Ok. The two cores extra would definitely make a difference.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
M5 gaming
- Thread starter M4pro
- Start date
- Sort by reaction score
I could have replaced DS 2 with why Skull and Bones isn't on macOS.
Is it possible M5 has enough bandwidth but not enough core clocks to change things? Or maybe since it is mostly/all FP32 so the uplift isn't as great?
Reviewers tend to test with both and them comment on image quality as well. Same with Frame Generation.
The first is the same as memory bandwidth being overkill which I doubt. The latter is what I'm going with. But as @leman said earlier, all speculation.
Depends on the reviewers and their audience. Gaming-centric reviewers might, but lots of reviewers don't (I mean obviously since we are discussing a bunch of reviewers that didn't
).
Last edited:
Yeah non gaming-centric testers should just stick with the games chosen defaults. Makes things more easily comparable.The first is the same as memory bandwidth being overkill which I doubt. The latter is what I'm going with. But as @leman said earlier, all speculation.
Depends on the reviewers and their audience. Gaming-centric reviewers might, but lots of reviewers don't (I mean obviously since we are discussing a bunch of reviewers that didn't
Nothing too surprising. Does SPEC. They got similar CP2077 results to Ars and oddly NBC which all purport to come from different CP2077 versions, some native and some not. That part, while maybe not surprising since it confirms earlier results, is very confusing.
That's actually why posted the verge numbers, its comparing apples to apples. No hyperbole, no
exaggeration, just basic facts.
Yeah, but some folks rush to publish their reviews, sometimes corners are cut or they just want to bang out game numebrs. ¯\_(ツ)_/¯
I kind of wish JarodTech would get review samples. He normally tests notebooks so this would be right up his alley.
@crazy dave You were correct. Turning on SSRQ to Psycho yielded 37 fps like the Ars review. Must have been the old settings.
The MBP screen is 1964p so 1200p doesn't scale particularly well either. If the wanted to scale for the screen it would be 982p (but I don't think games offer this).
The Andrew Tsai benchmark looks good, but it's a bit worrying that the min fps is only 26 fps. Would be interested to see if dropping from 1080p Ultra with MetalFX quality to either 1080p Ultra with MetalFX balanced or 1080p High with MetalFX quality would enable a locked 40 fps
Check the "More space" option on the 14" MacBook Pro display setting, it's 1800x1169, which I agree is very awkward in terms of scaling but I'm assuming that's what they're running at when they say 1200p. Either that or all these reviewers have a 16:10 external monitor which has a 1920x1200 native or effective resolution. That's also possible - 16:10 is common. Probably the more likely now that I think about it.
I was afraid of that ... dang. I had hoped the new version that fixed that issue would've been out for long enough so that the bug would not affect reviews. Although, from a purely scientific standpoint, we get to learn that with that bug the raster performance of the native port and the xOver version are almost the same, within a couple of percent. Which is mildly interesting.
Last edited:
Just replying to my own post now I've watched a bit of the Andrew Tsai video. Looks like 1080p High with MetalFX quality (upscale from 720p) runs from the low 30s to mid 40s (so still not enough for a locked 40 fps) and 1080p Medium with MetalFX quality runs from low 50s to mid 60s. I didn't see a different upscaling factor (e.g. 540p to 1080p) demonstrated.
Sadly, 16:9 displays all but killed the much more useful 16:10 displays about 10-15 years ago
Do you have the binned M4 Pro or the full one? Maybe they have bin and you have full? Perfect scaling is rare but 20/16*37.6=47. Could be a coincidence? Feels very unlikely especially since NBC do specify that they have a 20-core Pro and got 39 FPS for the emulated version which lines up with the 37 FPS quoted by Ars. I don't actually think this is what is happening, I'm just triple checking - although even if I know what you have I still won't know what Ars has.
NotebookCheck claims, despite the labels, that they tested the Mac version of Cyberpunk 2077 (which makes their comparison to previous data, especially previous Mac data, moot but also confusing), so I've asked them to double check because then either the native port got near 0 improvement or the M5 GPU got near 0. But we're seeing consistent numbers even with Geekerwan and I have trouble believing that everyone has the 2.3 versus 2.31 patch. Something doesn't add up.
EDIT: Geekerwan didn't actually get the same, the 30FPS was the 1% lows, their average was 36FPS. Which makes a lot more sense:
EDIT 2: actually these might 1080p high
Last edited:
Actually I was wrong about Geekerwan, check my edit that literally just posted, they get 36FPS average for the M5! I was mistaking the 1% lows for the average. Unless I've screwed up again.
Okay I'm pretty convinced most of these outlets are using 2.3 rather than 2.31.
EDIT: unless Geekerwan tested on high not Ultra? I can't read the Chinese and am unsure if the autotranslation is saying 1080p high or 1080p Ultra. It might be high because the FPS for their M4 Pro mini is 54, even more than the 47 you quote. AGGGGGGGGHHHHHHHHH!!!!!!
Last edited:
I'd like to think as long as everyone is testing the same version of software their results should be easily repeatable (and comparable) since there is no such thing as silicon lottery with Apple devices.
The question is which version of the software is it? The first native version 2.3 with the bug or the recent patch 2.31? One would think it would be the latter ... but then these results make no sense unless I'm missing something obvious.
(Also there is minor silicon lottery amongst Apple devices usually on the order of about 6% or less, but, true, generally nothing substantial)
Have folks seen GPU clocks differ by up to that much? I mean yeah the CPU clocks may not be 100% equal for everyone, but when GPU bound that doesn't matter all that much (generally speaking).
It's not really top clocks, although maybe how long the computer stays there. You can check the 3DMark website for examples, e.g.:
Bottom scores are about 1577, top 1627, that's about 3%. Highest I think I saw was 6%. Can't remember which test.
Actually here's one:
Very bottom score I suspect guy had some background processes on, but the rest are within 6%.
Since 3dmark doesn't report clock speeds for Apple systems, I would argue that all the variations are probably due to background processes. Since it can't be proven otherwise 😅.It's not really top clocks, although maybe how long the computer stays there. You can check the 3DMark website for examples, e.g.:
Bottom scores are about 1577, top 1627, that's about 3%. Highest I think I saw was 6%. Can't remember which test.
Actually here's one:
Very bottom score I suspect guy had some background processes on, but the rest are within 6%.