z170 Extreme 7+ and Trident z 3200 o/c
Printed From: ASRock.com
Category: OverClocking(OC) Zone
Forum Name: OC Technical Discussion
Forum Description: OC enthusiast's tech talk
URL: https://forum.asrock.com/forum_posts.asp?TID=2713
Printed Date: 22 Nov 2024 at 2:57am Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com
Topic: z170 Extreme 7+ and Trident z 3200 o/c
Posted By: clubfoot
Subject: z170 Extreme 7+ and Trident z 3200 o/c
Date Posted: 23 May 2016 at 8:27am
I have a pair of F4-3200C16D-16GTZ modules in the second and fourth slots. No issues running the XMP profile, but 3466 causes errors in Prime95 mixed mode. On another note in AUTO and fixed 1.350v the BIOS reports 1.367v!
I see one site using 1.42 to achieve 3466 and 3600 but that seems really high.
An advice on what BIOS setting I should be tweaking to o/c these memory modules besides DRAM voltage?
BTW I've tried BIOS 2.6 and 3.0,...2.6 seems a little better and cpu is at 4.6 GHz.
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Replies:
Posted By: parsec
Date Posted: 23 May 2016 at 9:42am
If you check the Memory Support List for our board (you know I have one too, right?), you can find at the top of that page this note:
Note3: For 2pcs (DS), please use A1 and B1 slot.
Your G.SKILL F4-3200C16D-16GTZ TridentZ memory is a Double Sided (DS) type of DIMM, so you should try the A1 and B1 slots. I don't have any DS memory, so I've never tried this. I'd like to know how it works for you.
I was ready to say, clear the UEFI/BIOS after moving the memory to the A1 and B2 slots, but then I remembered you use two 950 Pros in RAID 0, and we both know what happens if you do that. Safe to try it without clearing the UEFI, but you might be stuck doing that.
I am using G.SKILL F4-3200C16-8GTZB TridentZ memory, which is 2 x 4GB DIMMs, Single Sided (SS.) HWiNFO64 shows a temperature reading for this memory. Nice warm day today, the memory is reading at 88° F. I wonder what yours is?
I've noticed a difference between the DRAM Voltage setting value, and the apparent actual value, as shown in the UEFI H/W Monitor screen, and in a monitoring program like HWiNFO64. I also get a reading of 1.367V for a setting of 1.350V. Running at 3200, I can set the DRAM Voltage to 1.330V, resulting at a reading of 1.352V. I've used this memory at 3200 below 1.350V, a setting of 1.320V shows as 1.344V.
Your DS memory many need more voltage than mine, so don't think you can reduce it and get the same results as I do.
Don't forget the basic reality known as the "silicon lottery", the differences in CPUs depending upon where they were cut from the large production processor die. Your CPU's memory controller may be better or worse than mine.
What are the values for VCCIO and VCCSA voltages you have now? The XMP profile, or Auto settings in the UEFI resulted in some (seemingly) high values for these voltages. VCCIO of 1.150V (reading of 1.176V), and VCCSA of 1.250V (reading of 1.280V), is what I get. These values may be needed to allow the memory to work with boards and processors of different quality. The concept (reality) of the silicon lottery really needs to be understood better by PC builders
I am able to reduce both of these voltages with the memory at 3200. VCCSA is set to 1.140V (reading of 1.168V), VCCIO is set to 1.100V (reading of 1.128V.) Again, you may or may not be able to use these voltages with your DS memory.
I've had my memory at 3600, DRAM Voltage set to 1.350V, with VCCSA and VCCIO set to 1.250V and 1.150V, respectively.
My other settings for 3600 are all timings on Auto, and MRC Fast Boot set to Disabled, to allow memory training to occur during POST. If you have the Fast Boot option set to anything besides Disabled, set it to Disabled to insure memory training occurs. I think I could not set MRC Fast Boot back to Enabled at 3600, after several POSTs. I can set MRC Fast Boot to Enabled, and Fast Boot to Ultra Fast, at 3200.
