this irresponsible amounts of mediocrity in new gpu
Well, it's more "an incomplete GPU never designed for use as a standalone solution", turned into a standalone solution because of greed/market desperation.
Knowing now what it really is, it's explains the design choices that seemed odd and unnecessarily dumb for a desktop GPU.
What it doesn't excuse is the price. The MSRP of $199 is insulting and inappropriate, even in the current market. $350 street prices are so ridiculous that you're genuinely better off getting a $175 GTX 970 on ebay
I agree, what I meant and what I was answering to is that given the similar number of compute units the ryzen 6000 APUs should perform quite well and give quite the uplift compared to previous generations.
I don't think there's any info on die size yet for Ryzen 6000, but going by 5000 series APUs which had a 180mm2 die and 4000 series which used 156mm2, and doing some napkin math with the Ryzen chiplet size of ~75mm2 and navi24 size of 100mm2 we'll be looking at pretty damn competent APUs. We can also see that any Ryzen 6000 laptop that ships with a navi24 dGPU will be a waste of silicon and power because it won't offer that much more than what the APU already had (unless we see the return of some crossfire scheme)
The APU will be power-limited to a cTDP of 28-45W most likely, and sharing that budget with the CPU cores and memory controller. So even if it has the full 12x RDNA2 CUs, it will be clocking pretty conservatively given that the graphics portion of the chip will rarely have an opportunity to use more than 50W under boost and closer to 30W sustained.
I wouldn't call Navi24 a waste of silicon. In a laptop it's designed to just be extra RDNA2 CUs that the APU has access to - with additional VRAM and additional cooling hardware because it's isolated silicon. It's highly unlikely that the Navi24 silicon will be boosting to 2.8GHz and is more likely to run at somewhere between maybe 1600MHz and 2200MHz if I had to guess. If it's being paired with 28-45W APUs, the extra Navi24 die will also likely be tuned to consume a similar TDP, perhaps in the region of 35-65W.
There will be a kind of crossfire scheme coming back into play for these, I can't remember which AMD presentation I saw it in - either the RDNA2 slides or Ryzen 6000 slides. It was tied in with a revamp of the dynamic power juggling systems that are only possible with an APU and GPU from the same generation and where AMD controls the entire platform (so APU, dGPU, chipset, motherboard implementation).
Nobody outside AMD has seen it active yet, but that's what AMD's goal is with their active-bridge patent that uses infinity cache to string multiple pieces of RDNA2 silicon together using a narrower bus without huge penalties. I have low expectations of the technology but remain hopeful that it's not the same driver/compatibility/performance disappointment that SLI/crossfire were; This multi-die approach has worked really well for AMD with Ryzen and Epyc. Hopefully they can work the same magic with their GPUs.
Getting near to RX 570 performance in an APU would be quite impressive, but judging by how far they have to push navi24's clocks to achieve this, it seems very unlikely. With fast memory, I think the 5700G is somewhere between the 550 and 560, so with much lower clocks to fit the power envelope, I'm not even sure the APU will be much faster.
I don't think the 12CU in the APU will ever have the power budget to stretch their legs but if we can get even close to a regular GTX 1650 or even beat a 1050Ti that's a pretty significant step up from what we currently have with Vega8.