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

Possibly bricked RX 5700?

Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
If you do, be sure to use flux.
Interestingly I was researching earlier and found a video exactly about that! Thanks for the reassurance! :)

still don't believe for a second you using this card is what borked it. I think the seller likely knew it was problematic but "hey it's working now so I'll just keep my fingers crossed it keeps working long enough to sell and wash my hands of it" and you were the one to happen to purchase it. Which if it is a standard 5700 with an XT BIOS could be a contributing factor....again...that is nothing you did wrong. You are only guilty of buying from a non-reputable/trusted source.

Yea I would agree if you are at your end of choices for troubleshooting it and the GPU is not useable, then anything is worth a short at this point. Good luck. If it doesn't work........Pour one out for our fellow GPU homie.....
Animated GIF
The beer is ready, along with a symbolic coffin for my 700€ (jk).

Thanks for the reassurance. Even thought I did submit a report to the police since this is a scam by all means in several ways (XT advertised, non-XT sold.. money back if not working well guaranteed...) I think I'm done with the guy and will just drop the charge.

Hopefully at least the online marketplace does something. The time and costs it can bring are not worth it the 700€. From a perspective of a part-time miner it sucks, but I'm fortunate to earn enough from work to afford making mistakes or I wouldn't risk it, so I'll just man up, take the loss and try to study better the cards I'm buying and stress test them properly.

Anyway, will update soon then!
 
Last edited:

eidairaman1

The Exiled Airman
Joined
Jul 2, 2007
Messages
33,518 (6.30/day)
Location
Republic of Texas (True Patriot)
System Name PCGOD
Processor AMD FX 8350@ 5.0GHz
Motherboard Asus TUF 990FX Sabertooth R2 2901 Bios
Cooling Scythe Ashura, 2×BitFenix 230mm Spectre Pro LED (Blue,Green), 2x BitFenix 140mm Spectre Pro LED
Memory 16 GB Gskill Ripjaws X 2133 (2400 OC, 10-10-12-20-20, 1T, 1.65V)
Video Card(s) AMD Radeon 290 Sapphire Vapor-X
Storage Samsung 840 Pro 256GB, WD Velociraptor 1TB
Display(s) NEC Multisync LCD 1700V (Display Port Adapter)
Case AeroCool Xpredator Evil Blue Edition
Audio Device(s) Creative Labs Sound Blaster ZxR
Power Supply Seasonic 1250 XM2 Series (XP3)
Mouse Roccat Kone XTD
Keyboard Roccat Ryos MK Pro
Software Windows 7 Pro 64
Interestingly I was researching earlier and found a video exactly about that! Thanks for the reassurance! :)


The beer is ready, along with a symbolic coffin for my 700€ (jk).

Thanks for the reassurance. Even thought I did submit a report to the police since this is a scam by all means in several ways (XT advertised, non-XT sold.. money back if not working well guaranteed...) I think I'm done with the guy and will just drop the charge.

Hopefully at least the online marketplace does something. The time and costs it can bring are not worth it the 700€. From a perspective of a part-time miner it sucks, but I'm fortunate to earn enough from work to afford making mistakes or I wouldn't risk it, so I'll just man up, take the loss and try to study better the cards I'm buying and stress test them properly.

Anyway, will update soon then!
Use ebay or amazon for pc parts
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
its-alive.jpg


well, kinda :D (TL;DR at the end @eidairaman1)

So... super nice guy from the hardware shop did the reflow (yes he used flux). Tried to POST with it after, nothing. Sad day I thought, gave up on the 700eur..

EXCEPT I'm a stubborn motherFer and decided to see if I got any results on my systems.

It still didn't POST, but I booted with it as secondary to see if I could still see the contents of the Bios and get a gpuz reading of the GPU after the reflow. Everything seemed still OK so I tried installing drivers. It didn't crash like before and AMD showed the correct model according to the bios in it. Still wouldn't post as primary PEG though!

One thing hasn't stop bothering me, maybe it's normal, but the bios chip has 1024kb but the supposed correct rom is 512kb.. So I decided to try to flash the bios it came with (the XT one) which was indeed 1024kb one last time (had done it before with no luck). IT POSTS! Now, this is still not a XT but it seems to not be the Sapphire Reference card either, at least I can't find a 1024kb version for the Sapphire card anywhere.

If anyone who knows 5700 PCBs well can take a look at the uploaded pics in the previous posts and see if they recognize it, I would really appreciate it to see if I can get this card the correct bios for once.

Didn't do any sort of stress tests yet, but I thought this was already worthy of sharing..! It POSTS and there's display output!
Will now see if it can handle 3D and post updates!


TL;DR:
- reflow done
- didn't post anyway as primary
- still posted as secondary and this time drivers installed! card recognized by radeon and not disabled by windows!
- still no POST
- flashed XT bios it came with, POST!!!
- now need some help figuring out which card this is, since no "Sapphire reference" 5700 out there has a 1024kb which seems to be what did the trick.

