RX Vega Owners Thread

When I tried different settings on my WC Vega64 the Power target at 0% kept the clocks in state6 and wouldn't touch state7, so it would only do 1660 and not the 1750 from state7. Vega is far too new and dynamic for the old ways, which I prefer. You may need to increase the power target to +20% at least and still lower voltage to get higher stable clocks even the stock clocks.

Thing is, even if I make P6 close to P7, it seems to go back all the way down to something else. Thanks, I'll try tinkering a bit more, but honestly this cooler is pretty much on the edge even at 0% PT. At 25% it's too loud for my liking.
 
I agree 100%..Set your state 6 to 1750Mhz (since these cards to not ever reach state 7) and set your Voltage to 1.05V with 50% Power Limit..You will see much better performance then stock with the same or a tiny increase in power usage. I am still humming along with my 80% PL and my Vcore @ 1.05V..Still seeing 1725-1750 depending on how stressful the load. I honestly couldn't be happier..

I have tried this, but I'm seeing a massive increase in power usage up to the 50% PT, and the performance is still below what it would be at standard voltage. In short, leaving clocks and volts at default and PT +50% leads to much better performance than undervolting and/or overclocking at PT 50%.

So I guess it's just me undervolting doesn't work for.
 
I have tried this, but I'm seeing a massive increase in power usage up to the 50% PT, and the performance is still below what it would be at standard voltage. In short, leaving clocks and volts at default and PT +50% leads to much better performance than undervolting and/or overclocking at PT 50%.

So I guess it's just me undervolting doesn't work for.
When I tried undervolting, it was crash and burn all the way. Not sure if the software is still messing things up for some of us. Since now I have power and the internet back I can start playing again.
 
When I tried undervolting, it was crash and burn all the way. Not sure if the software is still messing things up for some of us. Since now I have power and the internet back I can start playing again.
I will take the crashing! That means the undervolt is working. For me, it almost behaves like an overvolt+underclock.
 
x.9.2 so far is much more stable than x.9.1, which would BSOD crash from time to time. Been using the card nonstop since x.9.2 came out and so far all stable <knock on wood>
 
I will take the crashing! That means the undervolt is working. For me, it almost behaves like an overvolt+underclock.

What card did you have installed before your VEGA? If you had an AMD card, I HIGHLY recommend you DDU in safe mode and do a clean install of 17.8.2 or 17.9.2. I came from a pair of 290s and even though VEGA worked fine, WattMan was a bit funky. I had no idea how funky it was acting until I DDU'd and did a clean install of 17.8.2 (Ive been so busy the last 11 days I have barely touched my system so I have not tried 9.2 but I've read a lot of good things.)...

If you have a VEGA 56, I would also highly recommend you flash the BIOS from the VEGA 64...Everyone that has had issues with undervolting that I have read about are using the stock VEGA 56 bios. I know of two guys at OC.net that had their issues fixed by flashing to a 64 bios..

It really is zero risk, since you can flash right from Windows with the newest ATI Tool and the fact that the card has one of the BIOS locked, so even if you were to somehow mess up the flash you can flip the switch and reflash from stock. Oh, that reminds me, you are using the Performance BIOS right? The BIOS switch should be on the closest to the I/O bracket (toward the rear of the card)..
 
Tested again with CF working.

SuperCFOC_Small.png
 
What card did you have installed before your VEGA? If you had an AMD card, I HIGHLY recommend you DDU in safe mode and do a clean install of 17.8.2 or 17.9.2. I came from a pair of 290s and even though VEGA worked fine, WattMan was a bit funky. I had no idea how funky it was acting until I DDU'd and did a clean install of 17.8.2 (Ive been so busy the last 11 days I have barely touched my system so I have not tried 9.2 but I've read a lot of good things.)...

If you have a VEGA 56, I would also highly recommend you flash the BIOS from the VEGA 64...Everyone that has had issues with undervolting that I have read about are using the stock VEGA 56 bios. I know of two guys at OC.net that had their issues fixed by flashing to a 64 bios..

It really is zero risk, since you can flash right from Windows with the newest ATI Tool and the fact that the card has one of the BIOS locked, so even if you were to somehow mess up the flash you can flip the switch and reflash from stock. Oh, that reminds me, you are using the Performance BIOS right? The BIOS switch should be on the closest to the I/O bracket (toward the rear of the card)..

