@boombastik its a bug with one of the automatic pointers, you're better off inputting a manual value anyway and limiting auto voltages to no more than 1.15v to stop cards from cooking themselves anyway. Auto voltages are absolutely terrible on Polaris cards I have no idea how they passed any kind of QA test as those auto values will blast a card with 1.2v even if it only needs 1.1v or less.

Hello and I'd like to start by thanking you for this very thorough guide.

So I have an XFX RS RX570 8GB, the one with the shitty VRM, 2 BIOSes and Samsung(K4G80325FC) memory. I've been messing with the memory timings for about a week now, after reading your guide. I've been largely unsuccessful, as the 7-8 straps I've tried so far have either resulted in lots of errors in HWiNFO or no errors but black screens/freezes/artifacts during gaming.

Frustrated, today I decided to reflash the original BIOS, run oclmembench and use AMD Memory Tweak XL to check the load timings in real time. I also checked the timings for the idle 300mhz strap. This led to some interesting, yet annoying discoveries which I've outlined in the picture:


That is what happens during an oclmembench test done at stock settings(1150/2050). Clearly, the card isn't exactly following what the strap tells it to do. Worse, when I apply any setting in MSI Afterburner or OverdriveNTool, I notice the oclmembench results take a hit and if I check the real timings again, I notice that the top 2000mhz memory state now uses the ARB_DRAM_TIMING and ARB_DRAM_TIMING2 values from the bottom 300mhz state. These are very tight timings, like RAS2RAS 31, RP 13. So basically, since the card is idle when I apply settings in MSI afterburner, those 2 timing sections become "frozen" and stay the same no matter what strap the card is in. This also happens when I set anything in OverdriveNTool. This is likely why I'm either getting memory errors or crashes when testing modded straps, as I have set the card to 1750mhz in the BIOS. I then apply the modded timings to the 2000 strap and when I'm in Windows, I apply my 2100mhz overclock in MSI Afterburner. That ends up messing the ARB_DRAM_TIMING and ARB_DRAM_TIMING2 timings. The only way to avoid this is to do all my settings in the BIOS and never use any tools to alter them when in Windows. This is annoying since it means I could brick the BIOS if I try aggressive timings and I'd also be unable to pick different settings for each game(some stuff I play is old and really doesn't need 1400/2100 clocks).

So my questions are:
1. Should I worry about the usually messed up ARB_DRAM_TIMING and ARB_DRAM_TIMING2 timings? Out of the box, at 2050, the card scores 210GB/s in oclmembench and has no issues overclocking to 1400 at 1.125v and staying stable in games. No errors in HWiNFO either. This is despite the messed up ARB_DRAM_TIMING and ARB_DRAM_TIMING2 timings. It's not like I can do anything about this tho.

2. Is there any way to maintain the ability to change card settings in Windows without really screwing up the ARB_DRAM_TIMING and ARB_DRAM_TIMING2 timings? Or should I just settle on a default configuration and configure that via the BIOS?

Some more details about my card: ASIC Quality 73%. VDDCI is 0.838v under load.

that tweaker xl tool is unstable/broken? and it was also giving me troubles on rx560. using several oc tools at once will likely result in crashes. get rid of afterburner and use adrenaline's builtin wattman only

I only use the tool to check the timings. I didn't set anything in it. Wattman is worse than Afterburner in that trying 1400/2100 in Wattman results in the driver crashing. Only MSI Afterburner has given me stable overclocks, until I began playing with the memory timings. I doubt tweaker xl tool is broken as it seems to read everything correctly(I have to fix the fan profiles in the BIOS) and the drop in performance(in oclmembench) and occasional memory errors(in HWiNFO) coincide with tweaker xl tool detecting the 300mhz strap values for ARB_DRAM_TIMING and ARB_DRAM_TIMING2.

