Overclock.net banner
7701 - 7720 of 7760 Posts

·
Registered
Joined
·
95 Posts
I need some help, just got a new WD SN850 1TB @ M.2_1 socket
Interface speed is only gen4 x 2lanes @ about 50% performance



I had an article pop up in my news feed recently that stated WD is aware of the problem and will be releasing a firmware to fix it soon. It's a problem with these specific drives when running on x570 boards. I don't have a link but is probably easy enough to find if you Google it.
 

·
Registered
Joined
·
34 Posts
I had an article pop up in my news feed recently that stated WD is aware of the problem and will be releasing a firmware to fix it soon. It's a problem with these specific drives when running on x570 boards. I don't have a link but is probably easy enough to find if you Google it.
That firmware has already been released.

Regardless, one should always put their highest performance drive connected to the processor PCIE lanes rather than chipset. The chipset is of course shared bandwidth and won't necessarily be able to always reach highest performance in comparison to the dedicated link.
 

·
Registered
Joined
·
35 Posts
So I played around some more this past weekend with the 1900mhz issue. Still not change @1900 but 1933 posts with errors etc.1966 is way more stable but my Ram starts to become the issue, Latency way up and not quite stable.

I bit the bullet and got a 5900x coming, it may or may not make a difference (silicone lottery) but it was what I was originally trying to get back in Dec./Jan but only could find a 5800x which I will sell for what I can get. Hopefully I get lucky with this one!
Well my 5900x came in late last week but I had company until this afternoon so I couldn't get it installed until today. Now I am glad I did for sure, I loaded my old profile and immediately changed to 3800 and 1900 flck, with all CPU voltages on auto, posted right away and booted into windows fine! Now to run some tests etc but very happy with this so far! Might try to push to my memory max of 3866 for fun! Latency is just 2 ns slower than the fastest I have seen 62.7 but not sure if Bloatware affecting that.
2517428
 

·
Registered
Joined
·
907 Posts
Well my 5900x came in late last week but I had company until this afternoon so I couldn't get it installed until today. Now I am glad I did for sure, I loaded my old profile and immediately changed to 3800 and 1900 flck, with all CPU voltages on auto, posted right away and booted into windows fine! Now to run some tests etc but very happy with this so far! Might try to push to my memory max of 3866 for fun! Latency is just 2 ns slower than the fastest I have seen 62.7 but not sure if Bloatware affecting that.
View attachment 2517428
Nice rig, but "the fastest"? I have 55ns latency in AIDA64 at the same frequency, and afaik 55ns is nothing special. B-Die, granted, but shouldn't be such large difference.
 

·
Registered
Joined
·
4 Posts
View attachment 2512302

If the motherboard is a dark hero
I think it is efficient to use the PBO and Dynamic OC features
Depending on the CPU and settings, a high single frequency can be obtained My setting single 5250 run
First and foremost: thank you, safedisk.

Too much hate for a gentleman who low-key shared the recipe for fast single-core boost clocks using the Dark Hero or any other iteration of the Crosshair VIII, though no one noticed (or at least appeared to notice).

A once-dismissed feature is now working properly. This feature (a) likely works properly in BIOS version 3601 and (b) currently works properly in BIOS version 3701 (Beta). You can infer which feature is now working from Safedisk's HWiNFO screenshot.

Though I am still tuning, here are some screenshots of my own machine's recent/current performance using what I believe to be safedisk's recipe:

2517433


2517432
 

·
Registered
Joined
·
35 Posts
Nice rig, but "the fastest"? I have 55ns latency in AIDA64 at the same frequency, and afaik 55ns is nothing special. B-Die, granted, but shouldn't be such large difference.
Thank you and correction, I meant the fastest I have seen on my Rig with various configurations. I still believe it to be some of the Bloatware, I am forced to use Armoury Crate for RGB and iCUE for my headphones. Both are pretty bad and probably something to do with my occasional lag spikes though they are better since getting 3601.

