The RX Vega 64 Owners Thread

TonyTurbo78 · 17 Aug 2017 at 10:58

Kaapstad said:
The consistent performance he was getting was very impressive.

There are few Pascal cards that can maintain core clocks that good for very long.

I've been watching a few tear downs on YouTube from the main players, the general feeling is the cards and the components on the pcb have been built to a very high standard. I was shocked when I saw MDplatts bench and doubted that was reporting the number correctly. But the scores would suggest that's where it would be around with those high clocks.

MDPlatts · 17 Aug 2017 at 11:15

The fan is a little noisy when it gets upto speed but it is otherwise silent. It is currently loose inside the open case and facing a wall about a foot away so the noise and reflected noise will be higher than when inside a closed case where I would expect it to be inaudible. Its also on my desk/table next to me rather than on the floor at my left foot which will help too.

I've not played with the fan profiles - I've not raised the speed of the higher profile from approx 2K or so - it looks like it can go to 3K perhaps but I've not touched it.

LtMatt · 17 Aug 2017 at 11:22

Stable at 1750 Core undervolted by 0.075mv, makes it run a fair bit cooler and quieter at full load. Required fan speed to maintain 65c just over 30%. 0.100mv crashed immediately.

AmateurExpert · 17 Aug 2017 at 11:26

Chewie said:
I already have ideas about which fan to use, I'm more curious/clueless on how to do it as the fan cable is bundled with the cooling tubes and the connector is not visible/accessible.

Assuming you go down the fan replacement route, one thing you might test before opening up the shroud to disconnect the existing fan is to run both your new fan and the existing one in push-pull (new fan push, old fan pull) on the radiator; the improved efficiency might be enough to keep the old fan from spinning up high enough to make noise. If it works, it saves you some work.

Chewie · 17 Aug 2017 at 11:31

AmateurExpert said:
Assuming you go down the fan replacement route, one thing you might test before opening up the shroud to disconnect the existing fan is to run both your new fan and the existing one in push-pull (new fan push, old fan pull) on the radiator; the improved efficiency might be enough to keep the old fan from spinning up high enough to make noise. If it works, it saves you some work.

That's a good point actually, will definitely try that.

Chewie · 17 Aug 2017 at 11:35

AMDMatt said:
Stable at 1750 Core undervolted by 0.075mv, makes it run a fair bit cooler and quieter at full load. Required fan speed to maintain 65c just over 30%. 0.100mv crashed immediately.

Glad to hear your new card is working Matt.

Care to share how to do the undervolting and the best way to validate the result?

TonyTurbo78 · 17 Aug 2017 at 11:36

AMDMatt said:
Stable at 1750 Core undervolted by 0.075mv, makes it run a fair bit cooler and quieter at full load. Required fan speed to maintain 65c just over 30%. 0.100mv crashed immediately.

how are you measuring if your voltage changes are taking effect Matt? I'm not seeing any change on the wattage measure socket thing I've got plugged into the wall. With a 100mv drop sould be seeing 100 watts difference or so so it should be blatantly obvious. I don't even think my volt changes are taking any sort of effect.

Yaayuh! · 17 Aug 2017 at 11:39

AMDMatt said:
Stable at 1750 Core undervolted by 0.075mv, makes it run a fair bit cooler and quieter at full load. Required fan speed to maintain 65c just over 30%. 0.100mv crashed immediately.

That's good to hear.

Acesbomb · 17 Aug 2017 at 11:55

AMDMatt said:
Stable at 1750 Core undervolted by 0.075mv, makes it run a fair bit cooler and quieter at full load. Required fan speed to maintain 65c just over 30%. 0.100mv crashed immediately.

What are your stock clocks reading as? 1630?

LtMatt · 17 Aug 2017 at 11:57

Chewie said:
Glad to hear your new card is working Matt.

Care to share how to do the undervolting and the best way to validate the result?

TonyTurbo78 said:
how are you measuring if your voltage changes are taking effect Matt? I'm not seeing any change on the wattage measure socket thing I've got plugged into the wall. With a 100mv drop sould be seeing 100 watts difference or so so it should be blatantly obvious. I don't even think my volt changes are taking any sort of effect.

In the driver available on the website, voltage control is not working, but I'm using a slightly newer build in which it works. No ETA on its arrival I'm afraid, but it will come.