GPUZ with the XT bios and drivers installed:

gpuz5700xt.gif


Hopes are still none though, all monitoring software is showing the core clock between 5 and 25MHz, not a good sign. Let's run some tests x)

I'm speechless (printscreen came out black but the graphics were there, didn't crash)
1642169163874.png



core is super underclocked for a starter but will not slowly increase and post results :D what the heck
 
Last edited:

Tatty_One

Senior Moder@tor
Staff member
Joined
Jan 18, 2006
Messages
24,498 (4.19/day)
Location
Worcestershire, UK
System Name The Grey
Processor Intel Comet Lake Core i5 10600K
Motherboard MSI MAG Z490 TOMAHAWK
Cooling Artic Cooling Liquid Freezer II 280mm AIO + 4 Phantek 140mm case fans
Memory 16GB Patriot Viper Steel DDR4 4133Mhz @ 17-17-17-32 2T 1.45v
Video Card(s) Gigabyte RTX 2060 SUPER Windforce OC 8GB
Storage WD Blue SN550 1TB M.2 NVME//SkHynix 120GB SSD//CrucialM4 64GB SSD
Display(s) AOC Q2781PQ 27 inch Ultra Slim 2560 x 1440 IPS
Case Phanteks Enthoo Pro M Windowed - Gunmetal
Audio Device(s) OnBoard Realtek ALC1200-VD1 @ 5.1 surround
Power Supply Seasonic Core Gold 650w Semi modular
Mouse Coolermaster Storm Octane wired
Keyboard Element Gaming Carbon Mk2 Tournament Mech
Software Win 10 Home x64
It seems you are getting there. No I couldn't see any XT Sapphire reference designs either, plenty of Asrock's with Micron support, I think you mentioned you tried one of those in any case.
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
It seems you are getting there. No I couldn't see any XT Sapphire reference designs either, plenty of Asrock's with Micron support, I think you mentioned you tried one of those in any case.
non-XT that is! I mean, now that I know the GPU itself is good and that I have a store near me that can flash externally if needed, I'm up for trying suggested bioses. I've tried plenty reference model before, but none were 1024kb. Card was bricked though, but 100% sure it's not the Sapphire Reference bios, at least not the ones out there, they are all 512kb. Or maybe it's just in the wrong format, GPUZ also exports the XT bios as 512, while atiflash exported it as 1024kb.. confusing.

The GPU is handling 3d just fine, up to stock clock. It also mines fine BUT the mem temps are insane while mining, I have to stop it cause it reaches 100C in a few minutes (even with fan at 100%, which is one of the noisiest i've ever heard). I'll see if I can use it as a gaming gpu for more than a few days, if it lasts I'll probably buy a cheap-but-way-better heatsink for this card. The missing backplate is also playing a big role in the issue, pretty sure the mems dissipate a lot of their heat from the back of the PCB right?
 

Tigger

I'm the only one
Supporter
Joined
Mar 20, 2006
Messages
13,797 (2.38/day)
System Name Dirty Fecker
Processor Intel I7 12700K 5ghz-P 3.9ghz-E 1.25vcore
Motherboard Asus Rog Strix Z690-A WiFi D4
Cooling EK 360 PE+EK 280 CE, FLT 240 D5, EK supremacy classic, Generic GPU block
Memory Teamgroup 8 Pack Edition T-Force Dark Pro 2x16gb 3735 16-16-16-38 Bdie
Video Card(s) MSI GTX980TI Gaming 6gb, 1475/8000 water cooled
Storage Silicon Power Gen3x4 256gb M.2, WD SN850 1TB M.2 NVME Gen4x4
Display(s) MSI 32" Optix AG32CV 1080p 165hz 1ms
Case Lian Li o11 Dynamic XL Rog
Audio Device(s) onboard/Logitech Z4
Power Supply Corsair TX750M 80% Gold
Mouse Razer Viper
Keyboard Anne pro II
Software Win 11 Pro x64
Benchmark Scores do they matter
Interestingly I was researching earlier and found a video exactly about that! Thanks for the reassurance! :)


The beer is ready, along with a symbolic coffin for my 700€ (jk).

Thanks for the reassurance. Even thought I did submit a report to the police since this is a scam by all means in several ways (XT advertised, non-XT sold.. money back if not working well guaranteed...) I think I'm done with the guy and will just drop the charge.

Hopefully at least the online marketplace does something. The time and costs it can bring are not worth it the 700€. From a perspective of a part-time miner it sucks, but I'm fortunate to earn enough from work to afford making mistakes or I wouldn't risk it, so I'll just man up, take the loss and try to study better the cards I'm buying and stress test them properly.

