[TweakTown] AMD Radeon RX 6900 XT Liquid Cooled launched, faster than RTX 3090
113802
21 - 40 of 68 Posts
AMD used some unlisted 4900 RPM variant of the gentle typhoon ( only the 3000/4250/5400 rpm ones exist ) for the Vega 64 card.
System integrator products are always conservatively clocked and powered.
I wonder how much of the TDP is taken up by the 18Gb/s GDDR6...
Joined
·
49,299 Posts
It does not. I used to run Vega 64 at 400W with 120mm AIO and had no issues.
You must not have had one, it had a lower temperature cap to avoid evaporation which would cause it to thermal throttle sooner.
Memory in general is very sensitive to temperature, with the AIO I could only run it at ~1020Mhz or else it would slowly turn the screen black once it hit 55c+
My Titan RTX is in the same boat where once it hits a certain temperature the memory clock is unstable.
Here's my RX Vega 64 XTX with a waterblock at 400w running sustained at 1800Mhz with all the fans at 2150RPM. Same card that currently holds the #1 spot on FireStrike GPU score which I ran at 1827Mhz.
Joined
·
4,489 Posts
Depends on what your standards are. I had a 2080TI FTW3 (386w I believe?) on the 120mm Hybrid and it was great for benching, but sustained loads heat-soaked the rad and the card would eventually hit 60-65c. Same problem my 3080 450w on a 240mm rad suffers.. sustained loads soak the rad and water temp goes nuts. 65c isn't very good or what I aimed for.
Joined
·
49,299 Posts
60-65c bad?
Joined
·
4,655 Posts
That’s some big block, lopey camshaft unrestricted ‘Cuda power. The competition can only stare in awe.
Joined
·
14,666 Posts
Joined
·
4,489 Posts
Imo, yes. Hybrid coolers/AIOs are great for mediocre cooling while maintaining silence, but that's pretty much where they end. The air cooler that came with both my 2080TI FTW3 and 3080 FTW3 did roughly 65c on air, it was just louder. Hybrid/AIO is silent, which is great, don't get me wrong.. but cooling performance isn't very good.
GPUs benefit tremendously from lower temps, so imo, definitely the best thing to put under a full-cover block.
Joined
·
49,299 Posts
There is a difference between running 400w daily and 400w for benchmark runs. This is not the card if you want to run coo/quite/400w 24/7. custom water always required or that and even then, heat is heat. Dont want the GPU tp dump 400w when I am to game on a hot summer day.
I'm daily driving my 6900XT with 375W, +15%. It never draws anywhere near that when gaming, even with the core touching 2700mhz regularly and the VRAM maxxed out at 2150mhz. Core temp hovers in the low 40s (junction in the low-mid 50s) when gaming for a few hours, and that's in a loop with triple 360s, one of em a thicc boi. I think that a 120mm AIO would easily heat soak under those conditions, and probably start core throttling after a while, but that's a guess. I know my Radeon VII runs pretty hot too, in a similar loop at 350W. IMO, a 120mm AIO is just silly underpowered for this card if you plan on pushing the clocks a bit.
Joined
·
3,445 Posts
Out of the box my ftw3ultra 3080 was going 400W and boosting to just a hair above 2kMhz I dropped it down to 850Mv clocks went down to 1920 maximum which in gaming is going to be a super negligible hit to performance but i dropped 100W at full load and keep the card at 60% fans which is same noise level as the case fans and it doesn't go higher than low 60C at full load. Works for me. They definitely are pushing these cards harder than they're meant to be pushed as you quickly start hitting diminishing returns when it comes to power usage and performance. I always prefer the best balance i can find between power usage and perf since the electricity costs do add up. I used to run with power management at maximum in NVCP as well but went back to the default as the card was using an extra 60Watts at idle doing nothing because it wouldn't drop clock speeds.
Joined
·
6,327 Posts
Yeah, that's what I was wondering as well. If they binned this one better? Guess we'll see, if anyone snags one.
Joined
·
49,299 Posts
2850MHz/2400MHz
The main complaint was AMD's absurd max boost clock with Vega 10/20 and RDNA1 where it would never touch the speeds set for the highest p-state. Looks like they may have fixed it with RDNA2?
I could be wrong and it still behaves as the annoying max boost clock.
Joined
·
49,299 Posts
Its similar to Nvidia. You are power limited first, then voltage then temperature.
Just looked up a guide and it behaves the same as Vega/RDNA1. It still has that annoying max boost clock it never touches.
When eliminating both power limtiations/temperature. Nvidia cards easily hit the target clock I set while AMD cards always run ~50-100Mhz less.
Joined
·
49,299 Posts
Ah i see what you mean. i think Nvidia does not report target clock but actual clk. That is just the PLL trying to lock on. It will never hit the target clock.
21 - 40 of 68 Posts
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
-
?
