6800/6900 Overclock Results

Tried those settings in your screenshot kwikgta and my phantom locks up on the second test. So, we don't have identical twins. lol

That said, based on my testing today it seems this card tolerates lower voltage settings better than the nitro plus did.
1080v
+15%
2560/2660mhz
19813 on timespy graphics test

With some further adjustments and a memory overclock 20K+ doesn't seem far off.
 
Tried those settings in your screenshot kwikgta and my phantom locks up on the second test. So, we don't have identical twins. lol

That said, based on my testing today it seems this card tolerates lower voltage settings better than the nitro plus did.
1080v
+15%
2560/2660mhz
19813 on timespy

With some further adjustments and a memory overclock 20K+ doesn't seem far off.
I just ran DDU, pulled out the Sapphire 6900xt, popped in the Phantom , downloaded the latest Adrenaline, and put those settings in off the top of my head. I expected to lock up but it didn't and I just about shit my pants when it scored 20k. I tried to take it further but it locks. So i'm just gonna be happy with 20k :)
 
One thing I've noticed with timespy is that I can sometimes get it to finish a pass but it isn't consistently stable. Sometimes a second or third attempt results in a CTD or a lock-up.

Below are the settings I'm currently on that seem to work well in timespy. No memory overclock tried yet. 20K plus should be within grasp here. I did try these settings in TD2 this morning. I had a single CTD after 3+ hours however I'm not sure if it's due to the overclock or just the game crashing for some other reason. Regardless, I'm going to hold here for a while and run some further game testing then maybe next weekend if all is well move on to some memory overclocking. If I have any further issues, will bump up vcore to 1090-1100mv.
 

Attachments

  • 3dmark.PNG
    3dmark.PNG
    678.4 KB · Views: 0
  • wattman.PNG
    wattman.PNG
    138.4 KB · Views: 0
One thing I've noticed with timespy is that I can sometimes get it to finish a pass but it isn't consistently stable. Sometimes a second or third attempt results in a CTD or a lock-up.

Below are the settings I'm currently on that seem to work well in timespy. No memory overclock tried yet. 20K plus should be within grasp here. I did try these settings in TD2 this morning. I had a single CTD after 3+ hours however I'm not sure if it's due to the overclock or just the game crashing for some other reason. Regardless, I'm going to hold here for a while and run some further game testing then maybe next weekend if all is well move on to some memory overclocking. If I have any further issues, will bump up vcore to 1090-1100mv.
wattman is buggy. I was plugging in settings and was stable with 5 runs and an hour gaming. turned off pc, came back later and turned it back on and radeon settings wouldn't load. it just crashed. I've had to reinstall twice.
 
Oh hell yes! Very nice :)
I think I found the power limit on the reference boards. I have been using 325w ppt override in MPT and at +15% to that it was locking my system up half the time in the second graphics test during the heavily loaded portion (first half of test 2). Dropping the power boost to +12% keeps max GPU power draw at about 360W and eliminated the crash. Considering the pci-e + 2x 8 pin is spec'd for 375W max this makes sense, I know my motherboard is pretty sensitive if a card has the audacity to try and pull more than 75W from the slot. That said, these are much better behaved on the power draw ratio by source than my Vega 64 was.
 
R-Type:
You are on water right now is that correct?

With the settings I posted in #123 the wattman overlay was showing about a max of 330w during gameplay, 92c junction temps with fans spun up 80%.

EDIT: Loaded the driver released on 1/21 just now, rebooted, and with same settings as in post #123 am down by 700 points in timespy. WTF?
 
Last edited:
In your radeon settings do you have 10-bit pixel format enabled?

Settings->Graphics->Advanced->10-bit pixel format

If you enable this, and re-test, does your graphics score suffer?
 
R-Type:
You are on water right now is that correct?

With the settings I posted in #123 the wattman overlay was showing about a max of 330w during gameplay, 92c junction temps with fans spun up 80%.

EDIT: Loaded the driver released on 1/21 just now, rebooted, and with same settings as in post #123 am down by 700 points in timespy. WTF?
If you use MPT, you have to reapply the override anytime you update drivers.

Yes I'm on water, average temps during the run were 41C, junction 54C.
 
Your CPU score keeps increasing, obviously some undiscussed platform optimization happening in the background!

Nice score on the card, you think you're close to max for air?
Lol, I'm still on PB04 and LLC8 but on a newer bios than I originally had. Hopefully I'll get a day off work before long to play core optimizer. I think my card is maxed out on air.
 
Yeaah water block got here today. Hopefully this weekend I'll have time to get it mounted and tested but wifey has some honey doo's I owe her first for letting me splurge on the card and block.
 
