Review [Hardware Unboxed] Spider-Man Remastered 43 GPU benchmark

[tamz_msc]

Looks like AMD GPUs are CPU-bound and possibly the lack of PCIe lanes/memory bandwidth really hurts the low end AMD GPUs a lot.

 

[Leeea]

It is an interesting review!

The AMD "bottleneck" feels related to the real time shader compilation, which is something somewhat "unique" to this game. I suspect we will see it more with future console ports. Although, at over 100 fps not sure how much of a bottleneck this really is.


He did run it with a Ryzen 5800x3d, which is one of thee best CPUs for that game.


Be curious if he does a CPU follow up on it. Curious as to how the lower end units do, and if AMD pulls ahead as the CPU gets weaker.

 

[jpiniero]

Those Navi 24 numbers... woof. Really makes the 6600 look like a much better deal in comparison.

 

[Stuka87]

Those Navi 24 numbers... woof. Really makes the 6600 look like a much better deal in comparison.

It would be interesting to see an 8GB version tested. If it follows with other testing, those numbers would go way up.

 

[coercitiv]

Those Navi 24 numbers... woof. Really makes the 6600 look like a much better deal in comparison.

RX 6400/6500, GTX 1630 and Arc A380 are in a league of their own.

 

[DAPUNISHER]

Although I suspect we will see it more with future console ports. Although, at over 100 fps not sure how much of a bottleneck this really is.

Agreed, this game is perfectly fine at 60-75fps. Ray tracing is where it becomes useful to prospective hardware buyers. Otherwise it scales well.

 

[DAPUNISHER]

Started watching the vid, and he is already wrong. Claiming the street level is as demanding as swinging around is flatout incorrect. Swinging through the park and looking at the city from atop the highest buildings and jumping off is the most stressful. zWORMz shows this in gameplay over and over again on a bunch of different cards tested with either a 5700g or 5900x. Depending on if he thinks the APU will bottleneck a card or not.

This is the perfect example of how salty he gets when he can't run a canned bench. And throwing shade at Spiderman fans when you put mad hours into Fortnite is about the height of DERP. I am basking in the schadenfreude of him being forced to test a game by playing it. So what did he do? Test at street level LULZ!

 

[psolord]

Hey all.

I did a few tests of the first part of the game, to see how it runs on various systems, if anyone cares. Following vids not monetized. Just a hobbyist channel. Also note these are not exactly benchmarks per se, more like "how it runs" vids.

First on a 12400f+3060ti, straight very high preset, two vids one with RT and one without. These are recorded with an external recorder, so no performance hit. Also they are the only tests runs from a nvme. The other two run from plain old HDDs.

========

 

[psolord]

Then a quick test again at very high, on the old 1070 paired with a 8600k. Recorded with MSI Afterburner, fewer recording stutters compared to shadowplay, but a performance hit exists nontheless, so this vid is capped at 60fps, just to see if this system can manage it. It seems it does. The 1070 is aging even better than the 970, due to its vram mostly.

=======

And last but not least the age old GTX 970 paired with a Sandy 2700k. Again recorded with software, shadowplay this time, which causes many recording stutters unfortunately, but that's all I can do on that system. The 970 has big framebuffer troubles even at medium+fsr. The game does run quite well and it's even better without the recording actually. If you consider that this 11yo system executes the game, from a HDD and records on another HDD accessing both at the same time, its not bad at all.


Also color me impressed with fsr 2.0. It's the first time I saw it up close and it has very good quality. Much better than 1.0.

I bet the 2700k could hit 60fps even at very high with a better card, but that is a test for another time.

=======

Unfortunately the game runs very bad on the 7950 even at low. It hovers from 20-60fps and is quite unplayable during fights. I bet it would run better if there was a DX11 option, since the 7950 takes a 33% performance hit from DX12 maybe even more. I will try to retest if Nimez releases any newer driver, other than the official 22.6.1 AMD one that exists now.

That system is paired with a 2500k, which at the moments of 60fps, shows 50% load, so I guess even this old cpu could do OK with a better video card at higher settings. Maybe next week I will switch the 7950 with the 3060ti and see if the newer 12400f can help the old 7950 and the vice versa if the newer 3060ti, can help the older 2500k.

 

[psolord]

Also for the nth time, can we please remove the 2 media per post fml? We are in 2022 ffs, thank you.

 

[Tup3x]

RX 6400/6500, GTX 1630 and Arc A380 are in a league of their own.

With the upcoming drivers

(around 32min). Arc would perform way better.

For example from 13,5 fps to 58,9 fps (medium settings, 1080p) and now being faster than 6500 XT. Soo... There is potential in the hardware.

 

[coercitiv]

There is potential in the hardware.

The greater unknown is the potential in the software.

I'll take that nice performance uplift though, hope to see more of that in other titles before the official "western launch".

 

[Unreal123]

It is an interesting review!

The AMD "bottleneck" feels related to the real time shader compilation, which is something somewhat "unique" to this game. I suspect we will see it more with future console ports. Although, at over 100 fps not sure how much of a bottleneck this really is.


He did run it with a Ryzen 5800x3d, which is one of thee best CPUs for that game.


Be curious if he does a CPU follow up on it. Curious as to how the lower end units do, and if AMD pulls ahead as the CPU gets weaker.

Actually this game is 1:1 port of PS5 not like Days gone, God of War or Horizon Zero Down.

The interesting thing is that RX 5700 XT is on par with RX 6700 XT and RTX 2080 Ti is beating RTX 3070 TI in 1440P.

This means that this game love raw computing and Turing and Ampere has better computing than RDNA 2.