The RX Vega 64 Owners Thread

[Lokken86]

You have a great setting there. You just need to tweak it. Also it would be crashing if you keep P6 state at same voltage with P7.
It has to be lower.

Later tonight I will try to play with my Nitro+ and see at what point I can get 1600 at the lowest power available at 30% fan speed (watercooled) to give you an indication where to look at

You could yeah I'll certainly try but really I think the reference cards are often just not as good overclockers. I'm quite happy with it as it e.g Firestrike is up from 19000 to 20700 at less power then before.

 

[Panos]

You could yeah I'll certainly try but really I think the reference cards are often just not as good overclockers. I'm quite happy with it as it e.g Firestrike is up from 19000 to 20700 at less power then before.

Vega works in mysterious ways.

Really wish I had a reference one to do some testing with

 

[Lokken86]

Vega works in mysterious ways.

Really wish I had a reference one to do some testing with

It certainly makes you earn the higher speeds!

 

[Davedree]

You could yeah I'll certainly try but really I think the reference cards are often just not as good overclockers. I'm quite happy with it as it e.g Firestrike is up from 19000 to 20700 at less power then before.

Is that your graphics score? If so what clocks were you running. I have the same card it seems a bit low if honest

 

[Panos]

It certainly makes you earn the higher speeds!

I had more fun playing with the card settings than with the FuryX, GTX1080 and GTX1080Ti combined

 

[Lokken86]

Is that your graphics score? If so what clocks were you running. I have the same card it seems a bit low if honest

No that's the final score. The graphics score was 24000. I think the overall score should be more but the combined has decided to lower itself but that's another story ha.

 

[TonyTurbo78]

Is that your graphics score? If so what clocks were you running. I have the same card it seems a bit low if honest

Should be around 25k graphics score I think....mine on a average gpu overclock https://www.3dmark.com/3dm/23793903?

 

[Davedree]

No that's the final score. The graphics score was 24000. I think the overall score should be more but the combined has decided to lower itself but that's another story ha.

Ok no that's spot on then, was that holding a real 1550 1100hbm ?

 

[Lokken86]

Ok no that's spot on then, was that holding a real 1550 1100hbm ?

Yes that's holding 1550 throughout. Hbm I think I'd forgotten to increase past 1050 but it would maybe add another 100.

 

[Davedree]

Should be around 25k graphics score I think....mine on a average gpu overclock https://www.3dmark.com/3dm/23793903?

Yeah +1, out the box a reference v64 is around 20.5k. +50 power and stock clocks gets into the 24k. And the sweet spot on undervolt reference 1550-1600/1050-1100 hbm2 gets you into the 24.5-25k.

Yes that's holding 1550 throughout. Hbm I think I'd forgotten to increase past 1050 but it would maybe add another 100.

It's about 300points from hbm1050 to 1100 , so 2-3fps.

https://www.3dmark.com/3dm/28150459?
https://www.3dmark.com/3dm/28150428?

 

[Lokken86]

Yeah +1, out the box a reference v64 is around 20.5k. +50 power and stock clocks gets into the 24k. And the sweet spot on undervolt reference 1550-1600/1050-1100 hbm2 gets you into the 24.5-25k.


It's about 300points from hbm1050 to 1100 , so 2-3fps.

https://www.3dmark.com/3dm/28150459?
https://www.3dmark.com/3dm/28150428?

Cool my gpu score isn't lagging behind the Intel's too much then. I'll see if I can get 25k with the CPU cranked up a bit

 

[TonyTurbo78]

Yeah +1, out the box a reference v64 is around 20.5k. +50 power and stock clocks gets into the 24k. And the sweet spot on undervolt reference 1550-1600/1050-1100 hbm2 gets you into the 24.5-25k.


It's about 300points from hbm1050 to 1100 , so 2-3fps.

https://www.3dmark.com/3dm/28150459?
https://www.3dmark.com/3dm/28150428?

I really wanted to push and push and break 26k graphic score when I bought this card, but a few things happened when hunting a stable 1850mhz bench run that I bottled it and stuck with 1700 conservative lol.

 

[Lokken86]

I really wanted to push and push and break 26k graphic score when I bought this card, but a few things happened when hunting a stable 1850mhz bench run that I bottled it and stuck with 1700 conservative lol.

At 1700 surely it smashes 26k.

 

[Davedree]

I really wanted to push and push and break 26k graphic score when I bought this card, but a few things happened when hunting a stable 1850mhz bench run that I bottled it and stuck with 1700 conservative lol.

I know it's hard work as you have to dial it in on the voltage table otherwise the power limiter is so easily hit on 1080p firestrike gt1.Gt2 isn't so fussy. I had to use a LC bios to get there. But I haven't bothered since, too worried I was going to fry it.

 

[TonyTurbo78]

At 1700 surely it smashes 26k.

For my best run, as posted above, I think I had clock set to 1750mhz but 3dmark reckons it only hit 1682mhz which came out at 25.5k graphic score. I might have to redo them now I understand Vega so much better and drivers have come on a lot. But I reckon you'd have to get close to 1800mhz to get to 26k. I can't see any Vega 64 that break that number on the Firestrike Bench chart, not even Matt did it.

BIG EDIT...SORRY DAVEDREE, YOU BROKE 26K (just had another look). Quite surprised at that with 1722/1125. Fair play bud. Can we hit 27k lol.

 

[Lokken86]

For my best run, as posted above, I think I had clock set to 1750mhz but 3dmark reckons it only hit 1682mhz. I might have to redo them now I understand Vega so much better and drivers have come on a lot.

I'm still a Vega noob but one thing I have discovered is voltage plays a big part in clock stability. You probably know already but if you set Firestrike to window mode and run test one in a loop it's a great way to fiddle with settings and see what holds the clock speed best.

Then get back to us when youv cracked it

 

[TonyTurbo78]

I'm still a Vega noob but one thing I have discovered is voltage plays a big part in clock stability. You probably know already but if you set Firestrike to window mode and run test one in a loop it's a great way to fiddle with settings and see what holds the clock speed best.

Then get back to us when youv cracked it

yep thats how I always do it, best way to test for stability and peaks . Thanks anyway

 

[Lokken86]

yep thats how I always do it, best way to test for stability and peaks . Thanks anyway

Amazing that so many people and even reviewers aren't checking for these dips. Turbo mode even sets around 1550mhz on my card but it doesn't hold constant.

 

[Davedree]

For my best run, as posted above, I think I had clock set to 1750mhz but 3dmark reckons it only hit 1682mhz. I might have to redo them now I understand Vega so much better and drivers have come on a lot.

You could easily do maybe even 27k as you and the others are on water. On 18.5.1 you should see 26k at a held 1590mhz gt1, 1630mhz held gt2, hbm 1125.
https://www.3dmark.com/3dm/28150823?

Amazing that so many people and even reviewers aren't checking for these dips. Turbo mode even sets around 1550mhz on my card but it doesn't hold constant.

This has been the problem since forever with Amd unfortunately. Bar a few reviewers who truly know what they are doing, most just ran it out the box or failed to comprehend how to fiddle with the p-states, wondering why voltage and power limit doesn't get them anywhere.

 

[Sybra]

Trying to run crossfire vega 64, struggling to get it to work right.

Mostly play BF4 / Battlefront 2 on a 240hz monitor.

any advice on how to get it to run smoothly. Going from a 1080ti the video experience is far behind what I'm used to.

The video quality itself seems to lack and the smoothness of the frames seems a bit clunky? is that a way to describe it?

inb4 its the PSU. It's fine