Overclock.net banner

Status
Not open for further replies.
81 - 100 of 679 Posts

·
Mining the DB
Joined
·
2,336 Posts
@The Stilt

Someone asked I am too am interested, can you see any added advantage running a full custom waterloop with these chips given they are already so close to fmax?
Would the drop in temps allow for more vcore and clocks(50mhz+)?

Chips in question is a 3600X and 3700X
 

·
Registered
Joined
·
2,316 Posts
A question. I'm planning to order G-Skill TridentZ 3600 CL15-15-15-35 memory to go along with 3700X and aorus x570 elite, but I've read somewhere that ryzen doesn't go well with anything odd. Is this only for older ryzen or is this load-o-carp as usual? Just want to confirm before I have to send 200 euros worth of ram back and forth.
I have a 3600CL15 kit, but can't speak to how high it can go as I only have a 2700X. I can say it will get down to 3200 12-13-12-12 timings and very tight subs. That setup doesn't perform as well as 3600 14-15-14-14 and tight subs though.

Bearded Hardware just did a live stream with a 4000CL17 TridentZ Royal kit that he was booting at 3800 12-11-11-11 and tRFC 180. He wasn't testing stability, but the setup didn't crash and ran Aida and Cinebench. He was using a 3700X at 4.3GHz on an MSI X570 board.
 

·
Meddling user
Joined
·
7,423 Posts
@rdr09

Tanked up on caffeine and getting benching! :D


Maybe at some point :)

This time around thou, I might be able to do it without testing ;)
NP, will stay tuned ;) :thumb: .

@The Stilt Have You tried playing with BCLK ?? I'w not seen anything around on that topic.
No.
I never touch the BCLK on these platforms for following reasons: There is no need to and altering the BCLK can cause serious issues since BCLK isn't an independent domain.
I've lost data on a M.2 SSD by doing so, so until there is a clear benefit to be had, I never do it again.
I'd agree with The Stilt on this. I recently went NVMe and setup that had no issues BCLK OC with SATA SSD, would blank screen after UEFI splash screen even with 101MHz. Also Zeed see Johan's last sentence here.

@The Stilt

Someone asked I am too am interested, can you see any added advantage running a full custom waterloop with these chips given they are already so close to fmax?
Would the drop in temps allow for more vcore and clocks(50mhz+)?

Chips in question is a 3600X and 3700X
Checkout Der8auer's video on YT on 3000 series delid, etc, that well give you insight. Unless your going WC for say cooling additional, like GPU, for say improved noise aspect of cooling, etc all good, but MHz gains are not at all worth the outlay.
 

·
Premium Member
Joined
·
2,754 Posts
Discussion Starter · #84 ·
@The Stilt

Someone asked I am too am interested, can you see any added advantage running a full custom waterloop with these chips given they are already so close to fmax?
Would the drop in temps allow for more vcore and clocks(50mhz+)?

Chips in question is a 3600X and 3700X
Hard to say, but based on the known facts I would say no.
The main issue is getting the heat transfered out of the tiny 74mm² CCD. The actual cooling capacity of your cooling solution is not the issue here.
Maybe you can gain something (like the 50MHz you mentioned), but I definitely wouldn't increase the voltage based on the better cooling. The voltage is limited by the silicon itself, not by the thermal dissipation it results.
 

·
Meddling user
Joined
·
7,423 Posts
I confirm on R5 3600 retail I have full PBO settings access in UEFI based on what current UEFI 2406 AGESA ComboPi-AM4 1.0.0.2 allows :) . For anyone interested I have posted experience so far on ROG forum and will continue updates there on how things go.
 

·
Premium Member
Joined
·
20,361 Posts
I confirm on R5 3600 retail I have full PBO settings access in UEFI based on what current UEFI 2406 AGESA ComboPi-AM4 1.0.0.2 allows :) . For anyone interested I have posted experience so far on ROG forum and will continue updates there on how things go.
Nice. Got something to read.
 