they are all just frontends to set the desired clocks, and should not cause unstable overclocks, but the way each of them work may cause incompatibility between them. for example afterburner sets additional registry keys and soft powerplay table once you tick that OC checkbox which might set some bits different than adrenaline/overdriventool, but if you edit BIOS with all the required changes there should be no problem for adrenaline either..
about tewaeker xl, yes reads correctly, but sets timings that reset or autochange only seconds later - thus broken for whatever reason and this is where i stopped wasting time with it

Right, so I decided to go the BIOS tweaks only way and I'm glad I did. I applied the following settings in PBE:


I went for an overclock to 1350mhz for a start, since I know the GPU is stable at 1400@1125mv. Also, since I can't use MSI to set a +50 power limit, I upped the Max Power Limit to 187W and decreased the Power Control Limit to 2%. I chose 187 because when stock, the card is set at 125W and 50% Power Control Limit. 125*0.5=62.5. 125+62.5=187.5. So now the card defaults to what's basically stock and +50 power limit without me having to apply settings in MSI/Wattman, which always screws up the memory timings. S̶o̶ ̶n̶o̶w̶ ̶m̶y̶ ̶q̶u̶e̶s̶t̶i̶o̶n̶ ̶i̶s̶:̶
C̶a̶n̶ ̶I̶ ̶u̶s̶e̶ ̶a̶ ̶m̶i̶x̶ ̶o̶f̶ ̶v̶o̶l̶t̶a̶g̶e̶ ̶p̶o̶i̶n̶t̶e̶r̶s̶ ̶a̶n̶d̶ ̶s̶e̶t̶ ̶v̶o̶l̶t̶a̶g̶e̶s̶ ̶i̶n̶ ̶t̶h̶e̶ ̶G̶P̶U̶ ̶s̶e̶c̶t̶i̶o̶n̶?̶ ̶I̶ ̶h̶a̶v̶e̶ ̶6̶5̶2̶8̶6̶ ̶i̶n̶ ̶m̶y̶ ̶t̶o̶p̶ ̶s̶t̶a̶t̶e̶s̶ ̶w̶h̶i̶c̶h̶ ̶e̶n̶d̶s̶ ̶u̶p̶ ̶b̶e̶i̶n̶g̶ ̶1̶1̶5̶0̶m̶v̶ ̶a̶n̶d̶ ̶t̶h̶a̶t̶'̶s̶ ̶1̶.̶2̶v̶ ̶o̶n̶c̶e̶ ̶t̶h̶e̶ ̶d̶r̶i̶v̶e̶r̶ ̶g̶o̶e̶s̶ ̶c̶r̶a̶z̶y̶ ̶w̶i̶t̶h̶ ̶t̶h̶e̶ ̶o̶v̶e̶r̶v̶o̶l̶t̶i̶n̶g̶.̶ ̶S̶o̶ ̶w̶h̶a̶t̶ ̶I̶'̶d̶ ̶l̶i̶k̶e̶ ̶t̶o̶ ̶h̶a̶v̶e̶ ̶i̶s̶ ̶s̶o̶m̶e̶t̶h̶i̶n̶g̶ ̶l̶i̶k̶e̶ ̶7̶5̶0̶ ̶6̶5̶2̶8̶2̶ ̶6̶5̶2̶8̶3̶ ̶6̶5̶2̶8̶4̶ ̶6̶5̶2̶8̶5̶ ̶1̶1̶2̶5̶ ̶1̶1̶2̶5̶ ̶1̶1̶2̶5̶.̶ ̶W̶o̶u̶l̶d̶ ̶s̶u̶c̶h̶ ̶a̶ ̶s̶e̶t̶u̶p̶ ̶b̶r̶i̶c̶k̶ ̶t̶h̶e̶ ̶B̶I̶O̶S̶ ̶o̶r̶ ̶w̶o̶u̶l̶d̶ ̶i̶t̶ ̶b̶e̶ ̶f̶i̶n̶e̶?̶ ̶I̶ ̶c̶a̶n̶'̶t̶ ̶u̶s̶e̶ ̶6̶5̶2̶8̶5̶ ̶f̶o̶r̶ ̶1̶3̶5̶0̶m̶h̶z̶,̶ ̶s̶i̶n̶c̶e̶ ̶t̶h̶a̶t̶'̶s̶ ̶1̶1̶0̶0̶m̶v̶ ̶a̶n̶d̶ ̶I̶'̶m̶ ̶p̶r̶e̶t̶t̶y̶ ̶s̶u̶r̶e̶ ̶t̶h̶e̶ ̶c̶a̶r̶d̶ ̶w̶o̶u̶l̶d̶ ̶e̶n̶d̶ ̶u̶p̶ ̶b̶e̶i̶n̶g̶ ̶u̶n̶s̶t̶a̶b̶l̶e̶ ̶a̶t̶ ̶1̶1̶0̶0̶m̶v̶.̶ ̶
EDIT: I ended up modifying the BIOS with HxD using this post as a guide: https://www.overclock.net/posts/28663554/
The voltage still ended up spiking to 1181mv, so I've now manually set the voltages for all Pstates in PBE. Now it still spikes up to 1158mv, but during gaming I'm seeing mostly 1100 and 1125mv. Will now try to tweak the timings even tho I'm getting 220GB/s at 2050mhz. Might also try further undervolting to 1100 as everything seems stable at 1125. It seems that going for BIOS tweaks only without using OC software is entirely possible and not annoying as I previously thought. A bit like it is with CPU overclocking which I've always done in the BIOS. :D

