Help: BIOS settings wanted for better performance from a Z170 mobo system

[Dougal 165]

Working from memory, just cpu all cores, vcore and xmp profile. On phone now but will check my settings tmrw

[westman 257]

If you not want run aisuite , you run Hero VIII ?

This is a Asus app with light fotprint for OC in win.

I prefer it

http://www.mediafire.com/?6dezk90yig0s9om

[martin-w 3,256]

If anyone finds a way of keeping Asus AI Suite running, WITHOUT that damn onscreen graphic that Pete referred to, please post it, as I find Asus Fan Xpert brilliant for TOTAL fan and water pump control. MUCH better than Corsair Link.

Nice PC build Pete - almost the same as mine;-)

 

 

On mine it's a round device where you can select Fan Xpert etc. 

 

It's easy to get rid of, just right click and click exit.

 

To stop it appearing all together, even at startup... go to the right hand corner of the task bar, click the upward pointing arrow to reveal the icons, right click the Asus icon, and untick the "Asus Mini Bar". All done.  :smile:

 

Asus always allow such things to be removed.

 

 

 

Actually I just reinstalled it and re-ran it, but changing the strting settings a bit as another poster mentioned above. It went quite well, an it's given my what looks so far to be a stable 4.7GHz. I had to go back into the BIOS and reenable XMP mode again to get the memory speed correct, but apart from that I'm quite happy with it -- except for that non-removeable icon it sticks on the screen!

 

 

 

 

 

 

 

 

 

 

Pete... 5WO will tune the RAM also. So it's not surprising it wasn't the same as the XMP profile. 5WO in it's wisdom "may", theoretically, have been correct to do so, but yes, if you can indeed run with the XMP profile all is well.

 

I found the same, 5WO tuned the RAM down a bit, as the software deemed that to be the most stable configuration. I too then entered the XMP profile and found it stable. But it's not that the software was wrong technically, not according to the Asus engineers that coded it.

 

except for that non-removeable icon it sticks on the screen!

 

As I mentioned to Phil, right click the Asus icon on the task bar and untick "Asus Min Bar".   :smile:  All gone!

4.7 is a great result from 5WO. Max I achieved was 4.6. You should be happy with that.

 

What was your voltage and temp under load Pete, while running RealBench? Temp is VERY important. What kind of cooling? Temp use RealTemp. Voltage CPU-Z.

 

Re Sandra, I'm very confident all is well for you. I don't personally believe you have any issues. The fact that the benchmarks are inconsistent and read different values doesn't surprise me at all. 

 

As long as your temps are good and CPU voltage under load within normal range, that's fine. Asus and Intel recommend a max of 1.45 volts. AiSuite is designed not to exceed 1.44, and that will be an adaptive voltage... but do check the temp under load.

 

Manual overclocking [pretty easy] will enable you to lower the voltage and generate less heat. Hyper threading off will drop temp by a further 10 degrees. you must decide of course if you need HT or not.

 

Even though AiSuite has run stress tests automatically... I would still run ReaBench for an hour or so to be sure. RealBench is a very realistic stress test, it's not a synthetic stress test like Prime95 that bares no resemblance to how we run our PC's in everyday use.

 

Finally, don't be tempted to run Prime95... it's vicious these days as it runs AVX. If you ignore me and do run it, only run the blend test.

[westman 257]

If you want to do it manual , you can leave the most at auto

Set XMP , adaptive voltage , LLC ,

Good start is vcore 1.35-1.375v LLC at 5 , multi/uncore 46/45 then raise the multi\uncore from there.

Uncore 1-2 less then multi ( tumb of rule)

Initial a only test aida64 mem and run Cinebench R15 both multi and single tread test.

Then 1-3 hours Rog RealBench to se that a have it stable.

Here a easy bios setup with adaptive my CPU is CB15 stable up to 5.0ghz with this settings.

 

https://www.dropbox.com/s/nngde6cnyq6u209/Extreme%20VIII-bos.png?dl=0

[Pete Dowson 558]

Run some different memconfigs , only run Aida64

Run a Asus Extreme VIII all tests at 4.7ghz

Not now if it help

My eng is very bad but the screen talk for itself.

