Radeon Resizable Bar Benchmark, AMD & Intel Platform Performance

[tommybhoy]

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

3070 didn't even get an ultra+HD texture pack@1080p result never mind 1440p/4K, but got to be a problem with Hubs system eh?

Considering I saw a post someone payed (almost double)£865 for their 3070...

Steve:

It's quite clear now that Nvidia really messed up with the vram capacity of their initial Ampere gpus or at least they would have if the market didn't go absolutely bonkers

 

[gpuerrilla]

3070 didn't even get an ultra+HD texture pack@1080p result never mind 1440p/4K, but got to be a problem with Hubs system eh?

Considering I saw a post someone payed (almost double)£865 for their 3070...

This. Just so happens most of the reputable sites usually share much the same results/opinions i.e. techpowerup, oc3d, df, gamers nexus, Tom's hardware, guru 3d, hardware unboxed, kitguru, pcgh, DSOgaming and computerbase.

Certainly don't go by people who have no evidence to back up their research/claims
wink emote

 

[Nexus18]

I’m really only saying it won’t provide a negative, ie a -10% performance deficit. That’s why I maintain there’s no logical reason to disable it. It helps, or does nothing. It does not hinder. Not saying there will be huge gains s in those titles, I showed you that Apex example as proof that -10% is wrong, Apex sees no gain but no deficit either now.

Again not disputing any of that, hence why I am asking for proof from a 6800

None of this is a reason to ignore Resizable Bar, the fact that its "inconsistent across many games" is the reason HUB are giving to continue to ignore it, its a cynical and contrived excuse to exclude something that benefits AMD more than Nvidia. even his audience know it, they have been calling him out on it which is why he made two polls on it and despite over 70% say yes to Resizable Bar he still wont do it.

Its free performance, sometimes as much as a whole step up in GPU tier.

That post is nothing to do with saying why it should be disabled

On the topic of hub and their reasoning... Again, rewatch the video, their reasoning is pretty valid.

Also, it seems like they are now enabling it anyway i.e. see 3070 Vs 6700xt video, Sam/rebar is turned on throughout.

It's still either on or off

Yes there will be driver optimizations but there is know way in which Intel/AMD can make it better on their respective platform and hinder Nvidia/competitor without serious repercussion. Apart from that they would just shoot themselves in the foot doing it.

Not saying that intel or and are hindering it for Nvidia, just saying that when you have access to both "ends" in terms of drivers etc. A lot more can be done to make it work better on an all amd system.

Each vendor is responsible for their own optimisations. No one is going to do any additional work for another vendor.

Exactly and Nvidia have no access or rather not as much of an insight/control to the end to do with cpu, mb/chipset drivers/bios.

3070 didn't even get an ultra+HD texture pack@1080p result never mind 1440p/4K, but got to be a problem with Hubs system eh?

Considering I saw a post someone payed (almost double)£865 for their 3070...

Steve:

In terms of fc 6, iirc, it was only the 3080 I was questioning your experience with since I also have a 3080 and haven't seen the same issues you were posting about.

See hubs 3080 Vs 6800xt video as per link on previous thread where they do ultra settings and don't seem to have the single digit fps you experience. System issue?

Also not sure why you are quoting price as if to try and make a point of how poor the 3070/8gb vram value is? If someone chooses to overpay for "any" GPU, that's their fault, not the manufacturer..... Ps. Watch the full video for the 3070 Vs 6700xt, despite the 6700xt having more vram, the 3070 comes out 19% faster on average across 50 games at 4k and that's with no RT or/and only using RT in rt light titles

Guessing you can't read very well either.... "sites usually share much the same results/opinions", keyword being "usually" I.e. as in, not always the same....

