I got this X1950 XTX that is in physical "like new" condition. No missing components. It appears to be working perfectly in games and benchmarks BUT when the core reaches somewhere between 73c and 78c, it crashes. I ran R5MEMID and it passes perfectly.
The attachment 5.JPG is no longer available
The attachment 1.JPG is no longer available
The attachment 2.JPG is no longer available
Sometimes it will blue screen and sometimes it will go back to Windows with ATI's VPU recovery.
Interestingly your GPU-Z Screenshot shows the Fan not changing it's speed? It looks like it's fixed at 13%?
Btw. what a lovely card. Fond memories of me looking through magazines and imagining how awesome it would be to have this card. I had an X700 AGP at the time 😁
Can you monitor the VRM temperatures and VDDC externally?
Btw. what a lovely card. Fond memories of me looking through magazines and imagining how awesome it would be to have this card. I had an X700 AGP at the time 😁
Can you monitor the VRM temperatures and VDDC externally?
She’s beautiful!
I can use a thermal cam and possibly a multimeter. Will get back to you.
FYI, forgot to mention I replaced the core thermal paste and all the memory thermal pads but still got the same issue. I have some new cores on the way, just in case 😉
Interestingly your GPU-Z Screenshot shows the Fan not changing it's speed? It looks like it's fixed at 13%?
Btw. what a lovely card. Fond memories of me looking through magazines and imagining how awesome it would be to have this card. I had an X700 AGP at the time 😁
Can you monitor the VRM temperatures and VDDC externally?
Got some time to look at this again, I'm now certain it's a memory issue (or some issue with the power rail for the memory). I booted into Windows and immediately started getting artifacts on the desktop (without even running 3dMark and GPU cool). I pushed down on the top row of memory chips and I could get the artifacts to change.
The attachment Desktop.JPG is no longer available
I then ran R5MEMID again, and it passed.
The attachment R5MEMID_1.JPG is no longer available
I pressed down on the top row of memory cards and bingo I got artifacts on the screen
The attachment R5MEMID_2.JPG is no longer available
I ran R5MEMID again, finally got errors
1R5MEMID version 1.03, (c) Copyright ATI Technologies Inc, 2006 2Log file generation enabled to .\R5MEMID.LOG ... 3Extended logging enabled ... 4Reference data file (RDF) loading disabled ... 5No wait on error enabled (not effected by '-s') ... 6 [1] Mem production test FB001 7 8 Failing Bits 9 rk0 chH: 5 6 8 10 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 10 11 [2] Mem production test FB063 12 13 Failing Bits 14 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 15 16 [3] Mem production test FB064 17 [4] Mem production test FB065 18 [5] Mem production test FB066 19 20 Failing Bits 21 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 22 23 [6] Modified FB088 24 25 Failing Bits 26 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 27 28 [7] Modified FB091 29 30 Failing Bits 31 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33TEST RESULT SUMMARY: 34==================== 35R580/M59 (0x7240) detected. 36512M video memory. 37Test suite ran 1 of 1 times. 38Checking test status array ... 39[1 ] Mem production test FB001 : FAIL 40 Error ID G0FB096 41 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 42 Error count: total[29295760] chH[29295760] 43[2 ] Mem production test FB063 : FAIL 44 Error ID G0FB096 45 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 46 Error count: total[544048] chH[544048] 47[3 ] Mem production test FB064 : PASS 48[4 ] Mem production test FB065 : PASS 49[5 ] Mem production test FB066 : FAIL 50 Error ID G0FB096 51 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 52 Error count: total[36076] chH[36076] 53[6 ] Modified FB088 : FAIL 54 Error ID G0FB088 55 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 56 Error count: total[39245] chH[39245] 57[7 ] Modified FB091 : FAIL 58 Error ID G0FB091 59 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 60 Error count: total[43036] chH[43036]
zuldanwrote on 2026-05-03, 00:18:Got some time to look at this again, I'm now certain it's a memory issue (or some issue with the power rail for the memory). I b […] Show full quote
Interestingly your GPU-Z Screenshot shows the Fan not changing it's speed? It looks like it's fixed at 13%?
Btw. what a lovely card. Fond memories of me looking through magazines and imagining how awesome it would be to have this card. I had an X700 AGP at the time 😁
Can you monitor the VRM temperatures and VDDC externally?
Got some time to look at this again, I'm now certain it's a memory issue (or some issue with the power rail for the memory). I booted into Windows and immediately started getting artifacts on the desktop (without even running 3dMark and GPU cool). I pushed down on the top row of memory chips and I could get the artifacts to change.
The attachment Desktop.JPG is no longer available
I then ran R5MEMID again, and it passed.