 

https://www.dropbox.com/s/dz0owf6q7gjskbl/aida-6700.png?dl=0

 

I had a good look at your 4 results. It seems that the differences are really all in the individual memory, as follws:

 

54.3 GB/s, 34.9nS  with Membus 1800, ratio 54.3, 13-13-13-28-CR1

51.9 GB/s, 36,7nS  with Membus 1800, ratio 54.3, 14-14-14-55-CR1

46.9 GB/s, 39.2nS  with Membus 1600, ratio 48.3, 14-14-14-35-CR1

45/0 GB/s, 42.4nS  with Membus 1600, ratio 48.3, 20-16-16-35-CR2

 

I assume this was all with the same memory sticks. Was the last one above (top right in the picture) the "mormal" or default for them?

 

What can't be seen in the pictures is the make of memory, which might be important for overclocking, and the assorted voltages which affect the memory and which I assume had to be increased to get the better performance.

 

Many of the things I've read recently seem to be saying that overall better performance arises from better memory performance even if it means reducing core clock speeds to do this without damaging the memory from too high a voltage. Can anyone here confirm that? Should I be looking for an idiot's guide to overclocking memory on a Z170 motherboard?

 

To stop it appearing all together, even at startup... go to the right hand corner of the task bar, click the upward pointing arrow to reveal the icons, right click the Asus icon, and untick the "Asus Mini Bar". All done.  :smile:

 

Ah, I needed that too! ;-)

4.7 is a great result from 5WO. Max I achieved was 4.6. You should be happy with that.

 

What was your voltage and temp under load Pete, while running RealBench? Temp is VERY important. What kind of cooling? Temp use RealTemp. Voltage CPU-Z.

 

I don't remember at present and it's doing some other tests, in Prepar3D, for me at present. I'll check temperatures and voltage later today. I think the max temp was around 72. Voltage wasn't that high. Both seem lower than the last time I got to 4.7 -- the time which gave me blue screens.

 

The cooling is air except for the Corsair H100 water cooler on the processor. I once had a system built with full circulating water cooling (in fact it is still in use as my delevlopment and testing PC in my office, the one I'm typing on now, but not overclocked any more). I decided I didn't like it, too many pipes in the way and very difficult to change components.

Re Sandra, I'm very confident all is well for you. I don't personally believe you have any issues. The fact that the benchmarks are inconsistent and read different values doesn't surprise me at all.

 

 

Good. Time perhaps to look elsewhere for my Prepar3D stutters and hesitations. Must be one of the other processes. Elmination is the order of the day today ... and then trying to decide what to do about it once the culprit is identified! I already prune them as much as I can, and move much onto other PCs (there are 7 involved in my cockpit operations, though only 4 heavily).

 

Thanks

Pete

[westman 257]

My sticks is pretty good , XMP 3600 15-15-15-36 1T

My allday settings is 14-14-14-35 1t, memvoltage 1.375v

 

Its only to find good sticks a prefer samsung b-die chip , G-Skill doing good ones.

The CPU memcontroller is the problem ( for 24/7 i feel comfortable 1.4-145v on mems)

The memory chip (b-die) have no problem with 1.8-1.9v but put a lot stress on the memcontroller.

 

Asus Hero and Extreme VIII is good for 3600mhz over that it be tricky , better to run 3600 and tight the timings.

[martin-w 3,256]

 

 

I don't remember at present and it's doing some other tests, in Prepar3D, for me at present. I'll check temperatures and voltage later today. I think the max temp was around 72. Voltage wasn't that high. Both seem lower than the last time I got to 4.7 -- the time which gave me blue screens.

 

 

Thanks

Pete

 

 

 

Yep, that temp is fine. Same temp as me more or less in RealBench. With NH-D15S.

 

There is the RealBench benchmark as well, in addition to the stress test. You might like to try that.

[martin-w 3,256]

My RealBench score is...