Definitely a noticeable difference in smoothness with the little bump in fclk and new CPU. :)
 

·
Gamer and overclocker !
Joined
·
2,132 Posts
Well my 5900x came in late last week but I had company until this afternoon so I couldn't get it installed until today. Now I am glad I did for sure, I loaded my old profile and immediately changed to 3800 and 1900 flck, with all CPU voltages on auto, posted right away and booted into windows fine! Now to run some tests etc but very happy with this so far! Might try to push to my memory max of 3866 for fun! Latency is just 2 ns slower than the fastest I have seen 62.7 but not sure if Bloatware affecting that.
View attachment 2517428
take a look at windows event viewer if you don’t get « WHEA 19 » warnings with 3800/1900 and voltages on auto
Also at 3866…
 

·
Registered
Joined
·
907 Posts
I think I understand now why I got a idle reboot with 3702 and never had one on 3501/3601, minimum vcore voltage using PBO is now lower than I have ever seen before. All the way down to 0.248V! Earlier bioses was 0.8-0.9V at the lowest as far as I recall. No wonder the PC could suddenly shut off. I haven't tried the PSU Idle Current setting in the bios yet, it's still on Auto. Possibly it could affect things. I have a Corsair RM850x 2019.

I have a distant memory of early bioses had the same ultra low idle voltage, but it went away after a few updates.

I've only had one reboot thus far though.

The two CCD:s now idles at 2.5W, very low, I guess Asus implemented some sort of Energy saver that was absent before. As said, I use the same settings cross bios. The SOC still draws 16W and the full package 28W in idle, so maybe not that big of a saver, but still a saver.
 
  • Rep+
Reactions: CyrIng and RHBH

·
Registered
Joined
·
105 Posts
Samsung 980 PRO 2TB (M2_1, PCI-E 4.0 x4, 5900X)

2517545



ADATA XPG SX8200 Pro 1TB (M2_2, PCI-E 3.0 x4, X570)

2517546


The chipset M2_2 slot is so bad, SX8200 Pro is suposed to be 3.500/3.000MB/s drive.
 
  • Rep+
Reactions: GRABibus

·
Registered
Joined
·
134 Posts
The chipset M2_2 slot is so bad, SX8200 Pro is suposed to be 3.500/3.000MB/s drive.
ADATA not really reliable. Linus got video about it. ADATA swap it's components and there a lot revisions of one drive.

Video:
 

·
Registered
Joined
·
35 Posts
take a look at windows event viewer if you don’t get « WHEA 19 » warnings with 3800/1900 and voltages on auto
Also at 3866…
Been stable gaming @3800mhz, last WHEA error was 3/9/2021. Tried 3866 and post errors during memory training (not cpu locks like before). Tweaked ram voltage a little, posted and booted ok, quick stability test, whea errors crash etc.

Don't have time to tweak voltages right now so will have to do more tweaking later. Will be happy at 3800mhz I don't think it is worth the cost to get faster ram. I might get something with better timings (with Samsung B Die) down the road but happy for now. I need to apply a little more voltage for true stability so I will play around this weekend.
 

·
Gamer and overclocker !
Joined
·
2,132 Posts
SEAGATE FireCuda 520 2TB (M2_1, PCI-E 4.0 x4, 5900X)

Capture.PNG


Write values Q8T1 and Q32T1 should be more closed to 4400MB/s.
 

·
Registered
Ryzen 3950X > 32GB G.Skill GTZN 3600 > ROG Crosshair VIII > NVMe Rocket 1GB > Fractal Define 7 + S36
Joined
·
166 Posts
I think I understand now why I got a idle reboot with 3702 and never had one on 3501/3601, minimum vcore voltage using PBO is now lower than I have ever seen before. All the way down to 0.248V! Earlier bioses was 0.8-0.9V at the lowest as far as I recall. No wonder the PC could suddenly shut off. I haven't tried the PSU Idle Current setting in the bios yet, it's still on Auto. Possibly it could affect things. I have a Corsair RM850x 2019.

I have a distant memory of early bioses had the same ultra low idle voltage, but it went away after a few updates.

I've only had one reboot thus far though.

