Wednesday, August 2nd 2023
AMD Confirms New "Enthusiast-class" Radeon 7000-series Graphics Cards This Quarter
AMD CEO Dr Lisa Su, in her Q2-2023 Financial Results call, confirmed that the company will launch new "enthusiast-class" gaming graphics cards within Q3-2023 (any time before October). "In gaming graphics, we expanded our Radeon 7000 GPU series in the second quarter with the launch of our mainstream RX 7600 cards for 1080p gaming. We are on track to further expand our RDNA 3 GPU offerings with the launch of new, enthusiast-class Radeon 7000 series cards in the third quarter," she stated.
There are two distinct possibilities of what "enthusiast class" entails. The first and most obvious one, could be the introduction of the RX 7800 series, including the RX 7800 XT, which is expected to closely resemble the limited-edition RX 7900 GRE by the specs; but a less talked-about possibility could even be the RX 7950 series. In its testing, the RX 7900 GRE was found to offer raster 3D performance comparable to the previous-generation RX 6950 XT although with better ray tracing performance on account of improved Ray Accelerators, which would put it behind the GeForce RTX 4070 Ti that AMD is trying to compete with. This should mean that for AMD to have a compelling "RX 7800 XT" product, it should perform faster than the RX 7900 GRE (possible through higher clock speeds or a few more CU).
The Radeon RX 7950 series is an exercise at significantly shoring up performance over the RX 7900 series, by increasing clock speeds and power limits. AMD is probably hoping for the RX 7950 XTX to take a swing at the performance crown held by the RTX 4090, while the RX 7950 XT could get a little closer to the performance of the RTX 4080. The current RX 7900 XT already beats the RTX 4070 Ti.
The announcement could also be a hint at the likelihood of mobile versions of the RX 7900 series, given that AMD has already developed the mobile-friendly package that's found powering the desktop RX 7900 GRE. This package is physically smaller than the regular "Navi 31," has lower Z-height, and is hence optimized for notebooks. Its lower pin-count could indicate a narrower 256-bit wide GDDR6 memory bus, and fewer power pins to go with the lower power-limits.
Sources:
AMD Investor Relations, VideoCardz
There are two distinct possibilities of what "enthusiast class" entails. The first and most obvious one, could be the introduction of the RX 7800 series, including the RX 7800 XT, which is expected to closely resemble the limited-edition RX 7900 GRE by the specs; but a less talked-about possibility could even be the RX 7950 series. In its testing, the RX 7900 GRE was found to offer raster 3D performance comparable to the previous-generation RX 6950 XT although with better ray tracing performance on account of improved Ray Accelerators, which would put it behind the GeForce RTX 4070 Ti that AMD is trying to compete with. This should mean that for AMD to have a compelling "RX 7800 XT" product, it should perform faster than the RX 7900 GRE (possible through higher clock speeds or a few more CU).
The announcement could also be a hint at the likelihood of mobile versions of the RX 7900 series, given that AMD has already developed the mobile-friendly package that's found powering the desktop RX 7900 GRE. This package is physically smaller than the regular "Navi 31," has lower Z-height, and is hence optimized for notebooks. Its lower pin-count could indicate a narrower 256-bit wide GDDR6 memory bus, and fewer power pins to go with the lower power-limits.
102 Comments on AMD Confirms New "Enthusiast-class" Radeon 7000-series Graphics Cards This Quarter
They would have better yields and lower development costs, as they would have to develop only one "small" chip that would be scaled from low-end to high-end just by putting together 1, 2, 3, 4 of these base GPU/GDC, as already happens in ryzen and EPYC. The question is, If AMD could pull the magic out of the hat and make it work flawlessly (for gaming)...
The 6000 series of GPU is a mono chipset and if you take into consider that the GPU is getting close to 3 years old, it is still a viable option because of it's performance/value over the 7000 series.
IMHO. The reasons why they are going this chiplet route is.
1. They get a better yield on the Silicon Wafers, there by getting more components to make more GPU's over the 6000 series.
2. They can get a better yield on partial defects that are on the edges of the Silicon Wafers. It is possible to use some of them for the lower value video cards.
Oh they are making money off of this. Make no mistake they are making more profit on this over a mono chiplet, but it is at the expense of overall performance. Latency will always be there. It just depends on how much money AMD wants to put into their Video card Division and their willingness of take on Ngreedia.
Judging from current and past business practices, it looks like they enjoy... being 2nd best.
