The Radeon RX 6900/6950 XT Owners Thread.

[Nugsy88]

First time AMD card for myself, coming from a 1080. Just trying to do some OCing on the card and quite different, some guidance would be welcome. What tool should I be using and maybe a starting point, I have never undervolted etc

I am seeing in COD Warzone that my clock frequencies are boosting from 1700-2500 but its up and down like a yoyo and I feel I am getting inconsistent fps 160-260 on 1080p, card is running around 53 degrees. On previous Nvidia cards, once you found your peak boost it would stay at that boost consistently while under load.

Running 5600x
16gb 3600 cl14
Red Devil 6900XT
Rog Strix X470-F
850w EVGA NOVA G3

 

[LtMatt]

First time AMD card for myself, coming from a 1080. Just trying to do some OCing on the card and quite different, some guidance would be welcome. What tool should I be using and maybe a starting point, I have never undervolted etc

I am seeing in COD Warzone that my clock frequencies are boosting from 1700-2500 but its up and down like a yoyo and I feel I am getting inconsistent fps 160-260 on 1080p, card is running around 53 degrees. On previous Nvidia cards, once you found your peak boost it would stay at that boost consistently while under load.

Running 5600x
16gb 3600 cl14
Red Devil 6900XT
Rog Strix X470-F
850w EVGA NOVA G3

Navigate to the performance tab in Radeon Software, then Tuning.

Increase the power limit to +15 and click Apply.

Set memory to 2012Mhz, fast timings click Apply.

Under GPU Tuning, enable Advanced Control and whatever is youe default Max Frequency, set the Min Frequency to 100Mhz exactly below it, Click Apply.

Test. This is a good starting point.

From here you can now choose to undervolt and save some power. At stock you should be able to reduce voltage to around 1.025v - 1.100v. Your mileage may vary.

 

[Besty]

It does if you use the app

for forum BB code links ?

 

[Besty]

First time AMD card for myself, coming from a 1080. Just trying to do some OCing on the card and quite different, some guidance would be welcome. What tool should I be using and maybe a starting point, I have never undervolted etc

do what matt said above but the clock speed can jump around quite a bit, this is normal. be grateful if you can run timespy/firestrike and post the result in the official thread here.

 

[Space Monkey]

for forum BB code links ?

If you just need to host the photo to actively show on the forum, then yes, as I use it all the time myself, there's a trick to it that most seem to miss

 

[Jumper118]

Morning chaps

I think some crossfire is required

Gigabyte reference card and a Sapphire Nitro+ OC

 

[LtMatt]

Morning chaps

I think some crossfire is required

Gigabyte reference card and a Sapphire Nitro+ OC

Oof, look at you.

I think you'll struggle to find many games/benchmarks that work in CrossFire.

Are they for two different systems?

Interested to know your thoughts on both GPUs.

Also, what is the stock core clock for each GPU as shown in Radeon Software?

 

[Jumper118]

Oof, look at you.

I think you'll struggle to find many games/benchmarks that work in CrossFire.

Are they for two different systems?

Interested to know your thoughts on both GPUs.

Also, what is the stock core clock for each GPU as shown in Radeon Software?

i know time spy and port royal and super position will work. i am just testing time spy now. i bought the ref one, but then i saw someone had a nitro for sale, so i bought that was well. I'll probably end up selling one of them tbh. I just want to get the mega bench numbers while i have 2 as this kind of thing doesnt happen that often

Thoughts so far are, the ref card is much heavier and feels way nicer. the nitro is much bigger physically, but its 250g lighter. Its actually quite disappointing tbh. Even the 5700XT nitro is heavier. The nitro has a power limit of 300w when using the +15% whereas the ref card seemed to top out at 280w. I have used more power tool on the ref card though and it can easily pull 380w no problem.

The stock boost on the ref card is 2250 the stock boost on the sapphire card is 2285, in wattman ref says 2484 max and sapphire says 2494 max.

