Your Haswell-E max stable OC

[EarthDog]

I keep hearing that, yet nobody has said why AND proved it was harmful. P95, v28.5 has the latest instruction sets for Haswell CPUs. I think its in the AVX instruction sets? Not sure on that (dont quote me!). But running P95 v28.5 or AIDA64 with just FPU stresses the heck out of a CPU.

I'd like links on that mydog... that is the first I heard a reputable name attached to it at least..

That said, I have run it stock and overclocked on my haswell/haswell-e CPUs since I have had them with zero issues. Granted, on the 5820k it was run for a total of ~8 hours to test stability (4 hours small fft/4 hours blend). But on a 4770k and 4790k there were 10s of hours of P95 28.5 run on them with seemingly no ill effects.

 

[LiveOrDie]

OK thanks well its a bit late now i ran it on stock speeds for 30minutes to see what my water temps were like on the 1st setting Small FFT's? Think ill stick with aida64 for stressing seeing that program has been update to support H-E chips.

What about LinX with the newest intel Math Kernel Library ?

 

[Mydog]

I keep hearing that, yet nobody has said why AND proved it was harmful. P95, v28.5 has the latest instruction sets for Haswell CPUs. I think its in the AVX instruction sets? Not sure on that (dont quote me!). But running P95 v28.5 or AIDA64 with just FPU stresses the heck out of a CPU.

I'd like links on that mydog... that is the first I heard a reputable name attached to it at least..

That said, I have run it stock and overclocked on my haswell/haswell-e CPUs since I have had them with zero issues. Granted, on the 5820k it was run for a total of ~8 hours to test stability (4 hours small fft/4 hours blend). But on a 4770k and 4790k there were 10s of hours of P95 28.5 run on them with seemingly no ill effects.

Here's his reply to the question:

Praz nailed it really. The newer versions of Prime load in a way that they are only safe to run at near stock settings. The server processors actually downclock when AVX2 is detected to retain their TDP rating. On the desktop we're free to play and the thing most people don't know is how much current these routines can generate. It can be lethal for a CPU to see that level of current for prolonged periods.

As for the universal validity of various stability testing programs, that's a more difficult question to answer without using illustrations to simplify what occurs at the electrical level on some of the associated buses.

Being brief as possible and focusing on DRAM transfer as an example: Data is moved around the system in high and low logic or signal states. The timing of these systems and those that rely on them needs to be matched closely enough for data to be moved around and interpreted correctly.A burst of data may contain a series of 1s and 0s. The 1s pull more current as they require defined voltage level that is above 0. Each data pattern has a different effect on the timing margin. Some eat into the timing margin more than others (I may illustrate the theory of this in a future guide). If a given stress test does not generate patterns in a way that eats into the timing budget sufficiently to represent how the system is used, the stress test won't be as useful to the end-user.

That's why most stress test programs alternate between different data pattern types. Depending on how effective the rotation is, and how well that pattern causes issues for the system timing margin, it will, or will not, catch potential for instability. So it's wise not to hang one's hat on a single test type. Evaluate what your needs are from the system and try to run a variety of tools to ensure the system is stable in various ways. We also need to bear in mind that some stress tests only focus on a single part of the system, while others will impact multiple areas at once.

Seasoned users usually find a systematic way that leads them from stress tests that focus on individual areas to those that hit the entire system as part of their test regimen. Ultimately, this all comes down to what your requirements are and using enough testing to confirm reasonable stability for the system in its intended usage scenario.

We coded Realbench to generate stress with real-world apps. It's a useful tool for people that encode, render or crunch numbers with their systems. However, it's not the only method out there - there are many tools to evaluate system stability that are perfectly valid.

-Raja

Source: http://www.overclock.net/t/1510388/haswell-e-overclock-leaderboard-owners-club/2390#post_22900116

 

[EarthDog]

A big thanks MyDog... that is the first piece of concrete information I have seen on this issue.

 

[Jstn7477]

What type of stress testing are you doing to test what your temps are like?

I run World Community Grid distributed computing on my rigs, so 11 threads of that and an empty thread for F@H to feed the R9 290.

 

[EarthDog]

Not a thorough test.. but if it works for you......

 

[Mydog]

No overclock is a 100% stable like no system is, even at stock speed. All systems can have crashes due to hardware failure or bad coding in software

 

[Jstn7477]

Not a thorough test.. but if it works for you......

Nothing is a thorough test, but that is what runs 24/7 on the computer unless I am rarely gaming, so that is what I test with. The system has frozen and bluescreened many times in the past week, even froze 1.5 hours after taking my last screenshot. I backed the RAM down to 2666MHz and shall see if that helps.

