The RX Vega 64 Owners Thread

[TrixX]

Another quality reference cooler from AMD then! (In fairness I'm not a big fan of the Nvidia ones either)

To be fair while not the quietest cooler going it's actually good considering it's a reference cooler. A bit of tuning and you can get good performance reasonably quietly.

 

[shankly1985]

My gaming settings are
P6 1537 @1000
P7 1667 @1025
HBM 1110 @1025
50% power limit

This gives me 1610-1625 in games. I set my temp target to 72, this keeps the HBM under 80c (hot spot gets up to about 85). The fan runs around 2800-2900 but initially spins up to 3200 or so to catch the temp rising. GPU-Z gives the core power draw as 190-200 Watts. Performance is excellent. It took a bit of tweaking but I'm very happy with the results.

I use:-

P6 1557MHz @ 975
P7 1642MHz @ 1125
HBM 1100 @ 975
38% PowerLimit

I get ~1600-1625 in games. I found if I had P7 1657MHz, under certain circumstances of 3D load, it would over boost to a level where GPU crashed. So by bringing it down it solved that. I also found if I set 1100mV or 1112mV or 1125mV I got same ~1.075V when GPU loaded, but 1125mV made it more stable in various testing.

Thanks guys are you both running Air cooled cards?

 

[gupsterg]

Thanks guys are you both running Air cooled cards?

NP , mine's water cooled RX VEGA 64 AIR.

 

[r22snapper]

Yep, powercolour reference 64 air cooled

 

[Illuminist]

I just got a run on firestike of 24,045 graphics score using
1133 hbm @ 1133mv (;P)
1600 p6 @1133mv
1700 p7 @ 1200mv

was running at 1665
My best score yet
56 with 64 bios Air
+50%
temp
78 core
96 hotspot
84 hbm
Ignore this, i just beat it again*

 

[TonyTurbo78]

I just got a run on firestike of 24,045 graphics score using
1133 hbm @ 1133mv (;P)
1600 p6 @1133mv
1700 p7 @ 1200mv

was running at 1665
My best score yet
56 with 64 bios Air
+50%
temp
78 core
96 hotspot
84 hbm
Ignore this, i just beat it again*

Those are incredible temps for an air cooled ref card considering you're pushing masses of voltage at the card and overclocking the nuts out of it. Thats really aggressive for air cooled so fair play to you, got yourself a good one there overclocking that high and keeping temps in check with that ref blower. What fan speed you running for those results?

 

[Illuminist]

Those are incredible temps for an air cooled ref card considering you're pushing masses of voltage at the card and overclocking the nuts out of it. Thats really aggressive for air cooled so fair play to you, got yourself a good one there overclocking that high and keeping temps in check with that ref blower. What fan speed you running for those results?

It runs 100% fans after a few minutes, I beat this and got 24,629 graphics score with lower hbm

 

[Illuminist]

My overall score is massively let down by the cpu score, otherwise I think i would be in with a shot at the leaderboard.
I cant wait to get it on water and lc bios

 

[EsaT]

It runs 100% fans after a few minutes, I beat this and got 24,629 graphics score with lower hbm

Which isn't wonder when with such volts GPU only consumption (without memory, VRM losses and fan) is likely going way past 250W.

Comparing profile power limits to what TPU measured for card is good way to deduct roughly how much those make.
https://www.bit-tech.net/reviews/tech/graphics/amd-radeon-rx-vega-64-and-rx-vega-56-review/4/
https://www.techpowerup.com/reviews/AMD/Radeon_RX_Vega_64/29.html

 

[Illuminist]

Which isn't wonder when with such volts GPU only consumption (without memory, VRM losses and fan) is likely going way past 250W.

Comparing profile power limits to what TPU measured for card is good way to deduct roughly how much those make.
https://www.bit-tech.net/reviews/tech/graphics/amd-radeon-rx-vega-64-and-rx-vega-56-review/4/
https://www.techpowerup.com/reviews/AMD/Radeon_RX_Vega_64/29.html

This was purely to see how far i could push the card, i don't mind the watts, when its on water and LC bios i will undervolt it as well as push it.
My aim is really to get 1800 core and 1200 hbm, if i can do one run with that and record the firestrike score I will stop playing with it XD
I love this card so much it's so weird (like me) and fun to mess around with

 

[lordhawkwind]

I've got the Sapphire AC reference 64 card and my tweaks are as follows;

P6 1552
P7 1632
MV 975
HBM 975/ 1050mv (stock)
Power +50
Fan 2600 rpm.

Runs like a charm and the noise from the blower is not a big issue TBH. The core runs at a solid 1550 and the ram 975 which I think is pretty good for an AC card. In Wolfenstein II averaging well over 100 fps and often hitting my 144hz freesync ceiling on Ultra settings. Very happy.

 

[lordhawkwind]

This is a bit more like it! https://wccftech.com/amd-rx-vega-gets-22-performance-uplift-wolfenstein-2-latest-patch/ Now beating 1080ti woot.

Although I must say just went on Steam to test it out and it says my saves are from an older version of the game WTF. Only played it a week ago. It now wants to send me back to the start of the level which is a POS TBH. I thought Steam games were updated automatically?

 

[Hotwired]

Optimisation of game code seems to giving Vega some legs there.

 

[ICDP]

I've got the Sapphire AC reference 64 card and my tweaks are as follows;

P6 1552
P7 1632
MV 975
HBM 975/ 1050mv (stock)
Power +50
Fan 2600 rpm.

