• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Looking for volunteers with Fiji GPUs to figure out ASIC Quality

Joined
Jan 3, 2010
Messages
60 (0.02/day)
Processor AMD Ryzen 5 2600
Motherboard MSI B450M Mortar
Cooling BeQuiet! Dark Rock 4
Memory 16 GB DDR4 3333 MHz Kingston
Video Card(s) Sapphire RX Vega 56 Pulse
Case Fractal Design Define Mini C
Power Supply Seasonic Focus Plus 650W Gold
Software Windows 10 Pro x64
Thanks just tested it with my Fury OC:

asic.png
 
Joined
Jan 17, 2015
Messages
244 (0.14/day)
Location
Lurking over a keyboard ...
Processor Threadripper 1950X
Motherboard Asus Zenith Extreme
Cooling Custom WC
Memory G.Skill Trident Z 3200MHz C14 4x8GB (3400MHz C15 1T)
Video Card(s) GigaByte RX VEGA 64 Limited Edition
Storage Samsung EVO 840 250GB
Display(s) Asus MG279Q
Case Be Quiet Dark Base 900 (Mods: Front mesh panel, improved top panel for airflow)
Audio Device(s) On-board
Power Supply CoolerMaster V850
Mouse Logitech G700S
Keyboard Cherry MX-Board 3.0
Software Win 7 Pro x64 / Win 10 Pro x64


I would have assumed with 1.212V I'd have higher quality ASIC (High LeakageID)?

I currently don't have the other 3 Fiji cards I owned, but their DPM 7 VID on stock ROM / settings was:-

Fury Tri-X STD 1.243V @ 1000MHz DPM 7
Fury X 1.250V @ 1050MHz DPM 7
Fury X 1.187V @ 1050MHz DPM 7

I would have assumed from the information by the Stilt that the samples which were 1.250V or close to it would have been Lower ASIC Quality (Lower LeakageID).

I have another card coming within the next few days and will check and post screenie of that.

@Szaby59 are you able to do registers dump via AiDA64 when on stock ROM / settings? cheers :)
 
Joined
Jan 3, 2010
Messages
60 (0.02/day)
Processor AMD Ryzen 5 2600
Motherboard MSI B450M Mortar
Cooling BeQuiet! Dark Rock 4
Memory 16 GB DDR4 3333 MHz Kingston
Video Card(s) Sapphire RX Vega 56 Pulse
Case Fractal Design Define Mini C
Power Supply Seasonic Focus Plus 650W Gold
Software Windows 10 Pro x64
DPM0: GPUClock = 300 MHz, VID = 0.90000 V
DPM1: GPUClock = 508 MHz, VID = 0.93100 V
DPM2: GPUClock = 717 MHz, VID = 0.93700 V
DPM3: GPUClock = 874 MHz, VID = 1.03700 V
DPM4: GPUClock = 911 MHz, VID = 1.07500 V
DPM5: GPUClock = 944 MHz, VID = 1.11200 V
DPM6: GPUClock = 974 MHz, VID = 1.15300 V
DPM7: GPUClock = 1040 MHz, VID = 1.25000 V

Not just the DPM7 all are affected except 0 since from 1-7 the voltage is generated depending on the chip quality.
 
Joined
Jan 17, 2015
Messages
244 (0.14/day)
Location
Lurking over a keyboard ...
Processor Threadripper 1950X
Motherboard Asus Zenith Extreme
Cooling Custom WC
Memory G.Skill Trident Z 3200MHz C14 4x8GB (3400MHz C15 1T)
Video Card(s) GigaByte RX VEGA 64 Limited Edition
Storage Samsung EVO 840 250GB
Display(s) Asus MG279Q
Case Be Quiet Dark Base 900 (Mods: Front mesh panel, improved top panel for airflow)
Audio Device(s) On-board
Power Supply CoolerMaster V850
Mouse Logitech G700S
Keyboard Cherry MX-Board 3.0
Software Win 7 Pro x64 / Win 10 Pro x64
Cheers :) , the Fury Tri-X OC ROM is invalid for assessing stock VID DPM 7 IMO, as it's not at AMD stock clocks for Fury.

AFAIK the driver/ROM under EVV calculates VID per DPM at AMD clocks, deviating from it creates anomaly.

For example if I changed DPM 7 frequency on my Fury Tri-X STD to greater than 1000MHz I'd get 1.250V. In the case of Fury X cards which I had where DPM 7 @ 1050MHz lower than 1.250V increasing clock above 1050MHz resulted in 1.250V. The max EVV mode will make DPM 7 is 1.250V with clock changes from AMD stock.

*** edit ***

"Not just the DPM7 all are affected except 0 since from 1-7 the voltage is generated depending on the chip quality."

Yes, indeed :) .

Default GPU clock per DPM + LeakageID + other GPU properties.

From what the Stilt stated in the case of other gen ASIC was within LeakageID band there is also good / bad ASIC. For example we take two identical LeakageID ASIC, one good and other bad, the bad ASIC will end up with higher VID where as good ASIC lower VID.
 
Last edited:

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
19,864 (3.49/day)
Processor Core i7-4790K
Memory 16 GB
Video Card(s) GTX 1080
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 7
Everybody with Fiji: please download GPU-Z 0.8.8, go to the validation tab and click submit, no need to fill out anything.

This will record your ASIC quality, for comparison with everybody else's.

At the moment there are only a few Fiji's in there, so your relative score is unrealistic.
 
Joined
Jan 17, 2015
Messages
244 (0.14/day)
Location
Lurking over a keyboard ...
Processor Threadripper 1950X
Motherboard Asus Zenith Extreme
Cooling Custom WC
Memory G.Skill Trident Z 3200MHz C14 4x8GB (3400MHz C15 1T)
Video Card(s) GigaByte RX VEGA 64 Limited Edition
Storage Samsung EVO 840 250GB
Display(s) Asus MG279Q
Case Be Quiet Dark Base 900 (Mods: Front mesh panel, improved top panel for airflow)
Audio Device(s) On-board
Power Supply CoolerMaster V850
Mouse Logitech G700S
Keyboard Cherry MX-Board 3.0
Software Win 7 Pro x64 / Win 10 Pro x64
OK, cheers for info :) , I'll highlight to members on OCN owners club plus the Fiji bios mod thread to do validation ;) .
 

Eliovp

New Member
Joined
May 15, 2016
Messages
1 (0.00/day)
Here's some of mine. Both are Nano. (took them yesterday before others started to check theirs "the reason why the percentage is still so high compared to others").

As mentioned over @ OCN, Nano's seem to show lower Asic quality than Fury (x)'s..

Schermafbeelding 2016-05-13 om 19.39.18.png
Schermafbeelding 2016-05-13 om 19.40.39.png


Both cards run stable with an undervolt of -100mV at stock core clock 1000 but with the memory decreased to 300.

Greetings
 
Top