 

Image Editing          50809

Encoding                 180602

Open CL                 100032

Heavy Multitasking 146124

System Score         99385

 

17 6700K @ 4.6 GHz

GSkill Ripjaw V 3200 MHz

EVGA GTX 980 Ti Classified

[Pete Dowson 558]

If you want to do it manual , you can leave the most at auto

Set XMP , adaptive voltage , LLC ,

Good start is vcore 1.35-1.375v LLC at 5 , multi/uncore 46/45 then raise the multi\uncore from there.

Uncore 1-2 less then multi ( tumb of rule)

Initial a only test aida64 mem and run Cinebench R15 both multi and single tread test.

Then 1-3 hours Rog RealBench to se that a have it stable.

Here a easy bios setup with adaptive my CPU is CB15 stable up to 5.0ghz with this settings.

 

https://www.dropbox.com/s/nngde6cnyq6u209/Extreme%20VIII-bos.png?dl=0

 

Would I be better off overclocking the processors more, above my current 4.7 to your illustrated 4.9, or replacing the Crucial memory with, say, GSkill TridentZ 3600MHz rated as 15-15-15-35? I can get the latter easily enough from Amazon.

 

From several adviser's and results I see that often the memory is more likely to be the bottleneck, once you have a good processor speed (like 4.7 GHz?).

 

I've got a copy of the BIOS setting screens you posted, so thank you.

 

Regards

Pete

[westman 257]

 

Many of the things I've read recently seem to be saying that overall better performance arises from better memory performance even if it means reducing core clock speeds to do this without damaging the memory from too high a voltage. Can anyone here confirm that? Should I be looking for an idiot's guide to overclocking memory on a Z170 motherboard?

 

 

Thanks

Pete

I hope you understand my bad eng.

You dont need to rduce the core speed ,

Damaging the memory from to high voltage , with most memchips the CPU memcontroller die first.

 

Not seen any good guide for 24/7 memoverclocking yet.

 

Can dig in much deper but very very hard to find the right words.

 

Best chips ar the latest batch of G-Skill 4133 or galax limited 4133 , easy to reflash them to 3600 c13 or c14 1T 1.4v XMP ( do your own XMP profile )

 

Or buy 3600mhz C15 G-Skill and manual tight the timings to cl14

Would I be better off overclocking the processors more, above my current 4.7 to your illustrated 4.9, or replacing the Crucial memory with, say, GSkill TridentZ 3600MHz rated as 15-15-15-35? I can get the latter easily enough from Amazon.

 

From several adviser's and results I see that often the memory is more likely to be the bottleneck, once you have a good processor speed (like 4.7 GHz?).

 

I've got a copy of the BIOS setting screens you posted, so thank you.

 

Regards

Pete

Go for the G-Skill first , then when dialed them to C14 try 4.8-4.9 Its depending on the CPU .

 

Good luck

[Pete Dowson 558]

Or buy 3600mhz C15 G-Skill and manual tight the timings to cl14

 

Go for the G-Skill first , then when dialed them to C14 try 4.8-4.9 Its depending on the CPU .

 

Good luck

 

Thanks! I'll probably try this in the New Year.

 

Pete

[simmerhead 885]

Just in time to get ready for the 64 bit revolution of 2017! Good luck Pete, and may you have a wonderful holiday!

[lodestar 186]

 

I was really only looking for 1 figure for bandwidth and one for latency

I like this tool becouse you can realy test a lot of things. I've read some technical stuff about and this is how i understand it:

 

Idle bandwidth and idle latency are the best results your system can achieve - the most optimal scenario, when system is at idle. Loaded latency and peak bandwidth are the worst case scenario - high load, where bandwidth usage is at max, and that will lead to increased latency. You can play with CMD commands to test whatever you want, there is a list with all available commands in supplied PDF.

 

You can use CMD to run "mlc --loaded_latency" only. The first lines are the worst case scenario(high load) - this should be maximum bandwidth. For latency, first lines are the worst you can get under full load, and the best latency is in the last lines.

 

 

 

I still would love to know what bandwidth and latency figures you managed to get, though for comparison I suppose they'd need to be from the same tool -- i.e. Sandra or AIDA64, or the one you just pointed me to if you also tel me which figures I should look at or what parameters I should use.

 

I don't think this will be useful for you, becouse my system is prety old now - 2700K and 560ti. My bandwidth is limited by CPU, about 1/2 of what you can get with 6700K. 

