-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Radeon Resizable Bar Benchmark, AMD & Intel Platform Performance
- Thread starter Purgatory
- Start date
Caporegime
Not bad. It would be nice to have seen more games tested, especially as some of the ones they tested are known to provide little to no benefit and there are a lot of games that do benefit from SAM as discovered through my own testing.
However, they did at least include a couple of games which show good gains, for example up to 29% performance increase with a 6800 XT and a 5950X in Forza 5 when using SAM.
No negative to using SAM (as shown in this video) when using a 5950X (the HUB test rig), so I hope they decide to stop disabling it in future videos.
They should have also tested using a 6900 XT as well, since the 6800 XT is not the fastest GPU available to pair with the 5950X.
Still, a step in the right direction at least.
However, they did at least include a couple of games which show good gains, for example up to 29% performance increase with a 6800 XT and a 5950X in Forza 5 when using SAM.
No negative to using SAM (as shown in this video) when using a 5950X (the HUB test rig), so I hope they decide to stop disabling it in future videos.
They should have also tested using a 6900 XT as well, since the 6800 XT is not the fastest GPU available to pair with the 5950X.
Still, a step in the right direction at least.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 29,925
I think it's probably just launched as a PR move to be seen to match AMD with this feature and not really been worked on since.
That was kind of his point of this video... as he explained in the summary, in some scenarios it drops performance and in a few cases, there is no benefit and in a few cases it does improve perf. and it all comes entirely down to the setup and game i.e. it's not "consistent" enough to be included in benchmarks where they want to keep testing as uniform as possible without having to test the "unknowns" (at least for the time being), not to mention, you can see from their POV why it is a right PITA to test.
You can see why nvidia have a white list for this now.
This comment kind of summed it up well:
Kind of backs up my thoughts too, essentially vram amount has little to no play on how much of a benefit SAM/resize bar can provide, despite some saying, SAM/rebar benefits more with higher vram amount cards....
Caporegime
What their limited testing shows, is that there is no logical reason to disable SAM on their test setup (it's enabled automatically in the driver - they have to go out of their way to turn it off) when using an RDNA2 GPU and a 5950X. There is either large performance increases, little to no benefit. Overall, there is a benefit. There is no negative performance overall to using it.
They don't test games and graphics cards using previous generation AMD (Ryzen 3000 series) and Intel CPUs, where the gains are more hit and miss, and sometimes performance regresses (as shown in their video) depending on the GPU, resolution and game used. This would be a reason to keep it off if they did all their game/hardware benchmark videos using old hardware.
As you pointed out, Nvidia have a whitelist, so this add more weight that there is no reason to disable it.
Last edited:
IMO until more testing is done we don't know that, clearly he states that the 6800 (with more vram than the 6600) can benefit more from SAM than the 6600, which in turn benefits more than the 6500xt. So if vram makes no difference why does the 6500xt show little difference and the 6800 have more benefit?
https://youtu.be/FM-mDf0U38k?t=1059
A test of Far Cry 6 at 4k between the 6800 and 3070 with the high res texture pack and SAM enabled might help show the benefits of vram going forward. Although we already know that the 3070 runs out of vram at that rez with ray tracing enabled, so AMD including more vram in mid mid to high end cards already helps on some occasions
Last edited:
I can't. You should have watched until the end, because he also did testing on a 3080 and results were all over the place, with some notable performance regressions.
Seems like Nvidia's whitelist isn't worth the paper it's printed on if games present on it are losing 10% performance.
Caporegime
If only they would have tested a full suite of games like that on the 5950X and 6800 XT setup, some of those games see very big gains with SAM.
wow this doesn't look good
Not disputing that first point but how many people have the exact same combo as HU? Not many...
I think this approach is better, keep SAM/rebar for separate videos as proved by their video, it is a faff to see what works well and what doesn't and have the standard testing as the "base", which might be more relatable to the "masses".
True more testing does need to be done. As for why does the 6500xt show little difference compared to 6800xt, no idea, maybe difference in the hardware other than just vram? You can see in the video, the increase of rebar on isn't much different between the 6600 and 6800xt, in some cases, the 6600 even seen better increase in performance with SAM enabled than the 6800xt and that was in FC 6 too...
They tested FC 6 in the video and there was no benefit with SAM.
I can't. You should have watched until the end, because he also did testing on a 3080 and results were all over the place, with some notable performance regressions.
