Halo Infinite PC Graphics Benchmark: Over 30 Nvidia and AMD GPUs Tested

[Steve]

Today we're taking a look at Halo Infinite's graphics performance by testing over 30 Nvidia and AMD GPUs at three resolutions. We'll also take a look at quality preset scaling and VRAM requirements.

Read the full article here.

https://www.techspot.com/article/2382-halo-infinite-benchmark/

 

[Nobina]

RX 580 2124 FPS at 1080p. AMD fine wine indeed.

 

[Shadowboxer]

Seems in line with what I get on my RTX 2080 and 5800X. It’s bizarre, my CPU utilisation is reported at 55-65% but my brother who has an GTX1080ti and a 7700K is seeing similar frame rates with what’s here for a 1080ti and isn’t at 100% utilisation so I guess the game utilises bigger CPUs but doesn’t benefit from them.

It’s interesting how far from the 1080ti the 2080 is now. When the 2080 was released reviewed showed the 1080ti edged out the 2080. It’s something I’ve noticed as I live with my brother who has a 1080ti.

As I’m on a 144hz monitor I found best results from an unlocked frame rate, no minimum (DRS off), V sync off and I used Digital Foundry’s recommended settings which is mostly high with some at ultra and some at medium. This nets me about 60-70 in the open world and about 100 in the indoor sections at 1440p.

I just wish the game was longer, it’s so short it feels like I’ve played a demo of an open world Halo! I’d have rather had a normal campaign with chapters but I guess 343 don’t have the capability of doing that as well as Bungie so this is what we get.

 

[Nintenboy01]

Strange that the RX 580 does so poorly when the Xbox One X version with similar GPU horsepower and architecture edges it out quite a bit. Also strange that the performance difference between the 3070 and 3080 is quite small at 1080p and 1440p. In most games the 3080 is around 25-30% faster but here it's under 15%. Perhaps the game is a bit ROP-bound since the 3070 and 3080 do have the same number of ROPs (and the 3080 Ti and 3090 only have 16 more ROPs than the 3080)

 

[Hysc84]

My GTX 1060 6GB getting 40-60fps @ 1080p Low Preset

 

[Theinsanegamer]

I just wish the game was longer, it’s so short it feels like I’ve played a demo of an open world Halo! I’d have rather had a normal campaign with chapters but I guess 343 don’t have the capability of doing that as well as Bungie so this is what we get.

6 years of development, and all we got witha tiny campaign demo and 9 multiplayer maps. 343 quality, everyone!

Anyone who expects 343 to do well.....I mena, have you not been paying attention? After 4, 5, wars 2, the disasterous launch of MCC, and all their virtue signaling, it's obvoous 343 is staffed by morons.

Strange that the RX 580 does so poorly when the Xbox One X version with similar GPU horsepower and architecture edges it out quite a bit. Also strange that the performance difference between the 3070 and 3080 is quite small at 1080p and 1440p. In most games the 3080 is around 25-30% faster but here it's under 15%. Perhaps the game is a bit ROP-bound since the 3070 and 3080 do have the same number of ROPs (and the 3080 Ti and 3090 only have 16 more ROPs than the 3080)

Not really surprising. The game is horribly optimied. The game is made by 343, the literal interpritation of "a thousand monkeys and a thousand typewriters".

 

[Sausagemeat]

Strange that the RX 580 does so poorly when the Xbox One X version with similar GPU horsepower and architecture edges it out quite a bit. Also strange that the performance difference between the 3070 and 3080 is quite small at 1080p and 1440p. In most games the 3080 is around 25-30% faster but here it's under 15%. Perhaps the game is a bit ROP-bound since the 3070 and 3080 do have the same number of ROPs (and the 3080 Ti and 3090 only have 16 more ROPs than the 3080)

Im on an RX480 and if you enable resolution scaling I can get 50-60fps with the low setting at 1080p, although the game isnt all that smooth. I think the consoles must be going quite heavily on the resolution scaling to get the performance they are getting. But yeah Polaris is done, it really struggles to hit 60fps in quite a few modern games at 1080p now. Its a shame, looking at benchmarks it seems that it doesnt really fair any better than a 1060 despite having more memory, I guess we have the answer to that question, no, we did not need more than 6GB of RAM back when we bought these parts in 2016.

