Discussion in 'Videocards - AMD Radeon Drivers Section' started by mtrai, Sep 30, 2019.
But... But.. But!! Whql
I told You
Give them week or two
It would be nice for Fury owners to get ris next. They are pretty much forgoten at this point.
i hope RIS will be available on RX460
470 and up for Polaris cards.
Similar DX11 performance to 19.9.2 WHQL (~27.7k FS), a notch better in DX12.
This is Beta (for now).
Tested RIS in F.Horizon 4 -> working like a charm, no FPS drop.
DX11 w/Reshade BFV & BF1 = Ok
DX11 w/Reshade GR WL = Ok
DX11 w/Reshade AC:O = Ok
DX9/10 w/Reshade HoMM VII = Ok
MtG Arena = Ok
All what matters to ATI Drv. devs is how many ppl have it.
AFAIK Polaris has the largest Fan base, next is Vega uArch then the rest follows.
Thats why Polaris & Vega will be around (Drv. & features wise) for some time.....
BTW Vega has RIS Via Rapid packed Math, similar feature is present (in different form) in FC5.
hey, i noticed my gpu clock has gone down to 1984 mhz from default 2029 mhz on my sapphire radeon rx 5700 xt nitro+ with driver v19.9.3. i did a clean driver install after using DDU. is anyone having the same issue? i didn't change anything except for switching to the new driver.
And? You lost 1 fps in games?
Hoh, I flashed another bios just to squeeze 1fps more. Its more important than you think (in a crazy mind) haha.
Having that minimum voltage below 750mV can cause stutters when GPU load is low, I would raise that to 750mV at least.
On first opening of Wattman and if the settings files are removed such as during driver cleanup it does a sort of optimization which I assume AMD has adjusted or are in the process of adjusting a bit so it might be lowering things down a bit despite the bios defaults so it could be due to that but if desired it should be easy enough to dial it back up again if the card can hit that target speed.
Nitro+ would have an additional fan and a bit of a different cooling block layout including a larger memory cooler resulting in a further temperature drop over the Pulse though both also adjust some of the stock components from AMD's default reference board like VRM's but almost every single non-reference 5700 or 5700XT hits the same performance level anyway so it's mainly down to temperature of the GPU and other components unless going with a water block from the stock blower fan and reference model.
Can actually see it changing now though then if that's what this is with changes in how this card uses a more fluid curvature instead of set performance states and the newer OverDrive v8 API being utilized and so it's visible directly in Wattman itself instead of as with Vega using monitoring software and finding out that the settings might not be respected at all or using different values from driver to driver, most of which is resolved at this point AFAIK other than how the fan curve on Vega is very steep or actually more like points or stepping rather than more granular so tuning that in can be a bit more work.
Curve here goes up to 2.1 somewhere Ghz and 1.260 somewhere for voltage with the defaults from 19.9.2 landing at 2.1 and 1.2v which going by reviews and tests the GPU's mostly all end up at a range around 1.9 to 2.0 Ghz anyway thus that's what I lowered it down to and then started lowering voltage by 25 and after landing at 1.05v that felt like a good initial value though there's room to fine tune both clock speeds and voltage once the drivers get a bit more mature.
Cursor corruption was linked to memory overclocking other than the little loading circle which is a driver thing although it seems not everyone gets it so not entirely sure what is causing it yet.
Interesting experience with how games and such are stable other than performance actually lowering down but the desktop and idle hits some issue but it's good to have that confirmed as well thus 880Mhz instead of 900Mhz for now though it's not much of a increase over stock but then it's also not much of a decrease over the previous and performance isn't really affected under normal conditions so I could probably leave it at default ha ha.
(Would like to have a lower setting as the minimum though to try and check how it affects performance if dropped to 750 - 800 Mhz instead but 875 seems to be the default and it's also still overclocked from the default specs then, curious.)
EDIT: Ah and so that then must be where the falloff point for voltage is far as some of this stuttering issue goes, stock is I believe 650mv or 625mv on the idle speed and lowest point of that graph.
(And then game mode and boost mode respectively I assume although low, mid and high clock states also works I guess as the card increases and decreases depending on overall load and other factors.)
nice uncalled for implications right there, bud. xD of course i can just set the clock and voltage to whatever works best for me.
i was just wondering why wattman randomly decides to change my gpu bios' settings all of a sudden with this new driver.
edit: grammar ._.
I'll check with my Pulse later today and report back.
first of all thanks for your reply in general. regarding the gddr clocks: maybe you can try the "more power tool" if you feel like giving a soft powerplay table a shot to reach those lower clocks in the highest state. i guess you've heard of that tool?
the graph is exactly what watmann set up like when i went into the custom mode. nevertheless my voltage never goes below the 750mV level. i think everthing below the lower and above the upper dotted lines gets ignored by default. at least voltage-wise. not sure about clocks. haven't tried that.
edit: actually what i said is not quite true. i just checked and my card idles at 725mV. =)
Yeah I've seen that power play edits can now make a good impact on the Navi GPU's so that is something I have planned to give a test as well if it can work not just to push power limits and other restrictions higher but also if it's actually possible to lower them down to find out what the impact is from that in terms of both performance and stability and in a way also power draw but memory isn't going to be too affected from this and it's more of a curiosity for where these start really drop and if the stock overclock AMD put really was worth it.
The GPU itself outside of benchmarks goes up and down maybe 6% or so if even that from trying to really push the settings if it can maintain a 2.0 Ghz boost or higher and then changes for DX9 to DX11 and then DX12 and VLK being more demanding and possibly needing other lower values or other changes to maintain stability. It's a good undervolter but in terms of general performance it's not exactly a overclocker but then the stock settings once again pushes the card pretty far as is so there's not too much headroom although there is some as the current soft power play tables and changes have demonstrated already.
EDIT: Also in general the actual performance gain will fluctuate a bit from the dynamics of how boost is working but while the hardware has changed a bit it does function similar to what one could get out of Vega.
Will be interesting to see what the higher end Navi models will be able to achieve too whenever those come out, might be able to push things a bit farther too unless they are also close the realistic limits of what can be gotten out of adjusting voltages or clock speeds further.
hey, what's the matter with that invalid value? does it render chill non-functional on some hardware setups or anything? i checked my registry and indeed i had the invalid value, but chill was working for me nonetheless (with an rx 5700 xt).
Keep it civil sparky.