Quote from: Bandwidth on September 24, 2020, 02:10:03
@Wiwid:

This is just a wild guess but I'm guessing Van Gogh will be released between Q1-Q2 next year. AFAIK, Surface Laptop 4 and Surface Pro 8 are being delayed  (according to rumours) until then aswell.

I had high hopes for Van Gogh, but if it's being paired with LPDDR4x ... is there any point? Sure at lower resolutions it'll be faster than Tiger Lake but as soon as you'll get to higher resolutions it'll bottleneck real fast. I feel the same Rembrandt.

Does it matter having a RDNA2 based 4 TFLOP gpu apu (similar PS4 Pro perf.) if it's gonna be stuck with 100 GB/s memory vs the 200 GB/s on PS4 Pro?

It's going to be bus width starved :(

Are you for real? Did you forget that the 6800xt is able to compete with the 3080 despite only using regular GDDR6 with a total bandwidth of 512GB/s vs the 3080 760GB/s, if AMD did include infinity cache in their RDNA2 APU then it will be the fastest iGPU on the planet period, nothing Intel plans to release in the near future can touch it.

Quote from: Bandwidth on September 24, 2020, 01:57:11
Does any know how much the max theoretical bandwidth of DDR5 / LPDDR5 is compared to DDR4 / LPDDR4x ?

Googling shows stuff like for per channel or module, instead of like what the total effective bus rate would be in a laptop with 2x16gb of dual channel memory.

I'm guessing it's something like around ~50 GB/s vs ~100 GB/s.

Still kind of disappointed, when you realize the 7 year old PS4 was doing 179 GB/s.

What I find impressive is the Intel-AMD collaboration, which led to those Kaby Lake G with RX Vega M GL graphics. They were doing, 179 GB/s years ago too.

And that extra bandwidth makes a difference, if you compare the benchmarks (at higher resolutions) on this site of Vega 8 (Ryzen 4000 Renoir) vs RX Vega M GL, despite both being rated at similar ~2 TFLOPs,  the Kaby Lake G chip is pulling twice the fps at 1080p and above resolutions.

AMD needs to go back to designing HBM APU's. As long as we're stuck on this DDRx and LPDDRXx standard we will forever be stuck with lack luster APU's.

AMD were so close, if they just had more design wins and kept making those HBM APU's...

HBM memories are too expensive for intergrated solutions, and don't compare PS4's GDDR5 with LPDDR5, they have a completely different purpose

Quote from: JayN on September 23, 2020, 23:27:26
Intel says there are 100 TGL laptop designs coming before end of 2020.

TGL has lpddr5 support already built in.

TGL already has  pcie4 is built in.

An 8 core TGL-H has already been confirmed.

TGL laptop chips already have a winning GPU, integrated WIFI6, Intigrated Thunderbolt 4, av1 decode, Optane support.

Looks to me like AMD is in catch-up mode.

AMD is in catch-up mode? Haha, really funny. First Intel has to deliver something. Then we can talk about who is catching up. Which in fact will still be Intel.

Intel can say a lot of things. In recent years it turned out to be just hot air. And that won't change anytime soon.

Tiger Lake has no winning GPU. Only 4-core Tiger Lake has a "strong" GPU. But I doubt it can come close to Van Gogh's RDNA 2 based GPU. 8-core Tiger Lake will have only one third of the EUs of the 4-core version. That GPU won't have any chance even against the old Vega based GPU of Cezanne.

And some features you mention are pointless. The market doesn't care about things like Thunderbolt or Optane. Everything else will have AMD in their upcoming mobile processors too.

@Wiwid:

This is just a wild guess but I'm guessing Van Gogh will be released between Q1-Q2 next year. AFAIK, Surface Laptop 4 and Surface Pro 8 are being delayed  (according to rumours) until then aswell.

I had high hopes for Van Gogh, but if it's being paired with LPDDR4x ... is there any point? Sure at lower resolutions it'll be faster than Tiger Lake but as soon as you'll get to higher resolutions it'll bottleneck real fast. I feel the same Rembrandt.

Does it matter having a RDNA2 based 4 TFLOP gpu apu (similar PS4 Pro perf.) if it's gonna be stuck with 100 GB/s memory vs the 200 GB/s on PS4 Pro?

It's going to be bus width starved :(

Does any know how much the max theoretical bandwidth of DDR5 / LPDDR5 is compared to DDR4 / LPDDR4x ?

Googling shows stuff like for per channel or module, instead of like what the total effective bus rate would be in a laptop with 2x16gb of dual channel memory.

I'm guessing it's something like around ~50 GB/s vs ~100 GB/s.

Still kind of disappointed, when you realize the 7 year old PS4 was doing 179 GB/s.

What I find impressive is the Intel-AMD collaboration, which led to those Kaby Lake G with RX Vega M GL graphics. They were doing, 179 GB/s years ago too.

And that extra bandwidth makes a difference, if you compare the benchmarks (at higher resolutions) on this site of Vega 8 (Ryzen 4000 Renoir) vs RX Vega M GL, despite both being rated at similar ~2 TFLOPs,  the Kaby Lake G chip is pulling twice the fps at 1080p and above resolutions.

AMD needs to go back to designing HBM APU's. As long as we're stuck on this DDRx and LPDDRXx standard we will forever be stuck with lack luster APU's.

AMD were so close, if they just had more design wins and kept making those HBM APU's...

Intel says there are 100 TGL laptop designs coming before end of 2020.

TGL has lpddr5 support already built in.

TGL already has  pcie4 is built in.

An 8 core TGL-H has already been confirmed.

TGL laptop chips already have a winning GPU, integrated WIFI6, Intigrated Thunderbolt 4, av1 decode, Optane support.

Looks to me like AMD is in catch-up mode.

I think Van Gogh is Athlon 4000G to be released Q1 2021 to test RDNA 2 on APUs before they start put it everywhere.

And what about Van Gogh?