Oh, and the card is now scoring between 220 and 225 in oclmembench(up from the 210 it used to score with the broken ARB_DRAM_TIMING and ARB_DRAM_TIMING2 values) :D

This is with the standard memory strap in the "silent" BIOS I'm experimenting with, so I'm very happy. HWiNFO is showing no errors as well.

For anyone who wants the stock timing for the 2050mhz strap, it is this:
777000000000000022CC1C00CE695D4CF0571218B90DE60B00400700240614207A8900A00300000013123740BE323E18

Thats not bad @TheVic1600, I don't have the timings to hand right now but I have an optimised set of Samsung timings that can probably get another 7-10GB/s out of the memory at the same frequency. Let me know if you want to try them and I'll dig them out.

@ket Yup, It'd be great to try those. I have been trying timings almost every day for a few weeks now and I've learned some things about what works for my card and what doesn't:

-I seem to be hitting a wall at 221GB/s at 2050mhz. I can get 225 and 227GB/s sometimes right after booting into Windows but subsequent tests always result in 221.
-2100mhz always results in artefacts unless I loosen the timings so much I end up scoring about 218GB/s. 2075mhz gave me just 1GB/s more but ended up crashing during gaming, so 2050mhz is max my memory can go.
-Sensibility to TW2R is huge. I can only lower TW2R from 31 to 29 if I loosen almost everything else and with that I only get 219GB/s, so I've given up on that.
-ACTRD and ACTWR seem to work at 19 and 18. Loosening those to 22 and 21 lets me go 4 on TRRD with 26 and 55 on TRP_RDA and TRP_WRA(tho I didn't test more than 1 game with that config, so I'm not completely sure that works)
-Memory errors only show up if I lower TW2R and they show up in the millions. Otherwise, instability manifests in the card either showing artefacts(Dirt 3's Benchmark seems t be great at bringing them when other games don't) or having a driver/system crash.

Overall, XFX's silent BIOS default memory timings are surprisingly good. I get the best stable performance when I take the default strap and loosen the timings you said don't affect performance at all or add stability. With that I can go 22 on the TCL(down from 24), but the performance seems to be the same(still hitting the 221GB/s wall). Interestingly, when I tried to overclock the core to 1400mhz(didn't work because of the undervolt), I managed to score a consistent 224-225GB/s with the same timings. Does the IMC also work better at higher core frequency? Maybe that's the key to going over the 221GB/s wall? I can go 1400 on the core, but it requires more voltage and I wouldn't want the card to become hot and noisy again, so I'd rather stay at 1350 and just accept I can't go over 221.
So yea it would be nice to see what your strap looks like, tho so far this "FC" Samsung VRAM seems to be noticeably worse than the "FB" ones. I'm still happy with the card at 1350mhz equalling a 1380mhz RX580 in 3DMark Fire Strike and Sky Diver. In Dirt 3's benchmark I'm up 10-11% on the card's stock result.

