Safe hardware PC builds list

simplex

Hey folks,

I suggest making a list of safe workstations, to share it with each other

Here is my safe workstation:

CPU: i5 9600kf ( all overclock options give strain )

MB: LGA 1151v2 gigabyte z390d with last F4 BIOS ( no difference in F2/F4 versions ). When you reset BIOS removing CMOS battery and power cable for 1-2 hour, all variables in UEFI also get clear ( no tails from previous grapfic cards ). I also succesfully tried gigabyte z390ud MB with last BIOS - same good experience

RAM: DDR4 2x 16gb 2133mhz (JEDEC default) patriot viper steel PVS432G320C6K. When XMP 3200mhz profile applied, I feel strain. When I use another JEDEC 2400/2666mhz memory ( CPU support upto 2666 JEDEC ), or set frequency 2400/2666/3000/3200 manually in BIOS, also feel strain

Graphics: asus dual gtx1060 6gb ( 86.06.0E.00.41 bios ) + nvidia 451.67 driver with "Display -> Adjust desktop color settings -> Use NVIDIA settings" checkbox ( reduce strain, after 460.xx version this option was removed and grapfic card became not calm ) + novideo_sRGB with clamped checkbox ( reduce strain, add "ColorspaceConfig" value in registry ). Uninstalling grapfic driver with DDU completely remove all tails from previous cards

All nvidia cards starting with rtx20 bother me. All new grapfic cards have Tensor/RT cores. Tensor cores can use float16 / mixed precision, which gives different results each time it calculated ( not static frames ). RT cores also use random sampling.

Monitor: 24" 60hz *VA benq bl2420z ( auo m238hvn01.0 / 6-bit+FRC / 6500k temperature panel ) with HDMI to DVI-D cable ( it passive, and work as pin cross-replacement ) or DVI-D to DVI-D cable, less strainy than DisplayPort cables I tried. Perhabs, monitor use DP to HDMI convertion, so you need to find which cableport is calmest in your monitor

OS: win10 enterprise LTSC x64 build 17763.1098. To get same version, you need to find w10 1809 ISO and install updates: KB4537480, KB4462930, KB4465065, KB4486153, KB4537759, KB4539571, KB4538461

Software: Firefox 133.0.3 with default settings except two option which disable colorconvertion in about:config

gfx.color_management.mode = 0
gfx.color_management.rendering_intent = -1

p.s. I feel strain on this equipment with options above, with all newer w10 OS ( 21h2, 22h2 ) or newer grapfic drivers

update part #1: amd 780m dithering research

1. AMD by default ( driver_default state ) use dithering in all color depth, according to monitor specs (EDID)

some evidence from linux core:

switch (bpc) {
 case 6:  
if (dither == AMDGPU_FMT_DITHER_ENABLE) { /* XXX sort out optimal dither settings */   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 0);
} else {   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 0);  
} break; 

case 8:
if (dither == AMDGPU_FMT_DITHER_ENABLE) { /* XXX sort out optimal dither settings */
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_RGB_RANDOM_ENABLE, 1); 
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 1);
} else {
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 1);
}  break;


case 10:  
if (dither == AMDGPU_FMT_DITHER_ENABLE) { /* XXX sort out optimal dither settings */
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_RGB_RANDOM_ENABLE, 1);   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 2);  
} else {
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);   
   tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 2);
}  break;

2. Comparing results, I found 780m set exact same registry settings in driver_default (0x0000C900) also as in dith8_no_frame_rand (0x0000C900)

3.AMD use same dithering settings in vega6, vega7, 780m. Here is list of possible values:

0x0000C900 default

0x00002023 fm6

0x00002000 fm8

0x00002000 fm10

0x0000E123 dith6

0x0000E900 dith8

0x0000F100 dith10

0x0000C123 dith6_no_frame_rand

0x0000C900 dith8_no_frame_rand

0x0000D100 dith10_no_frame_rand

0x00002001 trun6 <<<

0x00002011 trun8

0x00002023 trun10

0x0000E111 trun8_dith6

0x0000E921 trun8_dith8

0x00002011 trun8_fm6

0x0000E900 dith8_fm6

0x0000E121 trun10_dith6

0x00002021 trun10_fm8

0x0000E921 trun10_dith8_fm6

0x0000F100 dith10_fm8

4.Kawamoto's ditherig do not set 0x0 values when you press "Disable all dithering" in his app

