M2 Max Gfx bench ?

[l0stl0rd]

I just saw this.

GFXBench - Unified cross-platform 3D graphics benchmark database

The first unified cross-platform 3D graphics benchmark database for comparing Android, iOS, Windows 8, Windows Phone 8 and Windows RT capable devices based on graphics processing power.

gfxbench.com

 

[AgentMcGeek]

Seems consistent with the keynote. Between 20 and 30% better than M1 Max. Not bad at all.

 

[ssn637]

Which benchmarks are the most relevant when comparing the M2 Max to its predecessor?

For example, 1440p Manhattan 3.1.1 Offscreen frame rates are 50% higher, so I suppose the Texturing benchmark would be more representative of improvements we would expect?

And would 96 GB RAM make a difference compared to the current 64 GB shared RAM maximum for GPU performance?

 

[leman]

Which benchmarks are the most relevant when comparing the M2 Max to its predecessor?

For example, 1440p Manhattan 3.1.1 Offscreen frame rates are 50% higher, so I suppose the Texturing benchmark would be more representative of improvements we would expect?

There is probably no good answer to your question. These benchmarks are also fairly old and primitive by modern standards. Frankly, I don't think that much can be learned by looking at them. These ridiculously high frame rates mean that the GPU is doing very little work per frame, so the performance-limiting factors could be things like synchronisation or frame scheduling, which will show themselves very differently in a real-world game or app.

For general gaming benchmarking, right now there is only 3D Mark Wild Life Extreme, plus a handful of games (BG3, Metro, Resident Evil). To get a better picture one would ned to look at that.

And would 96 GB RAM make a difference compared to the current 64 GB shared RAM maximum for GPU performance?

If you work with scenes that need this much RAM, sure. But what game would that be?

 

[diamond.g]

For general gaming benchmarking, right now there is only 3D Mark Wild Life Extreme, plus a handful of games (BG3, Metro, Resident Evil). To get a better picture one would ned to look at that.

Anything that is repeatable is ideal.

 

[l0stl0rd]

Now this is interesting 14,5 =14” and 14,6 = 16”

 

[diamond.g]

View attachment 2145195Now this is interesting 14,5 =14” and 14,6 = 16”

higher clocks on the 16"

 

[MayaUser]

There is probably no good answer to your question. These benchmarks are also fairly old and primitive by modern standards. Frankly, I don't think that much can be learned by looking at them. These ridiculously high frame rates mean that the GPU is doing very little work per frame, so the performance-limiting factors could be things like synchronisation or frame scheduling, which will show themselves very differently in a real-world game or app.

For general gaming benchmarking, right now there is only 3D Mark Wild Life Extreme, plus a handful of games (BG3, Metro, Resident Evil). To get a better picture one would ned to look at that.



If you work with scenes that need this much RAM, sure. But what game would that be?

do you think these M2 Pro/Max scaling to be better than M1 Pro/Max thanks to the TLB buffer increase?

 

[MayaUser]

higher clocks on the 16"

and not on the 14" i wonder?
On the M1 pro and M1 Max the 14" had the same cooling system, same exact weight, but now it seems it could different since the M2 Max 14" weights more than the m2 Pro. Probably next week we will see

 

[l0stl0rd]

do you think these M2 Pro/Max scaling to be better than M1 Pro/Max thanks to the TLB buffer increase?

From the numbers that showed up I don’t think that it scales any better then before.

I had made an estimate of the M2 Max based on the old scaling and the M2 Pro score and I was of by about 900 points for the 14,6.

 

[diamond.g]

and not on the 14" i wonder?
On the M1 pro and M1 Max the 14" had the same cooling system, same exact weight, but now it seems it could different since the M2 Max 14" weights more than the m2 Pro. Probably next week we will see

IIRC on the M1 line the 14" was capped at a lower Ghz than the 16".

 

[MayaUser]

IIRC on the M1 line the 14" was capped at a lower Ghz than the 16".

yes, i saw that m1 max 14" draw less W compared to the 16"
Maybe here it will be closer if Apple indeed place a slightly better cooling for the M2 Max

 

[ssn637]

Could those metal scores be for the 30-core GPU?

 

[MayaUser]

From the numbers that showed up I don’t think that it scales any better then before.

I had made an estimate of the M2 Max based on the old scaling and the M2 Pro score and I was of by about 900 points for the 14,6.

still on graphics it seems they scale better
25.000 points between M1 pro and M1 Max while 34100 points are between M2 pro and M2 max

 

[leman]

do you think these M2 Pro/Max scaling to be better than M1 Pro/Max thanks to the TLB buffer increase?

Do we know that the TLB was increased? I am not aware of any scaling issues with M1 to begin with (aside Ultra in Blender, but there can be multiple reasons for that) or what TLB might have to do with it. Apple claims that the L2 has been increased, that will definitely have a non-trivial impact in GFXbench.