Anyway, will update soon then!

non-XT that is! I mean, now that I know the GPU itself is good and that I have a store near me that can flash externally if needed, I'm up for trying suggested bioses. I've tried plenty reference model before, but none were 1024kb. Card was bricked though, but 100% sure it's not the Sapphire Reference bios, at least not the ones out there, they are all 512kb. Or maybe it's just in the wrong format, GPUZ also exports the XT bios as 512, while atiflash exported it as 1024kb.. confusing.

The GPU is handling 3d just fine, up to stock clock. It also mines fine BUT the mem temps are insane while mining, I have to stop it cause it reaches 100C in a few minutes (even with fan at 100%, which is one of the noisiest i've ever heard). I'll see if I can use it as a gaming gpu for more than a few days, if it lasts I'll probably buy a cheap-but-way-better heatsink for this card. The missing backplate is also playing a big role in the issue, pretty sure the mems dissipate a lot of their heat from the back of the PCB right?

Shame you don't have a custom loop, you could probably easily get a nice full cover for that now pretty cheap
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
Shame you don't have a custom loop, you could probably easily get a nice full cover for that now pretty cheap
I could have it for free even, the guy offered me the water block for free... But yeah, I live in a cold country, air is just fine even with OCs :D I know it doesn't look as pretty but saves me a few bucks and headaches :D

BTW if it helps, here's the XT bios it's currently working perfectly fine with (attachment), except for the mem temps..
The closest one non-xt I can find with 1024kb is this:

Update: opened GPU to remove thermal pads, they were too thin and weren't helping cause no decent contact was made. Plug it back in, bricked again..?? I'm quite confused, the card was behaving absolutely normal before and I did several reboots with it to ensure it wasn't just a lucky single POST. Heaven benchmark ran for 10 min straight with clock at stock settings too, not a single indication that it would brick right away again, no artifacts, no stuttering, nothing.

I'd understand if it would run the benchmark and die but this is just weird. Might do a full alcohol rub, maybe there's something conductive somewhere it shouldn't? wtf.
 

Attachments

  • 5700xt_original.rom
    1 MB · Views: 3
Last edited:
Joined
Mar 21, 2021
Messages
1,558 (5.06/day)
Location
Colorado, U.S.A.
System Name HP Compaq 8000 Elite CMT
Processor Intel Core 2 Quad Q9550
Motherboard Hewlett-Packard 3647h
Memory 16GB DDR3
Video Card(s) NVIDIA GeForce GT 1030 GDDR5 (fan-less)
Storage 2TB Seagate Firecuda 3.5"
Display(s) Dell P2416D (2560 x 1440)
Power Supply 12V HP proprietary
Software Windows 10 Pro 64-bit
Did you get both the RAM and GPU reflowed?

Short term recovery is typical of heating without proper reflow.
 
Last edited:
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
Did you get both the RAM and GPU reflowed?
Not sure, I only saw it from a distance while he did it.
Did you get both the RAM and GPU reflowed?

Short term recovery is typical of heating without proper reflow.
I understand that but typically also they won't go from 100 to 0 for removing and reinstalling the card..

I'm not superstitious but since it was on my wife's pc I was able to install drivers and POST earlier today, I went back there and it worked again. For some bloody reason her system doesn't crash installing drivers and as soon as it does the display becomes available via the card. It seems that as soon as the card gets some juice it starts working normally again... about 30sec after the drivers finish installing you can hear the fan changing speeds as Radeon software picks it up.

If this exact thing happened in any of my other 2 systems I would label as just pure random/luck and instable anyway. But it has worked 2 for 2 in my wife's and 0 for 3 in mine and 0 for 2 in my testbench. Weird as f.
 
Joined
Mar 21, 2021
Messages
1,558 (5.06/day)
Location
Colorado, U.S.A.
System Name HP Compaq 8000 Elite CMT
Processor Intel Core 2 Quad Q9550
Motherboard Hewlett-Packard 3647h
Memory 16GB DDR3
Video Card(s) NVIDIA GeForce GT 1030 GDDR5 (fan-less)
Storage 2TB Seagate Firecuda 3.5"
Display(s) Dell P2416D (2560 x 1440)
Power Supply 12V HP proprietary
Software Windows 10 Pro 64-bit
I understand that but typically also they won't go from 100 to 0 for removing and reinstalling the card..

A bad solder joint can do that
 
Joined
Feb 20, 2019
Messages
3,561 (3.33/day)
System Name Bragging Rights
Processor Atom Z3735F 1.33GHz
Motherboard It has no markings but it's green
Cooling No, it's a 2.2W processor
Memory 2GB DDR3L-1333
Video Card(s) Gen7 Intel HD (4EU @ 311MHz)
Storage 32GB eMMC and 128GB Sandisk Extreme U3
Display(s) 10" IPS 1280x800 60Hz
Case Veddha T2
Audio Device(s) Apparently, yes
Power Supply Samsung 18W 5V fast-charger
Mouse MX Anywhere 2
Keyboard Logitech MX Keys (not Cherry MX at all)
VR HMD Samsung Oddyssey, not that I'd plug it into this though....
Software W10 21H1, barely
Benchmark Scores I once clocked a Celeron-300A to 564MHz on an Abit BE6 and it scored over 9000.
This might be a dead end but if you have it working properly in Windows try comparing the BIOS that works with the BIOS that it's supposed to have in RBE and MPT

I have 24 of your cards and used those tools to edit my BIOSes for mining. Between Christmas and New Years I had an issue across all rigs that was to do with a Microsoft driver for the 5700-series ignoring preferences to not update hardware drivers automatically, and being incompatible with the Radeon Settings, all whilst cluttering up the registry with repeated entries for GPUs.

