Overclock.net - An Overclocking Community - Reply to Topic
Thread: ROG Crosshair VII overclocking thread Reply to Thread
Title:
Message:

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in


  Additional Options
Miscellaneous Options

  Topic Review (Newest First)
07-13-2020 01:31 AM
oreonutz
Quote: Originally Posted by DDSZ View Post
Do you think it is safe to run DR JJRs >1.45V with fans on them?
Yes, more than safe. They don't even need the fans unless the Heat is messing with Stability. DDR4 (All DDR4, not just J-Die/JJR, B-Die, E-Die, etc...) is safe running at heat in excess of 70c. I would personally start to get concerned past 70c, but that is what they are rated to run at Spec. They are also more than safe to run voltages in excess of 1.5v, and with cooling you can run them up to around 1.75v without having any issues. DDR4 is hard to destroy with anything below 70c/1.7v. You are way more likely to run into issues with Stability (especially on Temp sensitive Silicon like B-Die) due to temperature or running higher voltages, before you ever come close to actually damaging the DIMMs.

Sources:
Spoiler!
07-13-2020 12:24 AM
oreonutz
Quote: Originally Posted by Keith Myers View Post
I have a point of contention. You only seem to have considered rendering and such for almost all core loading. But did you consider distributed computing at all where the cpu is primarily engaged in cpu only mathematical processing with some support of gpu mathematical processing also?

In my testing, I can never get much more than 4025-4150Mhz clocks on average across 90-95% of all cores. And the voltages on Auto or PBO are very high with accompanied temperatures.

I find a static all-core overclock to give much better performance at much lower voltages and temperatures. The amount of cpu time the cores spend on a tasks is also on the order of at least 30 minutes to as much as 8 hours. So much longer than a simple run through a benchmark or rendering test.

I have reviewed your paper but don't see anything of value in it for my use case.
I am in the same boat as Keith. I actually find the work that @tcclaviger does to be awesome, and the White Paper he wrote to be extremely informative. It's just that not everyone's sample is capable of recreating the same awesome results he is able to achieve with his sample.

I actually was quite happy at first, mimicking tcclaviger's settings and then dialing them in for my 3950x, because I was getting all core boosts to 4300Mhz (which still isn't as good as my Per CCX but damn near close) while providing amazing single core boosts up to 4.75Ghz. The problem was it turned out not to be stable. I worked for over a week on tuning the EDC, TDC, and PPT to achieve the same clocks, but remain stable, and no matter how I configured it, it just wouldn't stay stable. And the thing was, it was fine under heavy all core loads, or even heavy lightly threaded loads, just at some point, usually either while idling or watching a youtube video it would just crash, giving the watchdog or other hardware error. It seems the CPU would trip over itself while boosting during either idle or near idle workloads. And when raising the EDC high enough to no longer be exploiting the bug, the awesome clock results for all core would disappear.

Unfortunately AMD has yet to provide a method of control over PBO OC's that will allow you to tune exact voltages at exact clocks on exact cores, if I did have that level of control I could set up an excellent PBO overclock, but unfortunately the way PBO works, the Cores determine their own boost based on the parameters you set with PPT, EDC, and TDC, and none of my 3 samples will give me an all core boost I am happy with AND remain stable.

I do know that it is possible with some samples though, as I have a clients CPU, a recent 3900x Retail sample, that was able to hold an all core of 4350Mhz under a full AVX2 Load, Boost to 4.65Ghz for the Lightly Threaded Loads, and still remain stable through extensive stress testing. So I know that it is possible, but it is heavily dependent on the specific sample, so some people will be able to hit a good balance of High All Core and High Single Core clocks, while others even using the same exact settings on the same model Board, with all the same variables will not. It leads to some people swearing PBO is the answer, and others hating it, when really PBO can be both, it just depends on the specific silicon you received in your sample unfortunately.

I really hope that we get much more specific controls in PBO's next iteration. It would be nice to have controls over voltages used on individual core's clocks, and work load types. And then just like with normal all core overclocking, if the CPU can't hit it, then it crashes and you try something different, instead of letting the CPU control those variables itself. I understand why they do it that way, but for advanced users it would be nice to have further controls.

For instance, I would like to tune my sample to have cores 1 and 2 to hit 4.75Ghz at 1.46v during a single or dual core load, have cores 1, 2, 3, & 4 hit 4.5Ghz during a 4 Core load at 1.42v, and then to have an all core load at 4.35Ghz at 1.3v. These are all clocks and work loads that I have observed my CPU of being able to hit regularly under PBO, and while it will repeatedly give those voltages itself when using PBO, for the Single, Dual, and 4 Core Work loads, it unfortunately will drop the all core load to 4.1Ghz or lower at 1.3v, even though my sample can easily hit 4.35Ghz at 1.3v across all cores all day.

The situation is different on every CPU as to how PBO behaves, but being able to finely tune the CPU during a particular workload type would be the ultimate OC achievement for Zen, and allow us to really push these CPU's to their max.

