Overclock.net banner

7681 - 7700 of 8118 Posts

·
Registered
B550 AORUS MASTER, 3700X, 32GB g.Skill DDR4-3200 (@3800MT/s; 14-14-14-28), XFX RX 5500 XT
Joined
·
67 Posts

·
Registered
msi mpg B550 gaming edge wifi, 5600X, 3200c14, 2070S FTW 3 ultra, 8200SX pro 2tb nvme
Joined
·
240 Posts
So close to that 50 ns threshold....

6 months ago i could have cared less, so long as it worked fine, ran youtube/netlflix,
COD, CP2077, you know games and the normies use cases....

but now..... well, i got away from an actual addiction problem to become addicted
to numbers and benchmark scores...smh lol

2488695
 

·
Registered
msi mpg B550 gaming edge wifi, 5600X, 3200c14, 2070S FTW 3 ultra, 8200SX pro 2tb nvme
Joined
·
240 Posts

@lmfodor

You running offsets via Core Optimizer? I'd look at that first...
from what i could gather here the error (if im not mistaken)
is something to do with a drive.....

ctrl f the page and search VST

(im probably wrong) but it could be windows not allowing
ycruncher folder access at this part of the test...
 
  • Rep+
Reactions: TimeDrapery

·
Registered
B550 AORUS MASTER, 3700X, 32GB g.Skill DDR4-3200 (@3800MT/s; 14-14-14-28), XFX RX 5500 XT
Joined
·
67 Posts
from what i could gather here the error (if im not mistaken)
is something to do with a drive.....

ctrl f the page and search VST

(im probably wrong) but it could be windows not allowing
ycruncher folder access at this part of the test...

Oh, good look dude! I'm gonna bookmark that page... Thanks!

Maybe page file sizing?

Edit: Nope, not page file sizing 😂😂😂😂😂
 

·
Registered
Overclocking - latest technology
Joined
·
277 Posts
from what i could gather here the error (if im not mistaken)
is something to do with a drive.....

ctrl f the page and search VST

(im probably wrong) but it could be windows not allowing
ycruncher folder access at this part of the test...
Hi! for some reason when I ran y-cruncher all test, frequently crash on this test.. Now I went back to my stable timings 1.5T and the error disappeared

2488708
 

·
Registered
Joined
·
3 Posts
In my limited experience, high FCLK seems to be very dependent on the silicon lottery of your motherboard (in addition to getting lucky with your RAM and CPU).

I have a set of single rank Rev. E (3600c16) that I can run stable, WHEA-free at 4267c16, 2133 FCLK on an Asrock B450 Fatal1ty Gaming-ITX/AC motherboard.

Using the exact same components and just switching out the motherboard to an ASUS Crosshair VIII Impact (X570), the highest I could get stable was 3867c16, 1933 FCLK. No matter how much voltage tweaking and BIOS manipulation I tried (and the ASUS x570 board has wayyy more options than the Asrock B450), I was unable to get the system to post at 3933:1967 or higher. Which is bananas because the ASUS is a much more expensive, and supposedly higher quality, motherboard.

I suspect I was unlucky with my ASUS board, since there are others running the Crosshair VIII Impact at 2000+ FCLK stable, WHEA-free — and I know my CPU and memory can perform at that level. You may also have an unlucky board, since Veii (and I think some others?) are running 2000+ FCLK on the Asrock B550 Phantom Gaming ITX.
Yeah, I transplanted and tried everything in an ASUS ROG STRIX B550-I Gaming. It was about the same and just depended on what I plugged into the various main drivers of FCLK. I get it. :geek: Maybe I'll try my other set of the same RAM in my B550 Phantom Gaming ITX.
 

·
Registered
Joined
·
106 Posts
In my limited experience, high FCLK seems to be very dependent on the silicon lottery of your motherboard (in addition to getting lucky with your RAM and CPU)
Could it be coming down to different approaches to treat instability for different boards and agesa? After trying to get at least 1933 IF on my cpu (it works, but instant whea's at prime avx large plus gpu compute load), I don't believe there are samples around that can do rock stable 2000mhz. xD
 

·
Registered
Joined
·
14 Posts
Spent a bit of time finding a lower tRFC - my PC wouldnt POST at 230 (it posted at 235). I set it to 245 for some "headroom" and stresstested with memtest (2500mb*12, ~400%) and it seemed stable enough. Aida64 down to 55.2ns. I know my primaries arent that impressive but I tried 15-16-16 with GDM off/T2 and still would get errors after a while in memtest - for this i had secondary and tertierary on auto (so, very loose). I haven't tried higher than 1.5Vdimm tho. I also have not tried 16-15-15 or 16-14-14 or some weird thing like that (not sure its practical??).
2488736
 

·
Old crazy guy
Joined
·
2,772 Posts
I'm disappointed with my F4-4000C16-32GTZR kit.
Not only tRCDRD struggles below 16 but it crashes above 1.95V and it's impossible to get 3800CL12 stable.

Just ordered a F4-3200C14D-32GTZSK kit.
Hopefully it's a good bin like Infrared got, 4000CL12 at 2.0V:
2488741
 

·
Registered
Joined
·
225 Posts
Hi! for some reason when I ran y-cruncher all test, frequently crash on this test.. Now I went back to my stable timings 1.5T and the error disappeared
VST error can be from VDDG apparently.

Or on Zen+ I think it can be more generally fabric stability, but I don't think I actually got that one with much frequency on the 2700X.
 

·
Registered
Overclocking - latest technology
Joined
·
277 Posts
VST error can be from VDDG apparently.

Or on Zen+ I think it can be more generally fabric stability, but I don't think I actually got that one with much frequency on the 2700X.
Yes, every time I change my timings , in addition to running TM5, I run Y-C all test ... and if for some reason there is an error, it appears in that test. So now instead of running all the tests I try the VST first and if it works then I run all the tests. But I didn't know it was related to VDDG. I'm going to investigate a little more. Thanks!


Sent from my iPhone using Tapatalk Pro
 
  • Rep+
Reactions: TimeDrapery

·
Registered
msi mpg B550 gaming edge wifi, 5600X, 3200c14, 2070S FTW 3 ultra, 8200SX pro 2tb nvme
Joined
·
240 Posts
I'm disappointed with my F4-4000C16-32GTZR kit.
Not only tRCDRD struggles below 16 but it crashes above 1.95V and it's impossible to get 3800CL12 stable.

Just ordered a F4-3200C14D-32GTZSK kit.
Hopefully it's a good bin like Infrared got, 4000CL12 at 2.0V:
question to you...

how do you know when you have "such a good bin"
does tRAS come into play, would you think (image below)
would be a "good enough" bin to achieve this?

yes, depending on PCB which is A0 says TB but,
i have not taken the heat spreader off to check, and teamgroup
cant say what it is other than R2 spec which they claim they
dont use a standard normal PCB or some B.S. like that in an email i recieved when i asked about it.
2488840
 

·
Old crazy guy
Joined
·
2,772 Posts
question to you...

how do you know when you have "such a good bin"
does tRAS come into play, would you think (image below)
would be a "good enough" bin to achieve this?

yes, depending on PCB which is A0 says TB but,
i have not taken the heat spreader off to check, and teamgroup
cant say what it is other than R2 spec which they claim they
dont use a standard normal PCB or some B.S. like that in an email i recieved when i asked about it.
The XMP profile doesn't really tell much about the binning.
You need to test it in order to see what it can really do.

I think the old 3200C14 kits were good binning and A0/B0 PCB (or at least based on it).
From the spreadsheets and the HWBot submissions those are the best kits for low timings.
In theory A0/B0 should start struggling above 3600/3800 MHz but I see a lot having fantastic results at 4000 and above.
Going to do see very soon, this week I should get the kit delivered.

Yes most manufacturers don't use standard PCBs.
They slightly modify them or mix some parts.
Especially when they integrated RGB LEDs.
That's what I understood lurking here.
 

·
Overclock the World
Joined
·
2,488 Posts
Why rcdrd refuses to work at 16? :(
View attachment 2488751
SD, DDs are wrong and voltage high in combination with procODT
1-5-5-1-7-7 and sub procODT 40 , else disable RTT_NOM (voltage too low for /7)
But for sure lower procODT

You should increase RRD_ to 5-7-20 or change RTTs or lower VDIMM & disable RTT_NOM
Pick one of the 3 options
In theory A0/B0 should start struggling above 3600/3800 MHz but I see a lot having fantastic results at 4000 and above.
Going to do see very soon, this week I should get the kit delivered
A0's usually "should" fail beyond 4133 (4000 in reality) which is why many board partners beyond 4000 swap PCBs or make a custom design
* A1 should be similar 4133+ is A2 zone
Except for when you adjust RTTs
Please remember to adapt RTTs beyond 1.51v/1.52v, for the health of thekits

I suspect I was unlucky with my ASUS board, since there are others running the Crosshair VIII Impact at 2000+ FCLK stable, WHEA-free — and I know my CPU and memory can perform at that level. You may also have an unlucky board, since Veii (and I think some others?) are running 2000+ FCLK on the Asrock B550 Phantom Gaming ITX.
It's the I/O's fault & X570 PCIe + XHCI that causes WHEA issues
Sillicon lottery starts to be a thing beyond 2000 FCLK
All Vermeer have to run 2000+ on single rank

Boards PCBs even between Daisy-Chain & T-Topology start to matter beyond 4266MT/s
For up to 4 dimms and dual rank
As for my mention, 2133 is runnable and visually stable
But internally it throttles back 2ns
2100 is the optimal for now
2167 was not bootable but should be possible on Vermeer, as seen with Rev.E kits
2000 @ 14-14-14 seems to be the optimal between easy to run and reasable with PCB
4200 C15-15-15 is nearly identical but far harder to daily :)
Maybe 3800C13-13 could be an option, but th FCLK bump is worth it
2100 is complicated, you cut too deep into to boosting powerbudget
Likely fine in a 5900X but not so easy on a 5600X
120(122A) EDC cap is just too low for 2100 SOC requirements
 

·
Registered
Joined
·
273 Posts
If 1900 fCLK works straight out of the box, why would one need extra tweaking for > 1900?
Makes no sense. It does look like silicon lottery and agesa limitation.
 
7681 - 7700 of 8118 Posts
Top