The RX Vega 64 Owners Thread

[shankly1985]

@KentMan Over 18 hours in on BF5 with my VEGA 64 and not a single crash, plays excellent on ultra settings 1440p

 

[eeii]

@Illuminist

What’s your power supply rating. I’m feeling a crossfire itch coming on

 

[TrixX]

@Illuminist

What’s your power supply rating. I’m feeling a crossfire itch coming on

After reading what Illuminist said I'm feeling the same...

I do have a 1200W PSU after all

 

[Panos]

I am doing so every day, true there are a few games that don't support crossfire, but of the games i have i'd say 80% can use both cards, and in farcry 5 at 4k ultra with hd textures i get around 75fps.
With Elite i get 70fps on 5k with ultra everything ....

I read all the time people hating on crossfire "theres no support" "it doesnt work" "microstutter" I assume all these people have never ever seen or used it as I am having an awesome blast of a time running my games at 2080 ti speed and more for £700 total spend.

Once again, crossfire is awesome and does work.

Do you have Wolfenstein or other async compute supporting games? Because some like Wolf likes CF OFF and still is using both the cards via async compute at 101% scaling.

 

[george1976]

My Vega 64 Nitro+ works on a 750w Evga G2 with an overclocked i5 2500k without any issues, max power draw I can see after undervolting to 980 only, is around 210w except for furmark where 320w is the norm.

 

[george1976]

Just got my nitro 64. It's a BEAST in size! Few questions:

I've connected two PCI-e connectors to it. Hopefully the 850w superflower leadex platinum should be up to it?

Secondly, the small page in the box says something about fan USB connectors. There aren't any cables in the box. I'm searching for them in the mobo box but can't find any either. What are they for?

Appreciate your help.

Dont know anything about usb fans mate, you can connect external pwm fans o the card though, two of them

 

[Illuminist]

Roughly 80% of the games as of few months back, support Vega Crossfire. Either with native settings or creating custom profile yourself.
Some games that do not support CF, but support Async compute (Wolfenstein for example) with Crossfire off they scale on both GPUs also.
And those games scale at 100% some even higher (RE7 103%, Wolfenstein 101%), compared to around 85% scalling on most games.

However for CF to work you need to have good motherboard pushing at least 2 8x Pci-e, as CF since 2013 doesn't need external link.
The vast majority of motherboards (Intel) cannot do that hence "is not working".
on HEDT (X79/99/299, X399), or X370/X470 platforms there is no such issue though.

I have used this async on shadow of the tomb raider another great feeling when I didn't have to set anything in the amd crimson
I just changed the game to dx12 and then got 70+fps 4k all everything turned on

As for your motherboard comment, I have 16x and one 8x on my mb.
Already competed all the wolfensteins out I would give that a go too

 

[Illuminist]

After reading what Illuminist said I'm feeling the same...

I do have a 1200W PSU after all

I have a superflower 850 running 2 bad boys with no problem I made sure to undervolt them both and no go 1100 hbm 1700 core crazy on them both.
My aim would be to put the 56 under water like my 64 and get a fatter psu and hand them both balls to the walls for 4k 80/90fps and to do a small wee on 2080ti scores

 

[YB2ndBEST]

Decided to finally try some different bios and sad to say I'm better off on my stock bios. I tried using 4 different LC bios and they all technically worked fine, but the clocks to voltage ratio was not worth it. What's also odd is on all but the sapphire LC bios I couldn't even get the same clocks / voltage combo I use on my stock bios. If I loaded up 1040mv / 1650mhz then in game I'd get like 14xx clocks it was awful (except the sapphire lc bios). So naturally with all of them I attempted max oc and on xfx 8709 I think it was I got 1730ish in game stable but at 1250mv and with hbm at 1050 it was only a literal couple fps above my stock bios setup that pulls significantly less power. I'm ok with this, just posting my luck with bios flashing and stuff haha

So after messing with softpowerplay registry I've finally finished off with my stock bios and these settings with great fps and temps!

In game it holds an average between 1605 - 1620mhz with 1150hbm rock solid. This combo put me within only a couple fps of 1730ish/ 1075 on the LC bios (all bios I did the powerplay mod as well letting them pull whatever they need).

 

[george1976]

So after messing with softpowerplay registry I've finally finished off with my stock bios and these settings with great fps and temps!

In game it holds an average between 1605 - 1620mhz with 1150hbm rock solid. This combo put me within only a couple fps of 1730ish/ 1075 on the LC bios (all bios I did the powerplay mod as well letting them pull whatever they need).

What voltage you applied to get those clocks mate?
Asking because I get between 1590-1600 with 980mV and curious if takes much more voltage to get to your clocks.
Thx

 

[Davedree]

Decided to finally try some different bios and sad to say I'm better off on my stock bios. I tried using 4 different LC bios and they all technically worked fine, but the clocks to voltage ratio was not worth it. What's also odd is on all but the sapphire LC bios I couldn't even get the same clocks / voltage combo I use on my stock bios. If I loaded up 1040mv / 1650mhz then in game I'd get like 14xx clocks it was awful (except the sapphire lc bios). So naturally with all of them I attempted max oc and on xfx 8709 I think it was I got 1730ish in game stable but at 1250mv and with hbm at 1050 it was only a literal couple fps above my stock bios setup that pulls significantly less power. I'm ok with this, just posting my luck with bios flashing and stuff haha

So after messing with softpowerplay registry I've finally finished off with my stock bios and these settings with great fps and temps!

In game it holds an average between 1605 - 1620mhz with 1150hbm rock solid. This combo put me within only a couple fps of 1730ish/ 1075 on the LC bios (all bios I did the powerplay mod as well letting them pull whatever they need).

Good info there, I too messed around with the LC bioses over a year ago, I will have to check but I think the 8709 lc one got me the best results on my p64 ref (peasant cooler), but the stock 8730 was better for hitting the sweet spot of 1600 MHz.

Either way the scaling from 1600 MHz to 1700 is exactly the same as Fiji from 1125-1200mhz, it's like 100-150w for about 2-3fps.

 

[flea.rider]

so what do you think ? not got the water block on yet waiting for new paste ..

 

[LoadsaMoney]

 

[YB2ndBEST]

Good info there, I too messed around with the LC bioses over a year ago, I will have to check but I think the 8709 lc one got me the best results on my p64 ref (peasant cooler), but the stock 8730 was better for hitting the sweet spot of 1600 MHz.

Either way the scaling from 1600 MHz to 1700 is exactly the same as Fiji from 1125-1200mhz, it's like 100-150w for about 2-3fps.

Yea on the sapphire LC bios it would hold basically same as my stock bios and voltage setup so really could have just left it on there lol. It's ok though I'm very happy with the cards performance as I've got it now.

What voltage you applied to get those clocks mate?
Asking because I get between 1590-1600 with 980mV and curious if takes much more voltage to get to your clocks.
Thx

So if I'm not pushing my HBM to 1150 and mostly focused on core clock it will hold 1590+ in game at 980mv when I originally tested it on stock cooler. Leaving it at 1030mv (which actually pulls .990 - 1.0 when looking at the logs) I get around 1630ish in game with 1150 hbm locked.

 

[Figo]

Has anyone had a considerable uptake in reported clockspeeds with the latest Radeon optional driver?

Im now sitting just shy of 1,700Mhz but with VERY reasonable temperatures on the GPU / HBM package and not much - if any - increase in power draw?

also when using Radeon chill, I used to have fluctuating clockspeeds and average GPU utilisation of 90+% now the clock speeds are higher + more consistent, but its the GPU utilisation figure that varies?

Not done any game benchmarks to see if there is actually a difference, but when rending out a video there was about a 7% reduction in time and clockspeeds during that were reported at 1,700Mhz +

 

[Street]

I had a play with my reference Vega 64 over the weekend to see if I could undervolt/overclock and these were the results..



I started by bumping the power limit to 50% as this seemed to be what most people recommended, however this pushed my power usage and temps through the roof. With 50% the power draw seemed to always be around 280W plus for very little gains. I dropped this to 25% and started again.

After many runs I found a reasonable balance between power usage and speed with P6/P7 at 1537/1602 and the voltages at 950/975. I found if I put the HBM up to 1050 with 950mv, it dropped it back to 800MHz so I left that at 1025MHz. The clock seemed to hit 1475MHz consistently and the temperatures never went over 73C. I ran a few cycles of Time Spy and Superposition and they seemed fairly consistent too. It works out at roughly a 6% speed increase for 5% power saving. Not much, but it runs cooler and isn't as loud and so far hasn't crashed at all with benchmarking or gaming.

Couple questions, what is the hotspot temperature? As under a few runs I saw that getting up to about 107C! Also, is it worth messing around with the target temperature etc? I didn't really touch the P6 clock either, is it worth tweaking this or leaving it?

 

[Illuminist]

G

I had a play with my reference Vega 64 over the weekend to see if I could undervolt/overclock and these were the results..



I started by bumping the power limit to 50% as this seemed to be what most people recommended, however this pushed my power usage and temps through the roof. With 50% the power draw seemed to always be around 280W plus for very little gains. I dropped this to 25% and started again.

After many runs I found a reasonable balance between power usage and speed with P6/P7 at 1537/1602 and the voltages at 950/975. I found if I put the HBM up to 1050 with 950mv, it dropped it back to 800MHz so I left that at 1025MHz. The clock seemed to hit 1475MHz consistently and the temperatures never went over 73C. I ran a few cycles of Time Spy and Superposition and they seemed fairly consistent too. It works out at roughly a 6% speed increase for 5% power saving. Not much, but it runs cooler and isn't as loud and so far hasn't crashed at all with benchmarking or gaming.

Couple questions, what is the hotspot temperature? As under a few runs I saw that getting up to about 107C! Also, is it worth messing around with the target temperature etc? I didn't really touch the P6 clock either, is it worth tweaking this or leaving it?

The hot spot from what I remember is the point between the core and hmb on the die, it does get incredibly hot on the air cooler and at 110 it will crash most likely
You definately need to set the second line of the fan to 100% and the temp thingy on
This is why you need to undervolt if you use the 50%power limit,

I keep mine on p6 p7 stock clocks. But 1050 and 1100 for the volts.
Hbm 1050 and 1000v
Power limit can then be 30 - 50% depending on how well your system handles the heat

 

[Morve]

Anyone got any settings for a nitro+ undervolt

 

[VIPER258]

Been playing around with undervolting my vega 64 but it occurred to me that I don't have the slightest clue as to what my power draw is.

What would be the best software, or perhaps a device to measure how much power I am using in real time?

 

[flea.rider]

Been playing around with undervolting my vega 64 but it occurred to me that I don't have the slightest clue as to what my power draw is.

What would be the best software, or perhaps a device to measure how much power I am using in real time?

hwinfo .. great little program