Vega 56 boost clocks at 4k fluctuating.

[LizardTa]

I need a bit of clarity on an issue I can't get my head round with my new Vega 56 card.

I have UV/OCd it to get a solid 1610Mhz boost clock when playing on my ultrawide 2560x1080 monitor but as soon as I transfer to 4k I get wild swings from the 1300 to 1520 range but get nowhere near the 1600Mhz mark.

It seems especially bad on Far Cry 5. Just wondering if anyone can shed any light on this and if there is a reason for it? I was thinking it could be a CPU bottle neck at 4k maybe. Running a 1700X at 3.9 but I would have thought it was enough.

If anyone has got any info I would be greatful.

Cheers.

 

[W1zzard]

Does that fluctuation go away with lower clocks / higher voltage / higher power limit / higher fan speed?

 

[PHaS3]

higher power limit

Like W1z says, that's probably the thing. 4K will bottleneck on the GPU, less likely the CPU. Also, Far Cry 5 uses AMD specific features like Rapid Packed Math, and is quite a graphically intensive game. The card is probably throttling on power limit or temperature.

The boost clock is just that, boost. It unfortunately isn't 100% guaranteed that the card will always hit the boost clock. Conditions have to be ideal and criteria met, like GPU / VRM / HBM temperatures, power draw, voltage, workload, etc.

 

[LizardTa]

No the fluctuating only occurs when I run the games on my 4k OLED. I run the exact same Volts/Fan/Clocks etc profile on my monitor and it stays over 1600Mhz for the most part with a few drops down to maybe 1590.

Transfer over to the OLED and its all over the show. GPU utilisation is consistent on both monitors as well as I thought it may be the card down clocking as utilisation fell but doesn't seem to be the case.

Was hoping someone else had come across it and knew the cause.

 

[PHaS3]

The difference is the load applied to the card running 3840x2160. Increase in resolution = increase in load, which will increase temperatures and power draw.

 

[LizardTa]

Yeah this is probably the case. I thought that if the cards were both running 100% utiliation irrespective of resolution then they should match the boost clock behaviour.

Thanks for the input.

 

[W1zzard]

I thought that if the cards were both running 100% utiliation irrespective of resolution then they should match the boost clock behaviour.

No longer the case with Vega/Pascal