 

My AIDA64 benchmark: https://s28.postimg.org/n4hf2jy8d/cachemem.png

 

Sandra:  Memory bandwidth: 21.12GB/s

                 Memory Latency : 20.2ns

 

With intel memory latency check, using "mlc --loaded_latency", max bandwidth is 23.7GB, best latency is 53.20ns(pretty much like AIDA results).

 

 

 

I looked at that, but thought I'd try the utomtic method again first. This time, because I think of different choices made before starting it, it seems to have done very well -- and even the temperatures seem fine.

 

maybe they improved it now. However, i belive manual overclock will be more stable, and better for CPU. I've "studied" OC guides and forums for a month or two and choosed the best metod(in my opinion)after that. After more then 4 years, my CPU still uses same Vcore, and i tested it last week - same Vcore for 4.9 and 5.0GHz like 4 years ago. So, no degradation  :smile:

 

Definitely check your Vcore and temps under load, it's most important thing when overclocking. With Asus utlity, my vcore was over 1.7 for 4.2Ghz(under load), this is why i uninstalled it immediately.

 

I'm about to build new system soon, similar to what you have, so i am interested in your results. 

 

About P3D, i'm not using it becouse i was newer able to get same smoothness like with FSX 1/2 vsync and fps locked at 30. Maybe this is the problem, if i use any other vsync in FSX - i get stutters and scenery loading looks pretty bad, it remainds me of my P3D experience. Just a thought

[Pete Dowson 558]

My RealBench score is...

 

Image Editing          50809

Encoding                 180602

Open CL                 100032

Heavy Multitasking 146124

System Score         99385

 

17 6700K @ 4.6 GHz

GSkill Ripjaw V 3200 MHz

EVGA GTX 980 Ti Classified

 

I've now run RealBench, and get these figures:

 

 

Image Editing          235011

Encoding                 189299

Open CL                 83627

Heavy Multitasking 180064

System Score         155274

 

17 6700K @ 4.7 GHz

Corsair 3200 MHz

GTX 1080

 

With Benchmark, are higher numbers better or worse?

 

Definitely check your Vcore and temps under load, it's most important thing when overclocking. With Asus utlity, my vcore was over 1.7 for 4.2Ghz(under load), this is why i uninstalled it immediately.

 

Under RealBench, using RealTemp, my temperatures were MAX 80/76/71/73 (on the 4 cores), but it only hit highs like 80 very briefly now and then. It doesn't show average but I would say it was about 74 on Core 0.

 

The top voltage I saw was 1.43/1.44 mostly, occasionally hitting 1.49 briefly.

 

Pete

[westman 257]

I've now run RealBench, and get these figures:

 

 

Image Editing 235011

Encoding 189299

Open CL 83627

Heavy Multitasking 180064

System Score 155274

 

17 6700K @ 4.7 GHz

Corsair 3200 MHz

GTX 1080

 

With Benchmark, are higher numbers better or worse?

 

 

 

Under RealBench, using RealTemp, my temperatures were MAX 80/76/71/73 (on the 4 cores), but it only hit highs like 80 very briefly now and then. It doesn't show average but I would say it was about 74 on Core 0.

 

The top voltage I saw was 1.43/1.44 mostly, occasionally hitting 1.49 briefly.

 

Pete

Good score Pete

 

Image Editing , Martins score someting wrong , 255902

Encoding, average scores , 209962

Open CL, depending on gpu , Martin 980ti , Pete 1070? 1080-1070 Cards downclock the mems , my run with a 980 , 10032

Heavy Multitasking, Martins score to low someting wrong, Petes score ok, 198736

Systemscore, Martin should be 150000, Petes godd score, 171151

 

Pete, what is your LLC setting think it overshot

Vcore set in bios 1.43 someting

Think the voltage at 4.7 seems little high but can be the CPU

 

Iam not try to harm any , count this in i bench very much , comepeting, have very good binned CPUs and mem.

Martin my best score with 2cores HT enabled ( 2cores/ 4treads) 119017

Have you turbo set to 2 cores , something is wrong with your score.