MSI Gaming X Trio 6900XT Overclocking

[Haider]

Hi,

I am overclocking using Afterburner using +12 on power limit and 2720MHz on core clock: -



1.) I am concerned difference between the GPU temp and hotspot, is that delta ok?

2.) My GPU clock is not hitting 2720 even though the card does not lock up or anything - 3D Mark Speed Way. What is the effective clock?

Thanks
Haider

 

[EarthDog]

From hwinfo....

Effective clock. This method relies on hardware's capability to sample the actual clock state (all its levels) across a certain interval, including sleeping (halted) states. The software then queries the counter over a specific polling period, which provides the average value of all clock states that occurred in the given interval.

 

[DaveB]

Hi,

I am overclocking using Afterburner using +12 on power limit and 2720MHz on core clock: -
View attachment 365167


1.) I am concerned difference between the GPU temp and hotspot, is that delta ok?

2.) My GPU clock is not hitting 2720 even though the card does not lock up or anything - 3D Mark Speed Way. What is the effective clock?

Thanks
Haider

The Hot Spot temp on AMD GPUs is safe up to 110C, so 88C is well within the low normal range.

 

[Haider]

Thanks lads

 

[Haider]

I followed an under-volting video on Youtube: -

I used 1080 for voltage that brought my 3DMark->Speed Way score to 3921, the average is 4177. I reckon a little tweak to the memory will get me to average score...

 

[Haider]

So I did benchmarking (Speed Way) and logged the values I noticed GPU Power went from 312W to 317W, GPU clock was 2550s-2660s MHz (2557 average), gaming workload on the Gaming X Trio according to Techpowerup review is 357W, a stock RX6900 is 299W (Techpowerup)
Source: -
https://www.techpowerup.com/review/msi-radeon-rx-6900-xt-gaming-x-trio/35.html

I'm impressed with increase in speed plus lower power. The Gaming X Trio stock 2340 boost clock and a stock RX6900XT is 2250. A 9.67% increase over stock 6900XT in terms of GPU clocks for a 5.67 increase in power.

 

[Haider]

Did some undervolting/overclocking today Time Spy Extreme 4K: -



Considering stock is 1175mv but I put power limit upto +12% since v=ir, resistance of a circuit is constant the current must also decrease and since P=VI or P=Isquared r therefore if current decreased then power consumption decreased. Can I then ask why do OC guides always say put power limit up to 12%?

Next I'll try running it at 1000mV or 1V tom, i'm trying to hit the average 9321 on Time Spy

Any ideas why AMD put a lower clock speed with 15% more current? Could it be sample variation i.e. TSMC can't get smaller variation in chip quality?

 

[EarthDog]

Can I then ask why do OC guides always say put power limit up to 12%?

So the power limit ceiling raises. Typically when you overclock, you raise voltage and therefore use more power. You dont want the power limit getting in the way.

Any ideas why AMD put a lower clock speed with 15% more current? Could it be sample variation i.e. TSMC can't get smaller variation in chip quality?

I suspect you're correct. I dont imagine they have the time to vet every single die, but understand that they'll all work/be stable with w/e voltage. It's been that way as long as i can rememeber and why some overclock higher or lower at a given voltage. Same would go for undervolting and overclocking... each sample is different.

Did some undervolting/overclocking today Time Spy Extreme 4K: -

That's great... but... where'd you start, lol??

Where'd you get the 'average' score for that card?

 

[Haider]

So the power limit ceiling raises. Typically when you overclock, you raise voltage and therefore use more power. You dont want the power limit getting in the way.

I suspect you're correct. I dont imagine they have the time to vet every single die, but understand that they'll all work/be stable with w/e voltage. It's been that way as long as i can rememeber and why some overclock higher or lower at a given voltage. Same would go for undervolting and overclocking... each sample is different.



That's great... but... where'd you start, lol??

Where'd you get the 'average' score for that card?

I started at stock voltage 1175mV, boost clock is 2250MHz. MSI set it to 2340MHz. I set it to 2700MHz and have been lowering the voltage. I also upped the memory clock to 2150MHz from 2100MHz in MSI Afterburner. In the RivaTuner overlay I can see 2138MHz clock for memory. I never hit 2700MHz core clock. I have read that it best to set the base clock to 100MHz less than the boost clock you're aiming for The average is the average of all 3D Mark users with an RX 6900 XT though some may be running a custom water-cooled solutions. I remember vaguely so not a 100% sure but in previous versions of 3D Mark I think they they would split the results up for air-cooled and water-cooled setups...

Generally the 3rd party OC cards tend to have better chips, I think the AIBs cherry pick those for their OC cards...

 

[EarthDog]

Generally the 3rd party OC cards tend to have better chips, I think the AIBs cherry pick those for their OC cards...

I dont think they have much choice in the matter...but do bin what they get.

The average is the average of all 3D Mark users

If it's from 3dm, then it's across overclocked and stock so keep that in mind. I dont recall them splitting it up between cooling though.

 

[Haider]

Techpowerup managed to overclock the MSI Trio X RX 6900 XT to 2760MHz : -

 

[EarthDog]

I started at

I meant score, btw... you listed the voltages already.

 

[Haider]

So this where I started: -


And this is where I ended: -

10fps more for 0.175V less, considering I paid £630 new when the 7900XT/XTX were launched, it's done nicely. I reckon I've still got a couple of years for the 1440P/Quest 3 gaming...I told a friend to get one or a 6950 XT and he bought a 3070Ti from Facebook marketplace...