For my personal taste, setting all values to 0, not works at all. The best eye-settings for my 6-bit + FRC monitor, I currently found, is: 0x00000011. To set same settings, you need to:

A) Download ditherig

B) Open in amd64 folder database.csv, and replace in 89..94 lines from

0x00073900,0x00000000

to (8-bit truncate only)

0x0000C911,0x00000011

or (no settings at all)

0x0000C911,0x00000000

5.As many ppl there noticed, some apps reset dithering settings when using - yes, it is true. Here is apps I tested for AMD driver reset:

Reset driver settings to default: Chrome, MS Office, OBS studio, Zoom Workspace

Do not reset driver settings: DXO Photolab, Firefox, Telegram Desktop, WinRAR, Win10 notepad / calc / paint /photoviewer, Visual Studio 2022, MPC-BE 1.8.0

6.Without installed AMD driver, win10 with Microsoft Basic Display Driver also shows 0x00000000 (no dithering applyed). When you install any AMD driver, dithering settings changes to 0x0000C900 (driver_default). I feel strain even without AMD driver. It means, the strain issue is hardware-based, and cannot be controlled via driver (or registers) settings

Eye-strain: amd 780m gives me strain (mini-PC or laptop based). Not calm

update part #2: ASRock Arc B580 Steel Legend dithering research in win10 LTSC 21h2 19044.1288 and Intel Driver 32.0.101.6647 (all default settings)

Pixel Extension = Zero Extended
Allow Double Buffer Update Disable = Allowed (Default)
YUV420 Mode = Full blend

Dithering type = Spatial
Dithering enable = Enable
Dithering BPC = 6 bpc
Pixel Extension = MSB Extend (Default)
Allow Double Buffer Update Disable = Allowed (Default)
YUV420 Mode = Full blend

By default, this GPU doesnt use output dithering with 8-bit monitor connected. But it force Pixel Extension=Zero Extended key insted of Pixel Extension=MSB Extend (Default). The Iris XE in 13500h / 13700h iGPU I tested always use Pixel Extension=MSB Extend (Default)

Eye-strain: b580 gives me strain, similar to amd 780m. Maybe a bit softer. Not calm

update part #3: Intel UHD 730 grapfics ( i5 13400 ) dithering research in win10 LTSC 1809 17763.1098 and Intel Driver 32.0.101.6647 (all default settings)

YUV420 Mode = Full blend

Dithering type = Spatial
Dithering enable = Enable
Dithering BPC = 6 bpc
YUV420 Mode = Full blend

It mean, UHD730 use +/- equal register values to IRIS XE and Intel ARC GPU. By default, no output dithering applied with 8-bit monitor

Eye-strain: no issues

ensete

simplex There is no universal safe hardware list.

Different people are impacted by different things and identical models of hardware can have different components.

autobot

simplex That's a good idea. we already have a similar list on the wiki. with many people added hardware.

https://wiki.ledstrain.org/docs/review/hardware/

simplex

https://filetransfer.io/data-package/GaHolmgv

Here is comparison of pixel moving, 60fps record allow to reduce speed to 12.5% in your player to find same repeating frame-patterns I synchronized, better to navigate frame-by-frame (Ctrl + Right keys in MPC-BE player) to get this pattern:

Vertical
Horizontal
Vertical
Smooth

According to monitor's 60hz refresh rate, this is 15hz 4 frame cycle, and no difference in them between z390 or z690 motherboards (same driver version, same settings, same win10 1809 build, dithering in nvidia disabled via ColorControl, in z690 - via Ditherig)

But z690 gives insta eye-strain even in BIOS, 2022 or 2024 BIOS versions doesnt matter

simplex

Who knows how to reduce image sharpness in motherboard?

