Vega owners club

[MrGenius]

So what are your plans on protecting the small little transistors around the GPU package? I'd really hate to see the tiniest bit of LM squeeze out during assembly.

5 coats of clear nail polish already applied. The first thing I did when I got the card was pull it apart and do that. Then put some Conductonaut on with the stock cooler. Ran it for a few hours to see if it worked, and how bad the cooler still was even with LM. Still bad(IMO, mainly just too loud for serious overclocking). And I don't recommend doing it. Other than for a few shits and giggles. I knew it was going to end up bad eventually. When it started to "dry out" on the bottom of that copper vapor chamber. Which it did...eventually. Worked as good as it was going to for a few hours of messing around with it though. I'd be pretty nervous about putting LM on the non-molded package too. I don't know how bad of an idea it would be if any got on the bare interposer. Might be bad. I wouldn't want to find out. Somebody probably already knows if it is or isn't.

 

[fullinfusion]

5 coats of clear nail polish already applied. The first thing I did when I got the card was pull it apart and do that. Then put some Conductonaut on with the stock cooler. Ran it for a few hours to see if it worked, and how bad the cooler still was even with LM. Still bad(IMO, mainly just too loud for serious overclocking). And I don't recommend doing it. Other than for a few shits and giggles. I knew it was going to end up bad eventually. When it started to "dry out" on the bottom of that copper vapor chamber. Which it did...eventually. Worked as good as it was going to for a few hours of messing around with it though. I'd be pretty nervous about putting LM on the non-molded package too. I don't know how bad of an idea it would be if any got on the bare interposer. Might be bad. I wouldn't want to find out. Somebody probably already knows if it is or isn't.

cool, yeah the stock cooler copper base is really thin and I wouldn't doubt it dry out really fast... Wait is you card not a reference?

 

[erek]

Am I the only one on TPU that has a Liquid Cooled RX Vega64?

https://www.ebay.com/itm/RX-Vega-64...86&_nkw="rare"&_from=R40&rt=nc&LH_TitleDesc=0

 

[Flyordie]

https://www.ebay.com/itm/RX-Vega-64-Liquid-Cooling-Powercolor-8GB-2048-Bit-HBM2-RARE/302749850198?epid=27010584046&hash=item467d4c1a56:g:I8EAAOSwJ-pbCFjW&_sacat=27386&_nkw="rare"&_from=R40&rt=nc&LH_TitleDesc=0

Yea, I can't justify that price. It should be closer to $750 at the most.

 

[erek]

Yea, I can't justify that price. It should be closer to $750 at the most.

heh, was just looking at your System Specs and realized i needed to update mine to reflect the Vega, heh

 

[kurosagi01]

One thing i finally tried doing last month was replace the thermal paste on my vega 64 with artic-cooling and the temps did drop by 2-5 degrees on load while playing games with my current overclock setup.

 

[INSTG8R]

One thing i finally tried doing last month was replace the thermal paste on my vega 64 with artic-cooling and the temps did drop by 2-5 degrees on load while playing games with my current overclock setup.

I really want MX-4 mine but it’s less than a month old and I just don’t wanna mess with it just yet(well I do I really wanna see it naked) but it tops out at 75C so it’s not “terrible”

 

[kurosagi01]

I really want MX-4 mine but it’s less than a month old and I just don’t wanna mess with it just yet(well I do I really wanna see it naked) but it tops out at 75C so it’s not “terrible”

I just went for it despite having warranty still, giving by everyone opinion that the thermal paste is dry cheese in the thread lol. It was pretty scary though as I thought i might break it but I took my time and all my tools out prepared like magnetic tray and screwdriver set for any funny screws.

 

[INSTG8R]

I just went for it despite having warranty still, giving by everyone opinion that the thermal paste is dry cheese in the thread lol. It was pretty scary though as I thought i might break it but I took my time and all my tools out prepared like magnetic tray and screwdriver set for any funny screws.

Oh I’m not scared of doing it. I did my Fury I really just want to see it naked and take naughty pictures of it. Also it’s a bit of a pig to install as it comes with a “support bracket” that was a bitch to to get right.

 

[kurosagi01]

Oh I’m not scared of doing it. I did my Fury I really just want to see it naked and take naughty pictures of it. Also it’s a bit of a pig to install as it comes with a “support bracket” that was a bitch to to get right.

Well one of my screws wouldn't tighten down properly which had me a bit worried but its been fine so far *fingers crossed*.

 

[INSTG8R]

Well one of my screws wouldn't tighten down properly which had me a bit worried but its been fine so far *fingers crossed*.

I’m bored I may have a go at it today actually...also fingers crossed.

Okay sorry for the double post but needs it’s own. Temps are maybe 2-3 cooler(I’d really should stop monitoring the Hotspot it’s just alarming at 90C) but it’s definitely boosting consistently higher. Now for some porn! Also note the unused header that look melted but it’s just filled with some kinda glue.

 

[kurosagi01]

I’m bored I may have a go at it today actually...also fingers crossed.

Okay sorry for the double post but needs it’s own. Temps are maybe 2-3 cooler(I’d really should stop monitoring the Hotspot it’s just alarming at 90C) but it’s definitely boosting consistently higher. Now for some porn! Also note the unused header that look melted but it’s just filled with some kinda glue.

That head doesn't look right, better cool it down with some sapphire nitro.

 

[INSTG8R]

That head doesn't look right, better cool it down with some sapphire nitro.

Best guess is that header was for when the cooler had LEDs in it they stopped that with my revision. Only my backplate lights up

 

[kurosagi01]

Best guess is that header was for when the cooler had LEDs in it they stopped that with my revision. Only my backplate lights up

Meant to be a poor pun but nevermind lol, but yeah it does look like the header for the LED switch thats on reference coolers.

 

[INSTG8R]

Just wanted to add replacing my thermal paste with MX-4 easily lowered my idle temps 5C could even go as far as saying 7C quite pleased with the results.

 

[fullinfusion]

Just wanted to add replacing my thermal paste with MX-4 easily lowered my idle temps 5C could even go as far as saying 7C quite pleased with the results.

Thermal Grizzly paste my friend, that's good for a 10c drop at the least. I used to use MX4 but my eyes were opened after TG was used. Tomorrow Nail polish to coat the transistors and CL metal for the copper cooler.. I know It will dry out but a few times will saturate it, the Gallium will soak in and I wont have to worrie about it after that.. Vega 64 Water bios here it come!

 

[Flyordie]

Fullinfusion, the issue was in the driver. May works great. Running it now.

Also, will have an update to my system specs soon.

Just bought a Gigabyte Aorus X399 Gaming 7 and the 1900X. (Just to get me by till the 2nd generation 16 core models come out and go on sale. I'll be paying this 1900X forward to a friend who I am hoping will dump his intel platform and go to a Threadripper platform. lol.)

 

[kurosagi01]

What are people thoughts on the EK 240R kit on the vega64+ryzen 5? I'm heavily tempted to try it as my first entry into watercooling.

 

[londiste]

What are people thoughts on the EK 240R kit on the vega64+ryzen 5? I'm heavily tempted to try it as my first entry into watercooling.

It'll work fine. Note that this is an aluminum kit though, meaning two things:
- you cannot use copper parts in the same loop which limits your choices for expansion
- while price is awesome, the tradeoff is that performance (temperatures across the board) is in the vicinity of 10-15% worse compared to the copper parts

 

[Dinnercore]

I gotta join here too. My build has a Vega 64 from MSI. The crappy air boost OC, which I only bought for the reference pcb. Put a Heatkiller IV on it and threw some water at it. Then flashed LE-bios over and now I fiddle with it.



There are some problems I had to overcome, like that my card demands voltage and refuses to cooperate if it has not atleast 1.13V. I still think it´s not really the voltage in P6/P7 but in one of the lower states, since it´s not binned to run the LE-bios. I use the 'memory voltage', which actually seems to be a baseline voltage for the card instead, and if that´s at 1V I can lower my P6/P7 a little more.

Atleast I can feed it some electrons without it heating up too much. This was after 20 minutes of Heaven benchmark with an attempt to get a stable baseline for core clock and +20% powerlimit:

core clk // HBM clk // core tmp // hotspot tmp // HBM tmp // VR SOC // VR Mem // GPU load // GPU power draw // Mem. usage // Voltage

1707.0 MHz , 1015.0 MHz , 42.0 °C , 64.0 °C , 46.0 °C , 61.0 °C , 59.0 °C , 99 , 313.0W , 1260mb , 1.1875V
1668.0 MHz , 1015.0 MHz , 41.0 °C , 63.0 °C , 45.0 °C , 61.0 °C , 59.0 °C, 99 , 300.0W , 1256mb , 1.1875V
1698.0 MHz , 1015.0 MHz , 41.0 °C , 65.0 °C , 45.0 °C , 61.0 °C , 59.0 °C , 99 , 314.0W , 1256mb , 1.1875V

It slurps the power like a champ, but the clocks are not where they could be. Next problem I encounter right now is, that wattman or the card does not care to what I tell it to do. That the core speed bounces around a bit is normal. On my card however, it sometimes just overshoots into the 1800MHz range which results in a crash shortly after. Like I set P6 to 1702MHz and P7 to 1737MHz and it runs at 1750MHz with 1.2xV and the powerlimit maxed out. This seems stable for a minute or two and then I suddenly see the speed jump for no obvious reason to 1820 and it crashes.
I´m still trying to figure out why it does that and how I can prevent that. Because I think atleast 1740ish I can get to work, if not for that sudden overshoot.

Oh and in non-3D workloads Vega behaves a lot better (that GPUPI result is still crappy tho, and I find it a bit funny that the last v2.4 produces faster results by about 0.3s):

 

[TheoneandonlyMrK]

I gotta join here too. My build has a Vega 64 from MSI. The crappy air boost OC, which I only bought for the reference pcb. Put a Heatkiller IV on it and threw some water at it. Then flashed LE-bios over and now I fiddle with it.



There are some problems I had to overcome, like that my card demands voltage and refuses to cooperate if it has not atleast 1.13V. I still think it´s not really the voltage in P6/P7 but in one of the lower states, since it´s not binned to run the LE-bios. I use the 'memory voltage', which actually seems to be a baseline voltage for the card instead, and if that´s at 1V I can lower my P6/P7 a little more.

Atleast I can feed it some electrons without it heating up too much. This was after 20 minutes of Heaven benchmark with an attempt to get a stable baseline for core clock and +20% powerlimit:

core clk // HBM clk // core tmp // hotspot tmp // HBM tmp // VR SOC // VR Mem // GPU load // GPU power draw // Mem. usage // Voltage

1707.0 MHz , 1015.0 MHz , 42.0 °C , 64.0 °C , 46.0 °C , 61.0 °C , 59.0 °C , 99 , 313.0W , 1260mb , 1.1875V
1668.0 MHz , 1015.0 MHz , 41.0 °C , 63.0 °C , 45.0 °C , 61.0 °C , 59.0 °C, 99 , 300.0W , 1256mb , 1.1875V
1698.0 MHz , 1015.0 MHz , 41.0 °C , 65.0 °C , 45.0 °C , 61.0 °C , 59.0 °C , 99 , 314.0W , 1256mb , 1.1875V

It slurps the power like a champ, but the clocks are not where they could be. Next problem I encounter right now is, that wattman or the card does not care to what I tell it to do. That the core speed bounces around a bit is normal. On my card however, it sometimes just overshoots into the 1800MHz range which results in a crash shortly after. Like I set P6 to 1702MHz and P7 to 1737MHz and it runs at 1750MHz with 1.2xV and the powerlimit maxed out. This seems stable for a minute or two and then I suddenly see the speed jump for no obvious reason to 1820 and it crashes.
I´m still trying to figure out why it does that and how I can prevent that. Because I think atleast 1740ish I can get to work, if not for that sudden overshoot.