Yeaah water block got here today. Hopefully this weekend I'll have time to get it mounted and tested but wifey has some honey doo's I owe her first for letting me splurge on the card and block.
Very nice, we’re open for new members in the 21k Timespy club. :)
 
And the 6800XT 20K club on air. ;)
Ah right, that too. Sorry, I didn’t see you standing over there behind that giant fan curve. ;)

This actually reminds me I should start aggregating scores in the OP for people new to the thread. I’ll setup a spreadsheet and then post the variables for people to PM me.
 
Ah right, that too. Sorry, I didn’t see you standing over there behind that giant fan curve. ;)

This actually reminds me I should start aggregating scores in the OP for people new to the thread. I’ll setup a spreadsheet and then post the variables for people to PM me.

I didnt have fans at full speed, max was 65%. There wasn't a large gain in temp drop with high fan speeds since the main issue was power or voltage.

Can't wait to see this card under water.

You could edit your first post and put up a spreadsheet.
 
I didnt have fans at full speed, max was 65%. There wasn't a large gain in temp drop with high fan speeds since the main issue was power or voltage.

Can't wait to see this card under water.

You could edit your first post and put up a spreadsheet.
Exactly the plan, and I was just teasing. :)
 
Got my Nitro+ this morning. It weighed in at 2lb 11oz, my reference was 3lb 4oz. The reference card is SOLID. Just popped it in and plugged in the same exact OC setting as my reference and this is what I got (all background processes running still):
1611880034925.png

Pretty solid score. Some comparison shots.
1611860893465.png

and height:
1611860953999.png
 
Last edited:
Very nice, we’re open for new members in the 21k Timespy club. :)
could you be kind enough to post your MPT settings by chance? Im still on air with my Asus TUF 6900XT and have my "watts" set to 325w with MPT, with that said, I can only cool 347w (7% power limit) then hotspot gets to 100c+ and driver crashes card CTD will occur. Hoping when a company finally releases a FCWB for this card, i can tame it enough to be able to use the full 15% powerlimit to 375watts and see if I can finally break 21000 in TimeSpy! Pretty sure I can, atleast im hoping. Would just like to see what you are running. this is my absolute best score ive been able to achieve on air with MPT, however ive been yet to achieve it twice.... its like the moon and stars have to be aligned perfectly LOL!
uv mpt oc.png
 
Last edited:
Getting so close...

Asrock Phantom 6900XT
View attachment 323890
I can attest that on air, yes, it seems like 19.8k is easy to hit and 20k an epic milestone.

The biggest issue I've seen people have is setting their boost clock too high and then requiring excessive voltage to maintain it. I think 2650mhz @ 1060-1080mV is a good starting point for a 6900XT OC. If its stable and delivering average clocks ~70mhz below setting, then you have enough voltage and can increase boost frequency further. If you're crashing, you need more voltage. You'll find a pretty narrow mhz band where that voltage/frequency/score curve inflects. That's the holy grail...
 
Eye ball average around 2540MHz in the first graphics test and hovering around 2500Mhz in the second graphics test. I haven't got close to the actual max speed, but from what I tested having the high max speed pushes the clock generally higher.
 
Eye ball average around 2540MHz in the first graphics test and hovering around 2500Mhz in the second graphics test. I haven't got close to the actual max speed, but from what I tested having the high max speed pushes the clock generally higher.
It does to a point, though only in lightly loaded situations (and in games this may pan out more than Timespy due to their more lightly loaded nature). That said, lowering your peak speed to 2700 or below and dropping your voltage some (you'll have room to avoid a crash since you lowered the peak speed) should result in higher average speeds.

Consider this: When power limit is not an issue, the cards will self regulate to about 70-60mhz less than the boost frequency you set. If you're routinely seeing a bigger difference than that, you are power budget limited and a lower voltage or MPT will help. If your clocks are in that range, then increasing the target frequency makes sense.

What doesn't make sense is people setting their target frequency to 2700mhz at max voltage, getting 2400mhz actual, and thinking they've got an effective overclock. That's not an indictment of anyone on the thread, but every review of these cards took that approach and dramatically understated the OC potential.
 
I'm planning to play around with the clock speeds more this weekend. Last weekend I started playing with the min clock speed more and found that helped with Time Spy.

Probably do like you said and see if I can get the average clock higher by reducing the max clocks/voltage.

I want to play around with the utility to update the max power draw, but I am on a reference card so I'll need to be careful if I do.
 
In testing with TD2 this morning I've noticed that the voltage still ends up at or near 1150mv while in-game playing a control point. So despite being set for 1080mv in the driver it doesn't really maintain voltage that low in practice while under load. This also explains why junction temps seem sort of high to me at 92-94c even when the fans are spun up 75%. It's using power (320-330w), hence producing heat.

I'd like to lower the temps a bit and pull back so I can cut the fans down a notch if possible. Doing so would probably push me under 20k timespy though.
 
Ya'll tried the new afterburner beta with 6800 support?

I really don't like having the drivers do a reset and blowing away settings.......
 
Back
Top