Safe hardware PC builds list

simplex

glvn

gtx1060 - GDDR5 Samsung

rtx2070s - GDDR6 Micron

rtx2080s - GDDR6 Samsung

rtx3080, 3080ti - GDDR6X Micron

simplex

I changed the RAM in safe mini-pc (13700h + Iris Xe 96EU) and got eye-strain

The 13700h processor supports upto 5200 MHz DDR5

KVR48S40BD8K2-64 was installed (2x32gb, 4800 CL40, ?micron? chips) - no issues

I installed KF556S40IB-16 (2x16gb, 5600 CL40, hynix modules) which worked at 5200 CL38 - discomfort

qb74

simplex
I have a weird theory which you could try.
Could you try and run RAM testing software, such as Karhuu, MemTestPro, TM5 and see if you're stable in them for +16h?
Or if possible, use the same RAM but lower the frequency to 4800MHz.
You might be, as the overclock community puts it, "unstable" and it might influence your smoothness perception during use.

simplex

qb74 +16h?

too much, dude %)

mini-PC cant use edit timings, but my safe PC z390d+gtx1060 - can

I inserted rtx2070s ( which gives me strain after 2hours ~month ago after that I switched it to gtx1060 and get calm )

First of all, I test does CPU overclock matters, setting default (3.7ghz), then Gaming profile (4.4ghz), then Advanced profile (4.5ghz) - nothing has changed

Then I tested how DDR4 memory works in my z390 by default - at 2133mhz (15-15-15-36). I cant say it is paperlike, but 100% not strainy as xmp profile ( 3200 CL16-16-18-36 ) did

Then I set 3000mhz with defaulty (MB calc that) timings 21-22-22-50 - it was same as 2133

Then I set 20-18-18-36 keeping same 3000mhz - a bit worse

Then 16-18-18-36 @ 3000 - definently bad

Okay, I set 3200mhz as max supported memory freq - it was also good with 22-23-23-53

Then 20-18-18-36 - not as good

Then 18-18-18-36 - bad

Finally, I continue testing my win10 with auto-timings by MB (3200 22-23-23-53)

Here is bad timinigs causing eye-strain with 2070s:

Here is good, calc by MB (not paper but…. okay? still testing)

Imagine if memory can impact

qb74

simplex

Imagine if memory can impact

RAM can impact systems in ways most don't comprehend.
It can cause microstutters, which you can easily be experiencing, thus leading to your eyestrain
Think of it like this (in a rudimentary way):
XOC (term for extreme overclock - used to indicate unstable RAM behavior) => microstutter => need to refocus more often => eyestrain
(this is just a theory for the eyestrain part)

Your testing seems to somewhat confirm this even further.
Keep in mind, system can get corrupted if your RAM is unstable, which could (potentially) further increase eyestrain.

RAM at XMP can be unstable if paired with a mediocre motherboard and CPU IMC, You're only relatively "safe" at CPU-vendor specified speeds (which gets very hazy in DDR5-era CPUs)

Would be funny if people here have eyestrain due to unstable RAM.

EDIT: Would it be possible for you to use ASROCK Timing Configurator or MemTweakIt to showcase your entire timing list? (primary, secondary, tertiary)

Link for first utility: https://download.asrock.com/Utility/Formula/TimingConfigurator(v4.1.0).zip

simplex

qb74 MemTweakIt

Current ?safe? timings determined by MB

simplex

Ok I will look at them

the facts:

DDR4 3200@16-18-18-36 + 1060 (GDDR5) = good
DDR4 3200@16-18-18-36 + 2070s (GDDR6) = you have to be patient
DDR4 3200@22-23-23-53 + 2070s (GDDR6) = not paper but dont need patient
DDR4 3200@16-18-18-36 + 3080 (GDDR6X) = bad
z690 DDR4 3200@16-18-18-36 + 3080 (GDDR6X) = bad

win10 share RAM memory with GPUs. Perhaps, very different GPU and RAM memory speed/timings cause strain.

