734 is fine on mine if I set p7 to a lower clock, I settled on 1702 for now. 774 causes some sort of awful stuttering on my card.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
The RX Vega 64 Owners Thread
- Thread starter Kaapstad
- Start date
734 is fine on mine if I set p7 to a lower clock, I settled on 1702 for now. 774 causes some sort of awful stuttering on my card.
Associate
774 is only on MSI boards so far right? I'll see what it does.
Also, my Timespy score (with original air BIOS) seems to have gone down from around 7100 to 6300 ... I will crank up the fan and see what happens. Not having the GPU control its own fan is a PITA ....
Also, my Timespy score (with original air BIOS) seems to have gone down from around 7100 to 6300 ... I will crank up the fan and see what happens. Not having the GPU control its own fan is a PITA ....
Associate
It was on the MSI LC cards. I think Asder00 from Guru3D uploaded the one on Techpowerup.
Soldato
Friend of mine just got a Vega64 air cooled but only has a 650w PSU and considering the store page says 700w minimum, Would this cause any serious issues possibly ?
Associate
Depends on the rest of the system. If running Undervolted it shouldn't pull more than 220W on air anyway. If running stock it can pull upwards of 300W quite easily even on the 200W BIOS setting.
With a 650W PSU you'd want the rest of the system to be under 300W under load to stay in a safe power draw range.
With a 650W PSU you'd want the rest of the system to be under 300W under load to stay in a safe power draw range.
Associate
OK, good news (everyone)! 
I discovered the issue with my Eiswolf. The back plate had caught on the edge of the front plate causing uneven fitting of the backplate and presumably reduced pressure on the whole sandwich.
So with the fan on 45% (quiet but not inaudible) and the pump on 80% (silent, I will tune it and the fan later) I am seeing temps of 45, 75 and 50 (GPU, HBM2 and Hot) in Furmark at 1024x768/175fps, 98% GPU load, ambient of 21.6°C which looks a lot better.
Looped into an Eisbaer 280 I think the cooling will be good and the fans can run in near silence.
Timespy is back over 7000 (just) now. 1800x stock with RAM at 3200MHz, loose timings.
P.S. I was only able to put one of the two provided "Eiswind" fans on the EW rad due to case constraints.
P.P.S. I just noticed that my replacement Sapphire LC seems to have come with the 774 BIOS.
I discovered the issue with my Eiswolf. The back plate had caught on the edge of the front plate causing uneven fitting of the backplate and presumably reduced pressure on the whole sandwich.
So with the fan on 45% (quiet but not inaudible) and the pump on 80% (silent, I will tune it and the fan later) I am seeing temps of 45, 75 and 50 (GPU, HBM2 and Hot) in Furmark at 1024x768/175fps, 98% GPU load, ambient of 21.6°C which looks a lot better.
Looped into an Eisbaer 280 I think the cooling will be good and the fans can run in near silence.
Timespy is back over 7000 (just) now. 1800x stock with RAM at 3200MHz, loose timings.
P.S. I was only able to put one of the two provided "Eiswind" fans on the EW rad due to case constraints.
P.P.S. I just noticed that my replacement Sapphire LC seems to have come with the 774 BIOS.
Associate
huh... what kind of processor/motherboard are you using right now? Oh and if you get the threadripper I WILL BE VERY JELLY!
I managed to damage the fan wires on my Vega default cooler. Clumsiness over 9000
So now I'm stuck with a passively cooled RX Vega 64.
Removed shroud. -50% power level.
It still drives a three monitor setup. And can plays youtube and 1080p videos no problem.
I was looking to replace the cooler with a water kit anyway. But until then could anyone who replaced Vega cooling lend-lease me his default cooler?
So now I'm stuck with a passively cooled RX Vega 64.
Removed shroud. -50% power level.
It still drives a three monitor setup. And can plays youtube and 1080p videos no problem.
I was looking to replace the cooler with a water kit anyway. But until then could anyone who replaced Vega cooling lend-lease me his default cooler?