@TheVic1600 heres the timing strap if you send me over a copy of your vBIOS I can probably make some tweaks and changes to smooth some things out. Its not an exact science optimising memory on Polaris due to no silicone being created equal but with enough time, testing, and patience you can nail it down to 95% at least some cards might just need a tiny tweak left or right for absolute stability. Samsung memory really doesn't play nicely on Polaris it doesn't matter if you have FC or FB Sammy, I had a card with FB Sammy actually and that topped out at 2050MHz no matter what I did or tried, Hynix and Micron memory are very good though last RX590 I played with I got the Micron on that to 2375MHz with optimised timings, or 9500MHz, whatever way you prefer to refer to the memory frequency. I did enough to get like second or third in the Futuremark rankings with systems using a 2700X. I'll chuck the screens up as a few attachments in case you're curious as you won't find the results on Futuremark tables because I'm not the sort to partake in OCing in that way anymore. In case you are wondering the card is seen as an RX580 in the attached images because I modded the vBIOS to get around the driver level check so I didn't have to keep messing around with a software hack, it was a laziness thing lol.

Anyway, here's that timing strap for Samsung memory; 777000000000000022CC1C00106A4D4DC0691016378D160B0068C70014051420FA8900A00300000015153F48B1384C1A

There's a tiny amount of play in the timings but this is for long term stability reasons and because it just doesn't make any real difference to performance. If these timings aren't completely stable for you (they likely will be, I made damn sure of that for 95%+ of cases) change these timings: TRRD to 5 (big factor for stability) TCCDS to 5 (should help prop up minimum frames in extreme cases a bit but adversely might lead to a bit of extra microstutter) TRP_RDA to 27 (big factor for stability) TR2W to 29. These changes shouldn't be necessary but if they are start with just changing TRP_RDA to 27.

@ket Thanks a lot! Will give these a try right away and report later. In my previous post I mixed TW2R and TR2W. I meant to say TR2W. Really curious wether your strap will run as it is with a TR2W that's tighter than everything i've tried so far.

I'll post my BIOS here if you feel like taking a look over it. I'm pretty ok with it tho what I'd like to fix are the voltage overshoots that still happen. I've used HxD to set a VID of 1.075V and while it does stay between 1 and 1.1V most of the times, HWiNFO sometimes still shows spikes of up to 1.181V. The temps and overshoots are clearly lower than at stock tho, so it's only a minor problem. You mentioned in one post that XFX's BIOSes are kinda shit and I feel that's true. I've tried adding an offset to the NCP81022 but that section I'm supposed to edit just doesn't exist in my BIOS. Also one of the 4 parts I'm supposed to edit to change the VDDCI is missing. I've also had to fix a broken fan curve which had the card's minimum fan speed at 75%.

EDIT: Would also be nice if the card got recognized as an RX580 since I have to use that driver patcher to make it work. I have tried using HxD to do the driver bypass tweak in your first post, but it didn't do anything, the drivers still refused to work without the patch.

EDIT: Well, so far so good. oclmembench is now mostly 222GB/s with some 221 values. FireStrike score is just 1 point(14333) better than what I scored yesterday(14332) but I'm sure you've tested your strap for stability way more than I did. Also had no memory errors and no artefacts in 3 Dirt 3 benchmarks in a row so this has passed all my initial stability tests so far. Will later see how it performs in more demanding stuff like Doom 2016 and Wreckfest. The Dirt 3 benchmark scores are about the same as what I got yesterday in terms of avg FPS, but I feel there's a 1-2FPS improvement on the min FPS values. Great, because I care more about min FPS as my monitor can only display 60FPS so I'm well over that in most of the games I play. Will likely have a go at playing with the values you mentioned as well if something goes wrong.