The attachment R5MEMID_1.JPG is no longer available
I pressed down on the top row of memory cards and bingo I got artifacts on the screen
The attachment R5MEMID_2.JPG is no longer available
I ran R5MEMID again, finally got errors
1R5MEMID version 1.03, (c) Copyright ATI Technologies Inc, 2006 2Log file generation enabled to .\R5MEMID.LOG ... 3Extended logging enabled ... 4Reference data file (RDF) loading disabled ... 5No wait on error enabled (not effected by '-s') ... 6 [1] Mem production test FB001 7 8 Failing Bits 9 rk0 chH: 5 6 8 10 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 10 11 [2] Mem production test FB063 12 13 Failing Bits 14 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 15 16 [3] Mem production test FB064 17 [4] Mem production test FB065 18 [5] Mem production test FB066 19 20 Failing Bits 21 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 22 23 [6] Modified FB088 24 25 Failing Bits 26 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 27 28 [7] Modified FB091 29 30 Failing Bits 31 rk0 chH: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33TEST RESULT SUMMARY: 34==================== 35R580/M59 (0x7240) detected. 36512M video memory. 37Test suite ran 1 of 1 times. 38Checking test status array ... 39[1 ] Mem production test FB001 : FAIL 40 Error ID G0FB096 41 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 42 Error count: total[29295760] chH[29295760] 43[2 ] Mem production test FB063 : FAIL 44 Error ID G0FB096 45 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 46 Error count: total[544048] chH[544048] 47[3 ] Mem production test FB064 : PASS 48[4 ] Mem production test FB065 : PASS 49[5 ] Mem production test FB066 : FAIL 50 Error ID G0FB096 51 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 52 Error count: total[36076] chH[36076] 53[6 ] Modified FB088 : FAIL 54 Error ID G0FB088 55 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 56 Error count: total[39245] chH[39245] 57[7 ] Modified FB091 : FAIL 58 Error ID G0FB091 59 1024 x 768 - argb8888 (60 Hz): TEST FAILURE 60 Error count: total[43036] chH[43036]
…Show last 2 lines
61Failure detected.
Just asking the obvious question: If the errors appear on boot with no pressure applied, do you have any measures to test the card for 20s with no cooler on memory so you can push down only on channel H?
Just asking the obvious question: If the errors appear on boot with no pressure applied, do you have any measures to test the card for 20s with no cooler on memory so you can push down only on channel H?
When I pressed down on individual memory chips I couldn’t pinpoint one that caused the most disruption on the screen.
Brand new cores for this GPU have arrived. I’m tempted to replace the core and if that doesn’t work, reball the memory chips (new chip replacements are almost impossible to get, Samsung GDDR4 K4U52324QE - B609 0.9ns). You know I love replacing parts 😂
On a side note, the small rectangular heat sink covering the VRM section wasn’t screwed to the board (I must have unscrewed it in my last repair session and forgot about it), so when I lifted the card up the heat sink “slid” out and damaged a VT1115SF chip (https://www.aliexpress.com/item/1005010604836806.html), totally my fault. There is an error LED that light that shows up now where you plug in the PCIE connection. Hopefully the replacement part should be here in 2 weeks.
zuldanwrote on Today, 08:59:When I pressed down on individual memory chips I couldn’t pinpoint one that caused the most disruption on the screen. […] Show full quote
Just asking the obvious question: If the errors appear on boot with no pressure applied, do you have any measures to test the card for 20s with no cooler on memory so you can push down only on channel H?
When I pressed down on individual memory chips I couldn’t pinpoint one that caused the most disruption on the screen.
Brand new cores for this GPU have arrived. I’m tempted to replace the core and if that doesn’t work, reball the memory chips (new chip replacements are almost impossible to get, Samsung GDDR4 K4U52324QE - B609 0.9ns). You know I love replacing parts 😂
On a side note, the small rectangular heat sink covering the VRM section wasn’t screwed to the board (I must have unscrewed it in my last repair session and forgot about it), so when I lifted the card up the heat sink “slid” out and damaged a VT1115SF chip (https://www.aliexpress.com/item/1005010604836806.html), totally my fault. There is an error LED that light that shows up now where you plug in the PCIE connection. Hopefully the replacement part should be here in 2 weeks.
Ok not impossible if you happy to pay 🤣
I found that too 😝 but $10 USD a chip, that’s $80 USD to replace all the chips, ouch!
In theory you could remove the assumed chip. If it's really channel H, it'll probably work as bad as it does now, since it appears all 32 bits of it are affected. That would tell if your assumption is right. Not sure what path I'd go here.. having to swap the ASIC in hope it fixes it is also a bold move 😁
Ok not impossible if you happy to pay 🤣
I found that too 😝 but $10 USD a chip, that’s $80 USD to replace all the chips, ouch!
In theory you could remove the assumed chip. If it's really channel H, it'll probably work as bad as it does now, since it appears all 32 bits of it are affected. That would tell if your assumption is right. Not sure what path I'd go here.. having to swap the ASIC in hope it fixes it is also a bold move 😁
That is a good idea. I will try it.
I’ve swapped ASICS a few times with success. It is scary but the card is unusable at the moment. I also have another fully functional X1950 XTX so not too worried if this card becomes For Parts card.
