The RX Vega 64 Owners Thread

Phixsator · 17 May 2020 at 16:59

montymint said:
at this point it's either that my case is smaller or that the nitro's cooler is much better than the powercooler Red Devil, which is probably the case - I've had nitros in the past and they are solid.

His results is much better than what i can get with my Vega 64 Nitro+. I cannot keep mine stable at over 1610 with 1000mv and my hotspot is still around 94C with noctua and vortex fans in a carbide 540 air from corsair.

MCFC_ANDY · 17 May 2020 at 18:37

Phixsator said:
His results is much better than what i can get with my Vega 64 Nitro+. I cannot keep mine stable at over 1610 with 1000mv and my hotspot is still around 94C with noctua and vortex fans in a carbide 540 air from corsair.

starting to get up to yours though, just had an hr of mw multiplayer @4k and then furmarked it, fans at 60%. got upto 87degrees hot spot, not overly agressive oc, but getting there, but its rock stable, gaming was only around 70degree, furmark whacked it up

Gerard · 17 May 2020 at 19:20

Still considering moving my rad then i seen this video via bitwit, his results came out that the best orientation is the one i currently have, rad at the front and an axial cooler. Just suppose this shows how keek the strix cooler is.

MCFC_ANDY · 17 May 2020 at 21:54

updated

https://www.3dmark.com/3dm/46915156

MCFC_ANDY · 17 May 2020 at 23:47

well after a few hrs of playing around i pretty much found the highest settings to get through timespy, could maybe eek out a little more but 1767mhz and hbm 1120mhz is ok i guess, decent gpu score of 8244

Trick · 18 May 2020 at 14:14

montymint said:
how low have you manged to get your voltages?

Not that low
P4 = 1030
P5 = 1040
P6 = 1045
P7 = 1050

ram mv = 1040 and I'm not overclocking the ram beyond stock 945mhz, i seem to get driver crashes if I push the ram to 1000mhz without inc the mv, and I'm not bothered about another % or two for the extra heat and noise.

MCFC_ANDY · 18 May 2020 at 14:21

Trick said:
Not that low
P4 = 1030
P5 = 1040
P6 = 1045
P7 = 1050

ram mv = 1040 and I'm not overclocking the ram beyond stock 945mhz, i seem to get driver crashes if I push the ram to 1000mhz without inc the mv, and I'm not bothered about another % or two for the extra heat and noise.

isnt the memory voltage set in bios, ive tried no end of settings in wattman to up the memory, but nothing works past 1120, even if voltage is set to 1000mv or 1200mv seems to make no difference

SonicSW20 · 18 May 2020 at 14:41

The voltage in the memory section is not HBM voltage. Its more of a "voltage floor" for the card. It should be set lower than what your P6 state is set to.

If you set it to 1200mV, and your P7 to 1100mV, your core will be locked to 1200mV.

Trick · 18 May 2020 at 15:14

yeah, that's what I read too, so I have it the same as P4 or P5

MCFC_ANDY · 18 May 2020 at 15:17

im not expert at this but, if your overclock temps are all fine and it fails the bench, is that the silicone limit? ive tried a loads of ways to get it above 1770mhz and 1120mhz memory but will not have it. i know it throttles at 85degrees but im only at 80

Trick · 18 May 2020 at 15:18

hotspot >85 starts to throttle the gpu?

MCFC_ANDY · 18 May 2020 at 15:20

Trick said:
hotspot >85 starts to throttle the gpu?

sorry thought it throttles at 85, not hotspot

SonicSW20 · 18 May 2020 at 15:22

MCFC_ANDY said:
im not expert at this but, if your overclock temps are all fine and it fails the bench, is that the silicone limit? ive tried a loads of ways to get it above 1770mhz and 1120mhz memory but will not have it. i know it throttles at 85degrees but im only at 80 degrees for hotspot

Not enough voltage to drive those clocks. Increasing voltage means more heat and the increase in voltage can cause stability. HBM voltage is fixed so whatever you get there is what it is. As the HBM is right next to the GPU core and contacting the same block of copper, pushing the GPU core creates more heat which will also increase HBM temps, so there's a balance to be had there.

