AMD reportedly EOLs its mainstream RDNA 2-based GPU.
AMD Reportedly Discontinues Navi 33: No More Radeon RX 6600-Series : Read more
AMD Reportedly Discontinues Navi 33: No More Radeon RX 6600-Series : Read more
The scheduled forum maintenance has now been completed. If you spot any issues, please report them here in this thread. Thank you!
Since 22nm, cost-per-unit-die-area has been going up rather than down, and for the last few nodes cost-per-unit-area has gone up VERY sharply. Cost-per-transistor has effectively plateaued to slightly increased (cost/gate has slightly reduced, but due to multiple gates per transistor becoming the norm rather than the exception for high speed transistors such as those in CVPUs and GPUs this is blunted). New processes offer improvements in energy efficiency, maximum overall performance (more transistors per die when you are reticule limited), and clock speeds, and (when mature enough to drop defect areal rates to parity with past nodes) offer greater die throughput via more dies per wafer in scenarios where wafer rate is the limit and demand exceeds supply - such as now.Meanwhile, has a smaller die size (204 mm2 vs 237 mm2) and is made on TSMC's N6 process technology (as opposed to N7 in the case of Navi 23), so it may well be cheaper to produce.
N6 is just refined N7, though, and while we don't have detailed information on precisely how much TSMC charges for the various nodes, N6 and N7 should be pretty close. N5 and N7 meanwhile are a big gap in cost. This was the whole reason for GPU chiplets in the higher performance RDNA 3 range.Since 22nm, cost-per-unit-die-area has been going up rather than down, and for the last few nodes cost-per-unit-area has gone up VERY sharply. Cost-per-transistor has effectively plateaued to slightly increased (cost/gate has slightly reduced, but due to multiple gates per transistor becoming the norm rather than the exception for high speed transistors such as those in CVPUs and GPUs this is blunted). New processes offer improvements in energy efficiency, maximum overall performance (more transistors per die when you are reticule limited), and clock speeds, and (when mature enough to drop defect areal rates to parity with past nodes) offer greater die throughput via more dies per wafer in scenarios where wafer rate is the limit and demand exceeds supply - such as now.
But the idea of "newer node means cheaper chips" has been dead for quite some time.
I don't think they plan on making a RX 7500.so inventories are probably high and can keep sales of cheaper cards until the RX 7500 arrives.
Eventually, APU's with larger iGPU's will take the price & performance class for the RX 7500 if you're in the market for level of performance.because they did not want to sell cards that have price lower than $200.
Even the 6500XT and the 6400 were "emergency" products that came out because there were no cheap GPUs on the market - these were originally mobile chips bolted on a PCB, and were decried for being feature-less and underperforming on PCIe 3.0. I mean, in many cases its predecessor the RX5500 was actually faster !I don't think they plan on making a RX 7500.
AMD Radeon Chief confirms Radeon RX 7000 (RDNA3) portfolio “is now complete”
Eventually, APU's with larger iGPU's will take the price & performance class for the RX 7500 if you're in the market for level of performance.