·
Untitled.jpg
Joined
·
1,044 Posts
I confirm on R5 3600 retail I have full PBO settings access in UEFI based on what current UEFI 2406 AGESA ComboPi-AM4 1.0.0.2 allows :) . For anyone interested I have posted experience so far on ROG forum and will continue updates there on how things go.
Wish my experience was going the same. 3700X on the X370-F with 5009 bios. Wont post with any changes at all to memory, has to be left to the 2133 c15 default. Elmor suggested something to do with boot voltage, but there's no setting for that on my board so I guess im SOL until a bios fix comes. The CPU is also hitting 1.5v at idle and I just don't know if that's normal or not.
 

·
Registered
Joined
·
659 Posts
Wish my experience was going the same. 3700X on the X370-F with 5009 bios. Wont post with any changes at all to memory, has to be left to the 2133 c15 default. Elmor suggested something to do with boot voltage, but there's no setting for that on my board so I guess im SOL until a bios fix comes. The CPU is also hitting 1.5v at idle and I just don't know if that's normal or not.
Apparently some people are able to run a 3900X on a B350.
https://www.techpowerup.com/review/amd-ryzen-9-3900x-tested-on-cheap-b350-motherboard/2.html
 

·
Newb to Overclock.net
Joined
·
4,729 Posts
Man, what a very clouded launch for an interesting design and affordable product.
 

·
Registered
Joined
·
405 Posts
My motherboard is an ASUS X370 CH6.

I have Vcore set to 1,25v in BIOS, and it also shows in BIOS during boot.

I don't understand why the VID values are 1,1v, while the Vcore value is still 1,25v?

 

·
Registered
Joined
·
77 Posts
Judging by the voltages alone, that isn't working properly either.
There is significant differences between the different core characteristics and not all of them should boost to the Fmax.
Likewise, not all of them should run at the same voltage either. No different to older CPUs, such as the 2700X. Only the best cores could boost to the Fmax (4.35GHz).

To me that looks like some kind of a bios hack, done by Gigabyte. Maybe disabled or misconfigured FIT controller.
Hi TheStilt, first and foremost, I respect the knowledge level you have and won't even want to enter into a discussion with you. As I said, have a lot of respect. About GBT hack, there might be a error on that, it's an AGESA 1002 prior to NPRP code (F1 XTREME) that is showing the 4.65 on all cores behaviour. Gurgstep also just used that same 1002 code (which it seems not to contain the NPRP code) and he's having similar results on R5 3600. You can check his post.

I guess I take for granted your conclusion on the Delta difference between best core and worst cores have a delta of up to 300 MHz (in your case, best core did not even reach 4.6 GHz). The behaviour of all cores reaching got something to do with 1002 non NPRP.

F5e is showing same behaviour as you posted, guess as you mentioned, 7nm process still immature and deltas are higher between best and worst cores in ST. As for voltage VID being fed on XTREME, F5e shows same behaviour.

P.D. Still figuring out the PCIe WHEA errors I'm getting on NVIDIA Cards on two boards and two samples I have.

With regards,
XanxoGaming
 

·
Registered
Joined
·
77 Posts
I confirm on R5 3600 retail I have full PBO settings access in UEFI based on what current UEFI 2406 AGESA ComboPi-AM4 1.0.0.2 allows :) . For anyone interested I have posted experience so far on ROG forum and will continue updates there on how things go.
Same behaviour I've experienced with AGESA 1002 (non NPRP version) on XTREME/MASTER/X470 AG7W on 3900X. Even F1 bios says 1003, according to GBT F1 is 1002 (probably typo error on website). I've tried Master (F4) and AG7W (F41a) boosting to 4.65 GHz on all core. Guess PBO is working or AGESA 1002 just behaves like that. Really hard to say, TheStilt mentioned only way to really measure power consumption need specialized tools and knowledge which escapes my reach (I admit).
 

·
Registered
Joined
·
33 Posts
The traditional Ryzen method for OCing was PBO with a slight undervolt. It seems that since you're hitting the FIT voltage issue and this is what is limiting clocks, have you tried playing with smaller negative voltage offsets while leaving power/current etc maxed out to see if it is able to boost any higher?