Also, Folding@Home has detected the finest instabilities with my R9 290's overclock (by looking in the logs and seeing where it encounters bad states and has to restart from checkpoint) and my R9 290 requires +81mV @ 1075MHz for problems I could not expose via gaming.

 

[Random Murderer]

Nothing is a thorough test, but that is what runs 24/7 on the computer unless I am rarely gaming, so that is what I test with. The system has frozen and bluescreened many times in the past week, even froze 1.5 hours after taking my last screenshot. I backed the RAM down to 2666MHz and shall see if that helps.

Also, Folding@Home has detected the finest instabilities with my R9 290's overclock (by looking in the logs and seeing where it encounters bad states and has to restart from checkpoint) and my R9 290 requires +81mV @ 1075MHz for problems I could not expose via gaming.

Glad to see I'm not the only one who tests their GPU overclocks with F@H, lol.
Using that exact method, I was able to find that one of my 7970s needed an additional 47mV for stability at the speed I chose(I like all my cards synced when running multiple cards), and like you said, this instability was not found when gaming.

 

[EarthDog]

I wouldn't touch GPU clocks with F@H until AFTER I looped heaven for hours... you really shouldn't test against F@H out of the gate... kind of ruins the project. Borking WU's, essentially intentionally is frowned upon in that community.

 

[cadaveca]

A big thanks MyDog... that is the first piece of concrete information I have seen on this issue.

It's the same as running IBT... when IBT/LinX first came out, we were told by Francios Piednoel (Intel employee, for those that don't know the name) to not use it as it was an internal tool that was meant to burn chips up. But users are still pushing the limits with it.

AVX testing is rather useless, unless you have an app that runs AVX instruction sets. As far as I am aware off, only a couple of renderers run AVX right now.


I end up using BF4 as the "end-all" stress test, combined with Cinebench, wPrime, SuperPi (32M flava), and a encode using Handbrake. 3DMarks added in for fun.

I wouldn't touch GPU clocks with F@H until AFTER I looped heaven for hours... you really shouldn't test against F@H out of the gate... kind of ruins the project. Borking WU's, essentially intentionally is frowned upon in that community.

They don't have an option for Stress-test only? That kind of sucks.

 

[LiveOrDie]

What about Asus RealBench ? It stresses your CPU on things you would normally use.

 

[buildzoid]

I just use IBT for all my stress testing.

 

[cadaveca]

What about Asus RealBench ? It stresses your CPU on things you would normally use.

Meh. ASUS page for it is down, and won't be back until Friday. I'd have to test it and see what it does, and how it compares to the other tests I use.

I just use IBT for all my stress testing.

That's not good enough for my own uses, but if it works for ya well, then that's great! It will shorten the life of your CPU, though, It's Intel Burn Test, which Intel said, will burn up your hardware.

Really though, I was given the impression that that was exactly what it was for, they let it run, and when hardware the hardware dies, they are given an idea about longevity under normal usage.

 

[dumo]

More test

5960X ES USA batch on single stage -40ish C evap. temp, -5C real temp and -11C bios

 

[LiveOrDie]

Meh. ASUS page for it is down, and won't be back until Friday. I'd have to test it and see what it does, and how it compares to the other tests I use.



That's not good enough for my own uses, but if it works for ya well, then that's great! It will shorten the life of your CPU, though, It's Intel Burn Test, which Intel said, will burn up your hardware.

Really though, I was given the impression that that was exactly what it was for, they let it run, and when hardware the hardware dies, they are given an idea about longevity under normal usage.

Thanks dave. Now i have no idea what to use i did like prime seeing it BSOD the fastest what should i use seems i can run my chip for gaming on as low as 1.24v but when stressing i need around 1.3v. And i can't go off reviews as most dont even say they stress the chips after overclocking.

 

[buildzoid]

Meh. ASUS page for it is down, and won't be back until Friday. I'd have to test it and see what it does, and how it compares to the other tests I use.



That's not good enough for my own uses, but if it works for ya well, then that's great! It will shorten the life of your CPU, though, It's Intel Burn Test, which Intel said, will burn up your hardware.

Really though, I was given the impression that that was exactly what it was for, they let it run, and when hardware the hardware dies, they are given an idea about longevity under normal usage.

Funnily enough the stress test and benchmark in Intel's Extreme Tuning Utility is LynX. IBT is just a much lighter weight skin for LynX than XTU. I don't run IBT very long just 10 to 15 min.

 

[LiveOrDie]

Would you call this stable? I'm still to see if this is 100% stable after that ill lower the voltage till the point its not stable. Question does adaptive voltage work the same i use to use it on my z87 rig and be able to swap from manual to adaptive in windows using the ai suite but it locks up my computer?

4.6Ghz on 1.3v

I tested this one below until i got a BSOD or a reset this was the lowest voltage i could go.

4.5Ghz on 1.235v

 

[Jstn7477]

You may want to try higher Stress Levels (the higher the RAM amount tested, the longer each test pass runs) if continuing to use IBT (I'm leary of using it now due to possible over-current degradation mentioned in this thread). 1.235v seems a bit low for 4.5GHz IMO, my 5820K seems to need 1.2V at 4125MHz to prevent random reboots, though I am testing with my RAM at 2666MHz now vs. 2750MHz as it seems much more stable. I am now testing 4.125GHz/1.2V core and 3.5GHz/1.25v ring on my 5820K and it was stable for 24 hours/11 threads of World Community Grid at those volts and 3.37GHz ring.

I'll keep you guys updated with my progress regularly, needless to say I am way more comfortable with this setup versus a week ago, and now have half decent cooling as well (84c max temp on an H100 with 2x2000 RPM cheapo Rosewill push fans and 2x 900RPM Enermax UCTB12s pulling).

 

[LiveOrDie]

You may want to try higher Stress Levels (the higher the RAM amount tested, the longer each test pass runs) if continuing to use IBT (I'm leary of using it now due to possible over-current degradation mentioned in this thread). 1.235v seems a bit low for 4.5GHz IMO, my 5820K seems to need 1.2V at 4125MHz to prevent random reboots, though I am testing with my RAM at 2666MHz now vs. 2750MHz as it seems much more stable. I am now testing 4.125GHz/1.2V core and 3.5GHz/1.25v ring on my 5820K and it was stable for 24 hours/11 threads of World Community Grid at those volts and 3.37GHz ring.

I'll keep you guys updated with my progress regularly, needless to say I am way more comfortable with this setup versus a week ago, and now have half decent cooling as well (84c max temp on an H100 with 2x2000 RPM cheapo Rosewill push fans and 2x 900RPM Enermax UCTB12s pulling).

After IBT i went over to RealBench and used Video Encoding to test and I'm up to 1.25v now for 4.5Ghz it put more stress on the CPU than IBT alos this is with 3000Mhz on the Ram, But i still need to do a few other tests to see if its 99% stable . Yes please keep me update on your progress. When you say Ring voltage this is the uncore right?

 

[Jstn7477]

After IBT i went over to RealBench and used Video Encoding to test and I'm up to 1.25v now for 4.5Ghz it put more stress on the CPU than IBT alos this is with 3000Mhz on the Ram, But i still need to do a few other tests to see if its 99% stable . Yes please keep me update on your progress. When you say Ring voltage this is the uncore right?

Yes, uncore. I also don't have an OC socket on my $300 Gigabyte X99-UD5 Wi-Fi, so I will likely hit a wall beyond 3.5GHz on the uncore. The automatic OC picked 4.3GHz core at 1.25V and it seemed stable, will get back to that core speed within a few days of testing I imagine. I also want to test >100MHz PCIe clock once stable as I had terrible issues even at 101 on the 1.25x strap, could have been RAM speed or something but I'll find out soon.

 

[buildzoid]

(I'm leary of using it now due to possible over-current degradation mentioned in this thread).

Intel's own stress test application and benchmark(XTU) for consumers uses LynX which is the same algorithm as IBT. So IBT is safe unless you think intel supports people frying their CPUs with an app that they sponsored.

 

[LiveOrDie]

Intel's own stress test application and benchmark(XTU) for consumers uses LynX which is the same algorithm as IBT. So IBT is safe unless you think intel supports people frying their CPUs with an app that they sponsored.

It depends if intel intended people to use the tool for stressing rather than just benchmarks.


Up to 1.27v it was unstable on 1.25-1.26v running realbench Heavy Multitasking benchmark.



I cant seem to set the adaptive voltage am i missing some think after i set it my computer won't boot is it because of the baseclock is on 125 and not 100?

 

[Mydog]

New "mainstream" memory arrived today, 16 GB Corsair Vengeance LPX 2800 MHz CL16-18-18-36 2T

First OC-test on them looks OK or??

 

[GhostRyder]

I plan on keeping mine after testing at 4.5ghz (5930K) for 24/7 use since that has a nice voltage point for me compared to 4.6 and 4.7 (Both of which I got stable). I may bump it to 4.6 if I feel the need for more power but right now I am happy and I want to prolong the life of my chip (MY voltage requirements after 4.5 get a little excessive).