Anyways, thats my book for the day...
07-12-2020 11:50 PM
DDSZ Well, I'm running 3103 and just for the test I did set 18-22-22-42-64 640 @ 1.45V for my dual rank JJRs, the highest I could get is DRAM = 4066 MHz and IF = 1900, didn't try tightening it, but it even passed 10 tm5 cycles of 1usmus config without errors.
Still can't manage to get 3800/1900 CL 16 stable
Do you think it is safe to run DR JJRs >1.45V with fans on them?
07-12-2020 06:39 PM
nick name
Quote: Originally Posted by kratosatlante View Post
you have the message bios recovery mode etc etc?
i haved 3 times, latest reflashed and get one time more, restart and work well for now
No, I didn't see anything like that. Nothing indicated a bad BIOS.
07-12-2020 06:23 PM
kratosatlante
Quote: Originally Posted by nick name View Post
Jesus Christ did I almost have a heart attack. Has anyone ever corrupted a BIOS in a manner that only the specific FCLK speed running at the time of corruption was what suffered after the corruption? Welp it happened to me and since the manner of the corruption seems now to only have impacted FCLK 1900 it was terribly hard to sort out.

It's been such a scramble for the last few hours that I can't even accurately recount what the actual f*ck I did to sort it.

I didn't assume BIOS corruption until after pulling and switching RAM kits that I ran at slower speeds to rule out RAM at which point I decided to try FCLK 1900 again. And that's when the ***** started again and made me think that can't be hardware so I re-flashed the BIOS.

And now -- everything is fine.

Now right after I get that sorted a file naming Windows 10 bug rears its head and destroys a 225GB file I downloaded on another machine.

Tonight is not my night.
you have the message bios recovery mode etc etc?
i haved 3 times, latest reflashed and get one time more, restart and work well for now
07-12-2020 06:14 PM
nick name Jesus Christ did I almost have a heart attack. Has anyone ever corrupted a BIOS in a manner that only the specific FCLK speed running at the time of corruption was what suffered after the corruption? Welp it happened to me and since the manner of the corruption seems now to only have impacted FCLK 1900 it was terribly hard to sort out.

It's been such a scramble for the last few hours that I can't even accurately recount what the actual f*ck I did to sort it.

I didn't assume BIOS corruption until after pulling and switching RAM kits that I ran at slower speeds to rule out RAM at which point I decided to try FCLK 1900 again. And that's when the ***** started again and made me think that can't be hardware so I re-flashed the BIOS.

And now -- everything is fine.

Now right after I get that sorted a file naming Windows 10 bug rears its head and destroys a 225GB file I downloaded on another machine.

Tonight is not my night.

Edit:
I think I downclocked the FCLK and RAM and still had the same problem until I switched RAM kits without changing BIOS settings. Perhaps it was hardware mounts, but after installing the different kit I had problems again at FCLK 1900.
07-12-2020 05:48 PM
kratosatlante
Quote: Originally Posted by crakej View Post
I've been playing around with this bios and various different settings, but I can't get anything like the performance I got from 2801, where I could run 3800:1900 and my 4400MTs XMP profile just worked.

I was also able to run at 4466MTs and 4600MTs, but can't do anything over 3733MTs now - not that it's awful performance, just could be (and used to be) better. Been up since 7am testing profiles that I entered by hand to no avail.

It could be that now I have 4 sticks it just won't do it, but it really should - at least the XMP profile. I may re-install 2801 and test with 4 sticks.


Same here new bios dont work xmp 4400 and 4266, cant boot 2 stiks 4000 only 3933(4stisks limit same), bios 2801 load easy xmp 4400, now have 4stiks viper 4400cl19, thanks for remember which bios work, if this week arrive ram cooler have time for some test
07-12-2020 05:14 PM
Keith Myers I have a point of contention. You only seem to have considered rendering and such for almost all core loading. But did you consider distributed computing at all where the cpu is primarily engaged in cpu only mathematical processing with some support of gpu mathematical processing also?

In my testing, I can never get much more than 4025-4150Mhz clocks on average across 90-95% of all cores. And the voltages on Auto or PBO are very high with accompanied temperatures.

I find a static all-core overclock to give much better performance at much lower voltages and temperatures. The amount of cpu time the cores spend on a tasks is also on the order of at least 30 minutes to as much as 8 hours. So much longer than a simple run through a benchmark or rendering test.

I have reviewed your paper but don't see anything of value in it for my use case.
07-12-2020 09:25 AM
tcclaviger Double post, can't seem to locate the delete post button lol.
07-12-2020 08:07 AM
tcclaviger I will include it, the relative performance would be effected for PB2/PBO results compared to static OC, favoring static.

EDIT: Revisions added, thank you for the input, attachment updated.
This thread has more than 10 replies. Click here to review the whole thread.

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off