Associate
Can you not just bolt on another fan and run it from the motherboard? That's what I have to do with the Eiswolf - the only real design flaw, they should include a mini-pwm to full size socket adapter. I have one on the main card in my 2 x 290x rig and it works a treat. Now to find some more.
https://www.overclockers.co.uk/ek-w...dator-fan-adapter-for-gpu-50cm-wc-9dd-ek.html
https://www.overclockers.co.uk/ek-water-blocks-ek-cable-pwm-fan-adapter-for-gpu-50cm-wc-9d9-ek.html
Doh and I just placed an order 10 mins ago without these!
https://www.overclockers.co.uk/ek-w...dator-fan-adapter-for-gpu-50cm-wc-9dd-ek.html
https://www.overclockers.co.uk/ek-water-blocks-ek-cable-pwm-fan-adapter-for-gpu-50cm-wc-9d9-ek.html
Doh and I just placed an order 10 mins ago without these!
Associate
@alec you can probably get hold of the Morpheus II kit for it as it fits according to the rumblings on Overclock.net.
As a side effect you can get much improved cooling via the huge fin space
http://www.raijintek.com/en/products_detail.php?ProductID=46
As a side effect you can get much improved cooling via the huge fin space
http://www.raijintek.com/en/products_detail.php?ProductID=46
Associate
- Joined
- 9 Aug 2016
- Posts
- 186
Hopefully I think I've found my sweet spot for my air cooled Vega 64.
P6 1552
P7 1632
Power limit +50
Voltage 975Mv
Fan 2500rpm
HBM at default 945 and 1050Mv
I get between 5-10% uplift over balanced, draw 60w+ less power and the card is now as quiet as my old Fury Pro used to be. That'll do nicely for me as I'm not trying to shoot the lights out I just want a card that performs well enough to push my 144hz freesync monitor without being too hot and noisy. During testing the card maintained a constant 1550 with no throttling and the GPU temp maxed out at 81c, HBM 84c and hot spot 87c. No crashes or any other issues at all. Tried to under volt to 950 but it just locked up on me after about 30 seconds. TBH not too bothered with achieving amazing scores in Timespy, Superposition or any other benchmarks as long a it runs my games well I'm a happy chappy indeed.
P6 1552
P7 1632
Power limit +50
Voltage 975Mv
Fan 2500rpm
HBM at default 945 and 1050Mv
I get between 5-10% uplift over balanced, draw 60w+ less power and the card is now as quiet as my old Fury Pro used to be. That'll do nicely for me as I'm not trying to shoot the lights out I just want a card that performs well enough to push my 144hz freesync monitor without being too hot and noisy. During testing the card maintained a constant 1550 with no throttling and the GPU temp maxed out at 81c, HBM 84c and hot spot 87c. No crashes or any other issues at all. Tried to under volt to 950 but it just locked up on me after about 30 seconds. TBH not too bothered with achieving amazing scores in Timespy, Superposition or any other benchmarks as long a it runs my games well I'm a happy chappy indeed.
Associate
@ lordhawkwind
dang dude that's really nice
. unfortunately my card will not stay stable under 1100 mV 
.. but im still happy with its performance.
dang dude that's really nice
. unfortunately my card will not stay stable under 1100 mV .. but im still happy with its performance.
OK, good news (everyone)!
I discovered the issue with my Eiswolf. The back plate had caught on the edge of the front plate causing uneven fitting of the backplate and presumably reduced pressure on the whole sandwich.
So with the fan on 45% (quiet but not inaudible) and the pump on 80% (silent, I will tune it and the fan later) I am seeing temps of 45, 75 and 50 (GPU, HBM2 and Hot) in Furmark at 1024x768/175fps, 98% GPU load, ambient of 21.6°C which looks a lot better.
Looped into an Eisbaer 280 I think the cooling will be good and the fans can run in near silence.
Timespy is back over 7000 (just) now. 1800x stock with RAM at 3200MHz, loose timings.
P.S. I was only able to put one of the two provided "Eiswind" fans on the EW rad due to case constraints.
P.P.S. I just noticed that my replacement Sapphire LC seems to have come with the 774 BIOS.
Thanks for posting and glad to see you got the temps sorted. I assume the 75 is your hot spot temp? can you tell me the voltage and clocks you got those temperatures at?
I was interested in getting the Eisbar Vega but as I had a Kraken G10 and a Corsair H110 at hand, I just saved the money and did the Kraken mod. Had to drill 4x holes at 64mm spacing but delighted with the temperatures and much quieter noise levels compared to stock.