And again, most of my posts in question to Tommy surrounding FC 6 was the 3080, not 3070 but nope, keep on being you

Edit:

I wonder if Steve left rebar off for FC 6 with the 3070 if he would have got the same fps issue at 4k given that is the only way I could get my 3080 to act the same (without fsr too), perhaps that's why the 3080 vs 6800xt didn't show any issues in FC 6 with ultra at 4k (as rebar was off for it iirc).. But I guess that requires too much thinking

 

[humbug]

@Nexus18

On the topic of hub and their reasoning... Again, rewatch the video, their reasoning is pretty valid.

Its not at all Valid, since when was free performance not valid? When it doesn't apply to Nvidia?

 

[LtMatt]

Don’t forget PCGH Nexus, they also had single digit fps, and computerbase and tech power up also noted bad frametimes, system issues too?

 

[gpuerrilla]

Waiting for his post on twitter calling Steve's out for fingers in ears from a professional debunker.

 

[Bill Turnip]

Going back the SAM point and whether it should be tested or not, I think it's quite important to show both results - If I was looking to buy a gpu, sure I'd like to know what it's full potential is, but I'd be more interested what it would be likely to do to MY system. If I have an AMD processor and looking to buy a 6000 series, then brilliant, I definitely want to know the SAM results. If I have an intel processor, I'd need to know what the results were without it. As a reviewer, if I reviewed with SAM enabled by default, and people (some who really only look at frame rate numbers to decide a purchase) bought the gpu, put it in an Intel system and didn't get the stated result... I can see a backlash to them as well. They could be written off as an AMD 'shill' unfairly.

The argument seems to be that Intel and nvidia have the greater mindshare (I'd agree, but probably less so these days) but getting an all amd system is very appealing now. But having to convert a whole system to get the desired results is more expensive than just a gpu... Which is pretty damn expensive on its own.

SAM definitely seems to be better than the nvidia version (I've seen no real effect overall with reBAR), but I don't want to to see SAM only results, as I think this could potentially mislead people into buying an amd gpu, and then being a little disappointed if they don't get the advertised, SAM fuelled benefits. This could then have a knock-on affect of the perception of an amd gpu, which would be hugely unfair.

It is important to show both sets of results for an informed purchasing (rather than showing the gpu in the best possible light) decision. I don't know about SAM, but reBAR is not a straightforward thing to apply if you don't know what or where you're looking for it.

 

[gpuerrilla]

Good points Bill.

Just to expand on your ethos there, AMD have released features that benefit any vendor such as people with a 1060 being able to use FSR (GPU agnostic). One of the considerations is that they tend to sway to open methods and can prolong the lifespan of a component like in earlier example. Intel and nVidia tend to do the closed shop route and then after time/pressure reluctantly open up. The reBAR you mention was nvidia not wanting to miss out on a selling point, even though they knew it would not benefit many games for them. Maybe performance left on the table but needed more time investment.

 

[Nexus18]

@Nexus18

Its not at all Valid, since when was free performance not valid? When it doesn't apply to Nvidia?

Again, did you watch the video? If so, you would see depending on the setup of said following things:

- game
- cpu
- cpu brand
- gpu
- res.

That's even considering other things that "could" impact what happens i.e. PCI gen 3 or 4? Not to mention the huge range of MB....

It's not always clear cut free performance.....

And again.... it's a moot point as in the 3070 vs 6700xt video, they have enabled rebar/SAM so you'll need to find a new reason as to why they're nvidia shills now.

Don’t forget PCGH Nexus, they also had single digit fps, and computerbase and tech power up also noted bad frametimes, system issues too?

Not sure if you're referring to the 3080 or 3070 but either way.....

As you just said your very self when it comes to "old data":

That’s old data from when the product launched and they tested it I believe.. Using old BIOS, GPU/Chipset drivers etc. Look at the list of games they tested, it’s from their launch testing of the feature.

Techpowerup mentioned this, "does stutter sometimes on the 10 GB GeForce RTX 3080".... no mention of fps drops to single digits? Bit of a difference.

Again, I never really "debated" tommys 3070 in FC 6 at 4k60 with no FSR or max ray tracing, mainly because I haven't followed that card nor have much interest in it and most importantly...... as I have said all along for the 3070.... a 3070 hasn't got enough grunt for 4k60 at "max" settings i.e. settings will have to be reduced or/and FSR/DLSS enabled, which is shown perfectly in that 3070 vs 6700xt HUB video i.e. even though the 3070 is 19% faster on average across 50 games at 4k (and that's not including heavy RT settings/titles either....), it's perf. is still not good enough, same way even though the 6700xt has 12GB vram, the grunt clearly isn't there..... What I did question with regards to his 3070 was his experience/issues he mentioned in other games like metro and rdr 2 though as, correct me if I'm wrong but where are those sites saying the same for those games?

Waiting for his post on twitter calling Steve's out for fingers in ears from a professional debunker.

zzzzzzz

Going back the SAM point and whether it should be tested or not, I think it's quite important to show both results - If I was looking to buy a gpu, sure I'd like to know what it's full potential is, but I'd be more interested what it would be likely to do to MY system. If I have an AMD processor and looking to buy a 6000 series, then brilliant, I definitely want to know the SAM results. If I have an intel processor, I'd need to know what the results were without it. As a reviewer, if I reviewed with SAM enabled by default, and people (some who really only look at frame rate numbers to decide a purchase) bought the gpu, put it in an Intel system and didn't get the stated result... I can see a backlash to them as well. They could be written off as an AMD 'shill' unfairly.

The argument seems to be that Intel and nvidia have the greater mindshare (I'd agree, but probably less so these days) but getting an all amd system is very appealing now. But having to convert a whole system to get the desired results is more expensive than just a gpu... Which is pretty damn expensive on its own.

SAM definitely seems to be better than the nvidia version (I've seen no real effect overall with reBAR), but I don't want to to see SAM only results, as I think this could potentially mislead people into buying an amd gpu, and then being a little disappointed if they don't get the advertised, SAM fuelled benefits. This could then have a knock-on affect of the perception of an amd gpu, which would be hugely unfair.

It is important to show both sets of results for an informed purchasing (rather than showing the gpu in the best possible light) decision. I don't know about SAM, but reBAR is not a straightforward thing to apply if you don't know what or where you're looking for it.

Exactly, well said. This is exactly what HUB are getting at It's an enigma and not everyone has a 5950x and 6800/6900xt.....

Hence why it is better to keep a separate video on SAM/rebar.

 

[Game]

it does seem a lot of faffing around for maybe a gain maybe a loss

 

[Bill Turnip]

Again, did you watch the video? If so, you would see depending on the setup of said following things:

Hence why it is better to keep a separate video on SAM/rebar.

No, not a separate video. I appreciate it might make videos a bit longer, but it is important to keep the information together. Any potential step up through SAM may indeed influence someone to change a whole system if it means they hit their desired targets. The more places you have to go for information the easier it is to lose.

Possibly keep the EXPLANATION for how SAM works on a different video, I get that, but the actual results should be side by side with the non-SAM results. Same with reBAR.

 

[Nexus18]

it does seem a lot of faffing around for maybe a gain maybe a loss

TBF, with amd, it's not a faff at all unless you have to update your bios and toggle the feature on. More of a faff on Nvidia.

No, not a separate video. I appreciate it might make videos a bit longer, but it is important to keep the information together. Any potential step up through SAM may indeed influence someone to change a whole system if it means they hit their desired targets. The more places you have to go for information the easier it is to lose.

Possibly keep the EXPLANATION for how SAM works on a different video, I get that, but the actual results should be side by side with the non-SAM results. Same with reBAR.

Well yeah same video ideally and just have a different section, this would work well in the case of just one game benchmark video but not for large scale game benching.

We already have plenty of sites who have rebar enabled all the time so it would be nice to have somewhere to see non rebar performance as most of the sites do you use a 5950x too, again very few people have these chips.....

Maybe Matt can get someone in amd to do an in-depth look at Sam/rebar and what's really required to get the best from it as would love to know why some games see no difference, is there something game developers need to do on their side?

 

[Bill Turnip]

Good points Bill.

Just to expand on your ethos there, AMD have released features that benefit any vendor such as people with a 1060 being able to use FSR (GPU agnostic). One of the considerations is that they tend to sway to open methods and can prolong the lifespan of a component like in earlier example. Intel and nVidia tend to do the closed shop route and then after time/pressure reluctantly open up. The reBAR you mention was nvidia not wanting to miss out on a selling point, even though they knew it would not benefit many games for them. Maybe performance left on the table but needed more time investment.

Very true. reBAR seemed quite a good idea, but in my real world case at least, nothing. I'm trying to convince myself it will be an added benefit for when direct storage and RTX IO become a thing, but I will wait and see. Making FSR available to all is a very good thing, and AMD are definitely on the front foot with that, but is it more effective with AMD than not? If the improvement is the same for both vendors, I see no need to test separately.

 

[varkanoid]

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

it does seem a lot of faffing around for maybe a gain maybe a loss

SAM is enabled or disabled in the AMD Driver Control Panel. Maybe Nvidia should follow suit ?

Hardly faffing around

 

[Bill Turnip]

SAM is enabled or disabled in the AMD Driver Control Panel. Maybe Nvidia should follow suit ?

Yes they should. They absolutely should. But it might not be quite that simple if you also have to have up to date motherboard drivers? Clutching here, I can't see any reason not to have it in the control panel.

 

[tommybhoy]

What's really interesting is VRAM usage. I measured well over 10 GB with the 4K, HD Texture pack, and ray tracing combo, which does stutter sometimes on the 10 GB GeForce RTX 3080, but runs perfectly fine with cards offering 12 GB VRAM or higher. I guess I was wrong when I said that the RTX 3080's 10 GB will suffice for the foreseeable future. That said, there's plenty of graphics options, and I wouldn't be surprised if a future patch addresses memory requirements.

Possible system error?

MY takeaway from the above quote is- tpu's w1zzard testing at launch showed then that MAX everything isn't happening for gpus with less than 12Gb evident in In tpu's performance charts with HD pack installed/max detail BUT RT'ing disabled.

 

[humbug]

I hate it when reviewers defend hardware vendors being tight.
I don't even understand what legitimate reason they could have for doing that, could anyone explain to me why i should be happy to settle for less?

 

[LtMatt]

Possible system error?

MY takeaway from the above quote is- tpu's w1zzard testing at launch showed then that MAX everything isn't happening for gpus with less than 12Gb evident in In tpu's performance charts with HD pack installed/max detail BUT RT'ing disabled.

Yup.

In other developments, it looks like this possible system error appears to be more widespread that first thought.

Note the computerbase chart regarding the 3080 and HD Textures.

Spoiler

PcGamesHardware found the same too.

Spoiler

 

[ASMB]

If SAM/re-sizable bar works best when it is properly implemented into a game by the game developer (like fsr being implemented into the game directly rather than using rsr at driver level) shouldn't we be encouraging game developers to properly implement SAM into a game so that we can all get free extra performance. After all I see many people call out a game for being poorly optimised, why shouldn't we consider if a newly released game has been well optimised for both re-sizeable bar and fsr, rather than let game developers get away with not programming for them correctly. For me the issue of hardware unboxed not having tested with SAM/FSR is that it takes pressure away from game developers to optimise for both. I would have thought they could do separate videos for gpu with native, SAM and FSR. They will get extra revenue for doing more videos and it will encourage game developers to get behind FSR and SAM, giving us free extra performance.

 

[gpuerrilla]

Yup.

In other developments, it looks like this possible system error appears to be more widespread that first thought.

Note the computerbase chart regarding the 3080 and HD Textures.

Are they in the "reputable source" list though Matt?

Possible system error?

I suppose its better than us plebs opinion right?