You can try to throw more voltage at the core but with Vega you'll hit thermal limits pretty quick. It's a balancing act.

MCFC_ANDY · 18 May 2020 at 15:24

SonicSW20 said:
Not enough voltage to drive those clocks. Increasing voltage means more heat and the increase in voltage can cause stability. HBM voltage is fixed so whatever you get there is what it is. As the HBM is right next to the GPU core and contacting the same block of copper, pushing the GPU core creates more heat which will also increase HBM temps, so there's a balance to be had there.

You can try to throw more voltage at the core but with Vega you'll hit thermal limits pretty quick. It's a balancing act.

yeah i found can push the core higher if reduce hbm to 1100, core will max out at 1768 if hbm is set to 1120

montymint · 18 May 2020 at 15:44

MCFC_ANDY said:
well after a few hrs of playing around i pretty much found the highest settings to get through timespy, could maybe eek out a little more but 1767mhz and hbm 1120mhz is ok i guess, decent gpu score of 8244

What voltage you got for that?

at 1000mV I've managed a quick and dirty 1700/1100 clocks that during the bench never got past 85c in the hotspot with 1899RPM on the fans. The score came in at 7760, which I'm happy enough with

MCFC_ANDY · 18 May 2020 at 15:51

Yeah the 1100/1700 hbm 1100 is pretty standard. I found to get above 1700 needed to push voltage to all way too1200mv 1770 is highest I got core. But may be a little wriggle room if u drop hbm

ShortWarning · 20 May 2020 at 09:22

Gerard said:
Still considering moving my rad then i seen this video via bitwit, his results came out that the best orientation is the one i currently have, rad at the front and an axial cooler. Just suppose this shows how keek the strix cooler is.

I skipped to near the end of the vid. This is basically the same set up as mine. I currently have cool hair coming in at the front and hitting my rad (which also means hot air is probably being pumped in) and fans extracting air at the top/back of the case.

Should I switch this round? Cool air coming in at the top/back then the rad fans set to pull the air out of the front?

Gerard · 20 May 2020 at 15:09

ShortWarning said:
I skipped to near the end of the vid. This is basically the same set up as mine. I currently have cool hair coming in at the front and hitting my rad (which also means hot air is probably being pumped in) and fans extracting air at the top/back of the case.

Should I switch this round? Cool air coming in at the top/back then the rad fans set to pull the air out of the front?

Probably depend on your temps currently.

ShortWarning · 20 May 2020 at 15:36

Gerard said:
Probably depend on your temps currently.

I don't think my temps are crazy (posted on the last page) but my GPU fans have to more or less run at 100% to keep them there.

Phixsator · 21 May 2020 at 07:07

So I've repasted my Vega 64 nitro+ with some Kryonaute and wow what a difference. My goal was to reduce the hot spot temp and I have. Hopefully, this will mean that the card can sustain it's maximum boost clocks a fair bit longer during long gaming sessions.

And may I add how a lovely experience it is to service the nitro+. Such well thought out design and ease of disassembly and reassembly.

Code:

Furmark pre repaste 10-minute run
GPU Temp: 76
HBM Temp: 85
Hot spot Temp: 97
Fan speed: 1850
------------------------------------------
Furmark post paste 10-minute run
GPU Temp: 73
HBM Temp: 79
Hot spot Temp: 82
Fan speed: 1646
------------------------------------------
Settings:
P7 Core: 1600mhz
P7 Voltage: 1025mv
Fan: auto
HBM clock: 975mhz
Voltage floor: 950mv
Power limit: +50%

Competitor rules

The RX Vega 64 Owners Thread

Pet Northerner