At 1000mV and ~1550 core (actual) 1100 HBM, ~215W, after a very long run in (1 hour +) with Witcher 3 at 4k.
Core ~48c
HBM ~50c
Hotspot ~65c
Last edited:
Soldato
Not really, often draws under 200w for casual use if using custom setup and 1000mv or less. 900mv would be about 140w but lower clocks, somewhere between those two its not troubling a good psu. People get a bit carried away with the doom and gloom, I dont like the default profiles otherwise its really ok. Overclock or setting records Im not sure that psu is enough but normal use, I would guess worries are overblown.
Just saw a 1050w psu go for 50 quid, seems like a bargain even if few years old. Hopefully Navi is not any more power I reckon not
It doesnt pick up the game profile settings if I launch via steam, I guess its not a big deal if I go via radeon to launch but maybe a desktop icon would be better
Last edited:
Associate
@ICDP
Yes, 75 is Hspot. Stock volts/clocks with furmark loading to 1350MHz/98%. Increasing the power ramps up the temps. 50/80/50 or so but the fps also goes up.
The highest I have seen the HS at now is 90 after a longer run of furmark stress and the room warmed up to 25c. So "acceptable" but I want it quieter ...
I'm now seeing ~7208 in Timespy without any effort.
The Eisbaer shipped today, I really feel the 120mm rad/single fan is only just adequate, so adding the 280 to the loop should help.
Yes, 75 is Hspot. Stock volts/clocks with furmark loading to 1350MHz/98%. Increasing the power ramps up the temps. 50/80/50 or so but the fps also goes up.
The highest I have seen the HS at now is 90 after a longer run of furmark stress and the room warmed up to 25c. So "acceptable" but I want it quieter ...
I'm now seeing ~7208 in Timespy without any effort.
The Eisbaer shipped today, I really feel the 120mm rad/single fan is only just adequate, so adding the 280 to the loop should help.
Associate
Update: Alphacool sent the Eisbaer by UPS express (despite what their web site said) so it actually arrived just now.
With luck I will be able to get the X62 pulled and the EB installed over the weekend, compare EB to X62 and then join the loops and see what happens to the GPU temps.
Once all that is done it will be time to pop the AIO in again and maybe the 2nd AC that is still an AC.
Anyone in Dublin want a Kraken X62 "cheap"?
Separately, regarding testing and general observations:
I updated to the WHQL drivers/clean install today and the Eiswolf card seems more stable with regard to random crashes and odd temps and such.
Regarding undervolting, while running a stress test, I set both HBM2 voltage and GPU voltage (whether those are what they are or not!) to 950mV in Wattman and saw ZERO difference to temps, voltage indicated (GPU-z) and power draw at the wall, all remained as they were at stock. However, I had the clock speeds set to 1400MHz core and 1100MHz HBM2 (my stress tester is an ether miner), so I don't know if maybe what is happening is the driver is seeing the actual clock as being in a P state lower/the same as than P6 or P7 and thus applying what is set (default) for that P state rather than what I have set (the same) for P6 and P7.
I then reset by applying balance profile again and this time I only get to adjust P7 and when I applied with lower volts it actually set it ... volts reported in GPU-z went down ~100mV as expected and power draw at the wall dropped 30 - 40W for the same M/H.
It seems to be that a) we should have access to all P states (all the time) when setting custom and that there is some bugginess going on whereby under some circumstances (yet to be determined) voltages are not applying, even though they do under other circumstances. The fact that sometimes P6 and P7 get sliders and at other times only P7, seemingly at random is rather odd/disturbing.
With luck I will be able to get the X62 pulled and the EB installed over the weekend, compare EB to X62 and then join the loops and see what happens to the GPU temps.
Once all that is done it will be time to pop the AIO in again and maybe the 2nd AC that is still an AC.
Anyone in Dublin want a Kraken X62 "cheap"?
Separately, regarding testing and general observations:
I updated to the WHQL drivers/clean install today and the Eiswolf card seems more stable with regard to random crashes and odd temps and such.
