News Lunar Lake iGPU benchmarks leaked on Geekbench 6 — comparable performance to Radeon 780M and Arc A380

[Admin]

Leaked Lunar Lake Geekbench 6 Vulkan benchmark results numbers show performance close to GTX 1650, RX 6400, and A380.

Lunar Lake iGPU benchmarks leaked on Geekbench 6 — comparable performance to Radeon 780M and Arc A380 : Read more

 

[Deleted member 2731765]

Since we are measuring the performance of an integrated graphics processor, you fail to mention which Xe2 iGPU has been tested here in your article.

We might be looking at the "ARC 140V" iGPU being tested in VULKAN API.

The system used for the tests shows 32GB of RAM with Windows 11 Enterprise OS,

It's MoP we are talking about here with Lunar Lake. Which means the platform was running 32 GB of on-package LPDDR5x memory.


EDIT:

For reference here is the rumored/leaked Lunar Lake lineup. Some chips are stilling missing from this list though. But more importantly, we get to know the different configurations of the Xe2 iGPU.

Lunar Lake "Core Ultra 200V" CPU Lineup:

Model	P-core Boost	E-core Boost	LLC	PL1	PL2	LPDDR5X-8533	GPU	GPU Clock	NPU	XMX
Core Ultra 9 288V	5.1GHz	3.7GHz	12MB	30W	30W	32GB/2R	Arc 140V	2.05GHz	48TOPS	67TOPS
Core Ultra 7 268V	5.0GHz	3.7GHz	12MB	17W	30W	32GB/2R	Arc 140V	2.00GHz	48TOPS	66TOPS
Core Ultra 7 266V	5.0GHz	3.7GHz	12MB	17W	30W	16GB/1R	Arc 140V	2.00GHz	48TOPS	66TOPS
Core Ultra 7 258V	4.8GHz	3.7GHz	12MB	17W	30W	32GB/2R	Arc 140V	1.95GHz	47TOPS	64TOPS
Core Ultra 7 256V	4.8GHz	3.7GHz	12MB	17W	30W	16GB/1R	Arc 140V	1.95GHz	47TOPS	64TOPS
Core Ultra 5 238V	4.7GHz	3.5GHz	8MB	17W	30W	32GB/2R	Arc 130V	1.85GHz	40TOPS	53TOPS
Core Ultra 5 236V	4.7GHz	3.5GHz	8MB	17W	30W	16GB/1R	Arc 130V	1.85GHz	40TOPS	53TOPS
Core Ultra 5 228V	4.5GHz	3.5GHz	8MB	17W	30W	32GB/2R	Arc 130V	1.85GHz	40TOPS	53TOPS
Core Ultra 5 226V	4.5GHz	3.5GHz	8MB	17W	30W	16GB/1R	Arc 130V	1.85GHz	40TOPS	53TOPS

 

[Deleted member 2731765]

However, you should note that these are just alleged test results. We cannot verify if these come from valid systems, especially as Intel’s partners are likely contractually bound to stay silent until after these chips officially launch

What an "absurd" assertion you made here.

Of course, there are early test results, but this latest benchmark was conducted on an Intel "reference evaluation platform/RVP" for LNL-M chips. This would also suggest that the chip is an engineering sample.

But assuming that these tests were not done on a 'valid' system is not correct. The GB entry clearly mentions Intel's RVP.

Model	Intel Corporation Lunar Lake Client Platform
Motherboard	Intel Corporation LNL-M LP5 RVP1

 

[aaronage]

For completeness:

Apple M1 (5nm, 2020)
OpenCL: ~20000
Metal: ~32000

M2 (4nm, 2022):
OpenCL: ~27000
Metal: ~46000

M3 (N3B, 2023):
OpenCL: ~30000
Metal: ~47000

M4 (N3E, 2024):
OpenCL: TBC
Metal: ~53000

 

[Deleted member 2731765]

Are you showing the advantage of comparing those Apple SoC chip scores with Lunar lake's igpu here?

 

[aaronage]

It’s a relevant comparison. Apple Silicon, X Elite should be included in these discussions in general.

Lunar Lake is supposed to be Intel’s answer to the base Apple Mx SoCs, it makes sense to include Apple numbers where possible.

The same thinking should be applied to AMD Strix Halo, too. It will be a high-end SoC with a large GPU and high bandwidth memory - it would be crazy to omit Apple’s Mx Max SoCs (the model for that kind of product) from comparisons.

 

[Deleted member 2731765]

Of course we should include other processors in the list, but the main issue is that there is only 1 entry for this processor.

So it is hard to take an average value for a valid comparison, and on top of that, this is an early engineering sample. So this Vulkan API score doesn't reflect the final performance.

Make a note though that we are dealing with the performance of only the igpu here, which is Xe2 Battlemage graphics in this case.

 

[TechyIT223]

AMD's AI 300's iGPUs have performance level around RTX 2050 GPU's range.

This Intel LNL chip performs around GTX 1650 range.

So they are all within 10% margin of each other.

 

[Notton]

Seeing these results, I want to see what X Elite and AMD Strix can do with memory on package.
The next obvious answer is a 192-bit or 256-bit mem bus.

 

[TechyIT223]

Do the X Elite Snapdragon processors too have memory on package? , 🤔

I don't think so