In theory, it could be, assuming the chip to chip latency would be low enough, which apparently is one of the big hurdles today for something like that when it comes to GPUs, from my understanding of it.
I guess this company is hoping to win that kind of business in the future.
www.techpowerup.com/311529/silicon-box-opens-ususd-2-billion-advanced-semiconductor-assembly-plant-in-singaporeLong term yes, but not today, due to the the points above. But it does really seem to be the way a lof of companies are heading, as it's simply not viable to make massive chips with low yields.
A large part of it will depends on AMD's (or whoever's) partners as well, as the chip packaging companies need to be able to deliver flawless chips on their end as well, which I believe isn't always the case today, as this is still relatively new technology at the level it's being done these days and even more so as it gets more complex.
We also seem to need better chiplet interconnects that can handles every growing bandwidths, without adding latency or some other issues.
We'll most like end up with a combination of this and 3D stacking, as long as the thermals can be controlled when 3D stacking is used.It's a step by step process, this was clearly a step to try a lot of things, that didn't quite pan out as planned, so back to the drawing board.
It took AMD a few generations with Ryzen as well to get it to where it is today and what it pressumably will be in the near future.
It's by no means magic and as pointed out above, a lot of it will depend on their partners to deliver packaging solutions that can handle the negatives from doing this well enough, the key one as you point out being latency.If it's so bad, why is Intel heading in the same direction as AMD?
The food bundle also includes 5 suvlaki pita
They could be launching a RX7950XTX with e.g. stacked LLC on top of the MCDs, but unless they're capable of significantly increasing the clocks on the GCD I really doubt that happening.
videocardz.com/newz/powercolor-leaks-radeon-rx-7800-xt-red-devil-navi-32-with-3840-cores-and-16gb-confirmed
Instead, we got N33 with dual-issue FP32 and the exact same amount of execution units as N23 while being >10% smaller, on a process change that shouldn't bring any area advantage.
So did the dual-issue FP32 bring massive a performance increase? Mostly no, because just a few instructions can use it so optimizations need to be hand-written and replace the game's shaders through drivers.
Did it come at some area cost that would have been better spent elsewhere? No really.
AMD returns to the RX 480/580/590 | RX 5700 XT type of releases with RDNA 4 o_O
videocardz.com/newz/amd-rumored-to-be-skipping-high-end-radeon-rx-8000-rdna4-gpu-series
We're also approaching a performance ceiling with there not being anything above 4k 144hz to realistically drive, not that current cards are able to do that consistently now but the point stands. A navi 43 being about the performance of navi 32 would be fine
The halo product is always extremely important.
Navi 31 can't drive some games at 4K@144, its ray-tracing performance is lackluster. AMD needs to do something to make more performant cards.
Or alternatively this could be 10% slower than the RTX 4060ti, but cost $330 and come with 16GB of vram with maybe a 8GB version existing as well and costing something like $300.
The 7800XT needs to be close in performance to the RTX 4070, maybe up to 15% slower, but come in with 16GB of vram and cost $500, if it can also beat the 4060ti 16GB edition by 20% margins all at the same $500 price it will be one of the best value cards.
If I was AMD these would be the cards and pricing:
RX 7500 at $170, 15% slower than the RX 7600, but also come with 8GB of vram and cost no more than $170.
RX 7600 at $250msrp, it's about 3% slower than the 4060 in rasterization at 1080p and 1440p, but at $250msrp it makes a lot more sense over the 4060 and is actually good value!
RX 7700 12GB at $300, 7-10% faster than the 4060 in rasterization. A decent 7-10% uplift over the 4060 and with 4GB more vram it's going to make a lot of sense and actually be good value, better value than the current RX 6700XT.
RX 7700XT 16GB at $370, 5% faster than the 4060ti, but comes with 16GB of vram. Slightly cheaper, slightly faster and with more vram its going to be the overall much better choice over the 4060ti.
RX 7800XT 16GB at $500, can be 10-13% or so slower than the 4070, but at $100 cheaper and with more vram it's going to be the best value card this generation.
RX 7900GRE 16GB at $600, generally 10% faster than the 4070 and with more vram makes it the better choice at $600.
RX 7900XT at $750 is already very good option and offers good value for once this generation, but it needs to become $750msrp, so that most of the models are sold at this price, right now only a few models sell at this price and they generally go back up after a while.
RX 7900XTX at $900msrp would actually be a solid high-end purchase and make wayyy more sense than the over expensive $1200 RTX 4080. At $300 dollars cheaper and generally 3-4% faster than the 4080 its going to be even better value and a much better choice.