Interesting. I might give the BIOS flashing a shot.
 
What card did you have installed before your VEGA? If you had an AMD card, I HIGHLY recommend you DDU in safe mode and do a clean install of 17.8.2 or 17.9.2. I came from a pair of 290s and even though VEGA worked fine, WattMan was a bit funky. I had no idea how funky it was acting until I DDU'd and did a clean install of 17.8.2 (Ive been so busy the last 11 days I have barely touched my system so I have not tried 9.2 but I've read a lot of good things.)...

If you have a VEGA 56, I would also highly recommend you flash the BIOS from the VEGA 64...Everyone that has had issues with undervolting that I have read about are using the stock VEGA 56 bios. I know of two guys at OC.net that had their issues fixed by flashing to a 64 bios..

It really is zero risk, since you can flash right from Windows with the newest ATI Tool and the fact that the card has one of the BIOS locked, so even if you were to somehow mess up the flash you can flip the switch and reflash from stock. Oh, that reminds me, you are using the Performance BIOS right? The BIOS switch should be on the closest to the I/O bracket (toward the rear of the card)..

This is what I did. Flashed the V64 BIOS, DDU in Safe Mode. Installed 17.9.2. The BIOS works, but exactly the same issue persists. I even tried two different V64 BIOSes, same story.

In fact, I don't even need to change the voltage. Just flipping the voltage adjustment switch from "manual" to "auto" means the clocks are lowered. It's basically power saving performance at balanced power draw. After that, it doesn't matter what voltage I input, it just leads to lower clocks. Yes, I have tried OverdriveNtool too, same issues.

At this stage, I have tried everything, same thing every time. I think I'll go back to BIOS #2 stock with a memory OC, I suppose that still has the best perf/W.
 
This is what I did. Flashed the V64 BIOS, DDU in Safe Mode. Installed 17.9.2. The BIOS works, but exactly the same issue persists. I even tried two different V64 BIOSes, same story.

In fact, I don't even need to change the voltage. Just flipping the voltage adjustment switch from "manual" to "auto" means the clocks are lowered. It's basically power saving performance at balanced power draw. After that, it doesn't matter what voltage I input, it just leads to lower clocks. Yes, I have tried OverdriveNtool too, same issues.

At this stage, I have tried everything, same thing every time. I think I'll go back to BIOS #2 stock with a memory OC, I suppose that still has the best perf/W.
You are power limited. Either further decrease voltage (including voltage floor), increase power limit or decrease clock.
 
You are power limited. Either further decrease voltage (including voltage floor), increase power limit or decrease clock.
No matter what voltage I input, 800, 900, 1000, 1050, 1150, 1200; it does it's own thing. I'm best off leaving it at Auto. So, in effect, I'm unable to undervolt or select my voltage. That is exactly my problem, that I'm not able to change voltage.
 
You are power limited. Either further decrease voltage (including voltage floor), increase power limit or decrease clock.

He definitely seems to be power limited...I am curious if it is an actual hardware defect. There is no way a fully functional core should be down clocking like that. I was power limited myself despite the massive thermal headroom I had until I modified the Power Play tables.

No matter what voltage I input, 800, 900, 1000, 1050, 1150, 1200; it does it's own thing. I'm best off leaving it at Auto. So, in effect, I'm unable to undervolt or select my voltage. That is exactly my problem, that I'm not able to change voltage.

Something is definitely wrong. You are using the performance BIOS correct?
 
got back from spain at 2am last night....and this morning this happened

Was super cautious mounting block, I don't think I've ever handled a card that I was so scared to mount a cooler/block on...the whole hbm thing makes it feel super fragile.

IMG_20170925_110958312.jpg
;):D

Die has epoxy. Does not undervolt as well as my first card, crashes under 1150 whereas previous card was content with 1050. HBM seems to do 1050 without issues. have not pushed core farther than stock. Loading around 45c or so, silence is bliss.
 
Last edited:
What card did you have installed before your VEGA? If you had an AMD card, I HIGHLY recommend you DDU in safe mode and do a clean install of 17.8.2 or 17.9.2. I came from a pair of 290s and even though VEGA worked fine, WattMan was a bit funky. I had no idea how funky it was acting until I DDU'd and did a clean install of 17.8.2 (Ive been so busy the last 11 days I have barely touched my system so I have not tried 9.2 but I've read a lot of good things.)...

