Hi

I have a card for repair with a symptom of bad behavior of GPU CORE phases. This card wakes up 50% of times but when it runs in Windows, only one or two phases work. The card goes dark when more power is needed.

I see proper pulses on the phase marked by me as "1" all the time but other phases are silent. Sometimes phases 2,3 or even 4 are alive. Pulses look perfectly good on the scope as far as shapes are concerned but they are out of sync, they are irregular and the temperatures of phases are different on the thermal camera.

I have read configuration by an original USB005
I replaced IR35217
I reprogrammed IR35217

Nothing changed !
The behavior is the same, absolutely no change, instable GPU CORE phases.
IR35217 is configured as 7+1 according to the configuration in USB005.
Maybe configuration file is corrupted?

Any idea?

При запуске до появления изображения как работают фазы?

"Грызные" импульсы при рассинхронизации фаз свидетельствуют о нерабочий одной или более фаз, но "чистые"... ?
Если ничего не помогает, я бы снял ДрМОСы и поменял местами все на все, так же пересадил бы дросселя (было пару случаев плохого контакта). Начал бы с дросселей - так проще.

Tere is only one working phase before screen shows up. One phase or sometimes two phases. I marked them 1 and 2. After the screen shows up there are the same phases running. PWM input pins are silent on not working phases.
Sometimes phases 3 and 4 start and pulses look perfectly good.

It's normal behavior, isn't it?
btw, usb005 is not programmer only. have you checked phase faults in powircenter while running Windows, idle & at load?

It is not a normal behavior. All other cards I have now in my hands, but from different manufacturers, behave differently. All phases start at once with smaller duty cycle but all phases work immediately from the start.

This Asrock behaves differently. All are based on IR35217, I have replacement chips directly from Mouser USA. USB005 does not show any faults. There must be something wrong in configuration or something else. What is responsible for ruch behavior that all phases start at the same time? Which configuration value?

Try to set 00 at 0x2C, 2D, 2E (See power saving section)

See also attached config for example (powercolor rx 5600 itx, 7+1)
---------- Добавлено спустя 41 минуту 50 секунд: ----------

have you done it without pci-e connection to motherboard?

> have you done it without pci-e connection to motherboard?

This card was connected through the riser, normally.

First "Read All Reg"
Second "Save Config File"

then I replaced IR35217, then I flashed it. Power off, power on and nothing really changed. The card works as before.

I will try registers tomorrow....

I mean the card should only get power (3.3V is enough) from riser, no communication with pcie host. Once you allow card (vbios, OS driver) to start - you'll get incorrect dump.

Hi

I did everything properly this time. I saved configuration properly, swapped controllers again.

After clearing 0x2C, 0x2D and 0x2E there is one phase as before but it stops running after 3-5 seconds and the VCORE disables itself.
I have 3 more programming attempts on the counter so I can test something

Zeroing these registers did not make it all phase mode. I programmed back the original file and it works as before.

Is there ahy differences (vs "first attempt" config)?

Maybe some additional mods are needed. See another config (gigabyte 5600 7+1) from Mafia67

When the riser is connected to the motherboard an only in bios I get the same data as when the card is disconnected. There is only user space an mfg space to be programmed for the first time and then only the user space.