RTX 30 Series Resizable BAR Support

Fingers crossed they bring this to 2000 series cards

 

Not happening. It's pretty much confirmed to be 3000 series only

 

This makes me sad, guess I'll have to sell a kidney to afford a 3080 then

 

Just to get the 2% improvement in 5 games from RBAR? lol

Can't imagine it ever coming to 2000 series, that's a fucking lot of VBIOS files manufacturers would have to go back and do for thousands of SKUs

 

Well, they do get the 2%+ from rebar and the performance the 3080 already has over the 2080ti for that kidney, worth it

 

If it nets me 165fps in destiny 2 at 1440p then sure it's definitely worth it

 

Interesting seems this feature causes performance regression in quite a number of cases, however on average 2% gains so still worth it.

 

Thank God, Resizeable Bar for CS:GO is disabled by default for Nvidia. You can enabled ReBar for unsupported games with NV Inspector if you want, but beware that you could have regression like shown in Linus's video. I'll just trust Nvidia to enable ReBar for the mean time, unless someone has down definitive test out there testing Rebar ON/OFF, before manually enabled it for that title.

 

Also confusing that it wasn't necessarily consistent across CPUs either. Like in some games the 5900X would do worse than the 5600X when running NVidia, but better when Radeon, but the reverse might be true in other games. And then that would sometimes also flip with a resolution increase.

I did my own testing (3 runs averaged) with RDR2 at 4K and just like TechpowerUp I found regression with reBAR on.

Might try F1 2020 and see if the mins improve.

 

Just a heads up, be careful with tests with RDR2 in-built benchmark, as for me, it's a little inconclusive, but the minimum FPS is lower with ReBAR, the it seems the 1% lows are better. There's something with ReBar, possibly at the start of a run that causes minimum recorded FPS, esp with how RDR2's benchmark records it, when in practice it's better across the board.

It may still overall be a regression, but worth checking whether the 1% or say 0.1% lows are actually lower or not, as the "minimum" reading is not reliable.

 

Yeah I concur, the minimums sometimes are 2fps, but the run was perfectly smooth so I've been ignoring the mins for RDR2.
Using a 3700X here, but about to switch over to a 5800X, so will be interesting to compare.

Average fps (average of all five passes per benchmark run) across three runs for me at 4k HDR (tweaked graphics settings):

No reBar:
82.639783
83.345056
82.581852

reBar:
81.562352
81.841075
81.612665

While the end results looks more consistent with reBAR, there is generally 1-2 fps swing between runs in the individual tests (passes) with or without BAR, so I wouldn't read anything into that.
When comparing the frametimes across all percentiles the runs things are really really close, but the runs without reBAR still prove ever so slightly better.

We're in the realm of 1% variance here, so that might be as simple as a boost clock dipped slightly between tests, it's so close I'm not even sure reBAR is working, despite checking in 4 places that it's 'enabled'.

 

Drinking game for the weekend. Watch this and take a swig everytime you hear "Resizable BAR".