This x10.

I tried the AMD "clean install" option which uninstall the drivers and reinstalls after a reboot... still was having the "Enhanced Sync" missing issue.

Only after a DDU in safe mode and a reinstall to that same driver level I just tried did the Enhanced Sync come back.
 
got back from spain at 2am last night....and this morning this happened

Was super cautious mounting block, I don't think I've ever handled a card that I was so scared to mount a cooler/block on...the whole hbm thing makes it feel super fragile.

View attachment 37589
;):D

Die has epoxy. Does not undervolt as well as my first card, crashes under 1150 whereas previous card was content with 1050. HBM seems to do 1050 without issues. have not pushed core farther than stock. Loading around 45c or so, silence is bliss.


Great that you have a working card again. Sucks that you didn't win the silicon lottery like your older card, but at least it has the epoxy this time. See if you can push the HBM to 1100 or 1105 which seems to be the ceiling.
 
He definitely seems to be power limited...I am curious if it is an actual hardware defect. There is no way a fully functional core should be down clocking like that. I was power limited myself despite the massive thermal headroom I had until I modified the Power Play tables.

Something is definitely wrong. You are using the performance BIOS correct?

Yup, performance BIOS. In fact, I tried flashing both V64 Performance and Power Saving BIOS. Though if I'm not mistaken, you can undervolt on the power saving BIOS too. Either way, I've tried both. At this point, it seems like I've tried everything :) I doubt it's a hardware defect, as everything works fine. It's just that I'm unable to adjust voltage manually. If I hit Manual, it picks up some arbitrary high voltage that's worse than "Auto". Irrespective of what voltage I actually enter.
 
Happy to report that my undervolting finally works fine with TRIXX 6.5 :) Stable at -175 right now. Yes, would love to see HBM voltage control in the future. But for now, this seems like the most reliable utility for Vega.
 
Very nice, only thing it does not have is HBM voltage control. The monitor is best I've seen as well, very easy to read even at 4K.

You guys are aware that what you call "HBM memory voltage control" in Wattman has ZERO effect on HBM voltage? It is actually a minimum core voltage. There is an interesting discussion on it over on Overclock.net.
 
Last edited:
Happy to report that my undervolting finally works fine with TRIXX 6.5 :) Stable at -175 right now. Yes, would love to see HBM voltage control in the future. But for now, this seems like the most reliable utility for Vega.

So glad to hear that!
 
Happy to report that my undervolting finally works fine with TRIXX 6.5 :) Stable at -175 right now. Yes, would love to see HBM voltage control in the future. But for now, this seems like the most reliable utility for Vega.
Oh great, I'm going to have to try this. My new vega can't umdervolt at all and I can't seem to get better than 1620mhz even the ugh it's at around 45c.
 
Oh great, I'm going to have to try this. My new vega can't umdervolt at all and I can't seem to get better than 1620mhz even the ugh it's at around 45c.

If you're on 17.9.1 or 17.9.2 I am hearing a lot of bitching about clock instabilities on various forums with these currently (where 17.8.2 was stable at a clock and now crashes).
 
If you're on 17.9.1 or 17.9.2 I am hearing a lot of bitching about clock instabilities on various forums with these currently (where 17.8.2 was stable at a clock and now crashes).
Mine hasn't seemed to change but then again mine pretty much was close to max from a slider point anyway.
 
Question to the Vega 56 owners - my current PSU is happily powering a GTX 780 which has a TDP of 250W. The PSU has a single PCIE power cable with a 6+2 and a 6 pin connector, which is what the GTX 780 requires. I want to upgrade to the Vega 56 which has a TDP of only 210W but requires a 8+8 PCIE connector.

I should be able to simply buy a 6pin-8pin PCIE adapter and use my current cable and PSU, correct? The same single cable that is powering the 250W graphics card will be used to power a 210W graphics card.
 
Happy to report that my undervolting finally works fine with TRIXX 6.5 :) Stable at -175 right now. Yes, would love to see HBM voltage control in the future. But for now, this seems like the most reliable utility for Vega.
Using TRIXX 6.5, 1170mv, crash and burn about 5 minutes into Serious Sam Last Hope VR, went back to stock volts no issues for over 3 hours before I stopped. Have not tried the other bios setting yet . . . hmmmm which one is the performance one?
 
You guys are aware that what you call "HBM memory voltage control" in Wattman has ZERO effect on HBM voltage? It is actually a minimum core speed. There is an interesting discussion on it over on Overclock.net.
So voltage adjustment is not voltage adjustment - no I did not know that, thanks.
 
Question to the Vega 56 owners - my current PSU is happily powering a GTX 780 which has a TDP of 250W. The PSU has a single PCIE power cable with a 6+2 and a 6 pin connector, which is what the GTX 780 requires. I want to upgrade to the Vega 56 which has a TDP of only 210W but requires a 8+8 PCIE connector.

I should be able to simply buy a 6pin-8pin PCIE adapter and use my current cable and PSU, correct? The same single cable that is powering the 250W graphics card will be used to power a 210W graphics card.

This will probably work:

Smitty2k1 You can get a molex adapter if you have some of those free on your PSU. Like this:
https://www.newegg.com/Product/Product.aspx?Item=N82E16812400034
Or the 6-pin to 8-pin on that same link I think might work.

Just be warned that pushing the limits of a PSU is never a great idea.
 
You guys are aware that what you call "HBM memory voltage control" in Wattman has ZERO effect on HBM voltage? It is actually a minimum core speed. There is an interesting discussion on it over on Overclock.net.
Yes, and that's why we're calling for HBM voltage control in Trixx. Currently the only way to do that seems to be BIOS flash.
 
Alright, I can get it stable at the stock 1630 and 1050 (so it maintains ~1590) at +25% and -70mv (1030). About the best I can get. If I bump up core it freezes in superposition, wawawa. Max temp of 51c core (stays in mid to high 40s typically) and hbm is upper 40s with a max of 54c. Radiator is DUMPING heat, happy I mounted it to the roof of my case. Going to try to get some more airflow over the card, wondering if maybe circuitry is hurting and getting too hot from less airflow.

Previous card definitely clocked better, was hitting close to the same clocks at +15% and -150mv. Guess I can't complain too much.
 
Alright, I can get it stable at the stock 1630 and 1050 (so it maintains ~1590) at +25% and -70mv (1030). About the best I can get. If I bump up core it freezes in superposition, wawawa. Max temp of 51c core (stays in mid to high 40s typically) and hbm is upper 40s with a max of 54c. Radiator is DUMPING heat, happy I mounted it to the roof of my case. Going to try to get some more airflow over the card, wondering if maybe circuitry is hurting and getting too hot from less airflow.

Previous card definitely clocked better, was hitting close to the same clocks at +15% and -150mv. Guess I can't complain too much.

You need to modify the power play tables man. With your level of cooling there is very little reason not too (although your core temps are much much higher then mine unless you are referring to the max hotspot temp which for me is 47C after 3-4 hours of stress testing)...I am running @ 1.05 with 80% PL limit and never see it drop below 1725Mhz. It usually stays right @ 1750Mhz according to RT OSD.

EDIT: sabrewolf732 , how much rad are you running? I am running a 360 (60mil thick) up top and a 230mm up front with twin 200mm fans in push/pull. Your CPU should be pulling much much less power then mine lol..I idle @ 144W~168W depending on where I have my fans set and draw ~220W+ with an AVX load @ 5ghz...With VEGA @ 80% PL I am drawing ~510W from the wall according to my UPS. So that means VEGA is drawing ~241W after accounting for my PSU's 94% efficiency rating @ 500W+..
 
You need to modify the power play tables man. With your level of cooling there is very little reason not too (although your core temps are much much higher then mine unless you are referring to the max hotspot temp which for me is 47C after 3-4 hours of stress testing)...I am running @ 1.05 with 80% PL limit and never see it drop below 1725Mhz. It usually stays right @ 1750Mhz according to RT OSD.

EDIT: sabrewolf732 , how much rad are you running? I am running a 360 (60mil thick) up top and a 230mm up front with twin 200mm fans in push/pull. Your CPU should be pulling much much less power then mine lol..I idle @ 144W~168W depending on where I have my fans set and draw ~220W+ with an AVX load @ 5ghz...With VEGA @ 80% PL I am drawing ~510W from the wall according to my UPS. So that means VEGA is drawing ~241W after accounting for my PSU's 94% efficiency rating @ 500W+..

