Discussion in 'Videocards - AMD Radeon' started by OnnA, Oct 29, 2020.
@SmokusPokus you can set VRAM settings to Fast but maximum frequency leave to 2100Mhz.
As for later, seems to me doesn't matter. For ram, do as Spajdr said. There is no sense to push VRAM to 2150, it hampers performance.
AAND. U can try switch card bios to "Rage Mode" !!!
Anyway, I accidentally switched it (while unpacking) so I do not know, what position it was on beginning and it does 2500Mhz(2455-2555Mhz) non stop with +15 power limit with 1130mV, no crashes, all flays. Good performance. 270-330W area.
But, set to 1130mv GPU-z still says 1.15v so dunno if it matters.
Alrighty will try it out. If I remember correctly, the BIOS is on default set to "performance" once. Switching it again will put it on "quiet" BIOS. Since I haven't switched anything, I guess I'm already in the "rage mode"
I still want to find out whether the restarts were because of unstable overclock or something else has caused it.
EDIT: By default it is set to "balanced" mode.
Rage mode isn't all that, slight increase to power and a mild overclock from the bios data the 6800XT and 6900XT has, I think you can do much better with manually adjusting the settings.
EDIT: It's kinda similar to the Extreme mode or whatever Vega had but instead of just burning extra power you do get some scaling and benefits from it but it's pretty low as it's fairly conservative.
Just going to borrow this image via Hot Hardware...
It's more for pointing out the higher overclock maximum on the 6900 but there's the various modes or bios readings for how these are applied might differ on some GPU models but yeah a bit more power letting the GPU boost slightly higher in certain situations.
Which would be all of them I suppose with the 6900XT especially since it's kinda power limited, feels like it should have had a little bit extra over the 6800XT but for some reason the two are almost identical sans the disabled cores which limits the GPU a bit as a result.
What I Iearned so far with infinite numbers of hours with tweaking of the 6800XT is that if you push the voltage too high in Wattman settings then the GPU is unstable.
I am currently using 2650MHz@965mV with 15% power limit.
I have with MPT the max GPX voltage set to 1120mV. I have also set in MPT power limit GPU to 275W.
This is less than what MERC319 bios in performance setting does. I think MERC319 has this at 290W if you use the performance bios.
It is working without any issues in stress test like Port Royal.
I played many hours of Assassin's creed valhalla and GPU clock was hovering between 2590-2615MHz all the time.
In Valhalla the GPU alone power consumption is in range of 220W to 250W. In TimeSpy GPU power goes at second GPU test up to 316W which is maximum that I have set it to be able to go to.
Guys, if you use HWiNFO and have been suffering from sudden reboots with WHEA crashes in Windows, take a look at this thread I created yesterday:
There, a group of affected users and myself, are working with the creator of HWiNFO to get to the bottom of it. So far, it seems to be related to the GPU sensors for RDNA2 in HWiNFO, and we're even testing a new beta to try and see if it solves the problem for us.
I'm letting you know in case there may be someone else around here affected by the same problem.
Thanks a lot for info.
I use HWINFO with 6800XT very often.
Luckily I did not experience this issue so far. I hope it remains like that.
I have version 6.42.
Last time I used it was to log data on weekend while playing Assassin's creed valhalla.
You're welcome. For us it would seem stable while gaming sometimes as well, and the fastest way to trigger it was by testing RAM stability using software such as TM5 or Karhu RAM Test.
Asrock posted new bios updates for some of their 6800 GPU models.
Challenger Pro - OC
Phantom Gaming D (The ??? kinda name for a GPU is this anyway. ) - OC
Presumably they tweaked a few parameters maybe it's possible to load these into More Power and there's some visible tweaks that can be compared against versus the stock settings for more in-depth info of what was changed.
Hey can I see a link for this as still playing with mine and want to see what your card runs at.
Clock running at average of 2500ish 2675 set but scores almost 1k more gpu score.
This is strange my junction temp is 8-10c above the normal 68-70c at full load
If you're having junction/hot spot temps that bad, it's clearly a sign of bad airflow, it's that simple.
Re-configure some things within your case and get some proper air flow going through there, I mean crap, a junction temp of 30c + the GPU temp means the heat is just pooling there and not being dissipated....
Looking over at you mini case builders who think fitting beastly components in tiny cases is cute
Thanks, I will run with the side panel off and see what I can feel for airflow from the front 3 intake fans and 2 exhaust fans.
Although it running much better since I last posted with undervolted and average 2350 GPU clock and 2100 mem clock
Attached photo of system for reference.
Looks like the MSI Gaming X Trio version of the AMD 6800XT possibly others got a bios update which resolves whatever was causing random black screen issues.
EDIT: Guessing some setting was either reduced or stabilized in some other way as this isn't a generic AMD 6000 series bios update but specifically for a custom model and what that might be using over the stock version.
Could make for a comparison if anything in More Power could show what's different I suppose but eh it's a good fix whatever it does for anyone affected by this issue.
My black screen issues after a prolonged idle time with the monitor off (e.g. during the night) were related to the C49RG90 problem --- first part of the fix was delivered two drivers ago with the G9 fix, the second and final fix was in the newest February driver. Since then no issues.
Yeah that falls under the other ongoing issue I presume that several early displays on HDMI needed firmware fixes for full support with the new 2.1 version here though I am not sure how this works.
If it's gradual for AMD and NVIDIA GPU's or can be resolved from a single firmware update for the display or multiple depending on GPU vendor or even GPU models here (Not that much but the latest actually utilize 2.1 though.) and it also seems VRR adds another layer of problems to at least some LG and Samsung displays where several firmware updates have improved compatibility since these displays were released.
Haven't followed that much either but it sounds like it's getting better and newer displays and TV's should have a more robust initial implementation and support across devices utilizing this new version of HDMI going forward.
Then likely a repeat with this all over once Display Port 2.0 is available unfortunately.
Think AMD and NVIDIA have also attempted driver workarounds and fixes where firmware updates have been slow or non-existent but both take time but it's nice to hear it's getting resolved.
EDIT: Or if there's other issues, sleep mode and AMD including fast boot and hibernation had a few after all and the OS going into this virtual display mode when the panel is entirely off and how that wakes or resumes, trying to remember if that was a early Vega or Fury issue too now that I think about it some more.
managed to pick up a 6800xt today for my fiancees rig, any advice on tweaks or settings i may need? havent had an amd card since my old x800xt, cheers
May sound weird, but check SA/IO votages, I mean, maybe worth to bump em a little. Or AMDs equivalent.
And just in case, disable PCI express "link state power management" in power plans in power options.