% sure your hardware will never fail on you, you are left with a pretty modest overclock after it all. The capabilities of a CPU does change with time, so it is differently possible to have a rock solid overclock at one point, that doesn't hold up well 2 years later. Also if something in your PC ever starts acting up, you can't be entirely sure it doesn't have something to do with your hardware. Its for these reasons I stopped overclocking my system. The 20% increase in performance isn't worth the investment in time and money to do right. I wouldn't say though that torture test software is without value. If you want to find out quick if you harware is broken, prime95 or furmark will usually tell you in 10 minutes. Also, if you invest in adequate cooling and test thoroughly, an overclocked processor or graphics can be ever bit as stable as one that isn't, although less than 1% of all the folks that overclock properly test for stability.