If only there was a $200-$250 GPU I could upgrade to, as it stands id be lucky to get a GTX1650 for that money and that part is slower than my RX480!

 

[Theinsanegamer]

Im on an RX480 and if you enable resolution scaling I can get 50-60fps with the low setting at 1080p, although the game isnt all that smooth. I think the consoles must be going quite heavily on the resolution scaling to get the performance they are getting. But yeah Polaris is done, it really struggles to hit 60fps in quite a few modern games at 1080p now. Its a shame, looking at benchmarks it seems that it doesnt really fair any better than a 1060 despite having more memory, I guess we have the answer to that question, no, we did not need more than 6GB of RAM back when we bought these parts in 2016.

If only there was a $200-$250 GPU I could upgrade to, as it stands id be lucky to get a GTX1650 for that money and that part is slower than my RX480!

That's not super surprising, the 480 was more powerful then the PS4/xbone, but had a slightly lower core count then the xbox one x (same core count as PS4 pro.). It also had a ~200 MHz clock speed advantage.

Compared to the PS5 and xbox series consoles, the 480 is much weaker, and that is quickly becoming the baseline for future games. The last gen frequently ran at 30 FPS too, so that's the range the 40 is expected to hit now.

Of course, the difference between "ultra" and "high" in most games is near nill, and at high settings the 480 still averages 60 FPS at 1080p. It's getting up there, but its still holding up for a 6 year old budget GPU. Halo infinite is very poorly optimized, not really a great example to show off polaris' capability

If anything, this shows why so many people wanted a larger polaris chip, and that AMD really should have made a 3072 core version .

 

[grumblguts]

Something up with these results. just seems like AMD flag ship running the game twice.

 

[Shadowboxer]

6 years of development, and all we got witha tiny campaign demo and 9 multiplayer maps. 343 quality, everyone!

Anyone who expects 343 to do well.....I mena, have you not been paying attention? After 4, 5, wars 2, the disasterous launch of MCC, and all their virtue signaling, it's obvoous 343 is staffed by morons.

Not really surprising. The game is horribly optimied. The game is made by 343, the literal interpritation of "a thousand monkeys and a thousand typewriters".

I dont have high expectations of Halo, it was amazing when I was younger but then again so were a lot of games. The Halo formula is spent, so 343 could release a Halo game similar to previous games and they would be criticised for not being original and then they could make something like this open world thing and what they have done is broken the formula. Now all guns and vehicles must be balanced to be used anywhere, this means the tank is gimped otherwise you could just roll up anywhere in a tank and the game would be easy. The open world lacks focus, in previous games each level had you using different guns and playing in a different way but now you can complete the entire game using the same guns and employing the same strategy, it’s a bit lost. The part that suffers most is that the most challenging parts are the bosses, which in 343’s world means adding a health bar to a brute and giving them a powerful weapon, it’s boring and unoriginal. In older Halo games you had pinch points in the level to provide challenge, once again the open world nature of the game kills this.

But MS is a corporation full of old men in suits, most of which probably don’t actually play and understand games and I think they just see the Halo brand as a cash cow and brand ambassador and they would be right. This game will sell millions of Xboxes and game passes (and also apparently some lunatics are paying the full price at £50 on steam, which is more than the cost of 6 months of game pass).

I can’t say I haven’t enjoyed Halo infinite and I don’t personally feel entitled to a new Halo game that meets my personal standards. But I do share you sentiments about 343, Bungie did it much better.

 

[yukka]

I am playing Halo with a 3600 and a 3060ti @1440P. I get 100fps at low and 80fps at high. Switch between low for multiplayer and high for campaign.

 

[Skjorn]

Where's the optimized settings?

If we want more performance the article just says switch EVERYTHING to medium. I don't buy that. There's got be 2 or 3 settings that can be bumped from ultra to high and gain considerable performance back. Sure you tested 30 cards but just selecting a preset is kinda lazy and just testing for the sake of testing and not seeing what actually eats performance the most.

 

[Vulcanproject]

Everything about the performance of this game is weird. The CPU tests are weird. At first it appears that it is CPU intensive, but actually nothing much seems to be happening.

The game can load 16 threads and then when you cut it down to 12 threads, 8 threads or even 4 you don't see some major average framerate drop or even 1 percent lows. Not on the tests I have seen.

What exactly the game is doing with 16 fully loaded threads when it runs virtually the same with only 8 threads is a mystery right now.

 

[Guberian]

Search up how to improve fps on Halo Infinite. It basically boils down to doing some file editing in order to force set the frames to the refresh of your monitor. Why you have to do this manually? I dont know.

 

[QuantumPhysics]

THANK YOU Techspot.

Glad to see the 3090, 3080Ti and 3080 snag all the top 3 medals for 4K, 1440p and 1080p.

 

[Shadowboxer]

Where's the optimized settings?

If we want more performance the article just says switch EVERYTHING to medium. I don't buy that. There's got be 2 or 3 settings that can be bumped from ultra to high and gain considerable performance back. Sure you tested 30 cards but just selecting a preset is kinda lazy and just testing for the sake of testing and not seeing what actually eats performance the most.

Steve’s test isn’t much use to players but does show how cards are fairing against each other. Although, these tests always come up too early to be that useful, there will be a number of driver updates and revisions to the game that will change things around yet.

If you want optimised settings, check out Digital Foundry’s analysis, the complete list of optimised settings can be found at 17:02 in. But I urge you to watch the whole video. Kinda shows how this game is undercooked. What on earth were they doing for 6 years. The Titanic was built in just a third of that time and they had less money!

 

[nnguy2]

THANK YOU Techspot.

Glad to see the 3090, 3080Ti and 3080 snag all the top 3 medals for 4K, 1440p and 1080p.

What a weird thing to be thankful for.

 

[Mr Majestyk]

What a weird thing to be thankful for.

Indeed. The fanboy is strong with this one.

 

[passwordistaco]

The story for me is always the lows. That's where stutters happen.

 

[Vanderlinde]

We have all the cores these days on the planet and this game still cant be properly optimized.

 

[EdmondRC]

On 5800x 32GB RAM RTX 3080, latest drivers, have the game on gamepass PC, every time I turn on async compute the game crashes within five minutes. Anyone else having a similar issue?

Also, it performs poorly with async on while it's running, especially when traversing the open world. I'm reinstalling the game and clean installing video drivers I guess that's the place to start. Reviewers seem to have had much better luck with async so trying to figure out where the problem lies.

Update: No luck, async compute is either broken in the game or my GPU just doesn't like it. I think it is the game.

Anyhow, the game is not well optimized, but it's not completely broken either. Hopefully 343 will patch async and better optimize the game for PC CPUs. DLSS would be nice, but extremely unlikely of course, considering they would have to completely add it to an engine that is optimized for AMD and XSX.

 

[Giddel Mirabal]

Why didn't they test Vega 64 cards? I woulda thought it would be more relevant considering it was the flagship card for the RDNA platform and does perform well in 1080p and 1440p gaming (maybe comparable to maybe a 1080?) with 8GB of HBM. Maybe they didn't have Vega 64 card on-hand?

 

[nnguy2]

Why didn't they test Vega 64 cards? I woulda thought it would be more relevant considering it was the flagship card for the RDNA platform and does perform well in 1080p and 1440p gaming (maybe comparable to maybe a 1080?) with 8GB of HBM. Maybe they didn't have Vega 64 card on-hand?

I had both the Vega 64 and 5700XT, they comparable at 1440p with the 5700XT performing about 5-15% better on newer games.

 

[PEnnn]

So, the difference between an RTX 3090 and a 3080 is about 3-5 FPS??

Those gamers who forked over $1000 more for the 3090 must be feeling really smart right now!!

 

[b3rdm4n]

THANK YOU Techspot.

Glad to see the 3090, 3080Ti and 3080 snag all the top 3 medals for 4K, 1440p and 1080p.

When AMD can pull out a win, it's all handshakes and pats on the back, that that big greedy Nvidia! Call out an Nvidia win tho and look at the crap you get for it.