Runs like a charm and the noise from the blower is not a big issue TBH. The core runs at a solid 1550 and the ram 975 which I think is pretty good for an AC card. In Wolfenstein II averaging well over 100 fps and often hitting my 144hz freesync ceiling on Ultra settings. Very happy.

If you have HBM at 1050mV then your P7 mV is also 1050. HBM is a voltage floor that the GPU will not drop below in P7 state (and I think P6). The fact you have stock clocks on P7 and are getting 1550 core indicates you are not dropping under 1050mV. From my own testing on the AC V64 1050mV is getting just short of 1600 core.

Now I am not stating your OC is poor, or that you should not be happy. Just setting the record straight on a common misconception that regretfully seems to persist in relation to HBM "voltage" and RX Vega.

I'm surprised so many feel that 2600 or even higher RPM on the stock fan is not a big issue. More power to them if they do (pun intended). IMHO anything over 2300 is obnoxiously loud on V64 AC.

 

[TonyTurbo78]

It runs 100% fans after a few minutes, I beat this and got 24,629 graphics score with lower hbm

Very good, nice indeed. On side note, I managed around 26k graphics score when pushing 1800mhz so good luck. I'm also let down by ageing 3770k ddr3 1600mhz and z77 motherboard. Once I update rest of rig I'll take the bench more serious lol.

 

[r22snapper]

I'm surprised so many feel that 2600 or even higher RPM on the stock fan is not a big issue. More power to them if they do (pun intended). IMHO anything over 2300 is obnoxiously loud on V64 AC.

I have a quiet case and play loud games

The advantage to it being so tweakable is that I also have a quiet setting that keeps the fan below 2400rpm and still gives improved performance over the stock settings.

 

[gupsterg]

Just setting the record straight on a common misconception that regretfully seems to persist in relation to HBM "voltage" and RX Vega.

And let me help you so I may also straighten out the misconception and you may continue to straighten it out.

HBM mv in WattMan and OverdriveNTool has diddling squat to do with:-

i) GPU voltage floor limit.
ii) HBM voltage.

1st read this post carefully, noting how the PowerPlay is and how SOCCLK and MCLK have ucVddInd, do read note for this label per table as it differs in what it is.

2nd read this post.

@AMDMatt

I request AMD driver team resolve this borked implementation of HBM mV in WattMan. It is causing confusion and issue, rather than helping anyone IMO.

 

[ICDP]

And let me help you so I may also straighten out the misconception and you may continue to straighten it out.

HBM mv in WattMan and OverdriveNTool has diddling squat to do with:-

i) GPU voltage floor limit.
ii) HBM voltage.

1st read this post carefully, noting how the PowerPlay is and how SOCCLK and MCLK have ucVddInd, do read note for this label per table as it differs in what it is.

2nd read this post.

@AMDMatt

I request AMD driver team resolve this borked implementation of HBM mV in WattMan. It is causing confusion and issue, rather than helping anyone IMO.

Thanks for the links and info. The HBM mV entry in WattMan is indeed confusing, the only thing I could see (or perceive) it physically did was to set a lower voltage floor or budget. So the mV setting in Memory mV would override any setting in P-State 7 (and vice-versa) IF it was lower. For example

P-State 7 mV: .975
Memory mV: 1050
Would result in the GPU using 1050mV

P-State 7 mV: 1000
Memory mV: .950
Would result in the GPU using 1000mV

P-State 7mV: .975
Memory mV: .950
Would result in the GPU using .975mV

So whatever mV setting was the highest in these two setting is what your GPU uses. It is literally pointless setting these differently if your aim is to use the least volts possible. The absolute lowest I could reach was .900mV in P-State 6, 7 and Memory.

 

[gupsterg]

I am willing to spend time helping you improve your thread on OCUK. By this I detract no value from the share you did . It will be simple testing that I will guide, so you can see what I see, if you are willing to do it. I will post in there later today when home.

Yours is a good thread that I did ref. I also ref'd many others posts and experiences, who had had VEGA before me. I choose to do my own testing and path to gain better understanding.

VEGA is by far more advanced on clocks, voltages and how it determines what it does than any past AMD GPU I have experienced. You'll see a post in the OCN owners thread where I did some testing with Dead Space uncapped FPS and capped. This gave some real good insight into how VEGA is.

VEGA has 2 distinct approaches to clocks/voltage.

i) non ACG/AVFS, DPM 0 to 4, capped FPS Dead Space used this.

ii) ACG/AVFS, DPM 5 to 7, uncapped FPS Dead Space used this.

Same app, but GPU choose differing method, due to case usage. How it chooses is not based on app or FPS but a combo of these IMO.

When non ACG/AVFS occurs it will have a very flat voltage used. ACG/AVFS has more "bounce" for voltage.

 

[4K8KW10]

This is a bit more like it! https://wccftech.com/amd-rx-vega-gets-22-performance-uplift-wolfenstein-2-latest-patch/ Now beating 1080ti woot.

Although I must say just went on Steam to test it out and it says my saves are from an older version of the game WTF. Only played it a week ago. It now wants to send me back to the start of the level which is a POS TBH. I thought Steam games were updated automatically?

They warned us before launch - the drivers are not ready, the drivers are not ready. Looks like a proof - the software isn't ready for RX Vega 64.