I'm uploading a BF4 video where i give it a quick run out with the undervolt in place. I'll share that with you once it's uploaded. As it's 4K @60FPS file size is 2.2GB so may take an hour or more.

Here's some quick HWINFO64 stats while it's uploading.

Stock settings (1750Mhz/945Mhz) +50% PL

Undervolted -0.075mv (1750Mhz/945Mhz) +50% PL

The 0.075mv undervolt allows for about a 10% reduction in required fan speed to maintain 65C. 30-35% fan speed is required and that is silent for me.

AmateurExpert · 17 Aug 2017 at 12:05

AMDMatt said:
In the driver available on the website, voltage control is not working, but I'm using a slightly newer build in which it works. No ETA on its arrival I'm afraid, but it will come.

I'm uploading a BF4 video where i give it a quick run out with the undervolt in place. I'll share that with you once it's uploaded. As it's 4K @60FPS file size is 2.2GB so may take an hour or more.

Here's some quick HWINFO64 stats while it's uploading.

Stock settings (1750Mhz/945Mhz) +50% PL
{snip}

Undervolted -0.075mv (1750Mhz/945Mhz) +50% PL
{snip}
The 0.075mv undervolt allows for about a 10% reduction in required fan speed to maintain 65C. 30-35% fan speed is required and that is silent for me.

Thanks - I presume that with the undervolt you've also been able to see higher frequencies maintained (and showing up as better performance in the game) too?

LtMatt · 17 Aug 2017 at 12:06

AmateurExpert said:
Thanks - I presume that with the undervolt you've also been able to see higher frequencies maintained (and showing up as better performance in the game) too?

Not really, if you max out power limit maximum frequency state will be maintained regardless of available voltage used in Wattman.

Davedree · 17 Aug 2017 at 12:12

Matt I take it the vddc voltage reported in hwinfo doesn't talk to the i2c register yet,
as both screenshots show 1.356v

LtMatt · 17 Aug 2017 at 12:14

Davedree said:
Matt I take it the vddc voltage reported in hwinfo doesn't talk to the i2c register yet,
as both screenshots show 1.356v

Yes i mentioned this in my video, however my logitech headset decided to mute itself so no sound came out, Lol. It reads that value whatever change you make to the voltage.

I have internal tools which measure voltage, it is 1.2V at stock and 1.125v when undervolted, with minor fluctuations.

Acesbomb · 17 Aug 2017 at 12:15

AmateurExpert said:
I suspect there are a fair few PSUs out there which were never capable of delivering on their rated capacities, but which have not been exposed as very few people are able to actually test their PSUs with high loads.

Looks like all the power for the GPU goes through the 2 8 pins

Davedree · 17 Aug 2017 at 12:15

AMDMatt said:
Yes i mentioned this in my video, however my logitech headset decided to mute itself so no sound came out, Lol. It reads that value whatever change you make to the voltage.

I have internal tools which measure voltage, it is 1.2V at stock and 1.125v when undervolted, with minor fluctuations.

brilliant thanks for the info, as just wanted to know what stock voltage is,
thanks

LtMatt · 17 Aug 2017 at 12:16

Acesbomb said:
Looks like all the power for the GPU goes through the 2 8 pins

I believe only fan speed/pump goes through the PCI-E lanes for Vega, but this is based on HWINFO stats so this is a best guess.

orbitalwalsh · 17 Aug 2017 at 12:31

which ones yours?

https://www.overclock3d.net/news/gp...been_found_to_come_with_two_different_forms/1

Acesbomb · 17 Aug 2017 at 12:33

AMDMatt said:
I believe only fan speed/pump goes through the PCI-E lanes for Vega.

That would put the watercooled vega over the certified limit for 2 8pins then wouldn't it?, 300w. Seems a strange approach.

Mr_Clark1000 · 17 Aug 2017 at 12:35

So I want to water-cool with an AIO type unit. I don't want to mess around with a custom loop.

I see there is the Eismann wolfy thing and EK are doing some weird modular thing too? The problem is, the EK one is already 130 odd just for the block, so with the integrated profiled pump / rad, probably talking nearer 250 - 300 for a working setup. The Eissman one looks to be a lot cheaper.

Any other options? The Kraken would have been good, but does not fit

Competitor rules

The RX Vega 64 Owners Thread

Man of Honour