In the end I had to seriously lock down Windows Update, kill the softwaredistribution folder, DDU in safe mode, and install working Radeon Drivers from AMD all whilst offline to combat it. I'd still get sudden blackscreens on rigs where I was being lazy and didn't clear out the registry because the shitty driver Microsoft was trying to install messed with ultra low power states for the cards. It can be fixed after the fact by going through the registry mashing F3 and changing around 200 values but I wish I'd taken the extra 5 minutes to do it properly in the first place.
 

eidairaman1

The Exiled Airman
Joined
Jul 2, 2007
Messages
33,518 (6.30/day)
Location
Republic of Texas (True Patriot)
System Name PCGOD
Processor AMD FX 8350@ 5.0GHz
Motherboard Asus TUF 990FX Sabertooth R2 2901 Bios
Cooling Scythe Ashura, 2×BitFenix 230mm Spectre Pro LED (Blue,Green), 2x BitFenix 140mm Spectre Pro LED
Memory 16 GB Gskill Ripjaws X 2133 (2400 OC, 10-10-12-20-20, 1T, 1.65V)
Video Card(s) AMD Radeon 290 Sapphire Vapor-X
Storage Samsung 840 Pro 256GB, WD Velociraptor 1TB
Display(s) NEC Multisync LCD 1700V (Display Port Adapter)
Case AeroCool Xpredator Evil Blue Edition
Audio Device(s) Creative Labs Sound Blaster ZxR
Power Supply Seasonic 1250 XM2 Series (XP3)
Mouse Roccat Kone XTD
Keyboard Roccat Ryos MK Pro
Software Windows 7 Pro 64
Not sure, I only saw it from a distance while he did it.

I understand that but typically also they won't go from 100 to 0 for removing and reinstalling the card..

I'm not superstitious but since it was on my wife's pc I was able to install drivers and POST earlier today, I went back there and it worked again. For some bloody reason her system doesn't crash installing drivers and as soon as it does the display becomes available via the card. It seems that as soon as the card gets some juice it starts working normally again... about 30sec after the drivers finish installing you can hear the fan changing speeds as Radeon software picks it up.

If this exact thing happened in any of my other 2 systems I would label as just pure random/luck and instable anyway. But it has worked 2 for 2 in my wife's and 0 for 3 in mine and 0 for 2 in my testbench. Weird as f.
Well the problem is your specific parts
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
Well the problem is your specific parts
Hard to believe. I've done more than enough switcharoos to make sure it's not the case. Is it possible her MB is so old it doesnt have the security features to prevent the system from booting with a currupt pci-e device? (it's like a 4th gen intel MB).


This might be a dead end but if you have it working properly in Windows try comparing the BIOS that works with the BIOS that it's supposed to have in RBE and MPT

I have 24 of your cards and used those tools to edit my BIOSes for mining. Between Christmas and New Years I had an issue across all rigs that was to do with a Microsoft driver for the 5700-series ignoring preferences to not update hardware drivers automatically, and being incompatible with the Radeon Settings, all whilst cluttering up the registry with repeated entries for GPUs.

In the end I had to seriously lock down Windows Update, kill the softwaredistribution folder, DDU in safe mode, and install working Radeon Drivers from AMD all whilst offline to combat it. I'd still get sudden blackscreens on rigs where I was being lazy and didn't clear out the registry because the shitty driver Microsoft was trying to install messed with ultra low power states for the cards. It can be fixed after the fact by going through the registry mashing F3 and changing around 200 values but I wish I'd taken the extra 5 minutes to do it properly in the first place.
Appears simillar at first sight, but unfortunately the problem seems to be in the card and not the systems. You could always POST, the problem happened later in windows. But in my case it seems the card gets soft bricked because it won't even POST. I say "soft" only because so far my wife's pc magically recovers it. The automatic driver installation from windows actually works fine in her pc too, in the others its disabled already (testbench has a bunch of disabled windows features including auto updates, gaming PC has automatic driver updates off too).

The only non consipiracy theory appart from just luck together with still bad solder joints as @Andy Shiekh pointed out, is that maybe the card had a huge sag and with constant overheated it could have distorted the PCB ever so slightly to the point that it only makes proper contacy of slightly bent. I say this because the only difference in the way I tried things out in my wife's PC and the rest is that I was lazy and didn't screw in the card holder plate, I just used a riser and it was upright in the table, but risers can't hold a gpu upright by itself, so it gets a bit of an angle (about the same you'd get from GPU sag). Might sound crazy but it's the only 100% correlation so far.

