I meant Polaris BIOS Editor - NOT WattTool.
Hmmm - but this tables in polarisbioseditor and watttol look so similar. You say that they mean different?I think so - VDDCI would start at 950-ish, usually, for memory state 1.
Topic: RX470 4GB Custom rom (Read 10145 times)
I think so - VDDCI would start at 950-ish, usually, for memory state 1.
Is that VDDCI/AUX in the editor? No. That is MVDD(C) - and does nothing.
So why if I lower it in WattTool below 0.95 with 2200 memclock I have instability? And no problems with 0.95 or higherI meant Polaris BIOS Editor - NOT WattTool.
256k is fine. 512k is also fine. Also, just to help you guys out, the "memory voltage" in the VBIOS can be set for Polaris (even using the editor) but it does nothing.
AFAIK this "memory voltage" is memory controller voltage and changing it via WattTool is definitely have influence on stability - when I set it less then 0.95v my card don't want to work with high memory freq.
I change the gpu voltage with voltage offset option, not directly in table so I think it is not the dependency between gpu voltage and mem controller voltage that matters...
The editor changes MVDDC, which does nothing. You describe VDDCI/AUX - which feeds the memory controller, among other things.
Is that VDDCI/AUX in the editor? No. That is MVDD(C) - and does nothing.
See screenshot here (NSFW): https://ottrbutt.com/tmp/wolfamdctrl-vddci.png
256k is fine. 512k is also fine. Also, just to help you guys out, the "memory voltage" in the VBIOS can be set for Polaris (even using the editor) but it does nothing.
AFAIK this "memory voltage" is memory controller voltage and changing it via WattTool is definitely have influence on stability - when I set it less then 0.95v my card don't want to work with high memory freq.
I change the gpu voltage with voltage offset option, not directly in table so I think it is not the dependency between gpu voltage and mem controller voltage that matters...
The editor changes MVDDC, which does nothing. You describe VDDCI/AUX - which feeds the memory controller, among other things.
will post pic later
I don't doubt WattMan can change VDDCI - I was saying the Polaris BIOS Editor cannot.
30MH rom for Saphire RX470 4GB Elpida
https://bitcointalksearch.org/topic/29mh-rom-for-saphire-rx470-4gb-nitro-oc-elpida-1710085
any
custom firmware
for asus 470 4gb???
custom firmware
for asus 470 4gb???
Any Modded Rom for AMD Sapphire Nitro+ Rx 480 8GB OC for 100+ Sols/s on zcash
My GPUs are ellesmere with Samsung chip
How to lower the power from wattman builtin tool?
My GPUs are ellesmere with Samsung chip
How to lower the power from wattman builtin tool?
Does anybody have a stock BIOS collection for Sapphire RX470 4Gb reference models?
I ran into a rather strange problem:
- bought 5 of these cards (exact same P/N and SKU, roughly similar product numbers)
- saved the full 512Kb image of 1 card as backup (I know, I should have saved them all separately and compare the hashes...)
- flashed a modded 256Kb BIOS (I prepared this in advance, based on the image file on techpowerup)
- noticed something is very wrong with one specific card
- restored the backup BIOS image to all cards
- problem persisted, so I pulled that single card out and used the remaining 4 with my modded BIOS re-applied
- I got the card replaced under warranty (the new one has the same P/N and SKU) but I flashed it, and now it's dead (even after restoring my backup BIOS)
Yes, I know I was stupid I didn't make a unique backup of the replacement card's ROM after the first incident. It's very obvious in hindsight now...
Although, I did always check the memory type with GPU-Z and it always reported Hynix (that's why I was confident I really have identical cards).
Any suggestions?
I ran into a rather strange problem:
- bought 5 of these cards (exact same P/N and SKU, roughly similar product numbers)
- saved the full 512Kb image of 1 card as backup (I know, I should have saved them all separately and compare the hashes...)
- flashed a modded 256Kb BIOS (I prepared this in advance, based on the image file on techpowerup)
- noticed something is very wrong with one specific card
- restored the backup BIOS image to all cards
- problem persisted, so I pulled that single card out and used the remaining 4 with my modded BIOS re-applied
- I got the card replaced under warranty (the new one has the same P/N and SKU) but I flashed it, and now it's dead (even after restoring my backup BIOS)
Yes, I know I was stupid I didn't make a unique backup of the replacement card's ROM after the first incident. It's very obvious in hindsight now...
Although, I did always check the memory type with GPU-Z and it always reported Hynix (that's why I was confident I really have identical cards).
Any suggestions?
I did a lot of tests to figure out witch voltage is wich and got only confusing results. There is clearly some synchronisation, when you change one voltage the others are often modified too.
Does someone have a clear idea how the whole thing works?
Does someone have a clear idea how the whole thing works?
I think so - VDDCI would start at 950-ish, usually, for memory state 1.
Ok. Thanks for the screen and explanation. I meant Polaris BIOS Editor - NOT WattTool.
Hmmm - but this tables in polarisbioseditor and watttol look so similar. You say that they mean different? Is that VDDCI/AUX in the editor? No. That is MVDD(C) - and does nothing.
So why if I lower it in WattTool below 0.95 with 2200 memclock I have instability? And no problems with 0.95 or higher I don't doubt WattMan can change VDDCI - I was saying the Polaris BIOS Editor cannot.
Sorry, just updated the post above ... see pics. 256k is fine. 512k is also fine. Also, just to help you guys out, the "memory voltage" in the VBIOS can be set for Polaris (even using the editor) but it does nothing.
AFAIK this "memory voltage" is memory controller voltage and changing it via WattTool is definitely have influence on stability - when I set it less then 0.95v my card don't want to work with high memory freq.
I change the gpu voltage with voltage offset option, not directly in table so I think it is not the dependency between gpu voltage and mem controller voltage that matters...
The editor changes MVDDC, which does nothing. You describe VDDCI/AUX - which feeds the memory controller, among other things.
will post pic later
I don't doubt WattMan can change VDDCI - I was saying the Polaris BIOS Editor cannot.
Haleluja!
Something i've been shouting for weeks
Finally someone who is on the same line! 256k is fine. 512k is also fine. Also, just to help you guys out, the "memory voltage" in the VBIOS can be set for Polaris (even using the editor) but it does nothing.
AFAIK this "memory voltage" is memory controller voltage and changing it via WattTool is definitely have influence on stability - when I set it less then 0.95v my card don't want to work with high memory freq.
I change the gpu voltage with voltage offset option, not directly in table so I think it is not the dependency between gpu voltage and mem controller voltage that matters...
The editor changes MVDDC, which does nothing. You describe VDDCI/AUX - which feeds the memory controller, among other things.
256k is fine. 512k is also fine. Also, just to help you guys out, the "memory voltage" in the VBIOS can be set for Polaris (even using the editor) but it does nothing.
AFAIK this "memory voltage" is memory controller voltage and changing it via WattTool is definitely have influence on stability - when I set it less then 0.95v my card don't want to work with high memory freq.
I change the gpu voltage with voltage offset option, not directly in table so I think it is not the dependency between gpu voltage and mem controller voltage that matters...
does anyone ever tried this multiplier?
http://www.ebay.com/itm/PCI-e-Express-1X-to-3-Port-1X-Switch-Multiplier-HUB-Riser-Card-USB-Cable-/262501285031?hash=item3d1e4bbca7:g:zTwAAOSwa-dWhMvm
tks,
http://www.ebay.com/itm/PCI-e-Express-1X-to-3-Port-1X-Switch-Multiplier-HUB-Riser-Card-USB-Cable-/262501285031?hash=item3d1e4bbca7:g:zTwAAOSwa-dWhMvm
tks,
yes i have and i couldnt get it to work. it recognized the multiplier but i couldnt get it to recognize a vid card in the slot. even if its the only card on the pc. I think there is good reason more dont use these. There is a limit to the pcie lanes on a cpu and once you exceed that it will not post. I dont know why the splitter didnt work with only one card other than it has a proprietary chip onboard and maybe that doesnt like vid cards. ethernet adapter or modem etc mifght be ok, idk
Best Regards
d57heinz
does anyone ever tried this multiplier?
http://www.ebay.com/itm/PCI-e-Express-1X-to-3-Port-1X-Switch-Multiplier-HUB-Riser-Card-USB-Cable-/262501285031?hash=item3d1e4bbca7:g:zTwAAOSwa-dWhMvm
tks,
http://www.ebay.com/itm/PCI-e-Express-1X-to-3-Port-1X-Switch-Multiplier-HUB-Riser-Card-USB-Cable-/262501285031?hash=item3d1e4bbca7:g:zTwAAOSwa-dWhMvm
tks,
Wolf0, could you upload somewhere your custom rom for Sapphire rx 470 4gb reference please?
reference sapphire 470 4gb modded rom with the 1500memory strap timing set to 1625, 1750, and 2000: http://www.megafileupload.com/k2gu/470_4gb_sapphire_modded_bios.rom
I use -12.5% (1065)/850mv core and 2000/850mv mem for high mhs and power efficiency
Thank you for sharing !
What about the hash rate you are archiving with this ROM?
Also how is the power distribution with the reference 470?
Reference 470 has 4+1 phases
Reference 480 has 6+1 phases
For the 480 4 phases are supplied by the 6 pin and 2+1 phases are supplied by slot.
This COULD mean that the 470s all 4 phases get supplied by the 6 pin and only the +1 (= for vram) is supplied by the slot.
The pci-e slot power draw from the ref. 470s would in this case be way lower compared to the ref. 480.
If that is the case, then it is quite safe to use the USB risers. If not, it is still safer than the RX 480.
Wolf0, could you upload somewhere your custom rom for Sapphire rx 470 4gb reference please?
reference sapphire 470 4gb modded rom with the 1500memory strap timing set to 1625, 1750, and 2000: http://www.megafileupload.com/k2gu/470_4gb_sapphire_modded_bios.rom
I use -12.5% (1065)/850mv core and 2000/850mv mem for high mhs and power efficiency
Thank you for sharing !
What about the hash rate you are archiving with this ROM?
Also how is the power distribution with the reference 470?
Reference 470 has 4+1 phases
Reference 480 has 6+1 phases
For the 480 4 phases are supplied by the 6 pin and 2+1 phases are supplied by slot.
This COULD mean that the 470s all 4 phases get supplied by the 6 pin and only the +1 (= for vram) is supplied by the slot.
The pci-e slot power draw from the ref. 470s would in this case be way lower compared to the ref. 480.
Jump to: