Radeon "boost clock" discrepancy

[SimicEngineer]

I have the XFX RX 5700 XT RAW II (XFX[www.xfxforce.com], TechPowerUp DB[www.techpowerup.com], TechPowerUp BIOS[www.techpowerup.com]).

According to both XFX and TechPowerUp, it's rated for a 1755 MHz "game clock" and 1905 MHz "boost clock". However, I've seen it apparently boost to over 2000 MHz under heavy load, and the TechPowerUp BIOS entry has a boost clock of 2150 MHz.

Can anyone explain why there seem to be two different specifications for boost clock, and whether it's normal for the reported clock of a Navi GPU to ever exceed the manufacturer-advertised boost clock by default?

[_I_]

mfg factor oc
along with better cooling and power deliver to the gpu cores

[Omega]

Modern GPUs and CPUs have very inteligent boosting behavior, as long as the power and thermal budget are available it may boost above typical clock speeds.

[_I_]

also depends a bit on the silicon lottery
gpu mfg will bin them again after getting from amd/nvidia and gpu dies that can perform better will go to oc versions

[SimicEngineer]

This isn't advertised as an OC card, and advertised boost clocks of over 2000 MHz are pretty high-end for 5700 XT. Since XFX did have such a product (THICC III Ultra), it seems like I'd need to really win the silicon lottery to get a "core" product able to go that fast without any tweaking.

To give a little more background, the reason I'm asking is that I've seen an increase in GPU-related crashes over the past couple of months, and there are some indications that the Linux driver might have started improperly clocking some GPUs at some point (Gentoo wiki[wiki.gentoo.org], amdgpu bug[gitlab.freedesktop.org]). I'm trying to figure out whether this might actually be happening in my case, but most of the information about this is anecdotal/speculative and mostly focused on newer GPU generations, so right now I'm mostly just trying to understand what the numbers are supposed to mean.

[_I_]

they dont label it as 'oc' anymore, just a variant with better cooler

[Bad 💀 Motha]

It's boosting way past 1900Mhz and not having problems then why complain

[SimicEngineer]

It is having problems, I just don't know if they're related to the clocking. Getting sporadic "ring gfx_0.0.0 timeout", which can supposedly be caused by almost anything that makes the GPU stop running a shader task (shader compiler bug, driver bug, firmware bug, bad card, inadequate PSU, cosmic rays, etc.).

[PopinFRESH]

SimicEngineer eredeti hozzászólása:

It is having problems, I just don't know if they're related to the clocking. Getting sporadic "ring gfx_0.0.0 timeout", which can supposedly be caused by almost anything that makes the GPU stop running a shader task (shader compiler bug, driver bug, firmware bug, bad card, inadequate PSU, cosmic rays, etc.).

Not likely the boost clock unless you’ve manually applied an offset?

More likely a driver issue. Try wiping the driver with DDU and then doing a clean install of the latest driver package from the AMD direct download

[SimicEngineer]

So far I've only seen instability under Linux, not Windows. That might just be because I don't use Windows regularly anymore, though; I've run a few basic tests like Superposition and FurMark, but those don't necessarily cause problems under Linux either. It seems like it's more associated with switching between low and high load rather than the GPU simply being maxed out.

[Bad 💀 Motha]

Wipe GPU Drivers out using the latest DDU app; each and every time you wish to change GPU Driver. This alone can prevent many issues or put a stop to various issues when its a GPU issue most of the time.

Win10/11 do not have TDR configured correctly; by default MS has that set way too low. Use the TDR Manipulator app (from makers of DDU) in order to raise the timeouts for those TDR related values.