Question How does the Radeon 780M compare to low-end somewhat recent dedicated GPUs?

[Order 66]

I know that there was a thread that talked about the 780M in comparison to the 680M, but that was with timespy, not real-world gaming tests.

 

[Eximo]

Basically a desktop GTX770 / GTX680

Something along the lines of 1/3 - 1/2 of an RTX 4060 mobile or roughly equivalent to an RTX3050 mobile at base clocks or an RTX 2060 mobile running full speed.

Difficult to directly compare because cooling/power/clock speeds etc are kind of up to the laptop maker.

Either way, comparable to a more entry level GPU of today.

 

[Order 66]

Basically a desktop GTX770 / GTX680

Something along the lines of 1/3 - 1/2 of an RTX 4060 mobile or roughly equivalent to an RTX3050 mobile at base clocks or an RTX 2060 mobile running full speed.

Difficult to directly compare because cooling/power/clock speeds etc are kind of up to the laptop maker.

Either way, comparable to a more entry level GPU of today.

Still impressive considering it is integrated. I wish I had known about this back when I was looking for an upgrade from my i5 6500 and RX 550. The 780m is far more powerful than my RX 550. I am happy with my current system though.

 

[Eximo]

That isn't integrated graphics to be clear. 780M would still be considered discrete GPUs, as in separate from the CPU/SoC with their own memory.

Low end Radeon/Vega APUs and Intel HD/Iris would be called integrated. Now we have RDNA2/3 and Intel Xe integrated GPU cores. Though in reality those are more like discrete GPUs that share system memory as they are still discrete tiles that just happen to be directly next to the CPU.

780M would also likely have been an MXM card rather than soldered to the motherboard. Possibility of upgrading to something like the 980m depending on the BIOS.

10 series from Nvidia is when the Desktop and Mobile GPUs more or less started being interchangeable and almost always soldered on to the motherboard. The mobile chips are generally the better binned versions of the desktop chips and run at a lower clock speed. Also generally a small core count change, usually in favor of the mobile chips.

 

[Order 66]

That isn't integrated graphics to be clear. 780M would still be considered discrete GPUs, as in separate from the CPU/SoC with their own memory.

Low end Radeon/Vega APUs and Intel HD/Iris would be called integrated. Now we have RDNA2/3 and Intel Xe integrated GPU cores. Though in reality those are more like discrete GPUs that share system memory as they are still discrete tiles that just happen to be directly next to the CPU.

780M would also likely have been an MXM card rather than soldered to the motherboard. Possibility of upgrading to something like the 980m depending on the BIOS.

10 series from Nvidia is when the Desktop and Mobile GPUs more or less started being interchangeable and almost always soldered on to the motherboard. The mobile chips are generally the better binned versions of the desktop chips and run at a lower clock speed. Also generally a small core count change, usually in favor of the mobile chips.

Wait, so the 780M is dedicated? Techpowerup seems to think it is integrated.

 

[Eximo]

Probably a mis-use of the term on their part.

Integrated generally refers to GPUs that share system memory with the CPU.
Discrete, either soldered on or modular, refers to GPUs that are not part of the CPU and have their own video memory.

Looks like MXM fizzled out completely with the RTX 20 series in 2019. So no more upgradeable laptops for 30 series or 40 series. (Even then it was pretty rare, and was reserved for very high end laptops/mobile workstations)

 

[Eximo]

Though I don't see them mention integrated, actually they specifically call out MXM.

NVIDIA GeForce GTX 780M Specs

NVIDIA GK104, 797 MHz, 1536 Cores, 128 TMUs, 32 ROPs, 4096 MB GDDR5, 1250 MHz, 256 bit

www.techpowerup.com

 

[Zerk2012]

Wait, so the 780M is dedicated? Techpowerup seems to think it is integrated.

Anything with a M behind it should be a chip not built into the CPU.

 

[Order 66]

Probably a mis-use of the term on their part.

Integrated generally refers to GPUs that share system memory with the CPU.
Discrete, either soldered on or modular, refers to GPUs that are not part of the CPU and have their own video memory.

Looks like MXM fizzled out completely with the RTX 20 series in 2019. So no more upgradeable laptops for 30 series or 40 series. (Even then it was pretty rare, and was reserved for very high end laptops/mobile workstations)

I think we are both confused. I meant the 780M referring to the Radeon 780M

 

[Order 66]

Anything with a M behind it should be a chip not built into the CPU.

I am referring to the Radeon 780M

 

[beyondlogic]

I know that there was a thread that talked about the 780M in comparison to the 680M, but that was with timespy, not real-world gaming tests.

notebook tech shows some gaming tests

https://www.notebookcheck.net/AMD-Radeon-780M-GPU-Benchmarks-and-Specs.680539.0.html

it is a integrated gpu that uses ddr5 5600 memory
LPDDR5X-7500

AMD Integrated Radeon 780M 25% Faster Than RDNA 2 Predecessor

Early 3DMark Time Spy benchmarks of RDNA 3 iGPUs

www.tomshardware.com

 

[Order 66]

notebook tech shows some gaming tests

https://www.notebookcheck.net/AMD-Radeon-780M-GPU-Benchmarks-and-Specs.680539.0.html

it is a integrated gpu that uses ddr5 5600 memory
LPDDR5X-7500

AMD Integrated Radeon 780M 25% Faster Than RDNA 2 Predecessor

Early 3DMark Time Spy benchmarks of RDNA 3 iGPUs

www.tomshardware.com

That was the thread I was referring to. I just don't understand why the 780M has RT cores considering it is not powerful enough to do RT in the first place. I feel like the die space used for the RT cores would have been better used for more general-purpose cores to improve performance.

 

[Eximo]

Ah, the old case of the recycling of names between vendors.

It has RT cores because they are part of the design language of RDNA3. It would cost them way more money to design a GPU without them, then to just make a smaller version of a larger GPU.

 

[beyondlogic]

Ah, the old case of the recycling of names between vendors.

It has RT cores because they are part of the design language of RDNA3. It would cost them way more money to design a GPU without them, then to just make a smaller version of a larger GPU.

yeah its just a cut down die or a failed die from a higher part like most chips most chips are like that like the 2060 super is pretty much a slightly failed 2070.

its just a way to prevent e waste and mostly more profit. if they have like a few thousand of these chips lying around they will use them eventually. either in oems or a super series etc