Regarding undervolting, while running a stress test, I set both HBM2 voltage and GPU voltage (whether those are what they are or not!) to 950mV in Wattman and saw ZERO difference to temps, voltage indicated (GPU-z) and power draw at the wall, all remained as they were at stock. However, I had the clock speeds set to 1400MHz core and 1100MHz HBM2 (my stress tester is an ether miner), so I don't know if maybe what is happening is the driver is seeing the actual clock as being in a P state lower/the same as than P6 or P7 and thus applying what is set (default) for that P state rather than what I have set (the same) for P6 and P7.
I then reset by applying balance profile again and this time I only get to adjust P7 and when I applied with lower volts it actually set it ... volts reported in GPU-z went down ~100mV as expected and power draw at the wall dropped 30 - 40W for the same M/H.
It seems to be that a) we should have access to all P states (all the time) when setting custom and that there is some bugginess going on whereby under some circumstances (yet to be determined) voltages are not applying, even though they do under other circumstances. The fact that sometimes P6 and P7 get sliders and at other times only P7, seemingly at random is rather odd/disturbing.
Associate
In regards to the voltage testing, Wattman is incredibly unstable when doing so. If the voltages don't apply for whatever reason (best test is Superposition as you can see if you get stuck in P5 instantly) you don't know until testing.
Hence I would use OverdriveNTool (on Guru3D). It applies the voltages correctly though the Radeon driver can still go fubar at times. For consistency I only enter Wattman to enable Custom, then close it as it's now useless as far as settings go.
On my card there is a marked difference between 900mv and 1200 mv especially in power drawn and heat produced. At 1100mv for instance my power draw in GPU-z was around 340W which is about 400W total for the card running 1752MHz on the P7 state with an actual frequency of around 1630-1660MHz depending on load during the test and PSU draw was around 690W. When running my current main settings of 950mv on P7 I'm hitting around 190W in GPU-z which is roughly 240-250W overall for the card and roughly 530-550W from the PSU.
If you still think you are getting locked into P5 state or below, use ClockBlocker (also on Guru3D) and enable the rules to make it blocked to P7 when in specific programs or 3D applications. Only way I have found to guarantee it's not dropping states to P5 or below for testing.
As always your mileage may vary.
Hence I would use OverdriveNTool (on Guru3D). It applies the voltages correctly though the Radeon driver can still go fubar at times. For consistency I only enter Wattman to enable Custom, then close it as it's now useless as far as settings go.
On my card there is a marked difference between 900mv and 1200 mv especially in power drawn and heat produced. At 1100mv for instance my power draw in GPU-z was around 340W which is about 400W total for the card running 1752MHz on the P7 state with an actual frequency of around 1630-1660MHz depending on load during the test and PSU draw was around 690W. When running my current main settings of 950mv on P7 I'm hitting around 190W in GPU-z which is roughly 240-250W overall for the card and roughly 530-550W from the PSU.
If you still think you are getting locked into P5 state or below, use ClockBlocker (also on Guru3D) and enable the rules to make it blocked to P7 when in specific programs or 3D applications. Only way I have found to guarantee it's not dropping states to P5 or below for testing.
As always your mileage may vary.
Associate
Ta!
Really nice to hear things like this. I look forward to seeing how Wolfenstein 2 runs, since the partnership between AMD and Bethesda seems to be working well. I'm sure other developers will take note at the Forza situation as XB1X games become more commonplace.
Aside from the above, also wanted to say how much I'm enjoying my Vega 64 LC card from Sapphire (£640 version from launch day). It hasn't missed a beat, hasn't crashed, it's just been rock solid stable and has been producing silky smooth gameplay on my 27" 1440P 144HZ Freesync monitor (BenQ XL2730Z) without fail. All of this while being absolutely whisper quiet
My next upgrade will be a new monitor/GPU in a year or two. I'm waiting for a 40" 4K 144Hz variable refresh rate HDR monitor, so may be waiting some time. I hope Navi will be the GPU I choose to power it, fingers crossed AMD pull out all the stops for Navi, as I'll have to think long and hard between Freesync2/Gsync at the time.
Soldato
this latest driver and Vega64 is incredible for me here. Upping a video as we speak with an overlay, hitting about 1790/1100 solid and the performance is just great for Project Cars 2. 100fps+ (80ish full grid at start) @ 1440p, ultra everything with supersampling AA thrown in as well. I'll post it tomorrow, I'm seriously impressed.
Last edited: