@fapador here you go, done. see txt for full changes
sapphire micron mod.zip (108 KB)
Thank you very much for your time . I really appreciate your effort. I will give it a go and report back
@ket I flashed your bios and something is wrong I cannot enter my PC bios any more. I get a black screen and my keyboard freezes. When in Windows it works ok though. are you sure you used the file (with name stockbios.rom and not ellesmere) I uploaded on the second time and not the first one ? ( I also think that PBE or Srb might have something to do with this and affect the gop in some way). As I got a similar issue though not exactly same with my attempts.
Edit: I flashed back original bios(same as the ones provided in link in post) no issues⌠thats why I only prefer a mod that is purely edited in Hxd
@fapador I didnât use SRB at all or PBE, if your board is in UEFI mode that will be why you didnât get anything until Windows as I didnât do any GOP mods, and without a GOP mod an edited vBIOS has an âinvalidâ GOP. And yep, I used the stock vBIOS.
@ket Ah ok I wasnât aware of this I will try disabling uefi then. Thanks again
Also Any Idea what it would take to make gop valid again? I am completely out of my depth in such things.
@fapador Thre is a seperate guide here on win-raid on how to fix the GOP, @Fernando can probably point you where it is he knows this place like the back of his hand 
@ket Tested memory for 1hr with superposition and some msfs. No memory errors on hwinfo even at 2200 but I see some artifacts occasionaly. Any reccomendation which timing should I try to relax? Thinking of TRCDW TRCDR TRRD . Also I get memory efficiency ~225 Gb/s but after running benchmark and rerun poclmembench It drops to 200. I restart PC rerun poclmembench and goes back to 225Gb/s. Really considering upping VDDCI to .975 as I tried it before and dont think it will cause harm. Also older Rx480 seemed to run mem controller up to 1.1v without problems but I dont know If its the same component.
Thanks for your premature praise. >Here< is the related Guide.
You are praised because you deserve it @Fernando !
@fapador relax TRP_RDA to 29, that should iron things out. Bandwidth will drop a tad, but nothing major. RX400 cards were quite different to RX500 cards, the memory controller is part of the GPU die itself you definitely donât want to go above 1v but youâll likely see better stability with it at 0.9v as AMD have a nasty habit of overvolting things by a rather large margin. Eg; most RX580s come with a stock voltage of 1.15v regardless but pretty much every card doesnât need more than 1.1v including a healthy OC. If youâre content with say around 1.4GHz youâll likely find that will run on 1.075v or so which leads to quite a nice drop in temps and power usage and negligible performance difference.
@ket Tried TRP_RDA ->29 still artifacts. Even tried the stock value of 33 no difference. It must be another timing causing them
@fapador theres several main timings to adjust to clear up artifacting, 2200MHz is clearly going to be the absolute maximum for your card any timing adjustments you make to try and get it stable are likely going to perform worse than running the card with the timings I added @ 2150MHz but some other timings you can adjust are; TRP_WRA 65, TCCDL 7, TCCDS 6, TRRD 8. If those changes work you can then revert them one by one to my original values until you find the specific timing responsible. TCCDL and TCCDs should be the first two you start with reverting if those changes work as primarily they are good for correcting memory that has regular misfires with EDCs but the card isnât inherently unstable.
@ket found it. it was TCCDS changed it to 6 artifacting gone. Now if only I can resolve the gop issue, which judging by the guide seems too complicated to fix I will be a happy man. @Fernando Have you done anything like this before? any help will be appreciated .
@fapador you might want to also increase TCCDL, at least to 4, itâll help prevent EDCs at like zero performance cost, tends to help iron out microstutter, and amusingly AMD from what I deciphered set this value incorrectly it should be set to 4 anyway for stability at low memory frequencies like desktop usage.
Attempted 1400mhz at 1100mv yesterday and while the card posted some impressive LuxMark scores, it crashed about 2minutes into a game. 1380mhz at 1100mv was actually stable but resulted in rather high temperatures(74c). Went back to 1350mhz at 1068mv which cut temps by quite a bit and resulted in a ridiculously small 2fps loss(152 vs 150 in Dirt 3 benchmark) compared to 1380mhz. poclmembench scores were 226 at 1400, 223 at 1380 and 222 at 1350. So the fan mod did allow me to push to 1380mhz but the tiny gains definitely arenât worth it considering the higher temps and stress put on the VRM. Still happy I now have a well-performing and silent card. 
Not a bad score there @TheVic1600 but you have some catching up to do to reach me 1350MHz looks like the balance between speed and voltage required but Iâd lean more toward the shitty VRM limiting you rather than the core, a good VRM would likely see you to 1420MHz or so @ 1.1v.
Well, I do have 4 fewer compute units to work with. I think I might actually have the fastest 32-unit Polaris in LuxMark since I really canât find any 32-unit GPU above me in the rankings, itâs just 36-unit GPUs. Sadly 1400mhz wasnât stable but my 16,280 score at 1350mhz isnât much lower and still faster than any 32-unit GPU. Not bad for a card with such a crippled VRM. Also, I found this interesting thread which seems to point to the XFX RS RX480 having the same VRM as my card. The guy there ran the GPU at 1150mv and encountered heavy throttling at 83c and 133W. Seems to remind me of when I encountered throttling in Fire Strike at over 75c and 1087mv. I got rid of that by raising the TDP but then encountered throttling again at 1137mv.
@TheVic1600 doing some quick math the 570 is around 5% slower than a 580 stock to stock, so with highly optimised timings you should be able to put yourself in the region of 17,500 in Luxmark, due to your crappy VRM though limiting your core clock lets say about 17,100. I can help you get those last few FPS if you want as Iâve also got an RX470 that just so happens to not like going past 1320MHz on the core without a ton of volts and 2050MHz on the memory, but its Samsung. That link you found about the RX480 the likely explanation is that XFX used shitty FETs that werenât rated high enough so the safety kicked in when detecting more watts were being drawn than the VRM could provide without risk of blowing.
EDIT: On a similar note I made some adjustments to my RX470, reduced memory controller voltage from the default 1v to 0.9v, increased vcore just a smidge (1.062v > 1.068v) as vcore was bottoming out a little more regularly than Iâd like, and without any surprise on my end at all the card is now running cooler, quieter, more power efficient, and more stable. In my case the 470 I have was somewhat of a problem child it was even getting EDCs at completely stock settings without any vBIOS mods at all, literally as stock as stock can be, but now all of those issues are cleared up. I would attempt to try and get the core or memory up but the card is in a case that while well ventilated isnât particularly spacious and its an old FX CPU that pisses out heat when under load like power consumption and heat donât matter in times when global warming and carbon footprints matter. Fortunately though the FX CPU I have happily runs 4.2GHz below stock voltage so I can at least keep the power consumption for it in check. End of the day, the 470 is pretty much fully optimised now, a tweak increasing TRRDS to 7 and TRRD to 5 might allow me to hit 2100 rather than 2050 but the relaxed timings, TRRD especially, would negate any benefit in the higher frequency, and if I needed to increase TRP_RDA as well that would really render the higher frequency pointless, and possibly perform worse.
hello, thank you for this topik
I have a nitro + rx580 8go micron, and Iâm trying to increase its performance a bit.
I tried your bios, I actually have an increase in perf, it seemed stable under benchmark (unigine, 3dmark), but it was unstable on an OCCT stresstest
I think these are the timings can be too aggressive
I donât have the impression that my ram can go up very high in frequency
@briaa OCCT is poor software to use to verify stability Iâve literally had totally stock rock solid stable systems fail the OCCT âstabilityâ test in seconds I stopped using it probably two decades or so ago OCCT should be put in the same bracket as the Furmark murderware. A much better test for system stability is BOINC, its a scientific piece of software that hammers a system hard by crunching various kinds of data depending on what projects you choose to run. If you are BOINC stable, that is as good of a guarantee as what you get really. We would need to know more about your card to work out whats going on but as a start it is always recommended to use the original vBIOS for mods not one from a different batch, model, or brand.
hello, thank you for answering
here is my bios file.
I have a saphire nitro+ rx580 vram micron
occt gpu test with origin bios
my gpu donât go to 1411 mhz and throttle, i have changed today thermal grease and pads.
Ellesmere.zip (108 KB)