The two CCD:s now idles at 2.5W, very low, I guess Asus implemented some sort of Energy saver that was absent before. As said, I use the same settings cross bios. The SOC still draws 16W and the full package 28W in idle, so maybe not that big of a saver, but still a saver.
0.2V is indeed the lowest Vcore I have measured so far.
2517561

Voltage value is read from undocumented Boosted P-State SMU register of 3950X with BIOS 2206
 

·
Invalid Media
Joined
·
4,400 Posts
Samsung 980 PRO 2TB (M2_1, PCI-E 4.0 x4, 5900X)

View attachment 2517545


ADATA XPG SX8200 Pro 1TB (M2_2, PCI-E 3.0 x4, X570)

View attachment 2517546

The chipset M2_2 slot is so bad, SX8200 Pro is suposed to be 3.500/3.000MB/s drive.
...not sure if this helps, but I had my PCIe. 3.0 WD SN750 NVME in both M.2_1 and M.2_2 w/o significant differences (X570 Dark Hero). Presumably it also comes down to what else you're running in terms of PCIe etc re. lane count etc.

2517565
 

·
Registered
Joined
·
105 Posts
ADATA not really reliable. Linus got video about it. ADATA swap it's components and there a lot revisions of one drive.

Video:
Curiously mine has Micron 64-layer flash, which supposedly is the original release of the drive, I validated this by looking at the drive, the Micron logo and number "64" is clearly visible.

Source:

Edit: This is the report I got from "smi nvme flash id"

Code:
v0.24a
OS: 10.0 build 19043 
Drive     : 1(NVME)
Scsi      : 1
Driver    : W10
Model     : ADATA SX8200PNP                         
Fw        : 32A0T54A
Size      : 976762 MB [1024.2 GB]
LBA Size  : 512
AdminCmd  : 0x00 0x01 0x02 0x04 0x05 0x06 0x08 0x09 0x0A 0x0C 0x10 0x11 0x14 0x80 0x81 0x82 0x84 0xC0 0xC1 0xC2 0xE0 0xE4 0xE5 0xE6
I/O Cmd   : 0x00 0x01 0x02 0x04 0x05 0x08 0x09
Controller: SM2262 [SM2262AB] 
FW revision: 32A0T54A
ROM version: 2262ROM:SVN00235
Bank00: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank01: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank02: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank03: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank04: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank05: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank06: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank07: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank08: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank09: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank10: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank11: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank12: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank13: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank14: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Bank15: 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
--- Experimental ---
FlashID : 0x2c,0xc4,0x8,0x32,0xa6,0x0,0x0,0x0 - Micron 64L(B17A) TLC 512Gb/CE 512Gb/die
Channel               : 8
Ch map                : 0xFF
CE map                : 0x03
Pages/Block           : 2304
First Fblock          : 1
Total Fblock          : 504
Bad Block From Pretest: 12
Start TLC/MLC Fblock  : 29
DRAM Info             : [0x4F 0x52]
DRAM Size,MB          : 1024 (possible incorrect)
DRAM Bus,bit          : 32
DRAM Type             : DDR3
DRAM Vendor           : Samsung
UEFI OROM             : Disable
 

·
Registered
Joined
·
8 Posts
I had an article pop up in my news feed recently that stated WD is aware of the problem and will be releasing a firmware to fix it soon. It's a problem with these specific drives when running on x570 boards. I don't have a link but is probably easy enough to find if you Google it.
Yeap this is accurate:

... Western Digital localized the problem to certain X570 motherboards that have their PCIe maximum payload size (MPS) value set at 128 bytes. This dictates the maximum transaction layer packet (TLP) that goes through the PCIe controller, and a low MPS value cripples performance. ...
 

·
Registered
Joined
·
181 Posts
SEAGATE FireCuda 520 2TB (M2_1, PCI-E 4.0 x4, 5900X)

View attachment 2517557

Write values Q8T1 and Q32T1 should be more closed to 4400MB/s.
...not sure if this helps, but I had my PCIe. 3.0 WD SN750 NVME in both M.2_1 and M.2_2 w/o significant differences (X570 Dark Hero). Presumably it also comes down to what else you're running in terms of PCIe etc re. lane count etc.

View attachment 2517565
Yeap this is accurate:
This is my WD BLACK SN850 2 Tb (2 partitions) with the latest firmware (613200WD):



I think I have some problem...

Baio
 

·
Gamer and overclocker !
Joined
·
2,132 Posts
Deleted
 
7701 - 7720 of 7760 Posts
Top