Seems like Nvidia's whitelist isn't worth the paper it's printed on if games present on it are losing 10% performance.
This is quite old now but I don't think it has changed much since as iirc, nvidia did say they wouldn't enable it "officially" for older games.
The only one on your list which nvidia "did" officially enable rebar was for watch dogs legion, however, not long after release, they disabled due to the regression.
AMD also have shown results with "notable" performance regressions too, again, it's not clear cut on either side:
But as I have said many times, given amd have full control over chipset/bios drivers, I expect them to see much better gains given nvidia have no control over the chipset/bios drivers.
Associate
@LtMatt. You reckon you don't work for AMD anymore and have a 3090. So you could benchmark both with SAM enabled and disabled. Maybe do ray tracing too LOL. My BIOS is quite old so can't enable it anyway and my system is working fine, so not gonna update Bios which might cause issues.
There is a significant difference between a 6800 series down to the 6500xt, which might also account for those occasions when the 6600 has a bigger increase compared to the 6800. That is the amount of infinity cache. If the infinity cache is having a big hit rate on the 6800 compared to the 6600 then there will be less need to call on vram on the 6800 series (and so less use of SAM) as the required data is in the faster infinity cache.
Apparently no amount of improvement in SAM will help the 3080 in Far Cry 6, 4k native, high res, ray tracing. Think that win goes to the 6800xt, with or without SAM, https://youtu.be/mMIue2mGKNI?t=22
Caporegime
I sold the 3090 on the MM, but it doesn't support SAM, so I had to settle for using resizable bar.
I did enable it (ReBar) in Far Cry 6 and Dying Light (forced via Nvidia inspector) and it made no difference that I could detect in frame rates.
What possible difference could that make? It's either available to the graphics card/driver or not, Irrelevant of OEM of the card.
Unless both Intel/AMD restrict Nvidia GPUs throughput in some way. In which I'm sure they would be caught out fairly sharpish if they did .
It's called optimisation.
VRAM, CPU, GPU, RAM, storage drive etc. all work together hence why driver updates of any kind, be it GPU, chipset etc. or even just a game patch can provide improvement to performance. I expect there is a lot more to it than just simply adding "resize_bar=1" or "resize_bar=0" to turn it on and off...
This image demonstrates how they interlink quite well hence again why being able to have access/control to both ends could allow one side i.e. an all AMD powered system see more of a benefit:
PCI interface and it's drivers also plays a part in it too, some have said PCI 4 sees more of a benefit than PCI 3....
I suspect a large part of SAM/resizable bar seeing more of a benefit for AMD than for nvidia also comes down to the games being optimised for consoles.
It is certainly something that I would like to see one of the companies provide more of an in depth look at especially since that recent HUB video on it, why is it some games see no difference at all, some see decrease in perf. and some see a huge gain regardless of it being an intel, amd or nvidia system.....
Last edited:
Performance is negative in some games at some resolutions with some CPUs, including AMD CPUs and performance can be both up and down in the same game depending on your hardware configuration and game graphics settings
that's a big red flag, IMO sam/resize bar is another example of a complete mess of a feature - you may be happy, but the average user can't be expected to do their own benchmark and jumping in and out oft be bios trying to figure out if they should keep it on or off for every game they may play, that's way too much onus on the end user and it's unfriendly, this is why people go to consoles
Last edited:
Soldato
Caporegime
With older CPUs like Ryzen 3000, sure. With Intel CPUs? Maybe. No negative with Ryzen 5000/RDNA2 though, as I mentioned in my post.
I'll play devils advocate though for you. Let's image there is one or two games out there where FPS are 1-2 lower with SAM on when using Ryzen 5000/RDNA2, what about all the other games that are up to 29% faster (Forza 5) and the multiple games with 20% improvements, ala Forza 4, Halo, Horizon Zero Dawn, Valhalla etc?
You don't need to do any homework with Ryzen 5000 and RDNA2, don't worry.
If you recall, support for Ryzen 3000 was added much later by motherboard manufacturers with updated BIOS.I'll say it again, there is no logical reason to disable SAM with Ryzen 5000 and RDNA2. If you are using this configuration and can prove me wrong, please provide some clear examples, backed up with data Grim. I'll be waiting.
Caporegime
Oh god I’ve just seen that, only high settings too and that 3070 is choking.
I wonder why HUB don’t use the ultra preset for Far Cry 5 but they do for other games.
@tommybhoy