It's finally weekend, I'll do a bunch more tests cause this is driving me insane

PS: not sure if this is a symptom or completely normal, but when the card is bad and I plug it into the "miracle pc", the card sensors in GPUz show 3GB of mem being used, even though the card isn't doing shit (not even display). Once the drivers install and windows detects the card as a 5700, the mem usage goes down to below 10MB as expected. Thought I'd share in case it points out to a brickedvmemory controller or so..?
 
Joined
Feb 20, 2019
Messages
3,561 (3.33/day)
System Name Bragging Rights
Processor Atom Z3735F 1.33GHz
Motherboard It has no markings but it's green
Cooling No, it's a 2.2W processor
Memory 2GB DDR3L-1333
Video Card(s) Gen7 Intel HD (4EU @ 311MHz)
Storage 32GB eMMC and 128GB Sandisk Extreme U3
Display(s) 10" IPS 1280x800 60Hz
Case Veddha T2
Audio Device(s) Apparently, yes
Power Supply Samsung 18W 5V fast-charger
Mouse MX Anywhere 2
Keyboard Logitech MX Keys (not Cherry MX at all)
VR HMD Samsung Oddyssey, not that I'd plug it into this though....
Software W10 21H1, barely
Benchmark Scores I once clocked a Celeron-300A to 564MHz on an Abit BE6 and it scored over 9000.
Ah, okay.
That does sound well and truly like broken hardware I'm afraid.

Sucks to get ripped off.
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
Hard to believe. I've done more than enough switcharoos to make sure it's not the case. Is it possible her MB is so old it doesnt have the security features to prevent the system from booting with a currupt pci-e device? (it's like a 4th gen intel MB).



Appears simillar at first sight, but unfortunately the problem seems to be in the card and not the systems. You could always POST, the problem happened later in windows. But in my case it seems the card gets soft bricked because it won't even POST. I say "soft" only because so far my wife's pc magically recovers it. The automatic driver installation from windows actually works fine in her pc too, in the others its disabled already (testbench has a bunch of disabled windows features including auto updates, gaming PC has automatic driver updates off too).

The only non consipiracy theory appart from just luck together with still bad solder joints as @Andy Shiekh pointed out, is that maybe the card had a huge sag and with constant overheated it could have distorted the PCB ever
Yeah so the 1024kb XT BIOS that came with the card is an exact replica of the 512kb from Sapphire and other reference model bioses when it comes to configured memory vendors and their timings. The difference is only visible in More Power tools, the XT obviously has higher max clock settings. Interestingly, the XT memory voltage can go dow to 750 mV while the non-XT can only go down to 800 mV (highlighting in case someone thinks this could be A cause). So, to get a good MPT file I replaced the values from the XT bios with the non-xt, saved the MPT file. Iin RBE I will import the MPT file and also change the boot message so it's seen as non-XT too (non that I think that's the issue but why not). Let's see if flashing this bios brings some consistency.

And BTW I am now back on my testbench and the card is working perfectly fine again... Let's see how long until a crash + brick + miraculous PC loop comes around X)

Update:
Flashing the bios as described fails and corrupts it. Once more my wife's PC has no issues with it, even with a clearly corrupt bios and is able to flash it with via GUI and running normally again.

Can someone tell me what I did wrong? I literally just used the correct power and clock settings for the non-XT in MPT and imported that into RBE with the xt rom loaded (since the timings were the same and itºs the only 1024kb version) and changed the boot message to match the non-xt as well. Uploading both the XT bios that works and the one that doesn't after the MPT change. I really want to try to get this card with the BIOS it should be to see if the inconsistency in behavior improves.

On a separate note, can anyone with Reference 5700s confirm the bios chip size? Just curious if anyone else has a 5700 non-XT with a 1024kb bios chip. @Chrispy_ ?

Edit: I think I got it. with atiflash 2.93+ it does work. So far I never needed to use the + version in her machine, but I guess by flashing the non-xt i'm changing the P/N so it requires a force flash, which only works for 5700s especifically in 2.93+ (the latest + version doesn't allow force flashing).
Will now plug it back to the testbench and leave it running, let's see.