The issue could be extra sharp rendering, which made monitor's pixel inversion/dithering very visible, then eyes get tired

MrLoco

Sorry I don't really know. I'm just still fascinated that anything OTHER than the monitor could be the culprit.

simplex

MrLoco OTHER than the monitor could be the culprit.

Mr. SunnyCove from 4pda also mentioned there could be case when motherboard set dithering in monitor

For example, my monitor doesnt support 10 bit, but MB send unsupported control signals for it, or send 10 bit data, which cause "sharp" and bad monitor behavior

Same eyestrain I met when plug 2021 year monitor, xiaomi mi 27 2k... a little reduced feelings, but same pressure. Today I take a day without testing bad z690 coz I waked up with headache

simplex

Can new (2020+ year) motherboards make custom area screen refresh?

I noticed that problems appear during reading (static). And if you launch the calculator application and scroll it around the entire screen, the pressure on the eyes stops for about 3 seconds. So gamers can not detec problem at all

Are there utilities like screensavers that can emulate a small action across the entire screen without interfering with reading?

simplex

My z690 experience:

Turn off all ASPM in BIOS ( some of them enabled by default )
Turn off all SpeedStep and SpeedShift in BIOS
Disable Above 4G Decoding and ReBAR
Old nvidia driver 466.67 - only DisplayDriver, all settings by default ( ColorRange = Full, 8 bit, RGB )
Change ColorControl's dithering mode from Auto: Disabled to Disabled
Set LowLatencyMode = Ultra and Power management mode to Prefer Maximum Performance in nvidia Control Panel
Disable 10bit, 4:2:2, 4:2:0 support, bt2020 support of monitor ( stay only bt709 ) and delete all HDR mapping data in CRU
Set windows powerplan to Max Perfomance

This all significantly reduce eye-strain ( to be honest, between brows pressure )

Here is 8 settings I changed, lets discuss which is important and which is not. I also notice problem more clear when reading ( static background ) than watching vids

simplex

@"photon78s"

All videos are in 60fps and contain 60hz monitor recording. You need to see vids frame-by-frame to get FRC pattern. Interesting fact is: in all specs monitor is 8 bit, but in datasheet it send only 6 bit data into panel 🙂

6-bit data input: you can see only pixel inversion: https://filetransfer.io/data-package/eqe6MFii
8-bit data input: you can see 4 repeating patterns ( oldschool traditional FRC type, modern A-FRC is very different ): https://filetransfer.io/data-package/9aej86qS
8-bit data + spatial-temporal dithering via Ditherig: https://filetransfer.io/data-package/l39tphOQ

Its hard to see FRC patterns in 8bit + dithering data, coz signal become very noisy. Imagine what happens with letters, one side of which takes only 1 vertical pixel row? They get chaotic non-static shift left-right, which overload eye nerves, to solve such problem ppl set lower resolution to get letter more wider/boldy and keep static elements inside of letter elements ( when edge elements of letters would keep dithering )

DisplaysShouldNotBeTVs

simplex interested in looking at these but they just expired today, can you reupload them?

simplex

In my safe gigabyte z390d motherboard, I am afraid to update BIOS to next version coz you cannot rollback to previos one after update. This update introduce capsule BIOS and re-bar ( CPU can access all GPU's videomemory ) function of nvidia. I am not sure update could be safe for me, at least z690 motherboard dont have it in patchnote so I think z690's BIOS includes re-bar by default

I got some stats here, in ledstrain, regarding user's issues with different laptops. If exclude pure display problem ( oled, amoled, t-pwm screen backlight tech ), mass problems starts with:

1) Intel 11 gen CPU's and newer: 8/9 gen is good - my own experience, no much complains regarding 10 gen of lga1200 but a lot bad reports regarding 11 gen lga1200 and 12 gen lga1700 ( my own experience ). Perhaps 10 gen is affected too

2) AMD 5800h and later series CPU's: 5600h of my mate is good, 4600h is also good ( my own experience )

simplex

6bit + FRC = 8 bit = 4 repeating pattern frames to get extra 2 bit color data

6bit + Hi-FRC = 10 bit = 8 repeating pattern frames to get extra 4 bit color data

At 60hz Hi-FRC very noticable, eye got strain, thats why manufactures increase framerate upto 120+hz to smooth effect. The question is: manufactures present 120hz as good for games and eyes, but can they "mask" false 8 frames?

At 120hz panel you get only 120/8 = 15 real frames, all others are "flickers"

At 144hz panel you get 144/8 = 18 real frames

Here is Hi-FRC monitor review:

"I worked on a laptop for about five years and had no problems with my eyes at all. My vision has always been very good, a little sharper in the distance. And so I wanted to buy myself a larger screen to make working with the code more comfortable. My choice fell on the Lg 27Ul850-W monitor, since it has Type-C for my MacBook. At first, I didn’t feel any discomfort while working, fatigue in my eyes began to appear after two weeks, but since I didn’t experience any problems with my eyes at all, I didn’t understand what was happening at all and chalked it up to general fatigue.

As a result, after two months of torment with settings, different calibrations, nothing changed, my eyes began to hurt wildly. What I have now: I can’t look at any screen at all, my eyes burn, my head hurts, etc. I can read the news for no more than five minutes. I went to the ophthalmologist and threw out a 10 for an examination. A diagnosis of accommodation was made and glasses were prescribed for unloading near vision; the glasses also have a blue filter. I tried to sit at this monitor with glasses on, the situation did not change, although the glasses really take off the load in close proximity and it became noticeably noticeable on the laptop. As a result, I switched back to my laptop, my eyes began to recover, and Irifrin was also prescribed for a month.

I just can’t understand what’s wrong with this monitor, I turned down the brightness and contrast, I did a pencil test and didn’t notice the pwm, I also turned on slowmo at 250 fps on the iPhone, I also can’t see the pwm, in short, it’s some kind of fantasy. My wife thought that I just had sensitive eyes, she worked on it for a week and in the end her eyes also hurt. In general, I do not recommend this model. After all this, I was generally disappointed in monitors from LG."

qb74

simplex

6bit + FRC = 8 bit = 4 repeating pattern frames to get extra 2 bit color data
6bit + Hi-FRC = 10 bit = 8 repeating pattern frames to get extra 4 bit color data

Are you sure this is the case? As far as I've read up:
6bit + FRC = 16.2m
6bit + HiFRC = 16.7m

As per http://www.lagom.nl/lcd-test/black.php at least.

At 60hz Hi-FRC very noticable, eye got strain, thats why manufactures increase framerate upto 120+hz to smooth effect. The question is: manufactures present 120hz as good for games and eyes, but can they "mask" false 8 frames?
At 120hz panel you get only 120/8 = 15 real frames, all others are "flickers"
At 144hz panel you get 144/8 = 18 real frames

Can you quote the source? I'm confused what you mean by this.

simplex

In specifications, they wrote 8bit panel as my Benq, but the truth is he 6bit + FRC ( 4 patterns ), in datasheet summary you can see " panel can display 16.7m of colors, 8bit input ", but in scheme I post above you can see only 6bit data sent to panel, keeping 8bit input

Who know model of real 8bit monitors? My experience talks, you can detect it only with microscope or telezoom camera lens keeping same FPS which panel refresh

simplex

When you look modern panels at low-freq refresh rate, some vertial rows all across the screen could be seen. I think this is vcom screen optimisation - in detail, it is green vertical subpixels in row. When you switch panel to 120+hz, those rows still there, but not visible for eye. Have a look scheme out of datasheet:

simplex

DisplaysShouldNotBeTVs can you reupload them?

which one?

simplex

All monitors were in 8bit 60hz mode with cleared EDID settings in z390 motherboard + rtx 3080 graphic card, video is 4k60p

!!! you need to x8 slowdown to see FRC difference !!!

AOC Q27B3MA - 2k60hz

https://cloud.mail.ru/public/g6vi/uG9gdWZ3r

benq bl2420z - FHD 60hz ( AUO m238hvn01.0 )

https://cloud.mail.ru/public/nDaD/yy71Zr9kK

xiaomi mi 27 2k 165hz ( AUO M270DAN02.B )

https://cloud.mail.ru/public/9w6Q/y7qSgnnvm

philips 275V8LA - 2k60hz ( TPM270WQ1-SG1G012 )

https://cloud.mail.ru/public/cqgm/xgmLS7Sgg

=======================================================

Upd:

And here is how simple dithering looks ( z690 + iGPU + benq in 6-bit mode )

https://cloud.mail.ru/public/anaU/kahSkroTa

And how classic 4 pattern FRC dithering look ( z690 + iGPU + benq in 8-bit mode )

https://cloud.mail.ru/public/H3BJ/7KjVMiJcA

=======================================================

Upd2:

3 video cards in z390d + benq 8-bit

The gtx1060 has the most visible FRC patterns in terms of sharpness, the rtx2/3 gen have stronger dithering (hides the FRC patterns) and overall image "blurriness"

https://cloud.mail.ru/public/dMBD/ZU1XN4T7e

simplex

qb74 Are you sure this is the case? As far as I've read up:

Yes

Moreover, I suppose most of 8/10bit monitors are 6bit+Hi-FRC, the bad is my camera is only 60p so I cant check monitor's FRC in 120hz mode to find 8 "blended" frames. On my benq I detected 4 FRC pattern frames, in xiaomi its hard to detect FRC patterns due it has A-FRC (not classical FRC patterns)

simplex

z690 + iGPU 6-bit vs 8-bit + benq

At 6-bit no FRC at all, only monitor's dithering

https://cloud.mail.ru/public/UhWz/kYGUfMPvr

Very interesting, I cant get same no FRC image when using gtx1060/rtx2070s/rtx3080 cards even forcing monitor to 6-bit. I get banding at wallpaper, but FRC still there with dGPU

I mean, this FRC patterns are always in 6 or 8 bit monitor mode -> https://cloud.mail.ru/public/dMBD/ZU1XN4T7e

qb74

I was not able to spot any flicker happening on the videos you've recorded.
It's a bit hard to capture this without a oscilloscope + probe, cameras do not suffice for this matter sadly.

simplex

qb74 I was not able to spot any flicker happening on the videos you've recorded.

Talking this video -> https://cloud.mail.ru/public/UhWz/kYGUfMPvr

You need to download video, then play video frame by frame - in bottom (8-bit label) part of the video you will see 4 FRC repeating patterns (vertical, horisontal, vertical, smooth) looking at taskbar, to the left position of the "up arrow" symbol.

In 6-bit there is no patterns, only monitor's dithering random noise on taksbar visible.

In 8-bit case, on the dithering noise layer you seen in 6-bit part, monitor add FRC layer with 4 patterns ( through the FRC patterns you can notice same 6-bit dithering noise structure )

qb74

simplex
Mate, do you not understand that my panel can dither and we cannot have the same experience as what you view IRL?
Oscilloscope + (fast) light probe is the objective way to measure this, not some random videos.

ensete
There is a "safe" gudeline I'd say for monitors.
But, ppl here just need to start evaluating biological factors instead though instead of blaming everything on the monitor.

A safe monitor consists of:

Having least amount of flicker possible (can be easily troubleshooted with a fast oscilloscope + light probe)
Highest refresh rate possible (sub-360hz is too blurry)
Highest resolution possible (if possible, opt for high refresh laptop panels due to PPI)
Does not use KSF phosphor / QDEF backlight (both seem to have reports of causing issues for people), instead employs regular LED, CCFL backlight or WOLED (seems to be fine for some people, only caveat is brightness dip at refresh rate cycle)
Avoiding high brightness (100-150 nits is fine for daytime use, personal preference territory)
Using warmer whitepoint (D55 / 5500k is nicer than 6500k or colder, but this is personal perference)

There is a gray area however, which is very individual + not a lot of info on this.

Matte/Glossy considerations (some work for ones, not for others)
Polarizer orientation (some orientations work better for ones, not for others)
Dithering technique used (temporal, simple temporal or static, I've seen reports of static being horrendous for some)

People here use extremely archaic & ridiculous panels imo.

There are outside factors such as:

Light in room (bulbs are notorious for flicker)
Biological/psychological factors (vitamin/mineral defficiency, stress levels, hormonal disbalances, diet etc.)
EMF's (yes, they play a major role in health)