Info on modifying powertables?
Running a 240 (no idea what rad, maybe an old black ice) and 2 quiet 120s. Those temps are peak per trixx hw monitor. After more playing around it seems like -115mv is stable and will maintain 1575ish, but it seems like if it tries to bump up past 1630 it crashes so its not hitting the final power state maybe? IDKSOLOL. Vega is complicated.
 
Vega doesn't ever hit the P7 state.. I can set my P6 to 1700Mhz and P7 to 1750Mhz, and it will stay @ 1700Mhz. If I set my P6 to 1750Mhz and P7 to 1755Mhz (P6 and P7 cannot be the same value on 17.8.2 or it gets stuck) it will hit 1750Mhz. Its funky, and I am not sure why AMD even included the P7 state. Maybe they figure we will all use LN2 to cool it during those long gaming sessions?:p:p:p:p

Anyway, for anyone interested here is the modified powerplay table: This will give you 142% PL, but do NOT go that high unless you are under a robust custom loop (I'd want at least a 240mm thick or 280/360 for the GPU alone). Here is the

To install this properly:

1. Reboot and DDU in safemode.

2. Reboot and install your choice of driver (you may need to use 17.8.2 at first but you can give 17.9.2 a try if you are already using it).

3. Open the zip file, and extract the file.

4. Double click on the file to add it to the registry.

5. Open Wattman, set your desired Voltage and crank that Power Limit Slider up as far as you want (up to 142%).

I would recommend starting with 75%, as you should see a nice healthy boost (since they GPU clock should stay around 1690-1715 on a so so card, or 1725-1750 on a good sample). I settled on 1.05V and 80% PL, since that gives me minimum frame rates of 75-80 @ 1440P all settings maxed in the games I am playing (mostly Maxed out GTA V with mods)...When I get back from my Vacay I am going to get a 1440P 144Hz or a 100Hz Ultrawide. Then I will crank that PL higher!!!:p

EDIT: In case you are wondering, this will work fine on a 56 flashed with the 64 bios. I am not sure if it will work on a stock 56 BIOS, but I believe it should since you are just modifying the PowerTables and how much current you pull at a given PL.
 
Info on modifying powertables?
Running a 240 (no idea what rad, maybe an old black ice) and 2 quiet 120s. Those temps are peak per trixx hw monitor. After more playing around it seems like -115mv is stable and will maintain 1575ish, but it seems like if it tries to bump up past 1630 it crashes so its not hitting the final power state maybe? IDKSOLOL. Vega is complicated.

I would recommend trying GPU-Z since it shows the core temp, and the HBM temp, and the GPU hotspot. Those temperatures still seem a bit high for 480mm of rad, but I know the older BI were a bit restrictive and you really need a fairly high end fan since they have a high FPI so its much harder to push cool air through them at lower fan speeds. I had a pair of the Quiet 120s in my old setup and I needed to keep my AP-15s @ 1100-1300RPM to get decent temps.
 
I would recommend trying GPU-Z since it shows the core temp, and the HBM temp, and the GPU hotspot. Those temperatures still seem a bit high for 480mm of rad, but I know the older BI were a bit restrictive and you really need a fairly high end fan since they have a high FPI so its much harder to push cool air through them at lower fan speeds. I had a pair of the Quiet 120s in my old setup and I needed to keep my AP-15s @ 1100-1300RPM to get decent temps.


Haha, I meant I have 1 radiator that being a single 240 with a set of quiet 120mm fans on it.

thank you very much for the powerplay table! Ill keep you posted!
 
Haha, I meant I have 1 radiator that being a single 240 with a set of quiet 120mm fans on it.

thank you very much for the powerplay table! Ill keep you posted!

Not a problem man, I hope it helps you eek a bit more clock speed outta your card. I'd be careful and keep your PowerLimit @ 75% max unless you replace that 240 with a super thick guy. I know that Ryzen is fairly cool running and low power, but if you have that CPU OC'd you aren't leaving a ton of cooling for the GPU. You are still better then the AIO 64 Version though, with a better pump so I think you will be fine @ 75-80%.
 
Back
Top