@TheVic1600 here is a tuned up copy of your vBIOS, check the included document for full details.

@ket Oh wow, that's a lot of changes, thanks a lot. Will try it out in some hours. :)

And yes, my card has a single 8-pin connector. I just have one question:

Is it safe to increase both TDP and Power Limit as long as TDC stays stock? No worries about the shoddy VRM? I saw some people on another forum calling the VRM as bad as the PowerColor one and that's the reason I went conservative on setting both. The card wasn't throttling at 150W power limit so that's why I left it like that.

Again, thanks a lot for the help.

Right, I got to test the BIOS and only had one memory error after playing some Dirt 4. The card was a lot more silent than usual, because I didn't use MSI Afterburner's custom fan curve but it also got pretty toasty(for my taste at least) at 83C in the most graphically demanding stage I played. The game ran better than usual tho, with fewer and smaller dips under 60fps. Also did a LuxMark test before flashing your BIOS and after and there was a gain of about 90points. I also remembered 1137mv should be enough for 1400mhz core, as I was hitting that with 1125mv previously so I went into PBE to set 1400 and also lowered the High PWM temp down to 75c. Before PBE I used HxD to disable Zero Fan at your suggestion, since I noticed I can't hear the fan spinning at 20% when the card is idle anyway. Oh and the Max Fan speed setting is broken in XFX's BIOS as if I set the card at 100% fan speed using Wattman, it reaches 4k RPM despite your 2.2k setting. No problem as I'm pretty sure 2.2k would not be enough to properly cool the card.

Sure enough 1400mhz ended up being stable and I now get 226GB/s in oclmembench. The High PWM adjustment helped too with the max temp now being 78C. At 1400, the LuxMark score jumped another 120points. Dirt 3 Benchmark ended up at 158/134, so my minimum FPS is now just 4FPS under the stock card's average FPS :D

I'm reading online that these cards are actually fine at 85-90c for even extended periods of time. Is that true? I was previously desperate to keep the card under 75, despite the jet engine noise necessary for that. If I can actually stretch to 85c, then 1400mhz at 1137mv is actually doable with low-ish noise.

EDIT: The card starts to glitch in Fire Strike as soon as the temperatures get into the upper 70s-lower 80s. Screen starts to randomly cut out(flickers to black) and I also see a few random textures appearing on screen when they shouldn't. It also begins to throttle at around 75(could have something to do with 'Target Temp' being 75 in the BIOS). I thought maybe it's my 1400mhz overclock but exactly the same thing happened at 1350. It got through the test at 1400mhz, despite the glitches and scored 14495, ~130pts over my previous best. Max temperature was 79C(since I adjusted the 'Max Temp' down to 75C from 85C). I cranked up the fan speed to 'jet engine' level and attempted another run at 1350 to see if this has to do with high temps. The throttling was reduced, the card only reached 72C and I was thinking this is a thermal issue until a black screen came. Note that this was a black screen in 3DMark. The system or drivers didn't crash. But yea, I guess it's time to loosen the timings.

EDIT 2: Managed to achieve stability in 3DMark. Started by changing TRRD(4>5) and TRP_WRA(53>55). At this point the bench still crashed, but at a later stage than before. Setting TRP_RDA(26>27) finally ensured a trouble-free run, despite an 83C temp. This was at 1350 so it remains to be seen what'll happen at 1400. The throttling is still an issue and I suspect the higher(1.137) auto voltage could be the culprit, as the card was previously holding a solid 1350mhz at 1.075V. Then again 1400mhz with throttling will still end up being faster than 1350 without throttling. Oh and oclmembench scores are exactly the same at these looser timings. Still getting 222-223GB/s at 1350core just as before.

