Halo Infinite GPU Performance

[EastCoastHandle]

https://www.computerbase.de/2021-12/halo-infinite-benchmark-test/

Haven't seen to many benchmarks for the campaign as of yet. This is the only one I've found so far. Was wondering how everyone else is doing with this title?

 

[EastCoastHandle]

GOW will support FSR and DLSS though. I haven't seen either in Halo Infinite yet. Although someone in the beta stated that Halo Infinite did have FSR support during the beta. So, I can only assume it will come out at some future point.

 

[EastCoastHandle]

Poor performance in the open world for me. Seem to match the computerbase results around 80fps max settings at 1440p.
I tried df optimized settings which is slightly better at 10-15 more fps but still not great. Doesn't feel smooth, I think it's some animations running 30fps.

Interesting, thanks for posting this information.

 

[EastCoastHandle]

Hardware unboxed (HUB), just like techpowerup (TPU), disable Smart Access Memory (SAM).
SAM adds 10-20% performance in this game from my testing.

I'm not sure why they choose to disable it when it is enabled by default and they use Ryzen processors.

Spoiler

SAM has been disabled for HUB since their first 6800 XT /6900 XT reviews, TPU the same. That affects all their results for every graphics card review and game benchmark they do. Something to remember when you look at their results but this is often overlooked.

Meanwhile, the following sites have RDNA2 ahead in Halo and they test the same open world demanding areas online (like HUB) which represent a worst case scenario. They are using the Steam version of the game according to the testing notes.

1. Sweclockers
2. PCGamesHardware
3. Computerbase
4. Kitguru

It is not clear if Kitguru used ReBar, but the others had SAM/ReBar enabled as there is no logical reason to disable it for AMD graphics cards.

Spoiler

Thanks for posting this LTMatt it explains the disparity from HUB bias for nvidia. I am sure that whatever re-test he claims to be making will only reinforce his own bias to make nvidia look good, as usual. But I certainly am not waiting for his results under baited breath. He's always been bias and that will never change. Anyone waiting for those results is like the old saying...fool me once shame on you. Fool me twice...well, you should know better by now but you know the rest of the saying.

Ist nvidiay photoshop GOW PC Specs to make people think that game only supports their gpus. Now you have reviewer(s) disabling AMD's SAM support to lower it's performance intentionally to make the competition looks faster then they actually are. I am willing to believe that he kept rebar support enabled for nvidia gpus in Halo Infinite and noticed very little performance improvement. So whenever he provides this update it will be from some other area of the map. But that's just my opinion. I wouldn't put it past him.

Another thing, according to these results from other websites he is clearly an outlier and he knows it. Probably for clicks. The guy has no credibility with me.

 

[EastCoastHandle]

Steve is doubling down. 'We are not going to test with SAM/ReBar enabled.'

That's unfortunate but not entirely unexpected. It's easier to double down rather than admit you got it wrong.

It means I can't reliably compare performance data using HUB, since I know that all of their RDNA numbers are wrong and below what they should be.

At least we have Computerbase and PcGameshardware who test using these standard features and they both do an excellent job, so its not the end of the world.

That's why I said I wasn't holding my breath for his followup review. The guy hates Radeon GPUs. It's not that he simply favor Nvidia. He simply despise Radeon. And he is implying that when he said he won't enable SAM support in future GPU benchmarks. He's not just doubling down. He really hate Radeon/AMD.



As you can see with Sam disable (even though he refuses to call it Sam) there is a huge performance loss. There is little performance found on nvidia gpus @1080p for example. Like I said in my prior post:

I am willing to believe that he kept rebar support enabled for nvidia gpus in Halo Infinite and noticed very little performance improvement...

Furthermore, we really don't know where he deviated from the test run. It's clear to me that he has. The 600 series does have a higher min. frame rate.

I do wonder if he simply took the worst runs to calculate FPS?

 

[EastCoastHandle]

A typical user tells him that the results are off. And this is his response. IE: "you are wrong"







This is what happens when a known benchmark organization gets involved. Yeah, he already had the results and had to come clean when he got called out. Notice Steve didn't tell him that his results were "apples to oranges" do you? CapFrameX results are still higher for the 6800xt at 1080p then Steve's results.

And, in his video he made it very clear that he will still keep SAM disabled in current and future GPU benchmark results. Even though he corrected his prior benchmark results by enabling SAM. It took CapFrameX to show the correct performance boost using SAM on Radeon Cards.


There is absolutely no reason why, after showing a huge performance gain with SAM enabled he still wants to disabled SAM support in future GPU benchmarks. Which have the potential to show Radeon GPUs in a bad light as exampled here. He also made it clear that if you don't like his decisions you can go to other websites. Which is something I've been doing for a long time. As he is already telling you he won't do GPU benchmarks correctly.

 

[EastCoastHandle]

I don’t necessarily think he hates either brand, I honestly don’t even though it is a strange decision. I remember not long ago people thought HUB had it in for Nvidia because he uploaded some game benchmarks with AMD ahead in a few games.

IMO, that's because it wasn't common for Radeon GPUs to be ahead in their benchmarks at that time. But I digress.

However, the issue here is a bit different though. He is fully aware that disabling SAM decreases performance substantially. Doing so makes a 6900xt slower then a 3080 in Halo Infinite, for example. Yet, hee was perfectly fine presenting this in his previous video. If it wasn't for CapFrameX I do not believe we would know for certain what was going on. Now we know for certain what happens in Halo Infinite when you disable SAM on Radeon GPUs.

But the other issue is the performance results of RTX with Rebar enabled. Even though there are some uplifts, depending on resolutions, still puts Radeon GPUs ahead performance wise. But a reviewer would need to make sure that SAM/Rebar are enabled (unless otherwise told to disable it do to performance degradation).

 

[EastCoastHandle]

I hate to say I called it, but well, I called it.

Computerbase, pchameshardware and sweclockers are the go to sites for graphics cards and benchmark reviews. The Germans are the best, as always. Way ahead of the rest at the moment.

Until HUB and TPU stop disabling Smart Access Memory, their results will always be inaccurate.

I see HUB getting called out for this on Twitter now, so I hope they take note and change their methodology going forward, since I generally like watching their videos. TPU I think are a long way off unfortunately.

I couldn't say it any better then this. I honestly don't understand why he is doing this. But I do understand that he is. And at the end of the day a person's actions are more important then understanding the actions.

HUBs testing section also brings up another interesting debate. They tested a worst case scenario in single player campaign. Nothing wrong with that per se, the problem is it's not reflective of 99% of the game, or multiplayer. Typically I'd say test a sequence that is largely representative of the game, rather than a worst case or best case scenario.

Ultimately though, I have no problem with the area they tested. They have to pick somewhere and all sites picked slightly different areas. It only raised a red flag for so many people as the results were totally different to everyone else's. As it turned out, this was mostly down to disabling SAM/ReBar and that is my main gripe with HUB and TPU.

By disabling SAM, you are hurting AMD performance in every graphics card and game benchmark they run. Between both sites, they run a lot of benchmarks. That means every single result from these outlets is inaccurate, that's a problem.

I know Nvidia choose which games and apps to enable/disable Rebar, so there should be no issue in enabling this feature for both. If it really does hurt performance in a particular title for a particular vendor, it would make sense to disable it for that game/benchmark for either vendor.

Due to driver updates, SAM does not hurt performance anymore. Furthermore, SAM can be toggled on and off via Radeon Software. I believe Nvidia offer similar functionality based on user feedback here. So, no logical reason to keep SAM/ReBar off - unless it negatively impacts performance or causes other issues.

There was another youtuber who also tested in this area. He mentions he was told about this area specifically but doesn't say by whom. So it's a well known portion of the map that taxes the GPUs the most. But who is specifically telling people this? Unknown.

But if they get more FPS with rebar on, why not having rebar enabled for Halo in the driver? The profile inspector is a 3rd party tool ( like Radeon Pro or whatever the name was a few years ago ). I am using the profile inspector on my old laptop because the control panel only works once.
It is a good tool but it is weird to use a 3rd party tool to enable rebar.

Res ipsa loquitur
It really does speak for itself once you look at before and after benchmark results IMO.

 

[EastCoastHandle]

Looks like HUB was called out by someone who he couldn't deny (CapFrameX). In which he was able to instantly provide SAM performance gains for 6000 series in his reply to CapFrameX. Indicating to me, at least, that he had those results. To only release a 2nd video only partly admitting that Radeon is faster in Halo Infinite. However, it's nebulous as to how he conducted those benchmark results at 4K. As he is still the outlier in which he has 6900xt behind at 4K. Even though the amount of enemies do not change in difficulty at lower resolutions (in which 6900xt won). That portion of his reasoning makes no sense to me.

Personally I find no value in HUB results. And it more clear why after this fiasco/debacle.
1) He hid the fact that he disabled SAM in Halo Infinite (I did not hear anything about SAM being disabled in his 1st video)
2) He already had performance uplifts available with SAM enabled when he replied to CapFrameX
3) He refuses to enable SAM support in future benchmark reviews. Putting the 6000 series at an artificially reduced performance scale.

 

[EastCoastHandle]

Don't worry, my minimum FPS is the same as his average, taking into account the 3-4 FPS cost of recording.

Thanks for the confirmation. But it's not a surprise to me when it comes to his review.

Thing is it's not even just harming amd but also harming Nvidia by not enabling rebar.

If you go back to that 1st video where SAM and Rebar (allegedly) was disabled. It clearly showed that a 6900xt was no match for a 3080 in Halo Infinite. It's all about presentation. And you would have to ask yourself would anyone care that Rebar was disabled on their 3080? IMO, no.

 

[EastCoastHandle]

Exactly, most people don't care, it's only amd fans getting their panties in a twist

That's why he got called out for it. The review show have been no different then any other review. Enable SAM/ReBar (disabled allegedly) and simply let the benchmark results tell who won. You don't need to cripple performance on the 6000 series. Which is why his review was/is an outlier.

In any case it's clear what he did. He will not enable SAM in future gpu benchmarks so it's really a moot point to trust his results moving forward.

I've just did a quick run though myself. And used Radeon's Performance Logging just to see what I get.

1440p Ultra
Outpost Tremonius


As you can see no were near "those" results.

 

[EastCoastHandle]

Just a quick run through with all the enemies at the end. Ran great

And, about that TPU benchmark review:

It's a typical open world scene, near one of the enemy bases, with a few enemies showing up towards the end of the run.
I'm at least halfway through the game and haven't really encountered "hordes" of enemies, so testing such a scenario would be cherry picking a worst-case.

https://www.techpowerup.com/forums/...enchmark-test-performance.289799/post-4667440

 

[EastCoastHandle]

If yopu enable AMD SAM then enable ReBar for Nvidia; if the gains are equal , then i must ask, why turn if off in the first place?

The gains are not equal.