I was reading a thread recently about extreme over clocking of DDR4 memory. Those guys don't blink at using 1.5V, which is the Intel absolute maximum. Some of them go beyond that voltage, but we are talking ~4000 on Skylake and ~3200+ on X99 boards. So ~1.4V is not crazy, but the VCCSA and VCCIO voltages are almost as important as the DRAM Voltage.
I'm not a memory OC expert at all, this is the basic stuff, tweaking your timings comes later.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 23 May 2016 at 10:51am
parsec wrote:
If you check the Memory Support List for our board (you know I have one too, right?), you can find at the top of that page this note:
Note3: For 2pcs (DS), please use A1 and B1 slot. Your G.SKILL F4-3200C16D-16GTZ TridentZ memory is a Double Sided (DS) type of DIMM, so you should try the A1 and B1 slots. I don't have any DS memory, so I've never tried this. I'd like to know how it works for you. >Yeah I saw that and tried the A1 and B1 and it worked there too, but I got to reading the manual for the umpteen time and moved them to its recommended slots :( Live and learn.
I was ready to say, clear the UEFI/BIOS after moving the memory to the A1 and B2 slots, but then I remembered you use two 950 Pros in RAID 0, and we both know what happens if you do that. Safe to try it without clearing the UEFI, but you might be stuck doing that. >Don't get me going on clearing CMOS, switching BIOSes or removing power from this board I've finally got it feeling solid and I'm not touching it,...I miss racing. We really need an RST update to 14.8
I am using G.SKILL F4-3200C16-8GTZ B TridentZ memory, which is 2 x 4GB DIMMs, Single Sided (SS.) HWiNFO64 shows a temperature reading for this memory. Nice warm day today, the memory is reading at 88° F. I wonder what yours is? >83 and 85! Which I thought was much cooler at 28.3 and 29.5 Celsius.
I've noticed a difference between the DRAM Voltage setting value, and the apparent actual value, as shown in the UEFI H/W Monitor screen, and in a monitoring program like HWiNFO64. I also get a reading of 1.367V for a setting of 1.350V. Running at 3200, I can set the DRAM Voltage to 1.330V, resulting at a reading of 1.352V. I've used this memory at 3200 below 1.350V, a setting of 1.320V shows as 1.344V.
Your DS memory many need more voltage than mine, so don't think you can reduce it and get the same results as I do.
Don't forget the basic reality known as the "silicon lottery", the differences in CPUs depending upon where they were cut from the large production processor die. Your CPU's memory controller may be better or worse than mine.
What are the values for VCCIO and VCCSA voltages you have now? The XMP profile, or Auto settings in the UEFI resulted in some (seemingly) high values for these voltages. VCCIO of 1.150V (reading of 1.176V), and VCCSA of 1.250V (reading of 1.280V), is what I get. These values may be needed to allow the memory to work with boards and processors of different quality. The concept (reality) of the silicon lottery really needs to be understood better by PC builders
I am able to reduce both of these voltages with the memory at 3200. VCCSA is set to 1.140V (reading of 1.168V), VCCIO is set to 1.100V (reading of 1.128V.) Again, you may or may not be able to use these voltages with your DS memory.
I've had my memory at 3600, DRAM Voltage set to 1.350V, with VCCSA and VCCIO set to 1.250V and 1.150V, respectively. >I 'll get these for you,...
My other settings for 3600 are all timings on Auto, and MRC Fast Boot set to Disabled, to allow memory training to occur during POST. If you have the Fast Boot option set to anything besides Disabled, set it to Disabled to insure memory training occurs. I think I could not set MRC Fast Boot back to Enabled at 3600, after several POSTs. I can set MRC Fast Boot to Enabled, and Fast Boot to Ultra Fast, at 3200. >Hmmm, MRC to disabled may be the trick to getting to 3600 stable! Same here I can run FAST and ULTRA at 3200,...haven't tried it at 3600, but I did have MRC on at 3466 and it would boot and run through Realbench but if I leave Prime95 running for over an hour I would lose a cpu core :( And when I saw what the reporting voltage was I did not want to go there.
