Benchmark Results of AMD Ryzen 9 7940HS APU with Radeon 780M iGPU

Published by

Click here to post a comment for Benchmark Results of AMD Ryzen 9 7940HS APU with Radeon 780M iGPU on our message forum
https://forums.guru3d.com/data/avatars/m/260/260828.jpg
I really want that the next gen of APUs to come with big 3d cache or some kind of HBM in the interposer
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
I expected more from rnda3 igpu. The old rdna2 680m isnt that much slower.
https://forums.guru3d.com/data/avatars/m/246/246171.jpg
Undying:

I expected more from rnda3 igpu. The old rdna2 680m isnt that much slower.
Not much you can do about it when you're bottlenecked by RAM. DDR5 is still somewhat young for mobile platforms, so I'm sure you're not getting anything with much better bandwidth than last-gen DDR4. I'm sure most of the performance improvement can be attributed to architecture and driver optimizations, since the specs on paper are overall the same but clock speeds are lower.
https://forums.guru3d.com/data/avatars/m/50/50906.jpg
Slap DDR5 6000 or even faster LPDDR5X, and this baby will fly.
https://forums.guru3d.com/data/avatars/m/266/266726.jpg
Undying:

I expected more from rnda3 igpu. The old rdna2 680m isnt that much slower.
given that rembrant already had ddr5 support, its not all that surprising because igps are still heavily bottlenecked by bandwidth, the lowly 8cu igp on renoir/cezanne gets still gets significant performance scaling from memory past 4800mt/s, with 12cus at higher clock speeds, the bottleneck is much worse.
https://forums.guru3d.com/data/avatars/m/63/63215.jpg
I'm treating this as a ROG Ally preview igpu-wise.
https://forums.guru3d.com/data/avatars/m/202/202673.jpg
Picolete:

I really want that the next gen of APUs to come with big 3d cache or some kind of HBM in the interposer
There's no point to that at all, it takes up half the die space, and costs increase dramatically, defeating the point of a graphics heavy APU on PC...speeding up the CPU won't help iGPU performance as it's too slow to begin with. The consoles seem to have some nice solutions (VRAM for the whole system was a nice start w. PS4), at price points you can't really replicate on PC.
https://forums.guru3d.com/data/avatars/m/196/196284.jpg
Texter:

There's no point to that at all, it takes up half the die space, and costs increase dramatically, defeating the point of a graphics heavy APU on PC...speeding up the CPU won't help iGPU performance as it's too slow to begin with. The consoles seem to have some nice solutions (VRAM for the whole system was a nice start w. PS4), at price points you can't really replicate on PC.
Including HBM actually makes sense and could improve iGPU performance dramatically since the iGPU would no longer have to deal with the memory bottleneck of using system ram. It would be an APU with dedicated VRAM. If done on a desktop processor, if the iGPU performance is good enough, it would alleviate the need for a dedicated GPU.
https://forums.guru3d.com/data/avatars/m/50/50906.jpg
sykozis:

Including HBM actually makes sense and could improve iGPU performance dramatically since the iGPU would no longer have to deal with the memory bottleneck of using system ram. It would be an APU with dedicated VRAM. If done on a desktop processor, if the iGPU performance is good enough, it would alleviate the need for a dedicated GPU.
I'm sure too much HBM would be necessary to be able to use it as "dedicated" and actually make a difference.
https://forums.guru3d.com/data/avatars/m/202/202673.jpg
sykozis:

Including HBM actually makes sense and could improve iGPU performance dramatically since the iGPU
It's not a $150-$250 retail PC part you want to put on store shelves just hoping it'll sell at all (like a RTX4080 trollollol). Some gaming laptop OEM should just do a request...gaming APU, 12-16MB Vcache...95W...anything... EDIT: Ah...the rumor mill is giving us Strix Point and Strix Halo now...well for 2H 2024 that is ...40CU should allow for 1080p laptop gaming at 4x-5x PS5 prices.
https://forums.guru3d.com/data/avatars/m/248/248627.jpg
I'm actually surprised they don't just make a different I/O die that can handle one cpu ccd as well as 1-2 gpu dies used on the current 7900xtx (each die alone has i think 1024 shaders) they could use all their currently used silicon instead of making specialized silicon other than having a different I/O die interconnect. Alternatively i don't see why they don't use added cache as it has proven to be quite effective so far in alleviating bandwidth constraints.
https://forums.guru3d.com/data/avatars/m/196/196284.jpg
heffeque:

I'm sure too much HBM would be necessary to be able to use it as "dedicated" and actually make a difference.
8GB of HBM would be just fine. We're not trying to feed the entire system. Just the iGPU.
https://forums.guru3d.com/data/avatars/m/50/50906.jpg
sykozis:

8GB of HBM would be just fine. We're not trying to feed the entire system. Just the iGPU.
Why not just buy a discrete GPU at that point? 8GB of HBM doesn't seem like it would come "cheap", that's for sure.
https://forums.guru3d.com/data/avatars/m/202/202673.jpg
There's today's MLiD rumor video, according to Tom, with 50%+ confidence, AMD will launch premium mobile APU's in 12-15 months topping out at - 40CU/(I assume 2,560 SP) RDNA3.5 GPU CHIP - Zen4/5 16 Core with chiplets, - 64MB L3, - 32MB 3D Vcache, - 256bit RAM 270GB/s... - drawing 120W. Spec table at 17m30.
https://forums.guru3d.com/data/avatars/m/50/50906.jpg
Texter:

mobile APU - drawing 120W
Wat? 😕
https://forums.guru3d.com/data/avatars/m/202/202673.jpg
heffeque:

Wat? 😕
It's a fully fledged GPU with two CPU chiplets, supposedly. SKUs range between 25W-120W.
https://forums.guru3d.com/data/avatars/m/50/50906.jpg
But why call it "mobile APU"? 120W is a desktop APU put on a desktop replacement laptop, not a mobile APU. What's keeping a 300W APU to be called "mobile APU"? It makes no sense.
https://forums.guru3d.com/data/avatars/m/202/202673.jpg
That's AMD's own traditional APU distinction. Mobile and Desktop processors. Laptop parts get M suffix graphics. Allows them to avoid confusion and launch different feature sets for different purposes, I guess.
https://forums.guru3d.com/data/avatars/m/268/268248.jpg
Am I the only one that thinks , dayum this apu does really well considering is using laptop ddr5? ! If you compare it to gddr .... And how fast dedicated GPUs drop when they have to use system memory this is impressive! 😛
https://forums.guru3d.com/data/avatars/m/165/165326.jpg
I love this APUs , currently i have an ASRock x300 mini with a 5700G and 32gb of ddr4 4000Mhz. I'm looking forward to the next big jump on performance , that's going to happen when they include infinity cache and ddr5 SO Dimm speeds go higher , faster and lower latency on the sff mobos ( eg. ddr5 SO Dimm 6800~7000Mhz ) , that will be my dream htpc 😀 , 1080p gaming machine.