Resizable Bar?

So nV is starting with a little handycap in the optimization department which should be a good reason to try and catch any low hanging fruits (flip in a switch if it proofs worthwhile) in order for benchmark scores to look a little better.

 

It would pretty much compensate for the performance loss of a HQ texture filtering default setting to match the quality of AMD's default setting since Vega.

 

I agree.

It has seemingly become even harder to enable it now on the user side because there are new keys related to that in the latest drivers. And it looks like FC6 has gotten some special treatment for rebar.

 

Do you know which keys? I enabled the same way I always have with halo. 3 keys in total need to bet set

 

My experience is the opposite. With Resizable BAR enabled, I see sizable improvements in performance in those five games.

 

No, they are still the same three keys: 0x000F00BA, 0x000F00BB and 0x000F00FF

 

I suspect that all of these are related to rebar in the latest drivers:

 

Looksl like In Hellblade: Senua's Sacrifice in DX12 mode game stutters rather heavily in at least two levels with ReBAR on: Bridge to Hel and one of the Odin's puzzle with masks. Can anyone else confirm?

 

its stutters for me with ReBAR on and off, doesn't seem to make any difference.

 

Well, it does stutter way less on those specific levels with ReBar off, so I was curious

 

Get a warning that R Bar is disabled in the motherboard bios but that is not true.
It was always running do not know why its broken now

 

Had the same issue and used some time troubleshooting, Astyanax had an idea that solved the issue.
A full reset of the motherboards BIOS, turn off the PC and use the Clear RTC Ram jumper (I suppose you still use that Asus Z390 A?).

I probably forgot to hit the BIOS reset button on my board with the last BIOS update and ended up with optimized defaults instead of a truly clean BIOS, which was enough to not enable Reziable BAR support for Windows even when it was enabled in BIOS.

 

No i am using the Asus Rog Strix Z590-E gaming
Mai by there is a button on the back ?

 

That board does come with a Clear CMOS button on the back and also a battery that is easy to lift up and set back into place.

 

Just an observation:

in my situation, it seems resizable bar gives minor (very minor) improvement in some games. In other games it introduces stutter.


However its still interesting and maybe peeks at insight for some other things.



The interesting part:
If I enable the resizable bar at a global level, AND turn shader cache OFF. The stutter that resizable bar introduced in non optimized games, seems to go away. The performance of theses games is NOT generally enhanced from this one action , but the introduced stutter from the Resizable Bar goes away. I realize that "that's not the way it works" applies.

The games that I am referring to are mostly mGPU but a few aren't:

mGPU
SOTTR
RTTR
Simper Elite 4
Zombie attack 4 (Game I'm currently playing)
Strange Brigade
red dead redemption 2 (game is flaky regardless)
Quake II RTX


Single GPU
Agony
Succubus
Control
Forgotten City
Necromunda
Chernobylite (Game I'm currently playing)
Metro Exodus enhanced edition
Resident Evil Village (Game I'm currently playing)


Side note:
Borderlands 3 has a driver default profile that enables Resizable Bar, this introduces stutter and hitching. Manually amending no resizable Bar in the profile of B3 resolved it. Resizable Bar destroys playability this game on my system.

Cyberpunk - has a driver default profile that enables Resizable Bar. I see no benefit and no detriment to resizable Bar under any condition

ROTTR- still has soviet installation Level still has issues of hitching since day one.

Metro Exodus enhanced edition - is the only game on the predefined 20 of the profile that showed any marginal improvement for the configuration change (about 1-3%) the rest are subjective and placebo. However it did seem to play "smoothly"

Strange Brigade was most improved player showing a 5-9% improvement (benchmarking is inconsistent with this title, that's why the wide range)

red dead redemption 2 - seemed to have 'less" issues but its frustrating non the less.


Other observations in having global resize bar on but Shader cache OFF: level loading times "seem shorter" tried timing it and results are subjective, but it "feels" faster. Then I thought about it, and this could be from just having shader cache off.

Games that I have injected enhancements into such as higher textures, update lighting or mods seemed unaffected, but then again these are much older games.

Stability on 466.77 standard drive was not an issue, but RDR2 was the only "unstable" title regardless.

Other observations in having global resize bar on but Shader cache OFF: level loading times "seem shorter" tried timing it and results are subjective, but it "feels" faster. Then I thought about it, and this could be from just having shader cache off.

All games played at 4k, NO DLSS, no vignetting, max eye candy, max shadows, no film noise, no motion blur, no dynamic resolution, and no DoF. Ray Tracing on max when supported. I have no page file (1.5TB ram), and I have the MS provided registry keys to maximize ram usage when page file is not enabled and the optimized feature mask settings. Running on Server 2022 Data Center as a workstation. When not in a benchmark, I like max eye candy, no DLSS, I play 4k, 10 bit, no HDR, G Synch, 120/144hz, Using 466.77 standard drivers. (metro exodus enhanced edition has severe issues on 472 and above), I do seem to prefer the 471.22 but it looses performance on mGPU titles that are DX12. 496/497dch introduced too many negative variables to the experience, I'm hoping nvidia resolves the trend..





Conclusion Resizable bar seems to benefit a few titles only -so far (in my situation). The negative impact on some titles stutter/ hitching/ longer load times seems resolved by turning shader cache off (I realize that it should not work that way). There is the placebo effect of faster load times. Launches seem faster (not placebo but that could be specifically from not using shader cache)

 

You've made an in-depth research, but i was enabled ReBar once on Z390 system (with matching BIOS) - stutterfest all the way in every game - disabled - marked ReBar as cr#p
Turned on once again on Z590 - less stutterfest, but also very uncomfortable and strange hitches and stutters here and there, disabled - marked ReBar as pile of cr#p, forgot about it lol

 

Watch Dogs Legion was the only game I had major stuttering issues with when using ReBar. That was more of a game bug though, if I remember correctly.

Otherwise, I haven't really been impressed or disappointed in ReBar. None of the other games whitelisted have any sort of noticeable bump in performance.

Assassins Creed Valhalla - one of the few games that is shown to have a decent benefit from ReBar, isn't doing much on my system. I've ran the benchmark multiple times with it on and off, my findings are only a reproducible 2 fps avg increase and 1% and 0.1% lows are slightly lower.

I might just turn it off, to rule it out as something that could cause issues with games.

 

If I forced resize bar on escape from tarkov would that red flag me with battlestate? I’d like to give it a try

 

So I have been running rebar since the first driver that supported it, Didn't really think I was having any issues - after this thread decided to disable it in my BIOS and give it a go, and can confirm it resolved a ton of stuttering issues, specifically in vulkan titles. much smoother now.

 

rebar, hags and even hvci enabled, no issues. they were enabled by default on a clean 11 install