Topic: AMD Mem Tweak XL - Read/modify timings/pp/straps on the fly - page 22. (Read 58831 times)

anyone found a fix for this?

"startservice() failed: this service cannot be started, either because it is
disabled, or because it has no enabled devices associated with it
     at WinFormsGUI.IOMap.LoadDriver()
     at WinFormsGUI.IOMap.Program.Main()"

Could you share you core/mem clocks and the whole rig power? Are the bioses flashed to 64?

I have 3 of those, for ETH mining, push -REF 65500.
=> 46 MH/S with much lower clock and much lower power (140 to 150W@wall).

I used to be able to do this but can no longer change on the fly after the Win10 update and a driver update.  I may try to go back to older drivers and see if that changes.  To me, that was the best part about the tool ... changes on the fly to see the effects.

In fact, once I run Claymore, I can't change any values even after I stop Claymore.  I have to reboot to make the values modifiable again.

This is Claymore 11.9, not the newer 14.x stuff that uses this same mechanism to adjust the card timings.  I can see where that might conflict.  But 11.9 was allowing mods during mining previously.

Yes. I didn't want to go into all that since this thread is for the Tweak tool, but I switched from "+" mode to "*" mode to get the best hr in TRM as they say to do in their readme with tweaked timings.

Also worth highlighting is that using the newer Teamredminer versions, you should set your overclocks, set your timings, and then run the auto-tuner for the algo you want to mine. Let it do its thing for a while and end up with a perfectly tweaked setup.

My Vega 56's (flashed to 64) are happily chugging XMR @ 2340 h/s each.

I actually had someone else tell me they could change the setting while mining too. For whatever reason (something with the drivers I assume), I could only change it before the miner started. I just thought I'd mention it for those that were seeing the same thing as me.

EDIT: I Just saw your edit, so it appears you can change on the fly with newer drivers, but might not work with older drivers.

This is not true. AMD driver 19.4.3. I change REF on my RX 580 cards during mining and it applyies and changing the mining speed.

Hey everyone,

I started messing with the Tweak tool yesterday, and after a day and half and lots of reading through this thread, I think I've found some good settings.

I have several Polaris (RX4xx/5xx) rigs with flashed timings, but still found that setting gave 3-5% increase on both CN/R and ETH on GPUs with Hynix memory. Elpida cards also saw a small increase (2-3%). Samsung with UberMix 3.1 timings did not get much of a benefit (maybe 10h/s on CN/R) and actually showed a few hardware errors in TRM, so I left the default setting on Samsung. Default value on all GPUs was 5.

Important: The setting MUST be applied before mining starts or the value won't change. Even if I closed the miner and tried to apply the setting, it would not take. However, applying via .bat after reboot before the miner started worked great.

Edit: The above method might only apply to older drivers (18.6.1, 18.5.1). It might be possible to change the settings on the fly with newer drivers, but I can't confirm since I'm not using them.

Average CN/R hashrate: 970+h/s. Average ETH hashrate: 29-30+Mh/s

Clocks: 1200 core / 2000 mem (+/- 100 depending on GPU)

System: Win10 (various versions from 1709 to 1903). Driver: 18.6.1 on all rigs
Miners used: Teamredminer v0.5.x, PhoenixMiner 4.2c

I also have one Sapphire Vega56 with Samsung (not flashed, stock bios) that I tested with. Core/Mem settings: 1407/975 @ 905mV. The only parameters I changed were: which I adapted from Mashy81's settings here. Thanks.
Below are the rest of the settings showing remaining stock settings:


With these settings I was able to get 2130h/s on CN/R (increased from 1980) and 43Mh/s on ETH (increased from 39.5) with the same miners above. Additionally, 2055h/s on CN_Haven.

Apologies if this has been posted in this thread (while I did read a lot, I didn't read every post), but I hope this is helpful for someone else out there.

EDIT: After running for a few hours, I got 1 HW error in TRM. So I changed " --RC 36 --RP 10 " to " --RC 37 --RP 11 ". Hashrate only dropped to 2110h/s. New settings:
I will update if I make any more changes.

I'm on the new Win10 May update and running Radeon driver 19.5.2.  I'm finding that I can run and apply AMDmemTweak changes prior to running Claymore 11.9.  However, once Claymore is running, changes in AMDmemTweak will not longer apply dynamically.  Occassionaly AMDmemTweak will throw an "application error" dialog box but otherwise continues to run.  I've turned off the new Tamper Control in Defender as well.

I can't run Radeon 19.6.2 because it throws a software error and exits when I attempt to enter the setting page.

A month ago, I had absolutely no issues of any kind running and applying AMDmemTweak settings.  I'm sure it's something in the new windows or newer drivers.

I'm using AMDmemTweak 0.2.3 BTW.

what exactly is Refresh here? the 5 value added newly? why wasn't it around before? how it links to other refresh timings "tRFC"... so increasing it (beside the strap being already customized for CN or ETH) tends to give better performance? I do mine CN Heavy variants still. let me know

Better late than never, this tool helps a lot for xmr miners. Great freeware tools are always welcome.

AFAIK it is built into the linux drivers now look up pp_od_clk_voltage

Hi I have quite a few different models of GPU is anyone selling timings for certain models ??
Or can I just take the straps that are used in moded bioses and decode and apply using amdmemtweak ?

Also I can see in my bios the list of possible memory chips models/vendors but is there any way to actually tell which model is installed ?

For Samsung i use 30 best stability, on Hynix in Claymore is 1 best performance but not so good stability

It can be increased up to 100+ for RX 5XX cards. But max stable amount depends on overclock and card quality. This parameter must be select individual for each card. Just need to try and if GPU start give incorrect shares lower REF parameter. Or increase it step by 10 and look into speed and stability...

Yes, increase REF to 12.

Does anybody tried on  Sapphire Nitro+ Rx 580 4gb?

What's your Temps? Strix arez or strix? What's your p6 state?

Thanks I'll try that