Oh and in non-3D workloads Vega behaves a lot better (that GPUPI result is still crappy tho, and I find it a bit funny that the last v2.4 produces faster results by about 0.3s):

I will add you when im at my pc later, im on phone atm.
Have you tried setting p6+7 identically my card tops out at 1750 /1100 on benches but will game at 1700/965 stable , personally I keep it at 1630/960 1.1v and 25% power slider just to keep power use down , chill works very well too imho.

 

[Dinnercore]

I will add you when im at my pc later, im on phone atm.
Have you tried setting p6+7 identically my card tops out at 1750 /1100 on benches but will game at 1700/965 stable , personally I keep it at 1630/960 1.1v and 25% power slider just to keep power use down , chill works very well too imho.

Thank you, and I have tried your suggestion in the past hour with some mixed results. It does run but for me I see not much difference in behaviour. Set it to 1752/1752 and 1.25V just to see where clock will settle and again it went far over that. Even in GPUPI it ignored the fixed 1750 and ran the same 1840 as if my clock target got some kind of override from the 1.25V voltage. Maybe there is some internal table where it checks voltage and applies a fixed clock based on that with only a little bias from the clock target we set. Like before I set 1802MHz in wattman with the result of max freq. hitting 1846, and now with setting 1752/1752 it 'only' hit 1830, with slightly lower performance then before.

For gaming I got my setting to save power too, 1587 at 1.1v, HBM 985. Power slider still maxed, ~200W. Core temp staying in a very comfortable zone around 32-34°C. If only that stupid ambient heat here would go away... 26°C room temp is the best I can do at night.

Still wonder about that overshoot. Vega usually stays below the clocks you type in wattman, atleast it does so on my gaming setting. Maybe it´s just a wrong readout or more a symptom of instability rather then the cause of it...

How exactly does chill work? Limiting framerate based on moving scene or user action input? I often play csgo and like to stay at 144fps. Got a 144hz freesync monitor. For csgo I don´t use chill but cap my fps at the refresh rate. It doesn´t take much to hit 144fps in that game, but I don´t know if I would like chill in titles like Dark Souls to cut my frames below 60.

 

[TheoneandonlyMrK]

Thank you, and I have tried your suggestion in the past hour with some mixed results. It does run but for me I see not much difference in behaviour. Set it to 1752/1752 and 1.25V just to see where clock will settle and again it went far over that. Even in GPUPI it ignored the fixed 1750 and ran the same 1840 as if my clock target got some kind of override from the 1.25V voltage. Maybe there is some internal table where it checks voltage and applies a fixed clock based on that with only a little bias from the clock target we set. Like before I set 1802MHz in wattman with the result of max freq. hitting 1846, and now with setting 1752/1752 it 'only' hit 1830, with slightly lower performance then before.

For gaming I got my setting to save power too, 1587 at 1.1v, HBM 985. Power slider still maxed, ~200W. Core temp staying in a very comfortable zone around 32-34°C. If only that stupid ambient heat here would go away... 26°C room temp is the best I can do at night.

Still wonder about that overshoot. Vega usually stays below the clocks you type in wattman, atleast it does so on my gaming setting. Maybe it´s just a wrong readout or more a symptom of instability rather then the cause of it...

How exactly does chill work? Limiting framerate based on moving scene or user action input? I often play csgo and like to stay at 144fps. Got a 144hz freesync monitor. For csgo I don´t use chill but cap my fps at the refresh rate. It doesn´t take much to hit 144fps in that game, but I don´t know if I would like chill in titles like Dark Souls to cut my frames below 60.

Chill seams to just restrain the gpu a little i have not used it at such high fps, could possibly cause issues ,your overshoot is weird, Vegas never done as it was told for me exactly but it's not been like that either, im on a stick air bios, perhaps its more pore asic aware or refined for.

