NVIDIA PCI-Express Resizable BAR Performance Test - 22 Games, 3 Resolutions, RTX 3000 series.

[chrcoluk]

Yeah the amd vs nvidia graphs are an eye opener, clearly amd did something else aside from just enable the pcie feature.

I am probably toggling CSM again on my next reboot, for my sharper bios instead of 0.5% performance boost.

 

[LtMatt]

There's a reason for that.

Beaten me to it, but glad someone finally did some testing and put this one to bed.

This is why it's good to have SAM globally on, because it provides gains in most games, and the amount of games where it hurts performance is low.

Now, if only some reviewers would start doing their graphics comparisons with it enabled for a change.

 

[chroniclard]

Looks to me AMD is winning on this tech at the moment, I have disabled rebar for now.

 

[gpuerrilla]

Not even updated the BIOS' yet which is why you need this type of testing to see if its worth all the hassle. OK I will get it done as its not too onerous a task but it has to yield improvements which so far seem to indicate its not worth rushing into just yet.

 

[Nexus18]

Looks to me AMD is winning on this tech at the moment, I have disabled rebar for now.

I would give it a go yourself tbh, seems like people are getting mixed results dependent on hardware combination.

Technically my 2600 and b450 tomahawk wasn't suppose to support it according to nvidia....... but it did and it worked beautifully, most noticeably on cyberpunk

Have kept it on with my 5600x too and no problems so far.

I think there are some setups like mine, where you "need" to also turn on the 4g memory decoding setting to get the best performance boost, didn't watch all of the above so no idea if hardware unboxed covered that?

 

[chroniclard]

I would give it a go yourself tbh, seems like people are getting mixed results dependent on hardware combination.

Technically my 2600 and b450 tomahawk wasn't suppose to support it according to nvidia....... but it did and it worked beautifully, most noticeably on cyberpunk

Have kept it on with my 5600x too and no problems so far.

I think there are some setups like mine, where you "need" to also turn on the 4g memory decoding setting to get the best performance boost, didn't watch all of the above so no idea if hardware unboxed covered that?

Had it on for a while but not really noticed as the 3080 at 1440p is plenty of grunt in everything I play. At some point I will do some off/on comparisons to actually check but for now I have disabled it.

 

[ltron]

Yeah the amd vs nvidia graphs are an eye opener, clearly amd did something else aside from just enable the pcie feature.

I am probably toggling CSM again on my next reboot, for my sharper bios instead of 0.5% performance boost.

Yes, I am very disappointed by these performance regressions particularly in the games that are not on the whitelist (it's supposed to be disabled, how is it lowering performance if it's disabled?). Nvidia's whitelist was supposed to avoid such problems. At least they saved me the hassle of updating my motherboard's BIOS (I already updated the 3080's VBIOS) and getting my RAM stable again.

 

[ltron]

Hardware Unboxed have been doing some sterling work recently, firstly on the DX12/Vulkan driver overhead issue and now on ReBar vs SAM. For the most part the rest of the tech press have been pretty silent on both issues, I wonder why.

 

[scubes]

I have to ask was it worth it?

 

[Vertigo]

So both mobo BIOS and GPU VBIOS upgrades required.

I come from the old-school of thought that you only update a BIOS if there's a good reason.

This doesn't qualify.

 

[Yaayuh!]

I've got to the point where I really can't be bothered to update my BIOS and punch in my configuration again and hope it sticks the same, just for 3%

 

[J1nxy]

Interesting thing to see this today, I ran the F1 2020 benchmark around Monaco last night in the wet with the optimal setting as defined by the Nvidia optimiser, and average about 83 fps, same test after upgrading the motherboard bios and 3090 bios this morning and see an average of 102 fps. Thats a pretty hefty bump in FPS, it was just coincidence that I ran the benchmark so just re-ran with exactly the same settings apart from the BAR setting and bios updates.

 

[ltron]

Interesting thing to see this today, I ran the F1 2020 benchmark around Monaco last night in the wet with the optimal setting as defined by the Nvidia optimiser, and average about 83 fps, same test after upgrading the motherboard bios and 3090 bios this morning and see an average of 102 fps. Thats a pretty hefty bump in FPS, it was just coincidence that I ran the benchmark so just re-ran with exactly the same settings apart from the BAR setting and bios updates.

Can you run the same test with ReBar disabled in the BIOS? That seems a much bigger increase than expected and I wonder whether it was the BIOS itself that fixed some problem.

 

[gbgt2]

So both mobo BIOS and GPU VBIOS upgrades required.

I come from the old-school of thought that you only update a BIOS if there's a good reason.

This doesn't qualify.

Same thoughts here. My system has been absolutely rock solid since I rebuilt it in December. I really can't be bothered dealing with issues for such a minimal gain

 

[J1nxy]

Can you run the same test with ReBar disabled in the BIOS? That seems a much bigger increase than expected and I wonder whether it was the BIOS itself that fixed some problem.

Looks like the BIOS changes have increased the none ReBAR FPS by about 10% and ReBar does the same so another 10%

 

[ltron]

Looks like the BIOS changes have increased the none ReBAR FPS by about 10% and ReBar does the same so another 10%

Interesting, thanks.