I did a quick test on the 3900X, using the brand new AGESA 1.0.0.3 w/ A & B patches and the newest chipset-drivers:

Core Fmax listing prior these changes:

Code:
Core 1 =  4575 - 4300MHz - 1.48750V (2*)
Core 2 =  4475 - 4350MHz - 1.48750V
Core 3 =  4525 - 4225MHz - 1.47500V
Core 4 =  4575 - 4275MHz - 1.47500V (1*)
Core 5 =  4475 - 4300MHz - 1.48750V
Core 6 =  4475 - 4300MHz - 1.48750V
Core 7 =  4375 - 4300MHz - 1.46250V (1*)
Core 8 =  4350 - 4300MHz - 1.46250V
Core 9 =  4350 - 4300MHz - 1.46250V
Core 10 = 4325 - 4275MHz - 1.48125V
Core 11 = 4325 - 4300MHz - 1.47500V
Core 12 = 4400 - 4300MHz - 1.47500V (2*)
Core Fmax listing after the changes: See above...

1* = The best core of the CCD, 2* the second best core of the CCD.

I'd say unless the factory V/F calibration is out of whack, which I really don't think it is, the only way AMD is going to increase the frequencies of the CPUs is most likely
allowing them to run at higher voltages (i.e. sacrificing reliability). The voltage figure seen for each core is the voltage the CPU asks from the VRM controller through SVI2 interface, when the specific core is stressed.
 

·
Meddling user
Joined
·
7,423 Posts
Nice. Got something to read.
NP, I got some bits done, more tomorrow over on ROG.

Wish my experience was going the same. 3700X on the X370-F with 5009 bios. Wont post with any changes at all to memory, has to be left to the 2133 c15 default. Elmor suggested something to do with boot voltage, but there's no setting for that on my board so I guess im SOL until a bios fix comes. The CPU is also hitting 1.5v at idle and I just don't know if that's normal or not.
Yours is 3700X, max boost is 4.4GHz, perhaps the reason for the experience your having. As the 3600 non X is 4.2GHz that maybe why I'm only seeing 1.3V. TBH dunno though, my 2700X even at stock could go circa ~1.45V on upto ~4.35GHz single/cycling core low load boost. I have no experience of that board, sorry. All I can say is cruising to 3533MHz using the :clock: The Stilt :clock: 3466MHz timings has been easy on C7H.

HWINFO idle data/Powerplan/OS/settings txt used in spoiler.


Hi TheStilt, first and foremost, I respect the knowledge level you have and won't even want to enter into a discussion with you. As I said, have a lot of respect. About GBT hack, there might be a error on that, it's an AGESA 1002 prior to NPRP code (F1 XTREME) that is showing the 4.65 on all cores behaviour. Gurgstep also just used that same 1002 code (which it seems not to contain the NPRP code) and he's having similar results on R5 3600. You can check his post.

I guess I take for granted your conclusion on the Delta difference between best core and worst cores have a delta of up to 300 MHz (in your case, best core did not even reach 4.6 GHz). The behaviour of all cores reaching got something to do with 1002 non NPRP.

F5e is showing same behaviour as you posted, guess as you mentioned, 7nm process still immature and deltas are higher between best and worst cores in ST. As for voltage VID being fed on XTREME, F5e shows same behaviour.

P.D. Still figuring out the PCIe WHEA errors I'm getting on NVIDIA Cards on two boards and two samples I have.

With regards,
XanxoGaming
At present I think I'm seeing what I'm seeing as perhaps using a low core count CPU, is on WC, but dunno really, all too fresh at present. Perhaps The Stilt will see something in the spoiler above containing info that I think you are referencing from my shares.
 

·
Meddling user
Joined
·
7,423 Posts
Now this I was not expecting Kahru RAM Test ACB 4.2GHz without OC on core :thinking: .

PE Default 3533S 1.025 0.985 1.35 0.675 PASS 750% room 26C.jpg

Settings used:-

View attachment 2406_R5_Base_3533S_setting.txt

Room ambient ~26C.

Under load VCORE I see on DMM ~1.297V - 1.369V, mostly ~1.3V, SOC is ~1.010V, VDIMM ~1.353V.
 

·
Registered
Joined
·
77 Posts
NP, I got some bits done, more tomorrow over on ROG.



Yours is 3700X, max boost is 4.4GHz, perhaps the reason for the experience your having. As the 3600 non X is 4.2GHz that maybe why I'm only seeing 1.3V. TBH dunno though, my 2700X even at stock could go circa ~1.45V on upto ~4.35GHz single/cycling core low load boost. I have no experience of that board, sorry. All I can say is cruising to 3533MHz using the :clock: The Stilt :clock: 3466MHz timings has been easy on C7H.

HWINFO idle data/Powerplan/OS/settings txt used in spoiler.




At present I think I'm seeing what I'm seeing as perhaps using a low core count CPU, is on WC, but dunno really, all too fresh at present. Perhaps The Stilt will see something in the spoiler above containing info that I think you are referencing from my shares.
I'm using AIO 280mm, seeing that result on AGESA 1002. Anyways, sending an e-mail soon to AMD just to clarify boost delta being that high on this gen between cores. Since we are small media outlet in Perú, it is hard to reach AMD directly (e.g. Robert Hallock or the person in charge of AMD technical USA). But yes, same boosts (higher VID, new PWM controller on XTREME) up to 4.65 GHz on almost all cores AGESA 1002 like you achieved on C7H.

I am being sent a C8H, testing some NVIDIA GPU issue with PCIe.
 

·
Premium Member
Joined
·
2,754 Posts
Discussion Starter · #97 ·
The traditional Ryzen method for OCing was PBO with a slight undervolt. It seems that since you're hitting the FIT voltage issue and this is what is limiting clocks, have you tried playing with smaller negative voltage offsets while leaving power/current etc maxed out to see if it is able to boost any higher?
I have, but cannot currently elaborate on that...
 

·
Meddling user
Joined
·
7,423 Posts
I'm using AIO 280mm, seeing that result on AGESA 1002. Anyways, sending an e-mail soon to AMD just to clarify boost delta being that high on this gen between cores. Since we are small media outlet in Perú, it is hard to reach AMD directly (e.g. Robert Hallock or the person in charge of AMD technical USA). But yes, same boosts (higher VID, new PWM controller on XTREME) up to 4.65 GHz on almost all cores AGESA 1002 like you achieved on C7H.

I am being sent a C8H, testing some NVIDIA GPU issue with PCIe.
Sorry I can not be of any use use other than share my limited experience on my current setup.

Voltage I'm seeing in SW and on DMM, idle or load, I'm pretty impressed with TBH. What is concerning me is why am I seeing ACB 4.2GHz in RAM Test, even though not on core OC.

That above run of RAM Test lasted 1100%, I manually stopped and trying 3600MHz with same timings.

Code:
[2018-11-17T19:54:18.7355837+00:00] License activated successfully.
[2018-11-17T19:55:06.6896294+00:00] RAM Test is up-to-date.
[2019-07-09T22:49:37.6416727+01:00] RAM Test is up-to-date.
[2019-07-09T22:49:52.6271620+01:00] Started testing 12642 MB with 12 thread(s).
[2019-07-09T23:11:41.0913115+01:00] Stopped testing after 0:00:21:48 with 1102 % coverage and 0 error(s).
 

·
Registered
Joined
·
1,158 Posts
Thanks for sharing,great information and this was classic.

At this point you should ask yourself if ASUS paid me off?
Everyone can be bought, its just the matter of the offered sum or bargain. Everyone claiming otherwise either lives in self-deception or frankly, is a moron.
I myself could definitely be bought. And rather cheaply too, I think. The thing is, just that at least until writing this, nobody has even tried to do so.
 

·
DRAM Obsessor
Joined
·
1,034 Posts
The traditional Ryzen method for OCing was PBO with a slight undervolt. It seems that since you're hitting the FIT voltage issue and this is what is limiting clocks, have you tried playing with smaller negative voltage offsets while leaving power/current etc maxed out to see if it is able to boost any higher?
I have, but cannot currently elaborate on that...
I feel like @The Stilt is saying that maybe we should try that but he cant actually say it bc Robert Hallock will find him and hit him in the knees with a lead pipe. That or theres even more weird behavior when doing it.
 
81 - 100 of 679 Posts
Status
Not open for further replies.
Top