 

[MrGenius]

(that GPUPI result is still crappy tho, and I find it a bit funny that the last v2.4 produces faster results by about 0.3s):

That'd put you in 10th place on HWBOT. I wouldn't consider that very "crappy" myself. A little room for improvement, but not bad. If you want to try getting those 0.3 seconds back you should run it with the latest blockchain driver. And/or on Windows 7 without HPET. I can almost guarantee you'll get better results.

EDIT: Also most(but not all) of the scores that are faster are using the 100M Batch Size. Looks like 5 of the 9 faster scores than yours on HWBOT are with Batch Size: 100M and Reduction Size: 256. It matters. That's all I know.

EDIT 2: A little update on mine. I modded the stock VRM heatsink and replaced all the thermal pads with Fujipoly SARCON® XR-m. Ungodly expensive SOBs(don't ask how much I spent on all those). But they're the best thermal pads money can buy(17 W/m-K). I've got it all put together without the cooler mounted. I'll insert a pic of what I did with that VRM heatsink here pretty soon. Then I'm gonna do the last step and mount the Morpheus II. I haven't been in any real hurry to get it all done. But I might be picking up a new 1440p 144Hz monitor in the next few days(possibly today). So that oughta give me a little more go to finally get it finished up and ready to play.

Got a little crazy with the hacksaw and drill(J-B Weld to the rescue!). But I think it'll work pretty good. Stuck on a bunch of the small heatsinks provided with the Morpheus over the MOSFETs. I intended on putting a few more over the chokes too. But I probably don't really "need" to. That and I'd have to cut them into various shapes and sizes to fit under the cooler. Which I hadn't fully anticipated. And is more of a PITA than I want to deal with at this point. Maybe later. Anyway, that's why there's still some areas with the black powder coat sanded off and left bare. Which shouldn't hurt anything leaving as is(might even help).

 

[Dinnercore]

EDIT 2: A little update on mine. I modded the stock VRM heatsink and replaced all the thermal pads with Fujipoly SARCON® XR-m. Ungodly expensive SOBs(don't ask how much I spent on all those). But they're the best thermal pads money can buy(17 W/m-K). I've got it all put together without the cooler mounted. I'll insert a pic of what I did with that VRM heatsink here pretty soon. Then I'm gonna do the last step and mount the Morpheus II. I haven't been in any real hurry to get it all done. But I might be picking up a new 1440p 144Hz monitor in the next few days(possibly today). So that oughta give me a little more go to finally get it finished up and ready to play.

Got a little crazy with the hacksaw and drill(J-B Weld to the rescue!). But I think it'll work pretty good. Stuck on a bunch of the small heatsinks provided with the Morpheus over the MOSFETs. I intended on putting a few more over the chokes too. But I probably don't really "need" to. That and I'd have to cut them into various shapes and sizes to fit under the cooler. Which I hadn't fully anticipated. And is more of a PITA than I want to deal with at this point. Maybe later. Anyway, that's why there's still some areas with the black powder coat sanded off and left bare. Which shouldn't hurt anything leaving as is(might even help).

Holy moly, I would love to see the results on your project. I can relate to 'uneconomical' measures for tinkering with hardware.

Thank you for the advice, actually 8th place is currently mine and you are right those results are not bad, but I feel I can get a bit more out of it. Limited currently by high ambient temps (1860MHz and 1150 HBM only crash at batch 14-16 when temps settle at the 50°C range), not being able to raise vcore without hardmodding it, stuck on win10 and it´s shenanigans till I finally finish my current 1050ti benching escapade, get the boot ssd out of that system and prepare it for the vega. It´s all those things that I want to push and just can´t atm because I don´t have the time (and the right weather lol) for it. So frustrating... And after all this build is not ideal for benchmarking a gpu, since low ipc cpu. Not really sure if I want to put that much effort just for GPUPI, on the other hand I really want to put that much effort into it...

Atleast I managed a 2nd place in 32B since not many make the effort to run that one