Это сообщение говорит о том что дрова не подошли.
То же пытаюсь найти решение запустить tserver для 6800 , про бывал драйвера обновить , но видимо моих знаний в линуксе не хватает для этого.

hello, I'm currently using a Python script to test the AMD cards, it's simple but it doesnt give me the exact chip failing as the Tserver, but I can run the script using the integrated graphic card and test the specific Video card using the memory address, for those cards that doesnt have video output. is it possible to use Tserver with running from the Integrated graphics to test AMD video cards conected to the PCIe?

Есть ли у кого инфа работает ли tserver на b450? У меня про попытки загрузить зависает

i tested it on x570 and works

---

Posted after 2 minutes 27 seconds:

if you can you should use an nvidia card to display the image (and disable internal amd card)

tserver repack (работает с любым количеством карт, также через встройку)
залазит на флешку 32гб или на хард.

3д режим тест графики работает. выкинул матс, матс лучше закинуть на отдельную флешку, ибо были траблы совместимости.
+ добавил вовнутрь инструкции на русском языке.
убрать инструкции для ламеров на автостарте:
rm /etc/update-motd.d/10-help-text
поддерживаются все карты от RX460 до радеона 7 включительно. хотя не факт. С большего проверил что мог...
в случае если есть что-то более древнее, юзайте agt в комбинации с dmgg.py

извините за мое незнание, но для чего нужна команда quickfmg? Я протестировал его здесь и увидел, что это своего рода тест рендеринга, но я не нашел журнала теста.

Чтобы не отходя от кассы проверить картинку под нагрузкой
+ Там можно запустить ./tserver -boartest=config или что-то такое, в общем просто пропиши ./tserver -boartest, он выдаст список возможных тестов и конкретно один из списка можно запустить, тот в свою очередь у тебя будет спрашивать какой конкретно тест запускать, помню был в списке и тест шины, и тест напряжений, и тест каких-то модулей гпу, и тест конкретно по перегреву, и срабатывания защит.

Тот репак что я замутил, по отзывам, работает хорошо на всех картах в списке, но на 6ххх серии пока никто не проверил. если у кого есть карта, можно попробовать сварганить тест, были идеи сделать универсальный tserver, но к сожалению версии самого выполняемого ./tserver разные... так что не прокатит...

Скачу в тест! У меня неисправный rx5700xt, который, как мне кажется, является неисправным банком памяти. однако заметил, что сидов в торренте мало ...

use second card connected to monitor, like nvidia or integrated graphics, tserver will test memory on AMD
dont forget to switch memory speed:
./agt -mem=300
300mhz default,
./agt -mem=1000
1000mhz slightly higher, sometimes it can show more errors.
to change gpu clock
./agt -eng=300
300mhz default
./agt -eng=800
off default
to run ./tserver with memory test:
./tserver -boardtest=memfa
to view result
pico memfa.log

sorry for only russian language instructions on my tserver-repack. i will add english one too later, when new 6xxx gpu's tserver will be leaked
also, sorry for torrent. it's slow but... like really good stuff always takes time to achieve.

also you can use both AMD gpus, one and secondary, you only need to add something like this:
./tserver -boardtest=memfa -d=1
or you can run
./tserver
and it will show you what it can do in english, also you can easilly understand what it is cappable off, like how to select cards, which tests to run and etc... also on other programms, like
./agt
it will show you list of it's abilities...
like cool feature
./agt -clkstatus
it will show your AMD gpu cloks and band


also, if current ./tserver can't run memfa, you can try
python3 dmgg.py
it is included on repack, older and newer versions.
and possibly, you will be able to tweak your card with any present ./agt

Hi, I have a XFX RX580 8GB that shows problem at memory speed from 363MHz to 499MHz, all the other memory speeds are ok.
I tested with "tserver" with memory speed set by "./agt -mem=...."
If I set the memory speed at 400 (or similar 462.5 or 487.5) the screen shows artefacts and memory tests is FAIL and also if I afterwards test at any other speed without a reboot.
If I set the memory speed at 400 (or similar 462.5 or 487.5) the screen shows artefacts, I change now the speed to any other value without doing any test, all the memory tests fail at any speed anyway.
If instead after a reboot I set any other speed (I tried 300, 500, 600, 900, 1000, 1125, 1250, 1375, 1500, 1625, 1750, 2000) the test is PASS and no artefacts.
I mod the bios with a slower stripe at 400MHz but it doesn't change anything.

I heard that Polaris reads some timing parameters directly from one of the GDDR5 memory chip, is it true? and which one is it?
I suppose that that chip contains corrupted data
Thank you !

try to clean out all waste from gpu and memory
then if it does nothing, try to flash proper firmware

which tserver it was? mine or else? i did have an issue similar, it was just some trash under memory chips

Hi Sparrow,
thank you for your answer

The PCB has been already cleaned 2 days soaked in IPA, I already tried several different Bios from Techpowerup and additionally I did a GPU reflow but the card behaves exactly as before.
I can run Furmark with mem @2000MHz without any problem but if I just type "agt -mem=363" artefacts everywhere, the behaviour is always the same and reproducible with exactly the same frequencies

For these reasons I really think that Polaris reads some timing data directly from the GDDR5 memory chips, but from only one chip or more, and which one? I suppose that there is a GDDR5 chip that contains corrupted data, is it possible?

i believe not. so, lets just try to build our own bios for your card with higher timings on certain memory clocks
first of all, can you start windows with that card bein secondary on system? with other card to use monitor. can you find it in device manager ?
and also, WHICH tserver did YOU use? mine self made repack, or something other ? like 500gb hard one with different hard partitions, or Vik-on's?
check out, all timings ARE located in firmware

.. yes, I can use the internal graphic card and I can find the RX580 in the device manager since the card also boots without any problem
I have the following testing software:



The gaming bios of the XFX RX580 GTX xxx (I could not attach the original file but the following one is equivalent):
https://www.techpowerup.com/vgabios/200 ... 192-171030
The card has problems with memory frequencies from 363MHz to 499MHz (means that if I set the freq = 362 or 500 it works, in between it doesn't)

wow, you did quite hard... i guess it took time a lot!

but yea, it's drivers on your pack of software is kind of a mess i must say, i better recommend you to create two different usb drives, one 32 gig with my firmware and second with just newest mats, it will work better.

as for the card... it's dead memory controller 100%, but you can just try reballing gpu or look up for broken smd's, but it's time consuming and may not give any results.

so, i better recomend you create new firmware bios. for clocks from 250 to 600, paste timings from 900

for flashing firmware you can use windows and ATIWINFLASH
or, option which i like the most, directly on a test bench with usb drive on HIVEOS


BUT NOTICE, IT HAS TWO DIFFERENT pages for memory!!!

check your's on your card, and what gpu-z says if it's different, so... we got the problem, if not, so try flashing custom firmware

... yes it took a lot of time but this is a hobby :)
I tried the bios mod as you wrote unfortunately the result is still exactly the same as before, I expected this because I tried already previously a slower strap for the 400MHz and didn't change anything.
I also think that the memory controller is very likely damaged, but I still hope that the Polaris gets some parameter for the memory controller directly from the memory chips as I saw in the past someone saying this on a youtube video that I cannot find anymore unfortunately. In this case the memory chip would be damaged of course instead of the gpu :)

Moreover strangely enough if with AMDMemoryTweakXL I disable the GPU states from P1 to P6 (only P0 and P7 enabled) the graphic card actually seems to work, at least for mining should work:o So another possible solution would be to disable P1 to P6 permanently from bios but I have no idea how.
Anyway thank you for your kind support for the moment, it will take time, I will let you know

hmm... i'd say try to tweak voltages, also, are they fine ? like if memory controller gets lower power on lower cloks, but way less then required
there is a dedicaded phase for memory controller inside gpu [it's often located near pci-e slot], try to play with it. might give some results

Привет,

впервые с тсервой. Я пытаюсь диагностировать xfx 470. Я получаю это сообщение, что не так?

никто не знает ? Я использую встроенную графику, потому что Ati не выводит изображение. какой-нибудь совет, пожалуйста?

Тестирую tserver'ом R9 390 8Gb (Сапфир)
Ошибка по каналу F0, сиротливая единичка. В третьем столбце, на 6-й позиции (скрин не делал).
Банки F0 F1 - пересадил (11-12 против часовой), ГПУ снимал, сажал снова (уже три раза).
Сигнальные линии, от контрольных пятаков до пятаков под чипом памяти и до пятаков ГПУ - прозванивал.
Единственное, БИОС не перепрошивал. Может там левый залит.

Собственно вопрос. Когда ошибка в таком виде (одна единичка) - что оно вобще означает?
Когда все "пятерки" - понятно. Не видит чипа. Когда "пятерки" в одном из столбцов - тоже понятно более менее.
А одна "1"? Помеха какая-то? Или всеж ГПУ отъехал в мир иной?

В диспетчере устройств две видео карты или только интегрированная?

I have the same output. My last post, tserver output.
viewtopic.php?f=48&t=118167

For me there is something wrong with all of this testing. It behaves differently when used directly with the card under test and when I use cheap Nvidia as a primary display and the same when I use dmgg.py. Sometimes there is a confusion, contradictory results. Sometimes dmgg says that all memory banks are dead when used with primary Nvidia, without it, when AMD is a single card it says no bad chips. Tserver is also unpredictable.

Нормальная история, большая часть карт именно так и диагностируется.
Я использую карточку Nvidia для проверки карточки AMD, чтобы Tserver не ругался. ибо если две карты амд, приходится:
./agt -i=1 -clkstatus
./tserver -d=1 -boardtest=memfa
Конечно так работает, но лишние 5 символов вводить лень
---------- Добавлено спустя 56 минут 47 секунд: ----------
Обновил репак Tserver
пофиксил размер, ибо получилось так, что не все флешки 32 гб на самом деле 32 гб.
встает как на жесткий диск, так и на флешку, 32гб и более.

Thank you very much for your work but I have a small suggestion

Could you please compress the image and add a date to the image? I would prefer dd and gzip instead of any third party bloatware.
It will reduce the download time by more than a half. Your packed image weights <12.8GB. In Windows RAR or 7zip is a better choice than a simple ZIP. And please add "mc" to the build.

Example of compression for Linux users


Example of decompression and creating USB or HDD/SSD at once.

thanks for advices. Will soon try it out. but i thought of making .iso, failed unfortunately...
also seeking for more software, like newest 6xxx cards. any other suggestions?

Hehe no more suggestions, just this size of the file and this commercial bloatware with reduced functionality and intentional slowdown I use dedicated SSD 120GB so I don't care about size, I can use command line parted to extend filesystem and install some additional tools from the repository. In general the MC should be installed in a package and that's all.
Thank you very much again for your work.

Iso is compatible with this and can be probably flashed by rufus.

gparted, yea, great thing.

Thanks comrade, will try, maybe monday, gotta dig some potato in village for the sake of mother Belarus!!!

This new image does not work, I used older 2 months old but version from yesterday and 13th reports partition as 82 swap. I think it will be better to make image as I described above with gzip/dd. Now I am writing the older image to confirm. But it was ok until today.

nope, my friend and also some strangers prooved working.

Maybe my old motherboard for tests? The older version works. I will put it on a mechanical HDD later and I will test. The older version is confirmed booting from SSD 120GB. There must be something wrong/incompatible anyway.