The Radeon VII Owners Thread

[L-Digital82]

Oh yeah just spotted that. Impressive, what did you set it at?

 

[Illuminist]

Oh yeah just spotted that. Impressive, what did you set it at?

2230 mhz and 1250 hbm with 0% powerlimit but the min and max sliders together on 2230

 

[eeii]

DDC pumps are smaller than the D5's mines been giving me sterling service for a couple of years. If the D5 is struggling the DDC will definitely will.

 

[Illuminist]

DDC pumps are smaller than the D5's mines been giving me sterling service for a couple of years. If the D5 is struggling the DDC will definitely will.

Thats weird as I have a Cheap XSPC res/pump combo that does the job magnificently giving me 30/40 core and 60/70 junction temps

 

[Chud_Jenny]

Gutted I've got a DDC pump and was really hoping this block was going to be good.

I think the block will be fine, I'm just pushing 3*360 1*120mm rads CPU and GPU blocks lol. My temps aren't terrible just will improve with a rethink on my flow.

 

[Illuminist]

I think the block will be fine, I'm just pushing 3*360 1*120mm rads CPU and GPU blocks lol. My temps aren't terrible just will improve with a rethink on my flow.

My lord that's a lot of rads would you not get better temps with less?

 

[Chud_Jenny]

My lord that's a lot of rads would you not get better temps with less?

I just added a 3rd 360mm in the bottom for a laugh. More rad area should help. A good rule is to have water in a rad and not in very long tubes as it can't cool in tubing.

 

[Scougar]

I just added a 3rd 360mm in the bottom for a laugh. More rad area should help. A good rule is to have water in a rad and not in very long tubes as it can't cool in tubing.

Did you add another pump to help with flow? 3 rads must be a lot of resistance.

 

[Chud_Jenny]

Did you add another pump to help with flow? 3 rads must be a lot of resistance.

No, I think the new thick 360 rad is what may of broken the camels back. My loop worked very well with my Vega ek block and without this new 360 rad. I'll do some trial and error and see how I get on. Worst case is going to a dual d5 bay res!? The resistance is more now for sure. I only ran at speed 2 before and get very good results.

 

[L-Digital82]

2230 mhz and 1250 hbm with 0% powerlimit but the min and max sliders together on 2230

Interesting, after reading this I had an experiment with Wattman settings as I have been using afterburner since it became compatible.

What is the deal with leaving power limit alone?

Using both sliders like you suggested I was able to set mine to 2175 on 1200mv, 2200 would instant crash no matter the voltage so I think I am at the upper reaches of my silicon but this is still a great chip. I left the power limit at 0 and hbm 1200

Running heaven this gave me a super stable clock speed around 2120 but when running a bench in 3dmark the actual performance was lower despite the higher, non fluctuating MHz. Like this the temps stayed around 80.
To get better performance with this new clock I had to pump the power level. With +20% it absolutely ripped and broke all my records on this card but the temps reached as high as 105 on water. I need to find a safe middle ground. What’s going on do you think?

 

[Chud_Jenny]

Interesting, after reading this I had an experiment with Wattman settings as I have been using afterburner since it became compatible.

What is the deal with leaving power limit alone?

Using both sliders like you suggested I was able to set mine to 2175 on 1200mv, 2200 would instant crash no matter the voltage so I think I am at the upper reaches of my silicon but this is still a great chip. I left the power limit at 0 and hbm 1200

Running heaven this gave me a super stable clock speed around 2120 but when running a bench in 3dmark the actual performance was lower despite the higher, non fluctuating MHz. Like this the temps stayed around 80.
To get better performance with this new clock I had to pump the power level. With +20% it absolutely ripped and broke all my records on this card but the temps reached as high as 105 on water. I need to find a safe middle ground. What’s going on do you think?

Interesting reading. That temp comes from loads of wattage use in this small package. There will be a sweet spot on pump speed that will show the limit of the heat transfer. Very like high core count Intel CPUs with big over clocks.

 

[L-Digital82]

Indeed. I’m more curious about the whole power limit thing. Illumi got his high scores with power limit at 0, i was able to get stable higher clocks like this but it didn’t translate into performance without the power limit being increased and the boosting the temp significantly.

 

[Chud_Jenny]

Indeed. I’m more curious about the whole power limit thing. Illumi got his high scores with power limit at 0, i was able to get stable higher clocks like this but it didn’t translate into performance without the power limit being increased and the boosting the temp significantly.

I ran my Vega undervolted to 1120mv with + 50 power limit @ 1677mhz liquid clocks. If you starve the card of power it will perform worse like you've seen. I'm going to try and run my vii around 1000-1020mv. I like getting more for less. The thing to do is find the sweet spot of the die, MHz vs MV and power needed. When you find that the temps will be very good. I could run my Vega @ 1632 default clock with 1100mv, so 20mv to get to 1677, and then it scaled poorly beyond that.
I'll get a grips on the behavior of the card under water and report back my findings next week. Got a ball valve arriving from ocuk Tuesday and awaiting new ram too so get back on track after that.

 

[Illuminist]

Indeed. I’m more curious about the whole power limit thing. Illumi got his high scores with power limit at 0, i was able to get stable higher clocks like this but it didn’t translate into performance without the power limit being increased and the boosting the temp significantly.

I think i read it on an overclocking thread on a different forum about the power limit, its so that the card does not boost resulting in greater stability, also when trying 2200 the junction reaches over 70.c and then it requires an extra 50mv to stabilise it I think i used 1250mv or 1270/1280 for that score. Also using the morepowervii or evenmorepowervii registry mods helped as they have higher soc and amps which help stabilise at those higher clocks without having to pump more watts with the power limit up

 

[L-Digital82]

Interesting. I will do some more tinkering this evening. I am going to try hit the sweet spot of performance without power limit boosting within normal tolerances based on what you have said. I’m a bit nervous about the custom tables and considering that my warranty is now dust due to putting a block on it, I don’t want to tempt fate and burn my chip with something too wacky. I will post back later

 

[eeii]

I think the block will be fine, I'm just pushing 3*360 1*120mm rads CPU and GPU blocks lol. My temps aren't terrible just will improve with a rethink on my flow.

My word sir that is radtastic

 

[Chud_Jenny]

Just one note for the members new to the water game. The sun has come out in the UK! Really it has! I think it followed me back from Majorca last week lol.
So just one thing to keep in mind is a higher ambient room temp will directly affect your loop temps. For example if you where 18C before and now 21C you'll more than likely notice 3C on top of the temps.

 

[Illuminist]

Interesting. I will do some more tinkering this evening. I am going to try hit the sweet spot of performance without power limit boosting within normal tolerances based on what you have said. I’m a bit nervous about the custom tables and considering that my warranty is now dust due to putting a block on it, I don’t want to tempt fate and burn my chip with something too wacky. I will post back later

I cast care to the wind and used the most powerful one, the evenmorepowervii_1433+++ one or whatever its called, it allows 2400 and 1500mv max's with 1400 hbm , but if i set the hbm over 1270 the screen went black until i lowered it again. i don't think it's possible to damage the card as there are hardware limits built into it and i tried +99% powerlimit 2400 core at 1500mv with 1250 hbm (for science!!!!) and it didnt explode it just crashed and restarted the pc back to defaults, at which point i thought... i wont do that again lol
It all comes down to the damn junction temps if they could be kept under 70.c the card would easily manage 2300+ core with whatever volts you throw at it as the custom VRMs on this beast are the best money can buy, but there are built in downvolting and underclocking safeguards that apply at 70.c 90.c and 110.c that prevent it performing the best it could

 

[Scougar]

I cast care to the wind and used the most powerful one, the evenmorepowervii_1433+++ one or whatever its called, it allows 2400 and 1500mv max's with 1400 hbm , but if i set the hbm over 1270 the screen went black until i lowered it again. i don't think it's possible to damage the card as there are hardware limits built into it and i tried +99% powerlimit 2400 core at 1500mv with 1250 hbm (for science!!!!) and it didnt explode it just crashed and restarted the pc back to defaults, at which point i thought... i wont do that again lol
It all comes down to the damn junction temps if they could be kept under 70.c the card would easily manage 2300+ core with whatever volts you throw at it as the custom VRMs on this beast are the best money can buy, but there are built in downvolting and underclocking safeguards that apply at 70.c 90.c and 110.c that prevent it performing the best it could

What's junction temps are you seeing on water?

 

[Illuminist]

My word sir that is radtastic

Hahha, that is why he is so chilled, he is far too rad for us