Discussion in 'Videocards - AMD Radeon' started by OnnA, Jun 11, 2019.
Another one (every ATI user is aware of this -> UV/OC)
120tW to 160tW for 28k 3dmark !
How good is RIS (CAS) in practice?
Spoiler: Image Quality, easy win for ATI
Looks like the texture quality is in high vs low.
Got to try RIS with BFV and The Division 2 in DX12, in BFV it looks not so good but The Division 2 made a big difference and looked incredible.
I was struggling passed few days with DXGI/DirectX fatal errors, and a flood of WHEA-Logger warnings in Event Viewer: http://prntscr.com/oexxs0
I bypassed the DXGI/DirectX errors by running my games in borderless window, as soon as i go fullscreen the games crash with this DX fatal error.
First i got these blackscreen/reboots due to RAM OC and now this.
Yea it really isn't loud if you set a fancurve and undervolt, my RX 5700 was very quiet out of the box but it got pretty hot so i set a slightly aggressive fan curve and undervolted and now it's not going over 67c.
But i can hear the fan a bit more clearer though, but it's not annoying like other heatsinks.
Guys, anyone here having RX5700 (non XT) with power table mod with decent cooler ?
How far could it be pushed . Guys, i have been surfing all over the internet and i am too inefficient to find a solid answer .
Is 2200mhz ( steady ) possible with this little champ?
I just found littletimyjoe's video with 1900mhz oc using power table mod and visiontech cooler slapped on it. But i feel like this gpu has more day to day ocerclock headroom left .
2100 is what I've seen through undervolting and having a sample good enough for boosting past 1900 - 2000 Mhz without throttling or requiring more GPU core power.
AMD might change Wattman parameters and characteristics through later driver releases, happened multiple times for Vega (19.6.1 seems to be another driver where it happened.) so if you are pushing the card that close to the limits remember to test and also re-test thoroughly as changes can make for instability.
EDIT: Ah figures, 2100 was the target speed and the actual would be hovering around 1900 - 2000 re-reading the reports, looks like it's doing pretty nicely but dropping below 1900 also sees a gradual performance dip though it is possible to go down to 1100mv or 1150mv perhaps even more instead of the default 1200mv and get a nice temperature reduction and less throttling from hitting the threshold though depending on GPU the max sustainable core clock speeds will differ a fair bit.
I assume AMD has a pretty wide range of chips again so they all pass at 1200mv but some can do with less and others can do with much less while retaining close to default speeds and thus boosting to more stable values instead of fluctuating.
Need to read up on what else Overdrive 8.0 could have changed if anything is available, driver situation is also a thing so who knows what it's going to be in a few months and what AMD might silently change.
I'm on Vega hardware so it's from what I am reading though for Vega the latest change seems to be more conservative on voltage trying to keep 20 - 50mv under the set amount thus 19.6.1 and newer leading to some stability reports with the card hitting too low values for profiles build on previous driver testing.
(And of course all of these changes are silent and not part of the release notes at all.)
Will be interesting to see continued testing with Navi here and what the 5700 and 5700XT will hit in a few months, tessellation tests also seem a bit under expected so possibly another driver shortcoming and something that will see a higher performance bump if fixed assuming it's a software issue in the first place.
(Or just go down to 32x in Catalyst Next instead of 64x since the result is still close to Vega 56 levels so it's not like Fury and earlier where it could be a bigger bottleneck.)
EDIT: Oh and the Signature edition cards or how they were called seem to be a bit higher clocked but retain similar voltage settings so they might be even tighter with limitations to undervolting and stability as a result unless manually clocked down a bit from what I am reading.
Basically the variant that looks like Swedish Postal service or local variant of mail delivery service that has deliveries down to a smashing success (Literally!) were the ones that handled shipping with some scribbled on signature from Lisa on it.
(UPS or what have you but apparently in this case it's part of the design.)
I got mine to 1800mhz on stock voltage with stock cooler. I'm actually very skeptical about Wattman's overlay as it reads only 65-68c on full load. On stock settings it hits 81c pretty quickly, but the fan speed is low.
But still such low temps with only a aggressive fan curve?
It seems stable so far, played 3 full rounds of conquest in BF4 and a little FFXIV without issues.
GPU Power increased from 110W to 131W but again with everything stock it maxed out at 153W?
UV to 975mV or 993mV
Well my RX 5700 XT arrived today...was delayed due to weather during shipping.
975mv seems to be stable.
Cool & Quiet... at 135tW
You've got Performance of 1700MHz Vega (165-200tW depends on chip & UV).
RX 5700 XT Navi Crosses 2.2 GHz Thanks to Custom SoftPowerPlay Table Registry-Mod
Igor Wallossek of Igor'sLAB Germany postulated a method by which an AMD Radeon RX 5700 XT "Navi" graphics card can be made to run at clock-speeds of over 2.20 GHz (engine clock), thanks to custom SoftPowerPlay Tables (SPPTs) deployed by modifications to the Windows Registry.
The AMD Radeon driver is designed such that it reads PowerPlay tables from the video-BIOS of an RX 5700-series graphics card the first time it's detected, and writes it onto the Windows Registry for quick-reference.
This is called a SoftPowerPlay Table or SPPT. It's the modification of SPPTs that allows you to manipulate the power limits of RX 5700-series graphics cards, and achieve higher engine clocks than the 2150 MHz engine-clock limit of the RX 5700 XT, which is set at just 1850 MHz for the RX 5700.
Wallossek's mod involves preparing your Windows Registry with a driver cleaner such as DDU, downloading and applying Registry files for various new power-limit targets you want.
The table below details the various power-limit and clock headroom on offer from each kind of registry file. There's also a registry file that cleans up your Windows Registry of any SPPTs, if you decide to roll-back your mod.
You can inspect a registry file by opening it in a plaintext viewer such as Notepad. Find links to the SPPT mods, and the Registry Cleanup in the source link below. You can also watch a video presentation by Wallossek in German language here.
You make any changes to your machine at your own risk, be sure to have proper custom cooling for your graphics card.
I'm already messing around with it, don't do this with the blower style heatsink lol it shoots up to 85c in minutes. I settled at 1850-1890mhz 1050mv, temps is stable now at 73-74c but i had to adjust my fan curve so it's pretty loud now lol.
Might be able to go higher without temps shooting over 80c.
Did you replace that graphite thermal pad for regular TIM?
Tomorrow I am replacing the stock blower with a
ARCTIC Accelero Xtreme IV and repasting with graphene paste and graphene pads. I can already easily hit just above 2100 with the stock blower.
Easily doing 28.5 k graphics..and well lets just admit my Physics and Combined scores are rocking considering I am on a 2700x with an X470 Motherboard.
No haven't replaced anything on it. But i'm thinking about replacing the cooler also with the Accelero Xtreme IV.
I have seen a video of it, what scares me is the plate that touches the core bending slightly.
Pictures of it: http://prntscr.com/oge9zj
Its normal I have the same cooler. Its just the plate bends a bit not the cooler, taht plate will act as a spring.
Really good cooler, my rx480 never goes over 60c and thats with 1.2v atm
Mortheus is a it better and fits vega also. Its a hard choise, if it was me atm mortheus is the one to go.