I was reading a thread recently about extreme over clocking of DDR4 memory. Those guys don't blink at using 1.5V, which is the Intel absolute maximum. Some of them go beyond that voltage, but we are talking ~4000 on Skylake and ~3200+ on X99 boards. So ~1.4V is not crazy, but the VCCSA and VCCIO voltages are almost as important as the DRAM Voltage. >Yeah,...1.42v is a long way away from 1.350v but we've seen that even the cpus can take allot of volts if they have good cooling and survive. I'm not a memory OC expert at all, this is the basic stuff, tweaking your timings comes later. > Wish me luck moving the DIMMs :)
|
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 23 May 2016 at 11:25am
Moved the DIMMs to A1 and B1 and turned off MRC,...no issues on XMP 3200.
DRAM AUTO 1.350v = 1.368v VCCIO AUTO 1.150v = 1.184v VCCSA AUTO 1.250v = 1.272v
Which BIOS are you using 2.6 or 3.0?
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 24 May 2016 at 7:47am
Using BIOS 2.6
Using A1 & B1 3466 at AUTO voltages ram temps during Prime95 blend 109F & 112F and passes Realbench 2.34,...no issues.
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: parsec
Date Posted: 24 May 2016 at 11:46am
I'm using UEFI version 3.00.
Really for no other reason than I feel I should try each UEFI version for the ASRock boards I have to see how they are, functionally. Nothing wrong with 2.60 except... well, as its description states...
Well, 3466 was easy enough, right? Some speeds like 3466, I believe, are more difficult to use. I bet you can get 3600 working as I did.
Your VCCSA and VCCIO voltages are almost identical to mine. You could try reducing them which might reduce the CPU temps a bit. I wonder if that would affect the DRAM temps?
Are those the DRAM temperatures during Prime95? Interesting they are that high, not that I really know better.
Now you've got me wanting 3600 again, gotta try that, I'll be back...
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: parsec
Date Posted: 24 May 2016 at 1:21pm
Back at 3600, no problem, first try:
Using Auto for all timings results in the same values for 3200.
My TridentZ memory has the cheaper Samsung "B die" memory chips, as far as I know. You should be able to get there with yours, unless the DS DIMMs are more difficult to get stable at higher speeds. There are zero DS DIMMs in the Memory Support list above 3400, but you've got that beat already!
HWiNFO should tell you what memory chips you have in that kit. Not in the monitoring screen, but in the system information screen, just click on the Memory entry in the device listing tree on the left. I'd be surprised if they are not Samsung.
The Number of Ranks value should be two for your memory, which is shown on the memory information display.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 25 May 2016 at 12:21am
parsec wrote:
I'm using UEFI version 3.00.
Really for no other reason than I feel I should try each UEFI version for the ASRock boards I have to see how they are, functionally. Nothing wrong with 2.60 except... well, as its description states... > Long story: I have an issue with the A BIOS chip that the B chip does not have with installing windows 10 rst driver and restoring an image. I'm not sure if it's the mb switch or the A chip itself but I've requested a replacement. Our boards are very finicky, cold shutdown, taking out video cards (xFire), having both windows install and driver usb drives connected,...You touched on it in another post where you were helping another user with APCI I believe not being initialized at least once.
Well, 3466 was easy enough, right? Some speeds like 3466, I believe, are more difficult to use. I bet you can get 3600 working as I did.
Your VCCSA and VCCIO voltages are almost identical to mine. You could try reducing them which might reduce the CPU temps a bit. I wonder if that would affect the DRAM temps? > I'll give this a shot next.
Are those the DRAM temperatures during Prime95? Interesting they are that high, not that I really know better. >Yes,...during P Blend.Now you've got me wanting 3600 again, gotta try that, I'll be back... |
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 25 May 2016 at 12:37am
parsec wrote:
Back at 3600, no problem, first try:
Using Auto for all timings results in the same values for 3200.
My TridentZ memory has the cheaper Samsung "B die" memory chips, as far as I know. You should be able to get there with yours, unless the DS DIMMs are more difficult to get stable at higher speeds. There are zero DS DIMMs in the Memory Support list above 3400, but you've got that beat already!
HWiNFO should tell you what memory chips you have in that kit. Not in the monitoring screen, but in the system information screen, just click on the Memory entry in the device listing tree on the left. I'd be surprised if they are not Samsung.
The Number of Ranks value should be two for your memory, which is shown on the memory information display.
|
It may be more difficult for me as my cpu is o/c to 4.6GHz with cache at 4200MHz. You were able to keep CAS at 16 in AUTO!? I've got to try 3600 with lower volts
http://s75.photobucket.com/user/clubfoot_2006/media/modules%20info.jpg.html" rel="nofollow"> http://s75.photobucket.com/user/clubfoot_2006/media/34664600.jpg.html" rel="nofollow">
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 25 May 2016 at 4:47am
No luck,...will not post successfully,...at least at auto volts. I have MCR off and even moved to BIOS 3.00. This config is very stable at 4.6 cpu and 3466 ram,...very happy and the board feels good and solid.
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: parsec
Date Posted: 26 May 2016 at 9:57am
clubfoot wrote:
No luck,...will not post successfully,...at least at auto volts. I have MCR off and even moved to BIOS 3.00. This config is very stable at 4.6 cpu and 3466 ram,...very happy and the board feels good and solid. |
I wouldn't feel bad about that either.
I revisited 3600 as you know, but while I said "no problem" it's just at the edge of stability.
I must keep the VCCSA voltage at 1.250V (actual according to the UEFI and HWiNFO, 1.280V) and VCCIO at 1.150V (actual 1.176V.) DRAM voltage must be set to 1.350V (actual 1.376V) too, or stress testing will fail. That's using the timings for 3200 (16 18 18 38 2T).
The TridentZ GTZ 3600 timings are 17 18 18 38 2T (my memory is the cheaper GTZB), and of course no idea what the rest are.
I'm not thrilled about the high VCCSA and VCCIO voltages at 3600, I can run 3200 at much less of each.
This 3200 TridentZ 4GB x 2 is going for $52 now at Newegg, so I can't complain. I really don't see the board as being the limiting factor.
Hey, why don't I try 3400 or 3466... well duh!
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: parsec
Date Posted: 26 May 2016 at 10:29am
Running 3466 now, seems Ok so far, not tested yet.
Set VCCIO to Auto, and I'm looking at 0.976V in HWiNFO right now! How can that be?
Question for you, how does the 3.00 UEFI seem for you regarding the freezing issue that can happen?
For me it seems better, possibly not perfect since I had a... hesitation I'll call it. I was moving around the UEFI screens and it seemed to freeze. But in a few seconds it responded, and I finished and exited fine.
It seems the freezes happen, for me at least, in two situations:
After adjusting a memory setting in DRAM Configuration, that is an OC or not stock/Auto value. That is of course before it is actually applied by a Save and Exit. Are memory settings applied immediately in the UEFI now? Very strange if true, I doubt it.
Moving around the UEFI screens very quickly. If I take my time, I have no problems.
Could be worthless observations, but that's what my experience is.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 26 May 2016 at 11:27am
Your memory timings are tighter than mine. IIRC mine are 18,18,18. That voltage is default 2133! You have a special ram kit. I flashed 3.00 from USB on BIOS #2 which I am using. I've never had a pausing issue in the UEFI screen,... Matter of fact it's better than my previous ASUS ROG HERO Vi for mouse control. The only feature missing from our BIOS is secure erase like in the Rog series and with the emphasis on SSDs on our boards it's sorely missed. This is my first ASRock in a long time since the 4core dual and it's a great board.