 

[Nugsy88]

@LtMatt Thank you for those tips

@Besty

Not ran firestrike or timespy before but this is my score after running firestrike with the settings LtMatt posted

 

[LtMatt]

i know time spy and port royal and super position will work. i am just testing time spy now. i bought the ref one, but then i saw someone had a nitro for sale, so i bought that was well. I'll probably end up selling one of them tbh. I just want to get the mega bench numbers while i have 2 as this kind of thing doesnt happen that often

Thoughts so far are, the ref card is much heavier and feels way nicer. the nitro is much bigger physically, but its 250g lighter. Its actually quite disappointing tbh. Even the 5700XT nitro is heavier. The nitro has a power limit of 300w when using the +15% whereas the ref card seemed to top out at 280w. I have used more power tool on the ref card though and it can easily pull 380w no problem.

The stock boost on the ref card is 2250 the stock boost on the sapphire card is 2285, in wattman ref says 2484 max and sapphire says 2494 max.

Thanks for the info mate.

Agree about the reference card. I had a Red Devil 6900 XT but sent it back to keep the reference 6900 XT.

I know the Red Devil would have been faster, but the reference card is just too nice to send back for me.

@LtMatt Thank you for those tips

Not ran firestrike or timespy before but this is my score after running firestrike with the settings LtMatt posted

That's a good score, well done.

 

[Nugsy88]

Settings used for previous score

2349 min 2449 max
2012 mem fast timing
power limit 15%
1100mv

same settings but max frequency 2700 produced this, maybe a lot more left in the tank? unsure

 

[LtMatt]

Great score, I can’t score this high in this bench but my system under perform in this particular bench.

check out the Firestrike Extreme thread and benchmark as I’ve posted a decent score there.

 

[Besty]

scores

Great stuff mate, if you click compare result online and share the link for the timespy score here from the 3dmark website we can tell you whether there is further to go. Based on your CPU the scores you get look bang on to me.

 

[eeii]

Morning chaps

I think some crossfire is required

Gigabyte reference card and a Sapphire Nitro+ OC

lol I cannot find one in stock

 

[Godhand007]

Happy new year guys. Hit a little bit of snag with my card. Was getting random PSU shutdowns even on stock (+ PL) during benchmarks. Trying various things today; Loaded optimized defaults in BIOS, No CPU oc and an MPT of 350W + 15% (just to test if my PSU is the issue), No crashes for about 2 hours in Port Royal loop and around 1 hr in OCCT with PSU tests. I think this might be due to a CPU undervolt I did a while back as I get that message during startup "The previous overclock settings have failed, system has been restored to it's default settings. Press F1 to enter setup. Press F2 to load default values and continue.". Can you guys help me determine whether my PSU is faulty or not? Already tried new PCI-E 8 pin cables and connected to different slots on my RM1000i. OCCT PSU test did not fail for around 40 minutes at which point I cancelled it.
In the mean time, try beating this score https://www.3dmark.com/pr/724618; only for the brave.

 

[Jumper118]

lol I cannot find one in stock

Yeah they aren't easy to find, you've just got to be in the right place at the right time.

 

[Godhand007]

It's fine wine, you need to look at it, then smell it, then you can drink it.

Wanted to ask a question about my system. I am facing a unique issue where I have set my monitor to turn off after 25 minutes and when I turn it on by moving mouse, my clicks on browser or anything else don't work for 1-2 seconds then everything is fine. Was wondering if this is related to AMD card as I don't recall such a thing happening with my 2080 ti. Has anyone else faced or is facing such an issue?

Was able to resolve this issue by switching mouse and keyboard to USB 3.0 ports. Just FYI for anyone who comes across a similar issue.

 

[Nugsy88]

@Besty

https://www.3dmark.com/3dm/55990125?

https://www.3dmark.com/3dm/55989543?

 

[Godhand007]

@Besty

https://www.3dmark.com/3dm/55990125?

https://www.3dmark.com/3dm/55989543?

Why are you guys not testing Port Royal given that it is the most demanding test for RX cards, is it not included in base 3Dmark?

 

[LtMatt]

Why are you guys not testing Port Royal given that it is the most demanding test for RX cards, is it not included in base 3Dmark?

No you have to pay for it from what i can see.