From testbench:
1642245684654.png
 

Attachments

  • 5700xt_moded_to_nonxt_with_powerplay.rom
    1 MB · Views: 2
  • 5700xt_original.rom
    1 MB · Views: 3
Last edited:
Joined
Feb 20, 2019
Messages
3,561 (3.33/day)
System Name Bragging Rights
Processor Atom Z3735F 1.33GHz
Motherboard It has no markings but it's green
Cooling No, it's a 2.2W processor
Memory 2GB DDR3L-1333
Video Card(s) Gen7 Intel HD (4EU @ 311MHz)
Storage 32GB eMMC and 128GB Sandisk Extreme U3
Display(s) 10" IPS 1280x800 60Hz
Case Veddha T2
Audio Device(s) Apparently, yes
Power Supply Samsung 18W 5V fast-charger
Mouse MX Anywhere 2
Keyboard Logitech MX Keys (not Cherry MX at all)
VR HMD Samsung Oddyssey, not that I'd plug it into this though....
Software W10 21H1, barely
Benchmark Scores I once clocked a Celeron-300A to 564MHz on an Abit BE6 and it scored over 9000.
On a separate note, can anyone with Reference 5700s confirm the bios chip size? Just curious if anyone else has a 5700 non-XT with a 1024kb bios chip. @Chrispy_ ?

All mine are 1024kb BIOS files, I use AMDVBFLASH to save and write BIOS files to the card. GPU-Z isn't recommended for 5700-series BIOS saving, IIRC there's an issue with the size of the files it saves. As for which version of AMDVBFLASH, it was whatever version was recommended by Igor's Lab. 2.93+ sounds familiar....

Here's dump of my 24 original BIOS files, I suspect they are all identical but feel free to try them. Sadly I didn't make a note of which ones are the Powercolor pallet of 20, and which ones are Gigabyte/Sapphire retail - I don't think it matters as all the reference cards are BBA and manufactured by PC Partner, regardless of who slaps their sticker on the fan hub.

Edit:
There are three distinct BIOSes in there, since they compress down to 3 unique filesizes. The only three worth trying (if you need to) are in Miner1 adapters 4, 5, 6:

1642251408829.png
 

Attachments

  • Original RX5700 BIOS Files.zip
    4.6 MB · Views: 4
Last edited:
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
All mine are 1024kb BIOS files, I use AMDVBFLASH to save and write BIOS files to the card. GPU-Z isn't recommended for 5700-series BIOS saving, IIRC there's an issue with the size of the files it saves. As for which version of AMDVBFLASH, it was whatever version was recommended by Igor's Lab. 2.93+ sounds familiar....

Here's dump of my 24 original BIOS files, I suspect they are all identical but feel free to try them. Sadly I didn't make a note of which ones are the Powercolor pallet of 20, and which ones are Gigabyte/Sapphire retail - I don't think it matters as all the reference cards are BBA and manufactured by PC Partner, regardless of who slaps their sticker on the fan hub.

Edit:
There are three distinct BIOSes in there, since they compress down to 3 unique filesizes. The only three worth trying (if you need to) are in Miner1 adapters 4, 5, 6:

View attachment 232481
Thanks for that, will try them and report back.

The memory heating problem is clearly a backplate missing issue. A simple small fan pointed directly at the back of the PCB makes the mem temps not go past 96-98. It's still shit, but it's getting better and well below the 105 max temp.

Also ordered new, proper thermal pads with different thicknesses and will try re-padding next week when they arrive. The only pads currently in the card are the memory chip ones which are clearly not doing shit, if anything they are making it worse. Everything else was removed to see if the mem pads would make better contact but clearly still not the case, the heatsink doesn't even get that hot where the mems are but the backplate does, so clearly the mems are dissipating more heat through the back than the heatsink, telling me the pads are shit.

All mine are 1024kb BIOS files, I use AMDVBFLASH to save and write BIOS files to the card. GPU-Z isn't recommended for 5700-series BIOS saving, IIRC there's an issue with the size of the files it saves. As for which version of AMDVBFLASH, it was whatever version was recommended by Igor's Lab. 2.93+ sounds familiar....

Here's dump of my 24 original BIOS files, I suspect they are all identical but feel free to try them. Sadly I didn't make a note of which ones are the Powercolor pallet of 20, and which ones are Gigabyte/Sapphire retail - I don't think it matters as all the reference cards are BBA and manufactured by PC Partner, regardless of who slaps their sticker on the fan hub.

Edit:
There are three distinct BIOSes in there, since they compress down to 3 unique filesizes. The only three worth trying (if you need to) are in Miner1 adapters 4, 5, 6:

View attachment 232481
Compared them with the one currently in my card, using `amdvbflash -biosfileinfo romname.rom`. They look exactly the same so don't think there's a point on flashing them.
What did you do to improve mem timings, is it just like the old Polaris cards? Copy 1500 timings to 1750 and above and done? I'm curious if better timmings will do any difference in the mem temps.
 
Joined
Feb 20, 2019
Messages
3,561 (3.33/day)
System Name Bragging Rights
Processor Atom Z3735F 1.33GHz
Motherboard It has no markings but it's green
Cooling No, it's a 2.2W processor
Memory 2GB DDR3L-1333
Video Card(s) Gen7 Intel HD (4EU @ 311MHz)
Storage 32GB eMMC and 128GB Sandisk Extreme U3
Display(s) 10" IPS 1280x800 60Hz
Case Veddha T2
Audio Device(s) Apparently, yes
Power Supply Samsung 18W 5V fast-charger
Mouse MX Anywhere 2
Keyboard Logitech MX Keys (not Cherry MX at all)
VR HMD Samsung Oddyssey, not that I'd plug it into this though....
Software W10 21H1, barely
Benchmark Scores I once clocked a Celeron-300A to 564MHz on an Abit BE6 and it scored over 9000.
5700 reference cards never had backplates. repadding is a good idea though, you want VRAM temps to remain under 95 ideally. I mine at 60C hotspot and 80C VRAM, and ETH mining is pretty hard on the VRAM. I've repadded a couple that had higher VRAM temps than the rest - used Gelid Ultimate 1.5mm but I didn't swap the hitachi GPU pad with paste; If you do that you should use 1.0mm pads on the VRAM.

Don't mess with the timings unless you're going to mine with it. The default 5700 BIOS is the most stable for gaming - mining timings are looser and worse for gaming by a long shot.
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
5700 reference cards never had backplates. repadding is a good idea though, you want VRAM temps to remain under 95 ideally. I mine at 60C hotspot and 80C VRAM, and ETH mining is pretty hard on the VRAM. I've repadded a couple that had higher VRAM temps than the rest - used Gelid Ultimate 1.5mm but I didn't swap the hitachi GPU pad with paste; If you do that you should use 1.0mm pads on the VRAM.

Don't mess with the timings unless you're going to mine with it. The default 5700 BIOS is the most stable for gaming - mining timings are looser and worse for gaming by a long shot.
Yeah currently the card has paste on the GPU, but I still think the pads are a bit too thin, maybe they are 0.5mm... it's not easy to tell the difference without a side by side comparison and they are old. Probably bad quality as well. I paid a bit more to get better quality than the regular Gelid ones (Cooler Master Thermal Pro), maybe the "Ultimate" ones are decent but the ones with fast delivery available were pretty shit. They are far from the best but are for sure better than what's in there. Seems they will arrive today so I can report back on the mem temps after applying them :)

I understand mining is a lot harder on the VRAM, at least ethash, I mined back in 2014 and 2018, which is why my testbench currently is a mining rig x) Have no use for it so why not a nice, roomy and open air testbench? :laugh: Thing is if the card proves to be as unstable and inconsistent as it seems, I'm not sure I want to leave this as a gaming GPU for my wife's and risk voltage issues Fing up other parts of the system, so I might actually see if it can hold mining since the now-testbench parts were all cheap except the MB and PSU and they have paid for themselves already, plus I have no other use for it as opposed to the wife constantly nagging that I broke her PC :laugh:

With that said, do you mod the timings the exact same way you would in Polaris GPUs? I wanna see if simply better timings at 1750mhz *default mem clock) helps or worsens the temps.
 
Joined
Feb 20, 2019
Messages
3,561 (3.33/day)
System Name Bragging Rights
Processor Atom Z3735F 1.33GHz
Motherboard It has no markings but it's green
Cooling No, it's a 2.2W processor
Memory 2GB DDR3L-1333
Video Card(s) Gen7 Intel HD (4EU @ 311MHz)
Storage 32GB eMMC and 128GB Sandisk Extreme U3
Display(s) 10" IPS 1280x800 60Hz
Case Veddha T2
Audio Device(s) Apparently, yes
Power Supply Samsung 18W 5V fast-charger
Mouse MX Anywhere 2
Keyboard Logitech MX Keys (not Cherry MX at all)
VR HMD Samsung Oddyssey, not that I'd plug it into this though....
Software W10 21H1, barely
Benchmark Scores I once clocked a Celeron-300A to 564MHz on an Abit BE6 and it scored over 9000.
ETH mining is looser timings, yeah. Run the VRAM as fast as it's stable and increase tref as far as you can without losing stability.


Memory Clock - tREF
1000 MHz - 3900 (Samsung & Micron)
1250 MHz - 4875 (Samsung & Micron)
1550 MHz - 6045 (Samsung & Micron)
1750 MHz - 6825 (Micron only)
1800 MHz - 7020 (Samsung & Micron)
1875 MHz - 7315 (Micron only)
2000 MHz - 7800 (Samsung & Micron)
2250 MHz - 8775 (Samsung only)
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
ETH mining is looser timings, yeah. Run the VRAM as fast as it's stable and increase tref as far as you can without losing stability.


Memory Clock - tREF
1000 MHz - 3900 (Samsung & Micron)
1250 MHz - 4875 (Samsung & Micron)
1550 MHz - 6045 (Samsung & Micron)
1750 MHz - 6825 (Micron only)
1800 MHz - 7020 (Samsung & Micron)
1875 MHz - 7315 (Micron only)
2000 MHz - 7800 (Samsung & Micron)
2250 MHz - 8775 (Samsung only)
Ok so proper thermal pads fixed the temps issue. Now same or lower than core while gaming and while mining core @ 62C, hotspot 74C, mems 78-82. Seems close enough to your numbers, so I didn't do an awful job with the mem paddings.

