6900xt turns on/off while gaming (solved)

So I got this red devil ultimate, only had a bit of time to play with it.
So far I tested with furmark:
2.5GHz 1150mv (1060mv read hwinfo) 320W read (390W measured at wall card only)
2.45GHz 1130mv (1009mv read) 275W read
2.4GHz 1120mv (930mv read) 245W read

I'm happy with the 2.4GHz, testing in forza 5 card sits at 200-210W, 55C/67C (gpu/junction)
Witcher loads the card basically almost like furmark, 235W 60/75C

Now my question is, I've seen the game stuttering from time to time, especially in witcher more often and looking at the graphs I can see the card going to 0% utilization for like half a second when it happens.
I doesn't reach the TGP/TDP (350W set in software), basically goes to 200-230W depending on game. Temperatures are OK. There is no issue with power delivery.

Any thoughts?

 

Do you get those issues with the card at default?

 

Yes still does that, GPU utlization 0% for 1 sec now and then.

 

It could be your PSU.
Can you test with a different one?

 

I increased the min clock to 2.2GHz and still does it but a lot less often.
I don't have a spare psu, running on a 850W gold+.
Should 230W card be a problem lol? I measured on the wall while gaming 320W whole system, 360W spiked.

 

The problem with modern GPUs is that they have very variable clock rates. So it demands that the PSU can adjust the power delivery really fast.

 

6900XT is a 300W card, reviews have shown that it can use 350W and can have transients higher than that, you probably wont see the transients at the wall.

 

Fair enough but it powered a similar load before, even greater than this without any issues running a nvidia card.

 

Yup it is as I measure myself up to 390W on the wall if left on default.
But even with UV and 270W on wall, rest of the system is 100W, PSU is 850W should be no problem.

 

Some PSUs aren't able to handle the wide swing of transient loads modern high powered GPUs and often trigger over current protections as a result, or aren't actually able to handle the load swings and end up dropping voltage/currant. What PSU are you using exactly?

 

And how many cables do you have connected to the card?
Do not piggy back cables. Use 3 separate PCI-E outputs from the PSU.

 

corsair rm 850 gold.
12v line drops to 11.98v when I have that 400W +/- gaming load.

 

I'm not. It's still a single rail psu.

 

It still baffles me how a UV card at 230W (270W wall) needs more than a 850W PSU.
My 3070 had a higher consumption than this and there was no issue at all.

Thing is the game doesn't crash nor the system.
It's just GPU utilization that goes to 0%.

I expect if there was a voltage failure for 1 sec the game to freeze or system crash or driver crash reboot etc.

 

I will try with more games to see if it's not game related, though it's weird

 

Lol another short update.
Looks to be from my undervolting, from I tested further on.
If I set anything lower than 1175mv in amd tuning I get those hiccups, regardless if I set max clock to 2200 or 2500.

I don't have alot of xp with amd cards, basically last one I had was an ati lol.
I've seen guides and videos with 6900xt UV going as low as 1050mv.

Is my card crap asic or it already drops vdroop too much so I can't UV as per other guides?

 

UV per se is chip(GPU) dependant.
If u have weaker chip, u need all voltage it supposed to have, I mean stock or just below. Good chips do better. Silicon lottery....

 

Though it seems You have resolved your issue I would like to mention tha, Its not so much that modern cards "need" it , its more that the psu are often pieces of crap. don't be deceived by branding, almost every vendor carries garbage from time to time. I remember when the fury launched, it had Extremely high load spikes, and there were quite a few psus that simply could not deal with it even if they were rated high enough. You can say that its the card's fault, but ultimately Its because those psus aren't as advertised. there is a big difference between 500w continuous rated psu and a 500w peak rated psu, thankfully there are fewer of those these days, but you'll often still see things like continuous X00w @ 25c, which isn't exactly realistic if its inside a computer case.

 

I have a Red Devil LE and my current UV is 1070mv stable currently.

Some good advice here and your problem could be PSU related but I'd say it's a long shot, simple test is reduce your driver max frequency to 2000mhz which you've basically done at 22000mhz to hit power draw about 250w well under your PSU. I was having instability and spikes with a CoolerMaster Vanguard 1000W, bought a new Corsair HX1000 and still had instability issues for a while.

I've found the card will not 100% lock to driver tuning settings, instead use a BIOS table power level range, so it will increase well past your max power limit to keep stability regardless of your custom settings in any app except MPT. So it's normal for power limit spikes over or under.

I had the following issues with instability that might help you.

1. Caused by a combo of mobo (1.2.0.4 A combo) PBO tuning and graphics driver auto tuning, if you have MOBO PBO tweaks make sure the driver CPU Global Tuning for CPU is set to Default and not Overclock CPU. System auto tuning must be set to Default, Your GPU should only be set Custom. I only tweak bios in the bios and not in Windows now, use latest 1.2.0.7 combo version even if beta.
2. Another Fix and test for above tho should be better in latest drivers now, can also try re-installing drivers with out RyzenMasterSDK.msi found in Packages\Apps\RyzenMasterSDK\ so there is no CPU tuning from the driver level at all, you'll still get all your GPU driver tweaking.
3. I also had two old crap SSD's causing instability doing my Instant Replay/Manual Recording/Instant GIFF recording, disable recording and test and change to a newer SSD drive/update or install latest codecs, I use HEVC which was only a problem in Windows 10.
4. Keep your GPU voltage to 1100mv @ 2514mhz during testing, your card should easily be able to UV stable to this IMO, the Red Devil cards have the best hardware headroom and max Power rating is 400w.
5. Don't rely on the software you're using to detect your settings, some drivers have broken app metrics including the Radeon software in previous cases, but if you're seeing in game lag spikes then that's obvious. I'm not having issues with 22.7.1 tho.
6. I also reset driver cache with each driver install found under Settings, Graphics and at the bottom of Advanced, I also run Windows Disc cleanup selecting Direct X Shader cache and use the obligatory DDU.

These fixed all my instability issues to date hope this helps.

 

@somemadcaaant
My card is not stable if I set less than 1175mv in the voltage slider.
Doesn't matter if it runs at 100W or 320W or any frequency (down to 2GHz).

What I run now is 2300min - 2400max , 1175mv.
On 100% load stays at 2350MHz @ 980mv, about 275W reported.

Doesnt work if I move the slider just a little to 1170mv, in GTA 5 which seems to be most sensitive it just spits DX error, despite the card runs at 185W in game. If slider is at 1175mv runs fine for just 5-10W more same clocks

Every game I tested so far at 1175mv runs fine, clock 2350ish, in radeon software says 2350MHz@980mv , consumption for card 180-270W depending on game.
Any milivolt lower than 1175 if I set on that slider GTA5 error, Forza 5 works down to 1130mv but with ocasional stuttering.