Overclocking 4670k on Asus Z97-A

If you run them several times, the benchmarks will slow down regardless... one has to be careful when benchmarking SSD's. Because of that reason (OCZ mentioned this to me) I will manually TRIM a drive (mine are empty, I reformat, same difference) before running another benchmark.

So be careful that is not part of what you are seeing too.

I havent ever seen AS SSD slow an SSD over time, to short a test. Although there are some benches like iometer that use huge portions of drive that definitely slow drive over time
But yeah, good idea to run trim between bunch of benchmarking to keep results on fresh drive.

But no that isnt part of what we are seeing. There is an entire thread on ocn.net with many people showing same thing on various motherboards and bioses from a year or 2 ago, when everyone was posting about it, mine are posted there as well. Other threads in other forums with same. Power savings enabled slow down 4k read and writes to varying degrees on most bioses/mobos unless you partially disable power states by using high performance, or use balanced performance but change cpu min power to 100%, or just enable c6/c7 but disable c1e/c3.

here is a an article from toms, titled, Does Power-Saving Technology Kill SSD Performance
http://www.tomshardware.com/reviews/ssd-hdd-power,2170.html

rge said:

Actually your 4.4ghz balanced with c states enabled took a 30% hit in 4K read speed with 20.45 along with corresponding .169 access time versus all the other runs were 28-30 with access times .113 to .119. Random 4k write was only about 10%. if you run them several times, you will see 4K read and write decreases are repeatable.

Click to expand...

Oh yeah, thanks for drawing that to my attention, obviously not getting enough sleep here

It does seem that various issues, including high DPC latency when transferring data over the NIC (such as when streaming video, which could result in stuttering), mean it's only sensible to run in High Performance mode all the time and just put the system to sleep when it's not being used.

Whilst there's some benefit from disabling EIST and C-states in the BIOS when using High Performance mode, I don't know if doing so might cause higher energy use during sleep and the benefit isn't particularly significant, so I think I'll just leave those enabled.

rge said:

Power savings enabled slow down 4k read and writes to varying degrees on most bioses/mobos unless you partially disable power states by using high performance, or use balanced performance but change cpu min power to 100%, or just enable c6/c7 but disable c1e/c3.

here is a an article from toms, titled, Does Power-Saving Technology Kill SSD Performance
http://www.tomshardware.com/reviews/ssd-hdd-power,2170.html

Click to expand...

Interesting article, thanks. It does say however 'The conventional power saving techniques, such as Enhanced SpeedStep—which reduces clock speed and voltage during idle or low load periods—showed the least impact. This is also the case for C1E mode (enhanced halt). Thus, we recommend keeping both of these enabled, as they contribute to lowering CPU temperature while reducing power consumption' and whilst the graphs clearly show that C1E has more impact than EIST, the relatively small effect they have might be considered an acceptable compromise in order to keep some power saving features.

I might do some more tests with only EIST and C1E enabled and in Balanced mode to see how they compare. It might also have some effect on the DPC latency problem.

yep, from the threads I have seen with variability between mobos/bioses, best thing is like you said, testing various settings on your own mobo/bios/ssd to find best compromise.

I've just got round to doing my own build (the previous one was for my Dad) with the same Mobo (and BIOS) and CPU and I'm not getting a very good bench with my Samsung 840 Evo 256GB SSD. My Dad's system has a Crucial MX100 256GB and to recap, this is what it got with AS SSD:



This is what I get with my SSD, running at 4.4Ghz in High Performance mode. I haven't disabled any power saving features in the BIOS yet but I don't think they do anything when using High Performance anyway.

My C (Windows) partition


My S (Games) partition


Whilst it does better for Seq. Write and 4K read, it's a bit worse for 4K write and significantly worse for 4K-64Thrd Write and Read Acc. time. Whilst the test is running, the 4K-64Thrd Write number occasionally hits 350-400 MB/s but most of the time, certainly for the latter half of the test, is around 70 MB/s or lower.

I was expecting the Samsung 840 Evo to be somewhat better than the MX100 but certainly not significantly worse. Anyone got any ideas what's happening here?

I also run the test in Samsung Magician at 3.8Ghz before overclocking and 4.4Ghz (both in High Performance mode) and got this:





I was quite surpised and disappointed at the poor Random Write (IOPS) result at 3.8Ghz but it seems fine at 4.4Ghz. Although I did Optimise the SSD before doing the latter test, so maybe that's the reason for the difference and not the overclock?

Looks like my 4670k isn't as good as my Dad's one sadly. At [email protected] it was fine with Aida64 stresstest but BSOD running ArmA3, just on the menus not even running a mission. Something like "WHEA_Uncorrectable_Error".

I bumped it up to 1.25v and it's fine now. I might be able to decrease it a bit still from there. Even though I'm using the superior Scythe Kotetsu HSF compared to my Dad's Hyper 212 Evo, temps were about the same at [email protected], so it'll be running hotter at 1.25v. I haven't bothered running Aida64 @1.25v yet though, so I'm not sure exactly how much higher but ArmA3 only pushed it to about 60c, so it's fine for normal use.

Damn thing just gave me "WHEA Uncorrectable Error" again, whilst at [email protected]. Surely I don't need to increase the voltage even higher for this basic overclock?

I was running DCS World at the time. Played ArmA3 for several hours the other day at these settings, which was what triggered the same error @1.225v.

Had some weird stuff happening shortly before it BSOD as well. First, I noticed that VAC (Voice Assisted Commands) wasn't working, so I opened the editor to check and the My Documents\VAC folder was completely empty. OK, I thought, at least I'm running Crashplan to backup my files, so I tried to restore with that but each time it would say completed but in the results it showed 0 restored files and indeed, nothing was restored. After rebooting following the BSOD, Crashplan worked fine to restore these files (it restores them to the desktop for you to move to wherever) but then I checked My Documents\VAC and all the files were already back again!

Doveman, I see you mentioned AIDA 64 was used to stress test it, how long was it run?

Only 30mins at a time. ArmA3 caused the BSOD within minutes of launching it though, whilst it was just sitting on the menu, not even playing the game, so it seems games (or some games at least) stress the system in ways that Aida64 doesn't and I need something that stresses the system in the same way as those games to check stability.

From my experience 30minutes is not enough of a stress test to confirm stability. Aida 64 is a good stress test though I have found instances where I can pass 3-4 hours of Aida and still have issues for what I use my pc for and that includes gaming. I find that in order for my rig to be 100% stable for what I do on it, I need to run Prime Blend for a minimum of 2 hours. Note though, I have also found that Prime Blend will heat the system up more then Aida 64. Therefore I may have to lower my Oc depending on my temperatures, I will also need more Cpu voltage at X oc in order to pass Prime Blend, compared to Aida. That said, I do not play Arma3 so you may need additional testing to see if it is truly stable. Though, I would be surprised if you passed 2 hours Prime blend and were still having issues. If you did it may be an issue with the program and not the OC. Which could be tested by setting things to default and see if the issue still occurs.

The guides advise to avoid running Prime with Haswell and to use Aida64 instead. Something specific about the way Haswell tends to overheat running those other stress tests as I recall. Aida64 seems sufficient anyway, as it pushes the temps about 10c higher than any normal apps/games do.

I've bumped up my Vcore from 1.25v to 1.26v and my CPU Input voltage from Auto (1.76v) to 1.8v now, so I'll have to see if that stabilises it.

As I understand, that WHEA error is sent directly from the CPU so I don't think it can be an issue with the program. Besides, it never crashed on my Phenom II X4 955 I was using before upgrading, so I'm pretty sure it's the OC that just needs tweaking.

doveman said:

The guides advise to avoid running Prime with Haswell and to use Aida64 instead. Something specific about the way Haswell tends to overheat running those other stress tests as I recall. Aida64 seems sufficient anyway, as it pushes the temps about 10c higher than any normal apps/games do.

I've bumped up my Vcore from 1.25v to 1.26v and my CPU Input voltage from Auto (1.76v) to 1.8v now, so I'll have to see if that stabilises it.

As I understand, that WHEA error is sent directly from the CPU so I don't think it can be an issue with the program. Besides, it never crashed on my Phenom II X4 955 I was using before upgrading, so I'm pretty sure it's the OC that just needs tweaking.

Click to expand...

Your choice to use what ever program you wish, I was just giving you my take on Aida 64. I have a 4770k and yes Prime 95 does heat it up more then Aida and also more then what I use it for on a daily basis. That said I need the rig to stay running without a hitch for what I use it for. I know if I pass 2 hours prime blend with my OC I will not have issues with the rig for what I use it for. That was not the case when I used Aida 64 to stress test it, I still had random crashes even after 3 1/2 hours of Aida. I do not think for a minute that 30 minutes of Aida is enough of a "stress test" to consider the rig stable.

As far as the program being an issue, I assume you did a re install of it? If so, you never know a file could be corrupted, it is possible. Though I would lean towards the OC not being stable.

Whilst I haven't had any problems at all whilst in long gaming sessions or general use but yesterday IE11 was doing weird things (I'm used to that though and normally use Chrome) and today, I've been trying to transfer a lot of files from one USB3 drive to another and it's crashed twice. The first time, I was comparing drives with FreeFileSync and I got a BSOD , so I dropped the CPU Multi down from 44 to 43. I restarted the data transfer but after an hour or so, I noticed the activity lights on both drives had stopped flickering and then I found I couldn't even wake the monitor up, so it had obviously crashed again silently.

I've got Intel Extreme Tuning Utility installed now, although I haven't used it so far but I'll drop the Multi down to 38x and see if it still crashes but it seems pretty bad if it can't even manage 43x at 1.26v (1.8v in CPU Input) doesn't it?

Well I got the same BSOD doing the folder compare with FreeFileSync even at 3.8Ghz, so I guess it's not the overclock that's the problem. The error is "Reference by pointer" and according to BlueScreenView it's caused by ntoskrnl.exe, on two of the three occasions anyway, on the third it was caused by fltmgr.sys.

I'm still having some issues and the odd BSOD.

Chrome is throwing up one or several error popups every time I launch it, which can't be clicked on to show details or close, so I have to kill it by right-clicking on it's icon on the taskbar and then do the same for the error popup. I installed the latest Windows Updates as I read a report that a recent one might be the problem (Chrome wasn't doing this about a week ago) but that didn't help. Running it in Windows 7 compatibility mode does seem to be helping now though.

Yesterday after I'd tried to launch Chrome and had this problem, I tried to launch ArmA3 and got a BSOD "Reference by Pointer".

I dropped my OC down from 4.4Ghz to 4.3Ghz and tested my RAM with memtest overnight.

Today, whilst doing the Windows Updates one crashed (I think it was Microsoft Mouse and Keyboard Center) and then IE crashed referencing flash.ocx.

I've taken some photos of my BIOS settings (sorry about the quality) and would be grateful if you could take a look and let me know if there's anything I should change.