Sadly the card can't handle more than 5min mining so far, and I'm not OCing, just undervolting the core as is usual for ETH mining. But at least it hasn't gone back to that bricked state in a while, windows catches the GPU error and restarts and PC still POSTS after. I didn't try more than 5 min gaming but it's probably going to be the same. If it is then @Andy Shiekh is probably correct and the relow was only partial and a one or many solder joints still remain improper, resulting in inconsistency. But If it is not then I don't know but probably the mem chips are fucked.
 
Joined
Mar 21, 2021
Messages
1,558 (5.06/day)
Location
Colorado, U.S.A.
System Name HP Compaq 8000 Elite CMT
Processor Intel Core 2 Quad Q9550
Motherboard Hewlett-Packard 3647h
Memory 16GB DDR3
Video Card(s) NVIDIA GeForce GT 1030 GDDR5 (fan-less)
Storage 2TB Seagate Firecuda 3.5"
Display(s) Dell P2416D (2560 x 1440)
Power Supply 12V HP proprietary
Software Windows 10 Pro 64-bit
Last edited:
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
I thought people under-clock for mining.

If the RAM was not reflowed, that might be worth a try.

The reflows I tried by hand seem to have worked
Video Card Reflow | Electronics Repair And Technology News Video Card Reflow (jestineyong.com)
but I would not do this by hand on a card worth more than $10
For ETH mining you underclock and undervolt the core yes. But the mems are hammered. Nevertheless, temps are not the issue anymore, unless it's a spot without a sensor. Might buy a heat gun and flux to try the mems myself, didn't know they were that cheap tbh. And yeah I know it's not advisable for a non-throw away card but what else would you advice? Don't know anyone who can do a proper rebail nor do I feel comfortable doing one.

ETH mining is looser timings, yeah. Run the VRAM as fast as it's stable and increase tref as far as you can without losing stability.


Memory Clock - tREF
1000 MHz - 3900 (Samsung & Micron)
1250 MHz - 4875 (Samsung & Micron)
1550 MHz - 6045 (Samsung & Micron)
1750 MHz - 6825 (Micron only)
1800 MHz - 7020 (Samsung & Micron)
1875 MHz - 7315 (Micron only)
2000 MHz - 7800 (Samsung & Micron)
2250 MHz - 8775 (Samsung only)
It's not much, but it's honest work.
survivor.jpg


So with the stock BIOS it seems to hold Heaven benchmark on Ultra 1080p for almost 20min (I stopped it but it didn't crash). But it couldn't mine for more than 5min. Followed a decent guide I found to use MorePowerTool (this didn't exist back in the polaris era, at least it wasn't really needed, so I wasn't familiar with it) and adjusted the timmings (just the 1500 timmings applied to 1750 and up, this card has micron mems). It seems to want to mine now for more than 10min, let's see how long. But this is already better than before. Will also test tomorrow 3d graphics for more than 20min. If it can handle graphics fine but not mining, that's fine by me.

PS: I know this card can do more for less watts, but I won't do anything that can cause the slightest instability until I see it running for a long time like this.
 
Last edited:
Joined
Mar 21, 2021
Messages
1,558 (5.06/day)
Location
Colorado, U.S.A.
System Name HP Compaq 8000 Elite CMT
Processor Intel Core 2 Quad Q9550
Motherboard Hewlett-Packard 3647h
Memory 16GB DDR3
Video Card(s) NVIDIA GeForce GT 1030 GDDR5 (fan-less)
Storage 2TB Seagate Firecuda 3.5"
Display(s) Dell P2416D (2560 x 1440)
Power Supply 12V HP proprietary
Software Windows 10 Pro 64-bit
I'd ask the guy who reflowed for you if he included the RAM
 
Joined
Jan 10, 2022
Messages
48 (3.69/day)
Location
Noord-Hollands, Netherlands
I'd ask the guy who reflowed for you if he included the RAM
Will definitely do that, thanks! Can only catch him in the store Monday now, but I will wait for that before I do anything else with the card. The heatgun and flux is ordered just in case though.. If I knew they were this cheap.. :)

ETH mining is looser timings, yeah. Run the VRAM as fast as it's stable and increase tref as far as you can without losing stability.


Memory Clock - tREF
1000 MHz - 3900 (Samsung & Micron)
1250 MHz - 4875 (Samsung & Micron)
1550 MHz - 6045 (Samsung & Micron)
1750 MHz - 6825 (Micron only)
1800 MHz - 7020 (Samsung & Micron)
1875 MHz - 7315 (Micron only)
2000 MHz - 7800 (Samsung & Micron)
2250 MHz - 8775 (Samsung only)
1hr mining and counting. Hard to believe but the better mem timings make it not crash.. yet at least. I'm not allowing myself to celebrate until this runs for 3 days stable :laugh:
 
Top