Do you have your CPU over locked? What speed are you achieving?
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: parsec
Date Posted: 26 May 2016 at 12:12pm
The PC may have booted with VCCIO at 0.976, but Firefox crashed on me when posting in this forum. So no way that can I run at 3466 with that VCCIO. Set it to 1.128V (actual) for now, survived a stress test. So no, not a special RAM kit, yours should have better chips.
The G.SKILL GTZ TridentZ models are more pricey than the GTZB model I have. Not by much anymore, due to Newegg's crazy pricing, but that model (GTZ) should be better.
My OC is 4.5GHz now, on an i5-6600K. Just using Auto (adaptive) voltage I'm at about 1.25V during stress tests. Cache is on stock (3.9GHz) for now, but I've had it higher in the past.
Really, never had a freezing/pausing issue in the UEFI? We see a fair amount of complaints about that with Skylake boards. Several different models. I've had it myself. Interesting, I wonder what the cause might be? What kind of mouse and keyboard do you use? Wired or wireless?
I've seen the UEFIs of some recent Asus boards. I'm surprised you say there aren't many missing features in our UEFIs. But I'm referring to the endless UEFI options those boards have, which IMO at least some are there only to make it appear to be sophisticated.
I helped a guy disable CSM on a ROG board, which I thought to myself at first, "come on, it's one option setting", like it is in every ASRock UEFI I've ever used (many!) I could not believe the nonsense it took to get CSM disabled in that UEFI. Cannot even remember all that had to be done.
Or options for versions of Windows? If there are underlying UEFI settings needed for an OS, fine, but why make us set them? What good does it do to be able to not enable something needed for an OS version?
But I also saw some options that would be nice to have. ASRock tends to put those only on boards that really ought to have them, like the OC Formula boards. Different philosophies of design, overall.
The secure erase (SE) feature is nice, I will give you that. I use mainly Intel and Samsung SSDs, and their SSD tools work pretty easily to perform an SE. Other SSDs have tools too, but to be able to do it in the UEFI on any SSD is a great feature. Did it let you SE an OS drive? I could see the usefulness of that, for installing an OS fresh on a SSD that just was SE.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 26 May 2016 at 4:13pm
I use a Logitech 502 wired mouse and an AZIO MGK1 wired keyboard connected to the USB 2 ports on the back I/O.
I'm at 4.6, 4.2 cache, level 2 LLC, 1.370 fixed on a 6700k. The chip will do >4.7 at level 1 LLC but I don't have enough cooling to push it comfortably. I should try default and see if I can hit 3600 memory? Even 4.5 at 3600 is nice.
I'm with you 100% against needless UEFI options that are unnecessary. Disabling the on board CPU gpu and audio were nice features too on my ROG board but they had no effect on overclocking,.... So not that important. I mentioned the SE because I can erase a boot SSD, only requirement was ACHI active and two reboots to disconnect the drive and erase,.... One at a time. And I could use UBU tool to update the BIOS with the latest IRST, Intel lan, CPU microcode, and patches. Can't do that with our BIOS :(
Were you successful in erasing your m.2 with the Samsung SE tool? No luck here. It would need to disconnect /unfreeze to erase.
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
Posted By: clubfoot
Date Posted: 29 May 2016 at 11:42am
I've been using Firefox for ages but recently it's started pausing, having difficulty with certain forum sites and distorting the desktop when playing embedded YouTube videos if I change the default video card memory settings! This behaviour doesn't happen in Chrome. Firefox is not what it used to be.
For final stability testing I always use Realbench 2.34 stress test and three loops of benchmark.
------------- https://valid.x86.fr/1tkblf" rel="nofollow">
|
|