@TheVic1600 see bottom of post for an updated vBIOS that should iron out the kinks, changelog in zip archive. I'd suggest uninstalling the driver with display driver uninstaller from safe mode then reinstalling it to make sure the driver isn't messing itself up which happens more regular than it should when you are changing memory timings. I'm shooting a bit blind when it come to the fan curve for your card so that might need tweaking still 1137mv I set as I'm unsure what vdroop is like on your card depending on the voltage regulator and if the phases are doubled or not vdroop can be as bad as about 60mv so I assumed worst case scenario here to be safe. The reason you can't hear the fan at 20% is because if your card is anything like the XFX RX590 I messed around with something is broken in the firmware or the drivers the fans won't actually spin at all until 35%, and if you set a lower RPM for the idle power limit it just breaks things even more. I'm going to point the finger at XFX for this one. Most GPUs, particularly Polaris, do run toasty but with some really good TIM, proper balancing of fan values and optimising that voltage to GPU frequency curve it is possible to have a fairly cool, power efficient, and quiet running card... even if its an XFX model :p

If you haven't already you might want to take a look at how the memory is being cooled on your card XFX cool the memory on the RX590 absolutely terribly (you can check my review of it for details) so its worth looking at your card. If you need to play with timings outside of what I've changed things to with this vBIOS the only options you'll need to try adjusting one at a time in order of what should be the most effective (and reverting the previous change) TRP_RDA 26 > 27, TRRD 4 > 5, TCCDS 5 > 6. Thats it, just those 3 timings, you can try to squeak TRP_WRA down to 53 if you are really looking for every last drop but 55 is the optimal value between stability and performance. In fact those screen blackouts you had by more than any other settings are caused by TRP_WRA being set too tight or TRRD, typically the former not the latter Samsung is REALLY good at holding a stupid tight TRRD. You might also want to try increasing VDDCI to 0.95v if you get memory errors before adjusting timings benefits with a lower or higher VDDCI is completely hit and miss sometimes slightly lower is better, sometimes not. The only constant here is that higher than 0.95v absolutely does not help with memory stability. Also I'm stupid and looked at the memory straps wrong in that other vBIOS, fixed that now.

@ket Thanks! I'll give it a go later and I'll also do a reinstall with DDU if I encounter problems again. As for voltages, the card is all over the place when I'm gaming/benchmarking. With your 1137mv setting I'm seeing anything from 1151 to 1056, with peaks close to 1.2V depending on the load put on the GPU. With a 1075mv setting I'm seeing between 1.1V and 1V with peaks up to 1.15. So it seems that for X max auto voltage, I'm getting X+15mv max and X-75mv min with X+70mv peaks. Vdroop is very hard to gauge in these conditions so I'll probably tweak the voltage myself and see if I can get rid of the throttling. I know the card didn't throttle at 1350mhz@1075mv, but it couldn't hold 1400mhz with that voltage, so maybe somewhere between 1137 and 1075 would be the sweet spot for 1400mhz.

As for the fans, I am 100% sure they are spinning@20% as I have checked with the flashlight on my phone and have seen them start to spin as soon as Windows gets to the login screen. HWiNFO also shows the GPU Fan at around 1k RPM when the card is idle and cold.

https://www.overclockersclub.com/reviews...gb_xxx_edition/
This is a review of my card where they take it apart. The memory cooling situation doesn't seem too bad if you ask me. They seem to be right under the heatsink under the fans. But then again I don't know much about GPU coolers. I imagine it would be better if they stuck some small heatsinks on the modules. Either way, the fact that the card passed the 3DMark test with looser timings despite being over 80C makes me think thermals aren't a big problem on this card right now.

As for VDDCI, if you look at the screenshots from my previous post, you can see that it's always locked at 838mv under load. I don't know if you adjusted it or not but I remember when I tried to change it, I only found 3 of those 4 B6 03 values I'm supposed to change. Also, while the first 2 were at the same offsets as the ones in your guide, the 3rd one was at a completely different one. Not wanting to risk frying my IMC, I left it as it is. Odd that it stays between 838 and 869mv when it's actually set at 950mv. Must be another great XFX BIOS "feature".