How to determine GPU timings to adjust RAM closer, to check theory?
How to disable GPU ability to eat/share RAM ?

qb74

simplex
~~Would you be able to list your whole PC hardware setup (peripherals included) and software stack? (GPU driver, windows version)~~ i'm blind, uve done so in OP
Which power supply are you using? Are you still using the hardware you've specified in OP? Which peripherals are you using?
When was the last time you reinstalled your OS cleanly?

GPU DRAM Timings have long since been locked down on Nvidia side.

moonpie

..................

simplex

qb74 Which power supply are you using?

before z690 sold, I tried another 750w PSU, then my 500w PSU which currently works with z390d - nothing change

OS installed in feb.2024 with blocked updates, 2x sata SSD + GPU + 1x m2 SSD available only

moonpie which can kick in VRR more often

in my BIOS (2019y), OS (2020 may), drivers (mid of 2021) - there is no VRR

It would be joke if DDR5@4800 dont have such things. Last laptop I tested was huawei d16 2024 with 13900h and IRIS xe (same as in mini-PC) but caused me strain after 20 min. Laptop also used OC ram - DDR4X @ 4266mhz

moonpie

................

jordan

moonpie wow I didn't know this. I remember someone a long time ago here said that running a system in 32bit mode or single channel ram would help strain which he thought was due to certain directX stuff not being enabled under those conditions.

moonpie

................

jordan

moonpie That's good to know. I'm assuming that applies to arc gpus too? I'll share this with someone I know who has issues with xe.

moonpie

.............

simplex

Guys, need advice.

My current safe asus laptop has lp173wf4-spf3 panel which I use at 80% brightness. After 11k hour of usage, it lost red color and got blue-green tint, so finally I found exact same new panel - lp173wf4-spf5. But first time I replaced it, I feel…. eye-strain ( not headache )

I measure it with Opple4 and found it has same modulation % depth and same brightness at 61% of brightness, keeping same results I have at 80% in spf3

spf3 at 80%

spf5 at 61%

spf3 at 90%

spf5 at 90%

Despite of spf5 have DC dimming, trinagles told it have ~1000hz smoothed PWM (4 triangles in 4 ms period)

Using same logic, spf3 have ~200hz PWM

The question is, is it possible to adapt for new panel? If I wait more few days, I can loose return period.

Each moment I tried to use SPF5, felt slight burning in the eyes, brghtness level doesnt matter. When I replace panel to SPF3, all became good instantly. Is is my nervous adaptation to the old panel that I have been using since 2018?

p.s. panels use same EDID. Datahseets told they 99% equal (same PWM freq limits and duty ratios, contrast, colors etc)

simplex

SPF3 v-com 2 images, play in circle to find pixel inversion type: 2 pixels shift horizontally by 2 pixels: img1, img2

SPF5 v-com 2 images, play in circle to find pixel inversion type: dot (chess) inversion, 1 pixel shift horizontally by 1 pixel: img1, img2

moonpie

...........

simplex

The only one change, similar to monitors I tested last summer, were dot (chess) pixel inversion in SPF5. I also was not able to use screens with such inversion type (they were IPS, VA, DC current with different "smoothed" PWM freq according to opple4, but all of them had "dot chess inversion")

gregtsakil9

simplex This is amazing what you are showing here .You just set the monitor at 45hz and you get this still image? I will try and see if that helps with my dizzines symtoms ..I also always feel that hdmi 144 feels better than dp 165 on my dell g3223d could their be differrences or not ? What is the solution of having both higher than 60 hz and also having a perfectly still image ?

simplex

gregtsakil9 You just set the monitor at 45hz and you get this still image?

Yes, in that monitor (xiaomi mi 27 2k) the v-com pattern activates at 46+ hz, but not sure its has same freq in your monitor…

TBH, same feelings I got - 144 at HDMI was better than 165 via DP, then I tried to clean EDID ( restrict to Windows to make extra color-convertinon, coz EDID had extra gamma curve which made image mode brighter and contrast )

Try to set 45hz, then 30hz via CRU to feel difference. Then try to experiment with EDID settings

p.s. the fact I found a bit after this post regarding 45hz - all extra color convertion in win10 also gives strain. Avoid ICM profiles and monitor's extra gamma settings writed in EDID (via CRU) - let win10 make its grapfic pipeline in Default colorspace, without convertions

« Previous Page Next Page »